Phosphines-Nitrogen-Phosphines Chelated CoCl2 Exhibits Potent Antitumor Activity in Pancreatic Cancer

Phosphines-Nitrogen-Phosphines Chelated CoCl2 Exhibits Potent Antitumor Activity in Pancreatic Cancer | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Article Phosphines-Nitrogen-Phosphines Chelated CoCl2 Exhibits Potent Antitumor Activity in Pancreatic Cancer Xuefei Guo, Yang Zhao, Xianle Rong, Xingyu Chen, Xiao Wang, Tian Liu, and 6 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7085332/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Pancreatic cancer is projected to become the second leading cause of cancer-related deaths globally by 2030, yet effective therapeutic options remain limited. Within the pancreatic cancer tumor microenvironment (TME), tumor-associated macrophages (TAMs) secrete interleukin-1 beta (IL-1β), promoting cancer progression while suppressing type I interferons (IFN-I), which is critical for tumor killing. Utilizing the convolutional neural network (CNN)-based DLINP model developed in our laboratory, we identified Co68—an effective metal catalyst featuring a Phosphines-Nitrogen-Phosphines (PNP)-chelated CoCl₂ complex—as a promising candidate to modulate innate immune responses. In animal models of pancreatic cancer, Co68 demonstrated superior antitumor efficacy compared to the STING agonist DMXAA and showed enhanced therapeutic effects when combined with PD-1 blockade. Single-cell RNA sequencing (scRNA-seq) revealed that Co68 reprogrammed TAMs to express interferon-stimulated genes (ISGs), attenuated pro-inflammatory cytokine secretion, and disrupted the IL-1β-PGE2 feedback loop, thereby facilitating the recruitment of NK and cytotoxic CD8 + T cells into the TME. Mechanistically, Co68 activated the IFN-I signaling pathway through the TLR4-TRIF-IFN-I axis and inhibited inflammation via the TLR4-SYK-STAT1 pathway. Collectively, these findings highlight the therapeutic potential of Co68, derived from PNP-pincer chemistry, to reshape immune dynamics within the pancreatic cancer TME, positioning it as a promising candidate for innovative immunotherapy strategies. Biological sciences/Immunology/Innate immunity/Pattern recognition receptors/Toll-like receptors Biological sciences/Cancer/Cancer microenvironment Biological sciences/Drug discovery/Drug screening/Virtual screening Biological sciences/Drug discovery/Target identification Pancreatic ductal adenocarcinoma (PDAC) Inflammatory response Innate immunity response Convolutional neural network Phosphines-Nitrogen-Phosphines (PNP)-pincer ligands TLR4 STAT1 NF-κB Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Highlights Co68 features a classic PNP-pincer ligand acting as an immunomodifier identified by CNN model DLINP Co68 reprograms tumor-associated macrophages within the pancreatic cancer microenvironment Co68 induces type I interferon via a novel binding mode with TLR4 Co68 demonstrates significant anti-inflammatory effects via the TLR4-SYK-STAT1 axis In brief A multiscale drug discovery platform integrating the convolutional neural network (CNN) model DLINP, multi-omics analysis, and animal models identified Co68—a compound featuring a classic Phosphines-Nitrogen-Phosphines (PNP) pincer ligand—as an effective agent with anti-inflammatory and anti-pancreatic cancer properties. Introduction Pancreatic cancer is often referred to as the "king of cancers" due to its insidious onset, poor prognosis, and high resistance to treatment 1 – 3 . Global statistics indicate that approximately 500,000 new cases of pancreatic cancer are diagnosed each year, with a dismal five-year survival rate of approximately 10% 4–6 . This malignancy is characterized by a lack of early-stage symptoms, which often results in diagnosis at an advanced stage when surgical intervention is no longer an option 7 . The inherent biological traits of pancreatic cancer, such as rapid tumor cell proliferation and extensive metastatic potential, further contribute to its aggressive nature 8 . Projections suggest that pancreatic cancer may become the second leading cause of cancer-related deaths by 2030 9 . Currently, treatment options are limited to surgery, chemotherapy, radiotherapy, and targeted therapies, including those targeting the epidermal growth factor receptor (EGFR), angiogenesis, and hypoxia 10 . However, these therapeutic strategies have had limited success in improving long-term survival, and substantial challenges remain in effectively treating this disease. The limited progress in pancreatic cancer treatment has intensified the focus on immunotherapy, primarily driven by the challenges posed by the complex and immunosuppressive tumor microenvironment (TME) inherent to this disease 11 – 13 . A hallmark of the TME is the intricate interplay among tumor-associated fibroblasts, immune cells, and tumor cells, which collectively facilitate immune evasion and complicate therapeutic approaches 14 , 15 . Tumor-associated macrophages (TAMs), the most abundant immune cell population within this microenvironment, play a central role in establishing a pro-inflammatory environment that promotes tumor progression 16 – 19 . TAMs secrete interleukin-1 beta (IL-1β), a key cytokine that drives pancreatic cancer initiation, progression, and metastasis 20 – 22 . Fueled by the positive feedback loop between IL-1β and prostaglandin E2 (PGE2), chronic inflammation has emerged as a critical driver of tumor development 20 , 23 . PGE2, secreted by tumor cells, enhances IL-1β secretion from TAMs, creating a vicious cycle that accelerates tumor progression and immune evasion, ultimately limiting the effectiveness of immune responses 24 – 27 . The reciprocal relationship between TAMs and pancreatic cancer cells, mediated through the IL-1β-PGE2 axis, underscores the pivotal role of TAMs in the progression of pancreatic cancer 18 , 20 . Targeting IL-1β and PGE2 presents a promising strategy to disrupt this inflammatory axis, offering novel therapeutic avenues to combat this aggressive and challenging cancer. Type I interferons (IFN-I), primarily interferon-alpha (IFN-α) and interferon-beta (IFN-β), are critical players in promoting anti-tumor immunity through multiple mechanisms, positioning them as valuable therapeutic agents in cancer treatment 28 – 30 . IFN-I enhances immune responses by activating natural killer (NK) cells and cytotoxic CD8 + T lymphocytes, which are essential for recognizing and eliminating tumor cells 31 , 32 . By boosting the cytotoxic activities of these immune cells, IFN-I modifies the TME and suppresses tumor growth 33 . Additionally, IFN-I improves the presentation of tumor-associated antigens by antigen-presenting cells, such as dendritic cells, thus promoting the development of adaptive immune responses 34 . Recent studies have further demonstrated IFN-I's ability to inhibit tumor cell proliferation and induce apoptosis, solidifying its potential as an effective cancer therapy 35 . Through the activation of specific signaling pathways, IFN-I can suppress tumor cell proliferation and initiate apoptotic mechanisms, making it a promising candidate for cancer treatment 36 , 37 . However, the TME is filled with factors that inhibit IFN-I signaling, including tumor-derived PGE2, which not only exacerbates IL-1β secretion in TAMs but also directly suppresses IFN-I signaling in TAMs 38 – 40 . This suppression reduces the expression of interferon-stimulated genes (ISGs) and diminishes the overall anti-tumor immune response 41 . Given the immune-suppressive role of IL-1β-secreting TAMs and the critical importance of IFN-I signaling in cancer immunotherapy, there is an urgent need for novel therapeutic agents capable of modulating the balance between inflammation and IFN-I signaling within the pancreatic cancer TME. Phosphines-Nitrogen-Phosphines (PNP) pincer chemistry, a subfield of coordination chemistry, focuses on the development and application of PNP pincer ligands—tridentate ligands that feature a central nitrogen donor atom flanked by two phosphine groups 42 – 44 . These ligands are known for their stability and reactivity, particularly in catalysis, and have been studied extensively in transition metal chemistry 45 – 47 . Recent advances in PNP-pincer chemistry have incorporated N-heterocycles into the ligand framework, leveraging the combined electronic properties of phosphines and N-heterocycles to enhance metal-ligand interactions 48 – 50 . This has led to the development of highly efficient catalysts for hydrogenation, C–H activation, and cross-coupling reactions 51 – 53 . Despite the growing interest in their catalytic potential, the biological activity of PNP-metal complexes remains largely unexplored, particularly in terms of their ability to modulate immune responses or serve as therapeutic agents. Machine learning (ML) techniques have revolutionized drug discovery by enabling the analysis of complex biological data and the prediction of drug efficacy and safety 54 – 56 . Advanced ML models, including deep neural networks, convolutional neural networks (CNNs), and support vector machines (SVMs), excel at identifying patterns within large-scale biological datasets, accelerating the drug discovery process 57 – 60 . In particular, deep learning frameworks such as TensorFlow and Keras have facilitated the development of sophisticated models that optimize various aspects of drug discovery, including molecular property prediction, de novo design, and retrosynthetic analysis 61 – 65 . These advancements have significantly enhanced the speed and efficiency of identifying novel therapeutic agents, including drug repurposing applications 66 . In this study, we developed a convolutional neural network model, DLINP, to screen small molecules that modulate both the IL-1β/PGE2 feedback loop and IFN-I signaling within the pancreatic cancer TME 20 , 38 . Our findings led to the unexpected identification of Co68, a PNP-pincer ligand complexed with CoCl₂, as a potent immune-modulatory agent. Co68 enhanced the IFN-I response and reprogrammed TAMs and tumor-associated neutrophils (TANs) to promote immune activation and reduce inflammation. Through single-cell RNA sequencing (scRNA-seq) and functional assays, we demonstrated that Co68 activates the TLR4-TRIF-IFN-I axis, which is essential for driving anti-tumor immunity. Moreover, we uncovered a novel TLR4-SYK-STAT1 signaling axis underlying the anti-inflammatory effects of Co68, distinguishing it from other immune-modulatory agents, such as DMXAA. Notably, Co68 is the first PNP-chelate identified to convert "cold" tumors into "hot" tumors by targeting TLR4, which has significant implications for overcoming immune evasion. This therapeutic strategy holds promise not only for pancreatic cancer but also for addressing immune evasion mechanisms in other malignancies. Our study demonstrates that Co68, a Phosphines-Nitrogen-Phosphines (PNP) chelated CoCl₂ compound, could serve as a novel therapeutic agent, offering new opportunities for cancer treatment by reshaping the immune landscape within the TME. Results Develop the DLINP model for screening potential immunomodulatory agents Excessive inflammation is detrimental to health, which could result in damage to multiple organs and promote tumor development 67 . For example, COVID-19 is characterized by a "cytokine storm," with elevated levels of pro-inflammatory cytokines such as IL-1β, IL-6, and TNF while concurrently suppressing the production of type I and III interferons (Figure S1 A) 68 – 70 . These findings suggest that targeting type I interferon (IFN-I) pathways may offer significant prognostic and therapeutic potential. Lipopolysaccharide (LPS) treatment in the RAW 264.7 macrophage cell line induced a robust upregulation of genes associated with both inflammatory and IFN-I responses, including Il1a , Il1b , Il6 , Ptgs2 , Ifnb1 , Ifit2 , Mx1 , and Oasl1 (Figures S1 B and S1C). These observations motivated us to identify compounds that not only promote IFN-I production but also suppress inflammatory responses. The approach workflow is summarized in Fig. 1 A. To achieve this, we developed the DLINP (Deep Learning for Innate immunity modulatory Potential) model, a convolutional neural network (CNN) designed to screen compounds from the CAS database, which contains over 127 million unique chemical entities. The screening revealed several compounds that both promoted IFN-I expression and suppressed inflammatory cytokine production. The top candidate compounds were further validated in vitro and in vivo, assessing their potential to inhibit NF-κB-mediated inflammatory pathways and enhance IRF3-mediated immune responses. "Co68" emerged as the most efficacious compound after comprehensive toxicological and functional evaluations. We subsequently investigated its molecular target and signaling pathways through which it activates IFN-I production. The antitumor efficacy of Co68 was further assessed in the Pan02 pancreatic cancer mouse model, employing single-cell RNA sequencing (scRNA-seq). Ultimately, we confirmed the broad anti-inflammatory signaling activity of Co68, positioning it as a promising candidate for therapeutic application in inflammatory diseases and cancers. To construct the DLINP model, we first extracted immune-modulating small molecules and their corresponding gene expression profiles from the L1000 and ChEMBL databases. As illustrated in Fig. 1 B, SMILES descriptors of each compound were converted into a one-hot encoded matrix, representing the atomic composition and chemical bonding features as binary values (0 and 1). Gene expression data associated with each compound were similarly one-hot encoded, with upregulated genes assigned a value of 1 and downregulated genes a value of 0, resulting in a compound-specific gene expression encoding matrix. The DLINP CNN model, consisting of one input layer, four hidden layers, and one output layer, was trained to predict the immune-regulatory potential of these compounds. The model’s performance was evaluated using confusion matrix analysis, Receiver Operating Characteristic (ROC) curves, and the area under the curve (AUC) metrics. As shown in Figs. 1 C and 1 D, the DLINP model demonstrated robust predictive capability, achieving an AUC of 0.91 and a true positive rate (TPR) of 93.11%, validating its accuracy and effectiveness in predicting immune-modulatory properties. Encouraged by these results, we applied the trained DLINP model to screen small molecules from the CAS database. Among the identified candidates, Co68 was predicted to have a particularly potent immune-modulatory effect. Its chemical structure, depicted in Fig. 1 E, combines the classical PNP structure commonly found in catalytic chemistry with cobalt dichloride, with the synthesis reaction outlined in Figure S1 D. We first assessed the cytotoxicity of Co68 in primary macrophages and various cell lines, including the Bone Marrow-Derived Macrophages (BMDMs), peritoneal macrophages (PMs), RAW 264.7, and J774A.1 (Fig. 1 F and Figure S1 E). While Co68 at a high concentration of 500 µM caused cell death, lower concentrations (< 100 µM) showed minimal effects on cell viability. Co68 was then found to induce the expression of Ifnb1 , with peak levels observed at 200 µM in RAW 264.7, BMDMs, and PMs, and at 100 µM in J774A.1 cells (Figs. 1 G– 1 H and Figure S1 F–S1G). These findings validated the predictive accuracy and efficacy of the DLINP model in identifying compounds capable of modulating immune responses. Time-course analysis revealed that Co68 induced an expression peak of Ifnb1 at 3 hours in RAW 264.7, J774A.1, and PMs, while in BMDMs, the peak occurred at 5 hours (Figs. 1 I– 1 J and Figure S1 H–S1I). Based on these results, the optimal concentration of Co68 for in vitro experiments was determined to be 100 µM for a 3-hour treatment period. Additionally, Co68 significantly induced IFNB1 production in the human THP-1 macrophage cell line across different subtypes (Figure S1 J). In vivo, Co68 administration at 10 µM for 24 hours significantly increased the expression of Ifnb1 and interferon-stimulated genes (ISGs), including Isg15 and Ifit3 , in spleen, lung, and liver tissues, as well as in peripheral blood mononuclear cells (PBMCs) from mice (Fig. 1 K, Figure S1 K–S1L), but did not affect heart and kidney tissues. Notably, Co68 showed instability when dissolved in DMSO. As shown in Figure S1 M, the activation of Co68 in RAW 264.7 cells was significantly diminished 3 days post-dissolution. Consistent with previous studies demonstrating that Co²⁺ directly binds to TLR4 to induce metal anaphylaxis 71 , we also observed that CoCl₂, a precursor of Co68, induced the expression of Ifnb1 , Il1b , and Il6 at a higher concentration (500 µM) following a 24-hour treatment. However, CoCl₂ did not trigger similar effects at the lower concentration (50 µM) for the shorter period (3 hours) as Co68 did (Figure S1 N–S1P). We further compared the gene expression profiles induced by Co68, CoCl₂, and LPS to investigate the differences in activation patterns. As shown in Figs. 1 L– 1 N and Figure S1 Q–S1R, Co68 induced a robust expression of Ifnb1 and Ifna4 after just 3 hours of treatment, significantly surpassing both CoCl₂ and LPS in these marker genes. In contrast, the expression of Il1b , Il6 , and Il1a was markedly lower in the Co68 group than in the CoCl₂ and LPS groups, indicating that Co68 activates a distinct gene expression program compared to both CoCl₂ and LPS. This observation highlights the unique activation pattern of Co68, which contrasts with the more classical pro-inflammatory response driven by CoCl₂ and LPS. Collectively, these findings demonstrate the potential efficacy and safety of Co68 in modulating innate immune responses, thereby validating the high performance and precision of the DLINP model. Co68 induces type I interferon signaling and suppresses inflammatory responses To gain deeper insights into the effects of Co68 on macrophages, we conducted bulk RNA sequencing (RNA-seq) on RAW 264.7 cells treated with Co68 for 3 hours. As shown in Fig. 2 A, a total of 472 genes were upregulated, while 153 genes were downregulated in the Co68-treated group compared to the DMSO control. As expected, Co68 treatment significantly upregulated the expression of interferons, ISGs, and major histocompatibility complex (MHC) molecules. Notably, genes such as Ifna2 , Ifna4 , Ifnb1 , Oas1a , Rsad2 , Ifit3 , Cxcl10 , Isg15 , Isg20 , H2-D1 , H2-K1 , and H2-T23 were prominently upregulated (Figure S2A). We performed a protein-protein interaction (PPI) network analysis of these differentially expressed genes (DEGs) using the STING database to elucidate the molecular mechanisms underlying these changes. The most enriched clusters of upregulated DEGs in the PPI network following Co68 treatment were predominantly related to innate immune responses, regulation of type I interferon production, and negative regulation of viral genome replication (Figs. 2 B and S2B–S2C), suggesting that Co68 acts as a modulator of innate immune responses. Gene Ontology (GO) analysis revealed that the upregulated genes in the Co68-treated group were enriched in pathways associated with defense responses to viruses, innate immune responses, cellular responses to interferon-beta, type I interferon-mediated signaling, and antiviral innate immune responses (Fig. 2 C). In contrast, the downregulated genes were linked to processes such as the positive regulation of cell migration, negative regulation of dendritic cell differentiation, apoptotic processes, ubiquitin-dependent endocytosis, and cell proliferation (Figure S2D). Additionally, protein analysis demonstrated significant phosphorylation of TBK1 and IRF3, further supporting the activation of key signaling pathways involved in innate immune modulation (Figure S2E). To evaluate the anti-inflammatory potential of Co68 in macrophages, we quantified the expression of genes encoding key inflammatory cytokines and enzymes, including Il1b , Il6 , Nos2 , and Ptgs2 , which are crucial mediators of the inflammation observed in conditions such as sepsis and tumor progression 67 . As expected, Co68 significantly reduced the LPS-induced production of IL-1α , IL-1β , Il6 , Nos2 , Ptgs2 , Csf1 , Csf2 , and Csf3 in RAW 264.7 cells (Fig. 2 D and Figure S2F). Interestingly, Co68 did not suppress TNF production; rather, it synergistically enhanced TNF levels in the presence of LPS (Figure S2G). A similar synergistic effect was observed for other immune mediators, including Ifnb1 , Cxcl10 , and Isg15 (Figure S2G). Co68 also exhibited broad-spectrum anti-inflammatory effects, significantly suppressing the expression of Il1b induced by various viruses such as VSV, HSV, and EMCV in RAW 264.7 cells (Fig. 2 E). In vivo, Co68 markedly reduced sepsis-related mortality following LPS injection into the peritoneal cavity (Fig. 2 F), further supporting its anti-inflammatory potential. To further elucidate the molecular mechanisms underlying these effects, we performed bulk RNA-seq to analyze the anti-inflammatory phenotype and pathways modulated by Co68. As shown in the heatmap (Fig. 2 G), Co68 prominently induced the expression of genes associated with IFN-I signaling, while LPS treatment primarily upregulated genes involved in the inflammatory response. Notably, Co68 induced the upregulation of several histone-coding genes, such as H1f5 , H2bc4 , and H2bc18 , a pattern absent in the LPS-treated group, suggesting that Co68 plays a specific role in the regulation of gene expression. The Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis of downregulated differentially expressed genes (DEGs) in the LPS-Co68 group, compared to the LPS group, revealed the suppression of several critical pathways, including the NF-κB signaling pathway, inflammatory bowel disease, rheumatoid arthritis, and the Toll-like receptor and NOD-like receptor signaling pathways (Fig. 2 H). Gene Set Enrichment Analysis (GSEA) further confirmed these results, demonstrating significant enrichment of the INFLAMMATORY_RESPONSE pathway within the downregulated gene set (Figure S2H). Transcription factor enrichment analysis, conducted using the iRegulon plugin in Cytoscape (v3.10.2), identified a significant enrichment of the NF-κB family transcription factor RELA (P65) among the downregulated genes (Fig. 2 I). Consistent with this, Co68 effectively inhibited LPS-induced phosphorylation and degradation of IκB-α, along with the phosphorylation of P65 and the production of key downstream enzymes such as iNOS and Cox2 (Fig. 2 J and Figure S2I). These findings indicate that Co68 specifically targets and suppresses NF-κB-mediated inflammatory signaling. Collectively, these results demonstrate that Co68 not only robustly activates the type I interferon signaling pathway but also exerts broad anti-inflammatory effects, thereby positioning it as a promising therapeutic agent for the modulation of inflammation. Co68 demonstrates potent antitumor activity in the Pan02 model Given Co68’s ability to robustly activate IFN-I and its broad-spectrum anti-inflammatory properties in macrophages, we sought to investigate whether Co68 could suppress TAM-mediated inflammation and enhance antitumor immunity within the pancreatic cancer TME by promoting IFN-I signaling. To explore this hypothesis, we established a murine pancreatic cancer model using Pan02 cells. As shown in Figs. 3 A- 3 C and Figure S3A, Co68 significantly inhibited tumor growth. This antitumor effect was found to be dependent on IFN-I signaling, as evidenced by the abrogation of Co68’s antitumor activity in Pan02 tumors implanted in IFNAR1 knockout (KO) mice (Figs. 3 D- 3 F and Figure S3B). Consistent with previous studies 72 , Pan02 pancreatic tumors are classified as "cold tumors", showing limited sensitivity to PD-1 antibody treatment (Figs. 3 G, 3 H, and Figure S3C). However, we observed that the combination of recombinant PD-1 antibody and Co68 significantly inhibited tumor growth compared to either Co68 or PD-1 antibody treatment alone (Figs. 3 G- 3 I). In contrast to PD-1 antibody treatment, DMXAA, a potent agonist of mouse Stimulator of Interferon Genes (STING), has been shown to inhibit Pan02 tumor growth significantly 73 . As depicted in Figures S3D-S3F, we confirmed the potent antitumor effect of DMXAA in the Pan02 tumor model in C57BL/6J mice. When comparing the antitumor effects of Co68 and DMXAA, we observed that Co68 exhibited a more pronounced antitumor effect than DMXAA (Figs. 3 J and 3 K). This enhanced efficacy of Co68 may be attributed to its additional anti-inflammatory effects, which were absent in DMXAA treatment. Taken together, these results demonstrate that Co68 not only exhibits superior antitumor efficacy compared to DMXAA but also shows enhanced therapeutic potential when combined with PD-1 antibody treatment. Moreover, we observed that Co68 not only significantly inhibited the growth of Pan02 tumors but also demonstrated potent antitumor activity in the MC38 mouse colon tumor model (Figures S3G-S3I), indicating that Co68 may offer broad therapeutic potential across various tumor types. Flow cytometry analysis further confirmed these results, showing a marked increase in the proportion of CD8⁺ T cells and NK cells, alongside a concurrent decrease in CD4⁺ T cells in the Co68-treated group compared to the PBS control (Fig. 3 L) in the Pan02 tumor. To further investigate the antitumor mechanisms of Co68 and to compare its superior efficacy with DMXAA in the pancreatic cancer context, we performed single-cell RNA sequencing (scRNA-seq) on the Pan02 tumor model following treatment with Co68, DMXAA, and PBS. As shown in Figure S3J, unsupervised clustering using t-distributed Stochastic Neighbor Embedding (tSNE) of tumor cells from the PBS and Co68 treatment groups revealed 25 distinct cell subpopulations. Subsequent annotation of these clusters (Figure S3K), based on the expression of subpopulation-specific marker genes, identified various cell types within the tumor microenvironment, including cancer cells, cancer-associated fibroblasts (CAFs), TAMs, tumor-associated neutrophils (TANs), T cells, B cells, NK cells, dendritic cells (DCs), endothelial cells, and epithelial cells. Figure S3K also revealed that Co68 treatment led to significant alterations in the relative abundance of TAMs, TANs, T cells, and cancer cells compared to the PBS control group. To further investigate these observations, we performed sub-clustering and annotation of these cell populations, which enabled us to assess the impact of Co68 on gene expression and pathway alterations within each subpopulation. The gene markers used for cell annotation are shown in Figure S3L. Unsupervised clustering using t-SNE and subsequent annotation of cell populations revealed a tumor microenvironment predominantly enriched in CAFs, with TAMs being the most abundant immune cell population (Fig. 3 M), consistent with previous reports 15 . We then compared the relative abundance of various immune cell subpopulations between the Co68 and PBS treatment groups. As shown in Fig. 3 N, Co68 treatment led to a significant increase in the proportion of CD8⁺ T cells and NK cells compared to the PBS control group, consistent with the flow cytometry results presented above (Fig. 3 L). In contrast, the proportion of the Cancer Cell 1 subpopulation was notably reduced. Differential gene expression analysis of the CD8⁺ T cell population in the Co68 group revealed upregulation of genes involved in T cell activation, lymphocyte differentiation and proliferation, response to interferon-gamma, and cytokine-mediated signaling pathways (Fig. 3 O) compared to the PBS group. Collectively, these results suggest that Co68 treatment not only increases the abundance of CD8⁺ T cells within the pancreatic tumor microenvironment but also enhances their activation, thereby strengthening their antitumor ability. Co68 reprograms TAMs within the pancreatic tumor microenvironment Notably, the proportion of monocytes, TAMs, and the TANs1 subpopulation was higher in the Co68-treated group compared to the PBS control (Fig. 4 A), suggesting that these cell types play crucial roles in Co68-mediated modulation of the pancreatic tumor microenvironment. Given the established involvement of TAM-derived IL-1β and pancreatic cancer cell-derived PGE2 in the initiation, progression, and metastasis of pancreatic cancer 20 , 26 , 27 , 38 , we focused our subsequent analyses toward investigating the effects of Co68 on TAMs. Consistent with our hypothesis, gene set enrichment analysis revealed that Co68-treated TAMs exhibited significantly increased expression of genes associated with immune response activation, positive regulation of innate immune responses, cellular response to type I interferons, pattern recognition receptor signaling, and negative regulation of inflammatory responses (Fig. 4 B). In contrast, PBS-treated TAMs were enriched in genes related to cellular responses to interleukin-1, regulation of mononuclear cell proliferation, and cholesterol metabolic processes (Fig. 4 B). Pseudotime analysis identified the TAMs1 subpopulation as the primary target of Co68-mediated reprogramming (Figs. 4 C- 4 E). Differential gene expression and gene set enrichment analyses of the TAMs1, Mono, and TAMs2 subpopulations showed that TAMs1 exhibited significantly increased expression of genes involved in the positive regulation of T cell activation, antigen processing and presentation, and lymphocyte proliferation (Fig. 4 F). Based on the expression of specific marker genes, we renamed the TAMs1, Mono, and TAMs2 subpopulations as IL-1β TAMs, HSP TAMs, and S100A10 TAMs, respectively, with IL-1β TAMs representing the Co68-responsive subpopulation (Fig. 5 G). Further differential gene expression analysis of IL-1β TAMs revealed substantial upregulation of several PRRs and ISGs, including Clec2d, Ifih1, Ly96, Cd14, Ifitm3, Cxcl10, Isg15, Oasl1 , and Rsad2 , in the Co68-treated group compared to the PBS control (Fig. 4 H). Gene set enrichment analysis of these differentially expressed genes demonstrated that Co68-upregulated genes in the IL-1β TAMs subcluster were significantly enriched in pathways associated with the activation of innate immune responses, while downregulated genes were associated with processes such as DNA replication, chromosome segregation, and cholesterol biosynthesis (Fig. 4 I). Cell-cell communication analysis using the CellChat R package revealed significant interactions between the IL-1β TAM subpopulation and Cancer Cell 1, which showed a markedly reduced proportion in the Co68-treated group compared to the PBS control (Fig. 4 J, Figures S4A and S4B). These interactions were notably stronger than those observed between Cancer Cell 1 and the TAMs2 or Mono subpopulations (Figures S4C and S4D). Further analysis of intercellular communication pathways between the IL-1β TAMs in the Co68 and PBS groups (Fig. 4 K) revealed that Co68 treatment significantly enhanced IFN-I-mediated communication between IL-1β TAMs and CD4 + T cells, CD8 + T cells, and endothelial cells (Fig. 4 L). In contrast, PBS-treated IL-1β TAMs exhibited strong IL-1-mediated communication with CAFs, Cancer Cell 1, Cancer Cell 3, monocytes, and epithelial cells (Figs. 4 K and 4 M). Consistent with previous reports 74 , 75 , we also observed that TAMs communicate with other cell populations in the TME through the SPP1 signaling pathway (Figure S4E). Given the significant alterations in neutrophil subpopulation proportions in the Co68-treated group (Fig. 4 A) and the robust intercellular communication between IL-1β TAMs and neutrophils as identified by CellChat (Fig. 4 J), we further explored the effects of Co68 on neutrophil reprogramming within the pancreatic tumor microenvironment (Figure S4F). Co68 treatment significantly increased the proportion of TANs1 cells while decreasing the proportion of TANs2 cells compared to the PBS control (Fig. 4 N). Differential gene expression analysis of the three neutrophil subpopulations revealed that Co68 treatment upregulated the expression of several ISGs in the TANs1 subpopulation (Figure S4G), including Ifit1, Ifit2, Ifit3, Rsad2, Parp14, Isg20, Irf7, Ifi204 , and Ifi47 . Gene set enrichment analysis demonstrated that the TANs1 subpopulation, with upregulated ISG expression, was significantly enriched in pathways related to the activation of innate immune responses, PRR signaling, regulation of innate immune responses, and response to interferon-beta (Figure S4H). These findings align with previous studies highlighting the enhanced antitumor activity of neutrophils with high ISG expression 76 – 78 , suggesting that the induced ISG⁺ TANs1 may also play a crucial role in the antitumor effect of Co68 on pancreatic cancer. To further investigate the differential effects of Co68 and DMXAA at the single-cell level, we performed scRNA-seq analysis on Pan02 tumors treated with either compound. Unsupervised clustering and annotation of the resulting data revealed distinct differences in cell subpopulation ratios between the Co68 and DMXAA treatment groups (Figs. 4 O and S4I) and compared to the Co68 and PBS groups (Fig. 3 M). Subsequent subclustering of TAMs based on marker gene expression identified four distinct subpopulations: IL-1β TAMs, S100A10 TAMs, HSP TAMs, and MKI67 TAMs (Figs. 4 P and S4J). Differential gene expression analysis of the IL-1β TAM subpopulation (Fig. 4 Q) revealed that Co68 treatment, compared to DMXAA, significantly upregulated the expression of several IFN-I-related genes, including Il1rn, Oasl1, Rsad2, Tnf, Cxcl10 , and Isg15 (Figs. 4 R and S4L). In parallel, Co68 treatment downregulated pro-inflammatory genes such as Il1a, Il1b , and Ptgs2 (Fig. 4 R). Gene set enrichment analysis of the upregulated genes in Co68-treated IL-1β TAMs revealed significant enrichment in pathways associated with immune response activation, regulation of innate immune responses, and tumor necrosis factor production (Figure S4K). These results suggest that Co68, unlike DMXAA, not only elicits a stronger IFN-I response but also attenuates IL-1-driven inflammatory signaling. Collectively, these findings highlight that Co68 enhances antitumor immunity by reprogramming TAMs and TANs within the pancreatic tumor microenvironment, modulating the balance between IL-1 and IFN-I signaling, and exhibiting superior therapeutic potential over DMXAA. Co68 activates type I interferon via a novel binding mode with TLR4 To identify the pattern recognition receptor (PRR) involved in Co68-mediated activation of interferon signaling, we initially employed pharmacological inhibitors targeting key innate immune pathways, followed by genetic knockout (KO) approaches. Inhibition of TLR4, TRIF, and TBK1/IKKe significantly reduced Co68-induced IFNB1 expression in both RAW 264.7 and J774A.1 cell lines (Figs. 5 A-B). These results were further validated using BMDMs from TLR4, TRIF, and IRF3 knockout mice (Figs. 5 C-D and Figure S5A), collectively confirming that Co68 induces IFNB1 expression via activation of the TLR4-TRIF-TBK1-IRF3 signaling axis. Previous studies have demonstrated that LPS-induced signaling through TLR4 requires the engagement of CD14 and MD2 79 . In contrast, Co 2+ directly binds to and activates TLR4 without the involvement of CD14 or MD2 71 . To investigate the roles of CD14 and MD2 in Co68-mediated TLR4 activation, we employed blocking antibodies specific to these molecules. As shown in Figs. 5 E- 5 G and Figures S5B-S5D, blocking antibodies targeting TLR4, CD14, and MD2 significantly attenuated the production of Ifnb1 and Isg15 induced by Co68. Notably, antibodies against TLR4 and MD2 nearly abolished Co68-induced expression of Ifnb1 , suggesting that TLR4 and MD2 play a more dominant role than CD14 in the induction of IFN-I. Furthermore, the transfection of plasmids encoding TLR4, CD14, and MD2 into HEK293T cells significantly induced the expression of IFNB1 and ISG15 upon Co68 treatment (Fig. 5 H and Figure S5E), further corroborating the critical involvement of these molecules in Co68-mediated activation of the IFN-I signaling pathway. To explore the direct interaction between Co68 and TLR4, we employed both the microscale thermophoresis (MST) and surface plasmon resonance (SPR) techniques. MST analysis revealed a dissociation constant (Kd) of 0.98 µM for Co68 binding to TLR4, which is lower than the dissociation constant for LPS binding to TLR4 (Kd = 4.37 µM) (Fig. 5 I and Figure S5F). Consistent with these findings, SPR analysis showed a Kd of 31.71 µM for TLR4-Co68 binding (Fig. 5 J and Figure S5G), compared to 147.64 µM for TLR4-LPS binding (Figure S5H-S5I). These results collectively demonstrate that Co68 binds directly to TLR4 with a more stable interaction than LPS, suggesting a distinct binding mode. Given the observed differences in the expression patterns between Co68, LPS, and Co 2+ , we hypothesized that Co68 exhibits a unique binding mode to TLR4 compared to LPS and Co 2+ . To further explore these differences, we employed computational approaches, including AlphaFold3 and AutoDock Vina (V1.2.5) for molecular docking studies. The crystal structure of Co68 was visualized using PyMOL (Figure S5J). Molecular docking with AutoDock Vina (v1.2.5) identified the most stable binding mode of Co68 to TLR4, involving interaction with the hydrophobic pocket of MD2 (Fig. 5 K and Figure S5K), with a binding energy of less than − 7 kcal/mol. As shown in Figs. 5 L and S5L, Co68 interacted with both TLR4 and MD2, with a stronger interaction observed with MD2. Further analysis of the interacting amino acids using PyMOL revealed that Co68 interacted with eight amino acids in TLR4: ARG434, MET358, ARG380, ALA382, PHE406, ASN407, ILE411, and LYS433 (Figs. 5 M and Figures S5M-S5N). These residues differ from those previously reported for TLR4-LPS interactions 79 , including ARG264, LYS341, LYS362, and LYS388. Additionally, chemical bonding analysis of Co68 and MD2 (Fig. 5 N and Figure S5O) predicted hydrogen bonds involving residues such as ILE32, ILE46, ILE52, LEU54, VAL61, LEU78, CYS133, and ILE153, which were distinct from the bonding pattern seen in Co 2+ binding to TLR4 predicted by AlphaFold3 (Fig. 5 O). To experimentally validate this distinct binding pattern, we introduced mutations in the eight predicted interacting amino acids of TLR4 with Co68. Luciferase assays revealed that mutations at F406A and K433R significantly reduced Co68-induced IFNB1 production (Fig. 5 P). Notably, the simultaneous mutations at F406A and K433R led to a more significant reduction in IFNB1 expression. Strikingly, these mutations did not affect the activation of downstream NF-κB signaling induced by CoCl₂ and LPS (Fig. 5 Q), further confirming that Co68 activates IFN-I signaling through a TLR4-dependent pathway that is distinct from the mechanisms of CoCl₂ and LPS. To further investigate the structure-activity relationship between Co68 and TLR4, we designed nine Co68 analogs with gradual substitutions in functional groups (Figure S5P). As shown in Fig. 5 R, the PNP structure, tertiary butyl binding to the phosphate radical, chloride ion binding to the Co²⁺ ion, and hydrogen atom binding to the nitrogen in the PNP structure were critical for Co68’s ability to induce Ifnb1 expression. These structural features are crucial for Co68’s capacity to activate the IFN-I signaling pathway. Furthermore, analogs of Co68 that significantly induced Ifnb1 expression in RAW 264.7 macrophages also demonstrated stable binding to TLR4, as validated by the SPR assay (Figure S6A-S6F). Consistent with these results, Co68 was unable to inhibit Pan02 tumor growth in TLR4 and IRF3 knockout mice (Figures S6G-S6L), further reinforcing the conclusion that Co68’s antitumor activity in vivo is mediated through the TLR4-TBK1-IRF3-IFN-I signaling axis. Taken together, these findings demonstrate that Co68 activates IFN-I signaling through TLR4, with a binding mode distinct from both LPS and Co²⁺, underscoring its unique mechanism during immune modulation. Co68 inhibits inflammatory responses through the non-canonical activation of STAT1 As demonstrated above, Co68 exhibits broad-spectrum anti-inflammatory effects and specifically induces the expression of several histone-modifying genes. Based on these observations, we hypothesized that a Co68-specific transcription factor mediates its anti-inflammatory activity. To identify potential Co68-responsive transcription factors, we conducted ATAC-seq analysis on samples treated with LPS, LPS-Co68, and Co68 for 1 hour, prior to detectable expression of Ifnb1 . Remarkably, Co68 exhibited a potent anti-inflammatory effect on LPS even at this early time point. ATAC-seq identified 50165, 41324, and 33316 accessible peaks in the LPS, LPS-Co68, and Co68 treatment groups, respectively (Figures S7A-C). The annotation of these accessible peaks with the ChIPseeker R package revealed that both the Co68 and LPS-Co68 groups exhibited a more significant number of accessible peaks in promoter regions compared to the LPS-only group (Figure S7D). Additionally, there was a higher density of transcription factor-binding sites within 3 kb of transcription start sites (TSS) in the LPS-Co68 and Co68 groups compared to the LPS group, suggesting an increase in chromatin binding sites for transcription factors induced by Co68 (Figure S7E). In contrast, peaks identified in the LPS group were predominantly located in distal intergenic regions (Figure S7D). As illustrated in the heatmap (Fig. 6 A), the Co68 and LPS-Co68 groups demonstrated a marked increase in TSS-proximal peak accessibility compared to the LPS-only group. Using the ClueGO plug-in in Cytoscape, we annotated genes nearest to the differential accessible regions (DARs) between the LPS-Co68 and LPS groups (Fig. 6 B), revealing significant enrichment in pathways associated with myeloid cell activation, response to interferon-gamma, macrophage migration, dendritic cell differentiation, T cell receptor signaling, and activated T cell proliferation. To predict potential binding transcription factors, we performed a footprint analysis using TOBIAS software (v0.17.0), which revealed distinct transcription factors between the LPS and Co68 groups (Fig. 6 C). As expected, NFKB-REL exhibited significantly more potential binding sites in the LPS group compared to the Co68 group, while IRF3 and EGR1 were significantly enriched in the Co68 group. Surprisingly, STAT1 was also significantly enriched in the Co68 group, even at the 1-hour time point before detectable expression of the upstream signaling molecule Ifnb1 (Fig. 6 C). The binding motif for STAT1, derived from the JASPAR database, is shown in Fig. 6 D. A heatmap of potential STAT1 binding sites revealed notably higher occupancy in the LPS-Co68 and Co68 groups compared to the LPS group (Fig. 6 E). Consistent with these findings, the binding motifs for STAT1 and EGR1, illustrated in Figures S7F and S7G, displayed markedly higher binding intensity in the Co68 group than in the LPS group. In contrast, the binding intensity of REL in the LPS group was stronger than in the Co68 group (Figure S7H). Track visualization using IGV software further confirmed these findings, showing stronger binding signals for genes such as Il6 , Il1b , and Tnf in the LPS group compared to the LPS-Co68 and Co68 groups (Fig. 6 G and Figures S7J-S7K). In contrast, genes such as Rsad2 , Ifi202b , Usp18 , and Ifit1 exhibited stronger binding signals in the LPS-Co68 and Co68 groups compared to the LPS group (Fig. 6 F and Figure S7I). These results were further corroborated by RNA-seq analysis following Co68 treatment for 1 hour (Fig. 6 H and Figure S7L), suggesting that STAT1 activation occurs independently of, and prior to, IFN-I signaling via IFNAR. To investigate the functional role of early, IFN-I-independent STAT1 activation by Co68, we first employed the STAT1 inhibitor fludarabine. Notably, fludarabine treatment abrogated Co68’s anti-inflammatory effects (Fig. 6 I). This finding was further validated by Stat1 knockdown using shRNA (Figure S7N), which similarly abolished Co68’s anti-inflammatory activity (Figs. 6 J and 6 K). To elucidate the mechanism through which Co68 activates STAT1, we examined the effects of inhibitors targeting TLR4, MyD88, and TRIF. Notably, only the TLR4 inhibitor significantly suppressed STAT1 activation, while MyD88 and TRIF inhibitors had no effect (Fig. 6 L). These results indicate that early STAT1 activation by Co68 is TLR4-dependent but MyD88- and TRIF-independent. To identify upstream regulators of STAT1 activation, we performed mass spectrometry analysis following immunoprecipitation (IP) of TLR4 (Figure S7M). This analysis revealed 20 kinases that were specifically upregulated 1-hour post-Co68 treatment, including SYK, a known direct upstream regulator of STAT1 phosphorylation (Fig. 6 M). Co-immunoprecipitation (Co-IP) experiments confirmed that TLR4 and STAT1 interacted with SYK only in the LPS-Co68 and Co68 treatment groups at 1-hour post-treatment (Fig. 6 N). Furthermore, Syk knockdown via shRNA (Figure S7N) and SYK knockout (KO) via CRISPR-Cas9 (Figure S7P) abolished Co68’s anti-inflammatory effects (Figure S7O, Fig. 6 O, and Figure S7Q). Collectively, these data demonstrate that Co68’s broad-spectrum anti-inflammatory activity is dependent on the TLR4-SYK-STAT1 signaling pathway. Discussion A growing body of evidence underscores the critical role of imbalances in innate immune signaling—particularly between type I interferon (IFN-I) signaling and inflammatory responses—in the pathogenesis and progression of a wide array of diseases 67 , 80 . The immune evasion strategies employed by pathogens such as SARS-CoV-2, coupled with the pro-tumor activity of tumor-associated macrophages (TAMs) within the pancreatic cancer microenvironment, highlight the centrality of this immune imbalance in both viral infections and cancer progression 81 – 83 . Rectifying these immune dysregulations presents a promising therapeutic avenue to improve clinical outcomes across diverse disease settings. To this end, we developed DLINP, a convolutional neural network (CNN)-based deep learning model specifically designed to identify small molecules with dual immunomodulatory activities. Through this platform, we identified Co68 as a new immune modulator, a cobalt-based metal complex synthesized by the reaction of a PNP-pincer ligand with cobalt chloride in tetrahydrofuran. We demonstrated here that Co68 robustly activates IFN-I-dependent antiviral responses while simultaneously exhibiting broad-spectrum anti-inflammatory effects, both in vitro and in vivo. Mechanistically, Co68 enhances IFN-I signaling via the TLR4-TRIF pathway, a critical cascade that initiates antiviral and antitumor immunity, while concurrently exerting potent anti-inflammatory effects through the TLR4-SYK-STAT1 axis. Notably, Co68 demonstrated substantial antitumor activity in the Pan02 pancreatic cancer mouse model, which is resistant to PD-1 blockade, outperforming the established STING agonist DMXAA. Moreover, Co68 exhibited significant synergy with PD-1 antibody therapy, promoting enhanced tumor suppression and improved therapeutic outcomes. Further mechanistic insights derived from single-cell transcriptomic analysis revealed that Co68 disrupted the IL-1β-PGE2 axis between TAMs and cancer cells—an essential pathway involved in tumor immune evasion. Co68 also substantially increased the abundance and functional activity of CD8 + T cells and NK cells, likely through IFN-I signaling induction in TAMs, thereby bolstering antitumor immunity. This study emphasizes the transformative potential of deep learning in drug discovery. Co68, identified through the DLINP platform, represents the first PNP-pincer ligand to exhibit significant immunomodulatory activity. The PNP-pincer ligand structure of Co68 makes cobalt chloride exert dual effects on immune signaling via TLR4-mediated modulation of transcriptional landscapes, suppressing deleterious pro-inflammatory responses while enhancing IFN-I signaling. This positions Co68 as a next-generation immunomodulatory agent. Importantly, the PNP ligand (as shown in Figure S1 D), used in the synthesis of Co68, does not activate IFN-I signaling in RAW 264.7 cells. Furthermore, cobalt chloride—one of the substrates used to synthesize Co68—only weakly activates the inflammatory response at high concentrations (500 µM) after 24 hours and does not induce significant transcriptional changes in macrophages within 3 hours of exposure. In contrast, Co68 activates IFN-I signaling at much lower concentrations (50 µM) within just 1 hour of treatment, with peak levels achieved by 3 hours. Co68 also exhibits potent anti-inflammatory activity against diverse stimuli, including LPS, VSV, HSV, and EMCV. Structure-activity relationship experiments demonstrated that the PNP ligand itself possessed broad-spectrum anti-inflammatory properties. Replacing the hydrogen atom bound to nitrogen in the PNP structure abrogates Co68's ability to activate IFN-I signaling. Additionally, substituting chloride ions in cobalt chloride with bromide or iodide significantly impairs Co68's capacity to activate IFN-I signaling. We hypothesize that the PNP ligand, when complexed with cobalt chloride, generated Co68 in a high-energy state, which endows the complex with unexpected biological activities, warranting further investigation to explore the underlying molecular mechanisms. As the research progresses, further studies on Co68’s pharmacokinetics, safety profile, and efficacy in additional disease models—such as psoriasis and SARS-CoV-2 infection—are necessary to assess its clinical potential. Future work will also focus on examining the role of histone-modifying genes, such as H1f5 and H2bc4, in Co68’s regulation of chromatin accessibility, providing deeper mechanistic insights into its molecular actions. We employed an array of pharmacological, genetic, and experimental techniques—including Microscale Thermophoresis (MST) and Surface Plasmon Resonance (SPR)—to confirm that Co68 directly interacts with TLR4, a member of the Toll-like receptor (TLR) family, which is key pattern recognition receptors (PRRs) involved in immune signaling 84 . TLR4 is particularly notable for its role in initiating immune responses. Previous studies have shown that TLR4 recognizes lipopolysaccharides (LPS) in conjunction with its co-receptors CD14 and MD2, triggering the activation of the NF-κB pathway through the MAL-MyD88 axis 79 , 85 – 87 . This cascade leads to the production of pro-inflammatory cytokines central to the innate immune inflammatory response. TLR4 also activates a distinct signaling pathway upon internalization into endosomes, triggering the TRAM-TRIF axis to stimulate IRF3 and promote IFN-I production, thereby enhancing antiviral immunity 88 – 90 . The dual roles of TLR4 in both inflammation and antiviral immunity make it a compelling target for therapeutic intervention, although its activation can also result in adverse effects, such as inflammation triggered by metal allergens like nickel and cobalt, or chemotherapeutic agents such as cisplatin, which activate TLR4 and exacerbate inflammation, potentially impairing patient outcomes 71 , 91 , 92 . Similarly, the spike (S) protein of SARS-CoV-2 has been shown to activate TLR4 93 , contributing to cytokine storms in COVID-19 patients, thus emphasizing TLR4's crucial role in immune regulation and its potential as a therapeutic target in a variety of diseases, ranging from infections to cancer and autoimmune disorders. Through AlphaFold predictions and molecular docking, we observed significant differences in the binding modes of Co68 compared to cobalt ions and LPS, with these findings further validated by point mutation experiments. Cryogenic electron microscopy (cryo-EM) is currently being employed to gain detailed structural insights into the Co68-TLR4 interaction. Notably, Co68 uniquely activates the TLR4-TRAM-TRIF-IRF3 pathway, bypassing the conventional MAL-MyD88-NF-κB cascade, thereby reprogramming TLR4-mediated immune responses. This unique mechanism may explain certain clinical observations in cancer therapies. For instance, paclitaxel, a widely used chemotherapeutic agent, targets TLR4, but its albumin-bound formulation (nab-paclitaxel) has been linked to enhanced tumorigenesis in specific clinical settings 94 – 96 . This discrepancy may stem from structural differences in their binding modes to TLR4, resulting in divergent effects on immune signaling and inflammation. A pivotal finding of this study is the identification of STAT1 as a critical transcription factor in Co68-induced early activation, which is independent of IFN-I signaling and essential for the compound's anti-inflammatory effects. ATAC-seq and footprint analysis revealed significant differences in chromatin accessibility between Co68- and LPS-treated cells, with STAT1 binding sites being notably enriched in the Co68 group. The early activation of STAT1, occurring prior to detectable Ifnb1 expression, was found to be a key factor in Co68’s anti-inflammatory activity. Furthermore, our data confirmed that TLR4 is involved in the activation of STAT1, with MyD88 and TRIF being dispensable in this process. Mass spectrometry identified SYK as an upstream kinase in the TLR4-SYK-STAT1 signaling axis, and co-immunoprecipitation (Co-IP) experiments revealed that SYK interacts with both TLR4 and STAT1 in the presence of Co68. These findings provide novel mechanistic insights into the TLR4-SYK-STAT1 pathway, which had not been previously described 97 – 99 . The TLR4-SYK-STAT1 axis emerges as a promising therapeutic target for inflammatory diseases. Although the precise mechanisms by which STAT1 mediates its anti-inflammatory effects remain to be fully understood, we propose that STAT1 activation may resemble the action of STAT3 in TLR4 signaling, where phosphorylation of serine 727 by downstream kinases such as TBK1 induces metabolic reprogramming and anti-inflammatory effects 100 . Further studies are needed to validate this hypothesis and deepen our understanding of the signaling dynamics at play. Co68 provides dual therapeutic benefits by concurrently activating the TLR4-TRIF and TLR4-SYK-STAT1 signaling pathways, thereby mitigating inflammation while enhancing antiviral and antitumor immunity. In the context of cancer, particularly pancreatic cancer, Co68’s ability to modulate the immune landscape by suppressing IL-1β signaling and enhancing IFN-I responses makes it a promising therapeutic candidate 20 , 38 . In this tumor type, TAMs are a major source of IL-1β, which drives tumor progression and inhibits IFN-I signaling through PGE2 produced by tumor cells, ultimately promoting immune evasion 101 – 103 . Compared to other antitumor agents, such as DMXAA, which also significantly inhibits the growth of Pan02 pancreatic tumor in mice 104 , Co68 induces TLR4 activation more effectively than DMXAA activates STING. DMXAA’s activation of STING leads to potent NF-κB activation 73 , causing substantial production of IL-1β and IL-6—cytokines that promote tumor growth and immune suppression. In contrast, Co68 attenuates IL-1β-mediated pro-inflammatory signaling, thereby reducing the tumor-promoting inflammation commonly observed in pancreatic cancer. Notably, Co68 treatment upregulated ISGs in specific subpopulations of TAMs, which facilitated the recruitment of CD8 + T cells and NK cells—key players in antitumor immunity 26 , 105 , 106 . A striking observation was the selective expansion of a neutrophil subpopulation enriched in ISGs. Given the growing evidence supporting the role of ISG-high neutrophils in antitumor immunity 76 – 78 , 107 – 109 , we hypothesize that this subset plays a crucial role in Co68’s enhanced therapeutic efficacy. However, further studies are required to fully elucidate the precise function of this neutrophil population within the TME. Additionally, Co68’s synergy with PD-1 blockade led to a more potent antitumor effect in the Pan02 model, suggesting that Co68 can overcome immunosuppressive barriers. This synergistic effect complements PD-1 antibody-mediated activation of CD8 + T cells within the TME, restoring immune surveillance through dual mechanisms 110 . Co68’s dual activity is particularly beneficial for “cold” tumors, such as pancreatic cancer and triple-negative breast cancer, which are often characterized by immune evasion 95 . Future studies are essential to evaluate the broader applicability of Co68 in other tumor models and further explore its clinical potential as an immune-modulatory agent for cancer therapy. In conclusion, Co68's unique mechanism of action, mediated by its PNP-pincer ligand structure, positions it as a promising therapeutic agent for targeting immune dysregulation in diseases such as cancers and viral infections. As research progresses, Co68 could provide innovative strategies for overcoming immunosuppressive tumor microenvironments, ultimately enhancing immune responses and improving therapeutic outcomes in cancer treatment. Co68 represents a critical step forward in the development of PNP-pincer ligands, opening new avenues for immune system modulation and offering potential solutions for treating cancers that continue to claim millions of lives worldwide. Materials and methods Reagents and antibodies Rabbit anti-COX2 antibodies were sourced from Wanleibio. Antibodies targeting LY96/MD2 (catalog no. 11784-1-AP), iNOS (catalog no. 80517-1-RR), TBK1 (catalog no. 28397-1-AP), NF-κB p65 (catalog no. 10745-1-AP), and Beta Tubulin (catalog no. 10094-1-AP) were obtained from Proteintech. Mouse monoclonal antibodies against Beta Actin (catalog no. 66009-1-Ig), TLR4 (catalog no. 66350-1-Ig), CD14 (catalog no. 60253-1-Ig), and the HA tag (catalog no. 51064-2-AP) were also purchased from Proteintech. Phosphorylation-specific antibodies, including rabbit anti-phospho-NF-κB p65 (Ser536) (93H1) (catalog no. 3033), anti-phospho-IκBα (Ser32) (14D4) (catalog no. 2859), anti-phospho-TBK1/NAK (Ser172) (D52C2) (catalog no. 5483), anti-phospho-IRF-3 (Ser396) (D6O1M) (catalog no. 29047), as well as antibodies for IRF-3 (D6I4C) (catalog no. 11904), Stat1 (D1K9Y) (catalog no. 14994), and Syk (D3Z1E) (catalog no. 13198), were obtained from Cell Signaling Technology. For secondary detection, horseradish peroxidase (HRP)-conjugated goat anti-rabbit IgG (H+L) (catalog no. SA00001-2) and HRP-conjugated goat anti-mouse IgG (H+L) (catalog no. SA00001-1) were both sourced from Proteintech. Flow cytometry antibodies included FITC-conjugated anti-mouse NK1.1 (catalog no. FITC-65138-25UG) and PE-conjugated anti-mouse CD8a (catalog no. PE-65069), which were procured from Proteintech, while APC-conjugated anti-mouse CD4 (catalog no. E-AB-F1097UE) was obtained from Elabscience. Recombinant mouse GM-CSF (catalog no. HY-P7361) and anti-mouse PD-1 antibody (catalog no. HY-P99144) were purchased from MedChemExpress. Recombinant human TLR4 protein (ECD, His Tag) was obtained from Sino Biological (catalog no. 10146-H08B). Cell Culture The cell lines RAW264.7, J774A.1, HEK293T, THP-1, and Vero were obtained from the American Type Culture Collection (ATCC). Bone marrow-derived macrophages (BMDMs) were isolated from the femurs and tibiae of 8-week-old C57BL/6J mice. The bone marrow was cultured in the presence of granulocyte-macrophage colony-stimulating factor (GM-CSF, RP01206, ABclonal) for 7 days to promote macrophage differentiation. Peritoneal macrophages (PMs) were similarly isolated from the peritoneal cavity and differentiated using GM-CSF for 7 days. HEK293T, RAW264.7, J774A.1, THP-1, Vero, and BMDMs were cultured in Dulbecco's Modified Eagle's Medium (DMEM; 03.1002C, EallBio), while PMs were maintained in RPMI 1640 (03.4001C, EallBio). All culture media were supplemented with 10% fetal bovine serum (FBS) and 1% penicillin-streptomycin to support optimal cell growth and maintenance. Virus Infection and Propagation Cells at 70–80% confluence were infected with the following viruses at the indicated multiplicities of infection (MOI): Vesicular Stomatitis Virus (VSV, MOI = 0.1), Herpes Simplex Virus 1 (HSV-1, F strain, MOI = 0.5), and Encephalomyocarditis Virus (EMCV, MOI = 0.1). Virus stocks were prepared as follows: the VSV Indiana strain, kindly provided by Dr. J. Rose (Yale University), was propagated in Vero cells. The HSV-1 strain 17, a gift from Dr. Zhengfan Jiang (Peking University), was cultured in Vero cells for viral propagation. The EMCV strain (ATCC VR-129B) was acquired from ATCC and similarly propagated in Vero cells. Mice and Construction of a Sepsis Model via LPS Injection Wild-type (WT) C57BL/6J mice were obtained from the Department of Laboratory Animal Science, Peking University Health Science Center. Tlr4-/- , Irf3-/- , and Trif-/- mice, all on a C57BL/6J background, were acquired from the Institute of Experimental Animals, Chinese Academy of Medical Sciences. The Ifnar1-/- mice were generously provided by Prof. Erol Fikrig (Yale University). Genotyping of these mice was performed using the following primers: Tlr4 KO (forward: 5′-TGCTCACACCATCATCAC-3′; reverse: 5′-CATGTACTAGGTTCGTCAGA-3′), Irf3 KO (forward: 5′-TCGTGCTTTACGCTATGCCGCTCCCGATT-3′; reverse: 5′-GAACCTCGGAGTTATCCCGAAGG-3′), and Ifnar1 KO (forward: 5′-CGAGGCGAAGTGGTTAAAAG-3′; reverse: 5′-AATTCGCCAATGACAAGACG-3′). All animal experiments were carried out in accordance with the Guide for the Care and Use of Laboratory Animals provided by the Chinese Association for Laboratory Animal Science. The protocols were approved by the Animal Care Committee of Peking University Health Science Center (permit number: LA 2016240). Mice were housed under specific pathogen-free (SPF) conditions at the Laboratory Animal Center of Peking University. Only male mice aged 6 to 8 weeks were used in the study. Lipopolysaccharide (LPS, HY-D1056, MedChemExpress) was dissolved in sterile PBS (P1020, Solarbio Life Science) to a final concentration of 20 mg/kg body weight. A total of 20 WT C57BL/6J mice were injected intraperitoneally with LPS using sterile insulin syringes. Following LPS injection, a subset of mice (n=10) received a 25 mg/kg dose of Co68 (200 μL in PBS) via intraperitoneal injection. The control group (n=10) was treated with an equivalent volume of PBS. Survival times were monitored and recorded following injection. Tumor Transplantation, Treatments, and Tissue Digestion Mice were subcutaneously injected into the right flank with 5 × 10⁵ Pan02 cells suspended in 100 μL PBS, unless otherwise specified. Tumor growth was monitored daily, with tumor size measured every 1–2 days. Tumor volume was calculated using the formula: volume = length (mm) × width² (mm²) × 0.5. Tumor-bearing mice were treated with 25 mg/kg Co68 or DMXAA via intraperitoneal injection every 2 days. For immune checkpoint blockade, tumor-bearing mice were administered 200 μg of anti-mouse PD-1 antibody (200 μL saline) via intraperitoneal injection on days 3, 7, and 11 following tumor inoculation. Control mice received 200 μL of rat IgG2a isotype (Clone 2A3, BioXCell) on the same schedule. Following euthanasia, tumors and the surrounding skin were excised. Tumors were minced and digested in RPMI 1640 medium containing 0.5 mg/mL collagenase D (11088866001, Roche) and 0.1 mg/mL DNase I (DN25-100G, Sigma-Aldrich) at 37°C for 1 hour. The resulting cell mixture was then filtered through a 70-μm cell strainer to obtain a single-cell suspension, which was subsequently transferred to flow cytometry tubes for further analysis. RNA extraction, reverse transcription and real-time quantitative PCR RNA extraction from cells or tissues following various treatments or infections was performed using TRIzol reagent (TIANGEN, A0123A01). The purified RNA was then reverse transcribed into cDNA using HiScript II RT SuperMix (Vazyme, R223-01). Quantification of target gene expression was carried out using SYBR Green qMix (Vazyme, Q311) in a quantitative reverse transcription PCR (RT-qPCR) assay. The relative expression levels of target mRNAs were normalized to the housekeeping gene Gapdh . Detailed primer sequences used in this study are provided in Supplementary Table S1. Luciferase Assay Cells were seeded into 6-well plates and cultured to 70–80% confluence. Transient transfection was performed using Lipofectamine 2000 (Thermo Fisher) according to the manufacturer's instructions. After 24 hours of transfection, the cells were treated with Co68, CoCl₂, or LPS at 37°C in a humidified incubator with 5% CO₂ for the indicated time points. Following treatment, cells were lysed directly in the wells using lysis buffer (FR203-01, Transgen), in accordance with the manufacturer’s guidelines. The resulting lysates were collected, and luciferase activity was measured using a TD20/20 Luminometer (Turner Designs). Total Protein Extraction and Western Blot Analysis Cells were lysed using RIPA Lysis Buffer (Strong) (HY-K1001, MedChemExpress), supplemented with a protease inhibitor cocktail (EDTA-Free Protease Inhibitor Cocktail) and phosphatase inhibitor cocktails (Phosphatase Inhibitor Cocktail I and III, both 100× in DMSO). Clarified cell extracts (10 to 30 μg) were resolved on SDS-PAGE gels and subsequently transferred to nitrocellulose membranes (FFN08, Beyotime). Following blocking, membranes were incubated with primary antibodies specific to the target proteins. The bound secondary antibodies were detected using the enhanced chemiluminescence (ECL) method (07.10009-50, EallBio). Cell Cytotoxicity Assay Target cells were seeded into 96-well plates at a density of 2 × 10³ cells per well in 100 μL of complete medium. Cells were treated with increasing concentrations of Co68 and incubated at 37°C in a humidified 5% CO₂ incubator for 48 hours. After the treatment period, 10 μL of CCK-8 reagent (40203ES60, YEASEN) was added to each well, and the plate was gently shaken to mix. The cells were then incubated at 37°C for an additional 1 hour. Absorbance was measured at 450 nm using a microplate reader to assess cell viability. Pharmacological Inhibition of Proteins To investigate the role of specific signaling pathways, cells were treated with a range of pharmacological inhibitors upon reaching 70-80% confluence. Each inhibitor was dissolved in anhydrous DMSO and diluted to the appropriate working concentration. The following inhibitors were used: C29 (10 µM, S6597, Selleck) for TLR2 inhibition, Procyanidin B1 (30 µM, HY-N0795, MedChemExpress) for TLR4 inhibition, MyD88-IN-1 (30 µM, HY-149992, MedChemExpress) for MyD88 inhibition, Pepinh-TRIF TFA (30 µM, HY-P2565, MedChemExpress) for TRIF inhibition, GSK8612 (5 µM, T5540, TargetMol) for TBK1/IKKε inhibition, MNS (10 µM, HY-78263, MedChemExpress) for SYK inhibition, and Fludarabine (10 µM, HY-B0069, MedChemExpress) for STAT1 inhibition. Following treatment, cells were incubated at 37°C for 3 hours. Control groups were treated with DMSO at a final concentration of less than 0.1%. This treatment scheme enabled the assessment of the contribution of each specific pathway to the cellular responses observed. Cloning and Construction of Expression Plasmids Mouse Tlr4 , Cd14 , and Md2 cDNAs were cloned into the pcDNA3.1-HA expression vector using a seamless cloning and assembly kit (Transgen, CU101-01). All expression constructs were generated using standard molecular biology techniques, and the coding sequences were fully verified by sequencing. Site-directed mutants were created using standard cloning procedures, with each mutant confirmed through sequencing. The Ifnb luciferase reporter and NF-κB luciferase reporter plasmids were generously provided by Professor Zhengfan Jiang from Peking University. The pCMV-VSVG and psPAX2 plasmids were obtained from Addgene. Gene Silencing of Stat1 and Syk using Short Hairpin RNA (shRNA) Gene silencing of Stat1 and Syk was achieved using the pLKO.1 plasmid vector, which contains EcoRI and AgeI restriction enzyme cutting sites. The shRNA sequences designed to knock down Stat1 and Syk were sourced from the Sigma-Aldrich online database (https://www.sigmaaldrich.cn/CN/zh/product/sigma/shrna), converted by the online tools MOI (http://www.fynn-guo.cn/seq_tool.php) 111 and are listed in Supplementary Table S2. Initially, three pairs of shRNAs targeting Stat1 and Syk were designed, and following validation, the pairs with the highest knockdown efficiency were selected for further experiments. The final shRNA sequences for Stat1 -shRNA-pair2 were: Sense strand : CCGGCCCTGAAGTATCTGTATCCAACTCGAGTTGGATACAGATACTTCAGGGTTTTTG Antisense strand : AATTCAAAAACCCTGAAGTATCTGTATCCAACTCGAGTTGGATACAGATACTTCAGGG The final shRNA sequences for Syk -shRNA-pair2 were: Sense strand : CCGGGCAGGCCATCATCAGTCAGAACTCGAGTTCTGACTGATGATGGCCTGCTTTTTG Antisense strand : AATTCAAAAAGCAGGCCATCATCAGTCAGAACTCGAGTTCTGACTGATGATGGCCTGC A nontargeting shRNA (scramble shRNA) from Sigma-Aldrich was used as a negative control. Transfections were performed at a final concentration of 10 nM using Lipofectamine® RNAiMAX Transfection Reagent (catalog number 13778030; Thermo Fisher Scientific), following the manufacturer’s instructions. After transfection, cells were split and selected with 5 µg/mL puromycin for two weeks. Stable clones were then isolated by limiting dilution and expanded for further analysis. CRISPR-Cas9 System for Syk Knockout Syk knockout (KO) RAW264.7 cells were generated using the CRISPR-Cas9 system. Guide RNAs (gRNAs) with high efficiency and specificity were designed using the SYNTHEGO online CRISPR design tool (https://www.synthego.com/products/crispr-kits). The oligos were annealed and cloned into the lentiCRISPR v2 vector, which had been digested with the BsmBI enzyme (NEB). To generate lentivirus, 293T cells were transfected with the following plasmids: lentiCRISPR v2 (2400 ng), packaging plasmid psPAX2 (800 ng; Addgene 12260), envelope plasmid VSV-G (800 ng; Addgene 8454), and PEI (1600 ng). The transfection mixture was incubated at 37°C for 72 hours, after which viral supernatants were collected and used to infect RAW264.7 cells in the presence of polybrene (Beyotime Biotechnology, China). Forty-eight hours post-infection, cells were refreshed with fresh culture medium and selected with 10 µg/mL puromycin. Three pairs of sgRNAs targeting Syk were initially designed (see Supplementary Table S2), and after validation, the pairs with the highest knockout efficiency were selected for further experiments. The successful knockout of Syk was validated by Western blotting. The primer sequences used for the knockout validation are as follows: Sense strand : CACCGUGAAGGGGUGCAGACAUGGC Antisense strand : AAACGCCATGTCTGCACCCCTTCAC Immunoprecipitation (IP) Assay Immunoprecipitation was performed using the ProteinIso® Protein A/G Resin (DP501, TransGen Biotech) according to the manufacturer’s instructions. Cells were lysed in NP-40 lysis buffer (HY-K1002, MedChemExpress), supplemented with protease inhibitors. The lysates were clarified by centrifugation at 12,000 rpm for 10 minutes at 4°C. Equal amounts of protein lysates were incubated with specific primary antibodies overnight at 4°C. Following incubation, ProteinIso® Protein A/G Resin was added and rotated gently for 4 hours at 4°C. The resin was washed three times with NP-40 lysis buffer to remove nonspecific binding. Immune complexes were eluted by heating the resin in SDS-PAGE loading buffer at 108°C for 5 minutes, and the resulting samples were analyzed by Western blotting. Microscale Thermophoresis (MST) Assay The interaction between TLR4 and Co68 was assessed using Microscale Thermophoresis (MST) at the State Key Laboratory, School of Pharmaceutical Sciences, Peking University Health Science Center. Recombinant human TLR4 protein was labeled with a fluorescent dye, and Co68 was prepared in a series of dilutions. The labeled TLR4 protein and Co68 were incubated together, and the mixture was loaded into capillaries for measurement. MST data were collected at 25°C, and binding affinity (K d ) was calculated using MO.Affinity Analysis software. All data analysis was independently conducted by our research team. Surface Plasmon Resonance (SPR) Assay SPR experiments were conducted using the S-Class label-free molecular interaction analysis system, in collaboration with Polariton Life Sciences, to evaluate the interaction between TLR4 protein and small molecules (Co65, Co66, Co67, Co68, Co69, Co70, Co71, Co72, Co73, and Co74). TLR4 protein was immobilized onto a C5 sensor chip via amine coupling, using a protein stock solution prepared at 10 µg/mL in 1X PBST (pH 4.5), with an immobilization time of 600 seconds. Small molecules were dissolved in 100% DMSO at 10 mM, then diluted in 1X PBST containing 5% DMSO before injection. Each analyte was flowed over the immobilized TLR4 at a rate of 30 µL/min, with association and dissociation phases set to 60 seconds and 120 seconds, respectively. Binding data were collected and analyzed to determine the kinetic parameters and affinities (K d ) of the interactions. Structure Prediction with AlphaFold The binding interactions of the TLR4-MD2 complex with the small molecule Co68 and the ion Co 2+ were predicted using AlphaFold3 112 . The amino acid sequences of TLR4 and MD2 were retrieved from the UniProt database and used to predict the structure of the TLR4-MD2 complex. The predicted structure was evaluated for accuracy based on the confidence score (pLDDT) and aligned with known crystal structures for validation. For molecular docking, the TLR4-MD2 complex structure predicted by AlphaFold was used as the receptor. The structure of Co68 was obtained from the PubChem database, and its geometry was optimized prior to docking. Co 2+ binding sites were analyzed separately by including the ion in docking simulations. Molecular docking was performed using AutoDock Vina (V1.2.5) 113 , and binding interactions were visualized and analyzed using PyMOL (V2.1) and Chimera (V1.18) 114 . Key residues involved in the binding of Co68 and Co 2+ were identified and compared to evaluate their interaction patterns with the TLR4-MD2 complex. RNA-seq and data analysis Total RNA was extracted using the high-throughput RNA extraction kit (TIANGEN, A0123A01). Quality control, library preparation, and sequencing were performed by Suzhou GENEWIZ Biotechnology company (https://www.genewiz.com.cn/), following established standard protocols. Data analysis was carried out in-house. Sequencing raw data in FASTQ format underwent quality control using FastQC (v0.11.9) and Trim-Galore (v0.6.4) software to generate clean data for downstream analysis. Trim-Galore was executed with the following parameters: trim_galore --gzip --trim-n --phred33 -j 7 --paired ${var}_1.fq.gz ${var}_2.fq.gz -o $wrk_dir/clean_result/. Clean reads were aligned to the mm10 reference genome using Subread (v2.0.0) with the following parameters 115 : subread-align -i $idx_dir -r $cle_dir/${var}_1_val_1.fq.gz -R $cle_dir/${var}_2_val_2.fq.gz -o $aln_dir/${var}.bam -T 30 -t 0. The gene count matrix was generated using featureCounts (v2.0.0) with the following parameters 116 : featureCounts -p -t exon -g gene_id -a $gtf_dir -o $cnt_dir/count_refGene $aln_dir/*.bam -T 29. The count data was normalized using the Fragments Per Kilobase Million (FPKM) formula. Differential expression analysis was conducted using the DESeq2 R package (v1.38.3), with the thresholds set at |log2(FoldChange)| > 2 and p -value < 0.05 to identify significant differentially expressed genes (DEGs). GO annotation and KEGG pathway enrichment analysis In this study, DEGs were subjected to enrichment analysis through Gene Ontology (GO) annotation and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis 117,118 . This analysis was performed using the clusterProfiler R package, specifically employing the enrichGO and enrichKEGG functions 118 . Alternatively, the online database DAVID (https://david.ncifcrf.gov/summary.jsp) (V6.8) was utilized for the same purpose 119 . Significance thresholds were established, with GO and KEGG terms having false discovery rates (FDR) less than 0.01 considered indicative of meaningful enrichments. Gene set enrichment analysis (GSEA) Gene Set Enrichment Analysis (GSEA) is a computational approach used to determine whether predefined gene sets exhibit statistically significant and coordinated differences between two biological states. Unlike traditional methods that focus solely on differentially expressed genes (DEGs), GSEA incorporates all genes in the analysis, regardless of their significance levels. In this study, GSEA was conducted using the clusterProfiler package, and the function gseaplot2 from the same package was employed to visualize the enrichment results 118 . Protein-protein interaction (PPI) network analysis Protein-Protein Interaction (PPI) network analysis is a computational approach used to investigate and map the interactions between proteins within a biological system. For this analysis, we utilized the online database STRING (https://string-db.org/) with default parameters 120 . The input consisted of differentially expressed genes (DEGs), and the output was a network representing the interactions between the proteins encoded by these DEGs. Visualization of the resulting PPI network was carried out using Cytoscape software (v3.10.2) 121 . Transcription factor enrichment analysis (TFEA) Transcription Factor Enrichment Analysis (TFEA) is a computational method used to identify transcription factors (TFs) that may regulate differentially expressed genes (DEGs). For this analysis, we utilized iRegulon (http://iregulon.aertslab.org/tutorial.html), a Cytoscape plug-in specifically designed for TFEA 122 . The input for iRegulon is a list of DEGs, and the output consists of a ranked list of predicted TFs based on enrichment scores. The results of TFEA can be directly visualized and further explored using Cytoscape's integrated visualization tools (v3.10.2). ATAC-seq We employed a high-throughput sequencing methodology based on transposase-mediated chromatin accessibility profiling (ATAC-seq) 123 . This technique utilizes the specificity of Tn5 transposase to selectively cleave accessible regions of chromatin. The transposase, preloaded with DNA sequence adapters, was incubated with isolated cell nuclei, enabling the targeted insertion of adapters into open chromatin regions. Subsequently, indexed primers were utilized for PCR amplification to construct sequencing libraries. The resultant libraries, following sequencing, provided comprehensive insights into DNA regions associated with chromatin accessibility. This experiment was performed using the Hyperactive ATAC-Seq Library Prep Kit (Vazyme Biotech, TD711), with sequencing services conducted by GENEWIZ Biotechnology company. (Suzhou, China). Data analysis of ATAC-seq For ATAC-seq high-throughput data, sequenced by GENEWIZ Biotechnology, raw fastq data underwent quality control using Trim-Galore (v0.6.4). Filtered reads were then aligned and quantified against the mouse genome mm10 using the bowtie2 aligner (v2.3.5.1) 124 , generating SAM or BAM files. The bowtie2 alignment parameters were: bowtie2 --very-sensitive -X 2000 -x $Bowtie2Index -1 $cln_res/${sample}_1_val_1.fq.gz -2 $cln_res/${sample}_2_val_2.fq.gz -p $PPN | samtools view -buSh -@ $PPN | samtools sort -@ $PPN -O BAM -o $aln_res/${sample}.sorted.bam. Sorting and indexing of SAM/BAM files were completed using samtools (v1.10) with the command 125 : samtools index -@ $PPN $aln_res/${sample}.sorted.bam. Peak calling was performed using MACS3 (v3.0.0a5) with default parameters 126 , yielding peak files in BED format for each sample. For downstream analysis, different strategies were applied to ATAC-seq data. ATAC-seq peaks were merged across all samples using the bedtools merge (v2.31.1) 127 function, followed by normalization with bamCoverage (v3.3.2) in the deepTools suite 128 using the command: bamCoverage --bam $var -o ${var%.*}.bw --binSize 100 --normalizeUsing RPKM --effectiveGenomeSize 2864785220 --ignoreForNormalization chrM –extendReads. Differential analysis for both ATAC-seq was conducted using csaw R package (v1.38.0) 129 , while peak annotation was performed using the ChIPseeker R package (v1.34.1) 130 . Finally, peak visualization was achieved using Integrative Genomics Viewer (IGV) (v2.17.4) 131 . Generation and Processing of Single-Cell RNA Sequencing (scRNA-seq) Data Tumor tissues were harvested from mice following subcutaneous implantation and processed into single-cell suspensions. Single-cell suspensions, library construction, and sequencing were performed by Genewiz, Azenta Life Sciences, adhering to the 10× Genomics Chromium Next GEM Single Cell 3′ Reagent Kits v3.1 protocol. Sequencing was conducted on the Illumina NovaSeq 6000 platform in paired-end 150 bp (PE150) mode. Genewiz provided the binary FASTQ files for subsequent analysis. Raw FASTQ files were processed using the Cell Ranger pipeline (v3.1.0) for barcode demultiplexing, alignment, and unique molecular identifier (UMI) counting, generating feature-barcode matrices. Further analysis was performed in R using Seurat (v4.3.0) 132 . The data were normalized, reduced in dimensionality, and clustered for cellular classification. Cell types were manually annotated based on canonical marker gene expression, and the results were exported for downstream analyses. Cell-cell communication was analyzed using the CellChat package (v1.1.3) 133 . Gene set enrichment analysis (GSEA) and gene set variation analysis (GSVA) were conducted using the clusterProfiler package (v4.6.2) 118 . Pseudotime trajectory analysis was performed with Monocle3 (v1.0.0) 134 . Marker gene expression patterns were visualized using dot plots, violin plots, and heatmaps. Preprocessing and Encoding of Drug-Gene Expression Profiles Data This study utilized a comprehensive dataset of gene expression profiles for small molecules, sourced from the Library of Integrated Network-Based Cellular Signatures (LINCS) project (https://clue.io/lincs) 135 , DrugBank 136 , and ChEMBL 137 , including chemical structure and bioactivity data. Strict data cleaning criteria were applied: molecules with fewer than seven replicates were excluded, and molecular SMILES were parsed using RDKit (v2024.03.5). The gene expression profiles for each molecule were averaged, disregarding variations in plate, dose, treatment time, and cell line. The focus was on landmark genes associated with NF-κB and IRF3, which are induced by lipopolysaccharide (LPS). This resulted in a final dataset comprising 7,861 valid molecules, which was split into training (5,502) and test (2,359) sets. The chemical structures of small molecules were represented as SMILES strings, which were converted into grammar trees and then transformed into one-hot encoded arrays. These molecular features were processed into fixed-length vectors for input into a convolutional neural network (CNN). Dimensionality reduction was performed using a variational autoencoder (VAE), producing the input matrix (X). Simultaneously, the gene expression effects of each molecule were encoded into one-hot arrays, with ‘1’ indicating gene upregulation and ‘0’ indicating downregulation, creating the output matrix (Y). Construction of the DLINP Model The CNN-based DLINP (Deep Learning Inflammation and Immunity Prediction) model was designed to predict the ability of compounds to inhibit inflammatory responses and enhance innate immune responses. The model was implemented using TensorFlow (https://www.tensorflow.org/) and Keras (https://keras.io/) to predict small molecule-induced gene expression profiles. The architecture of the model was designed to capture interactions between molecular features and their corresponding gene expression patterns. Each small molecule was represented as a feature vector of dimensions (4717, 6042, 1). The model architecture began with a Conv1D layer containing 32 filters with a kernel size of 3, applying the ReLU activation function to extract relevant molecular features. This was followed by a MaxPooling1D layer to reduce dimensionality while preserving critical information. A second Conv1D layer with 64 filters was added, followed by another MaxPooling1D layer to refine the feature extraction process. The output from these convolutional layers was flattened into a one-dimensional vector and passed through a Dense layer containing 128 neurons with ReLU activation. To prevent overfitting, a dropout rate of 0.5 was applied. The final Dense output layer, using a sigmoid activation function, predicted the expression status (upregulated or downregulated) of landmark genes. The model was compiled with the Adam optimizer and binary cross-entropy loss for binary classification. The dataset was split into training, validation, and testing sets in a 70:30 ratio. The training was conducted over 1,000 epochs with a batch size of 32, with 10% of the training data reserved for validation to monitor performance during training. After model training, compounds from the CAS database (https://www.cas.org/) were screened using DLINP. Evaluation of the DLINP Model Performance The performance of the DLINP model was evaluated using several key metrics, including the confusion matrix, precision-recall (PR) curve, receiver operating characteristic (ROC) curve, and the area under the ROC curve (AUC). These metrics offered a comprehensive assessment of the model's ability to predict gene expression profiles induced by small molecules, providing insights into its precision, recall, overall classification accuracy, and discriminatory power. The confusion matrix revealed the distribution of true positive, true negative, false positive, and false negative predictions, providing detailed information on the model’s performance. The PR and ROC curves, along with the AUC, illustrated the trade-offs between sensitivity and specificity across various thresholds, further aiding in the evaluation of the model’s predictive capabilities. Declarations Data Availability The RNA-seq, ATAC-seq, and scRNA-seq datasets generated in this study are available in the Gene Expression Omnibus (GEO) database under the following accession numbers: GSE288554, GSE288553, and GSE288555, respectively. Code availability The R and Python code used to generate the figures in this study will be made available upon reasonable request. Interested parties are encouraged to contact the corresponding author for access to the code. Statistical analysis All statistical analyses in this study were conducted using the Python package SciPy (https://pypi.org/project/scipy/). Results are presented as the mean ± standard error of the mean (SEM). The p values < 0.05 were considered statistically significant (*), p values < 0.01, and p values < 0.001 were regarded as highly statistically significant (** and ***). Acknowledgments We thank the support from the Peking University High-performance Computing Platform for providing the computing clusters to facilitate the data analysis procedure in this research. Author contributions Conceptualization, X.G. Methodology, X.G. Software, X.G. Validation, X.G. and Y.Z. and S.H. and X.C. Formal Analysis, X.G., and Y.Z. and F.Y. Investigation, X.G. and Y.Z. and and Y.S.Z. and T.C. Resources, F.Y. and Q.L. and X.R. and L.T. and X.W and X.W. Data Curation, X.G. Writing – Original Draft, X.G. Writing – Review & Editing, F.Y. and Y.Z. Visualization, X.G. Supervision, X.G, and F.Y. and Q.L. Project Administration, Y.Z. and T.C. Funding Acquisition, F.Y. Competing interests The authors declare no competing interests. Funding This work was supported by the Beijing Natural Science Foundation (Z210014), the National Key Research and Development Program of China (2021YFC2302602, 2020YFA0707800), the National Natural Science Foundation of China (31570891, 31872736, 32022028, 81991505, and 82201928), Peking University Clinical + X (PKU2020LCXQ009), the Peking University Medicine Fund (PKU2020LCXQ009), the Zhuhai Science and Technology Innovation Bureau (ZH22036302200063PWC to Z.Y.), the China Postdoctoral Science Foundation (2022M710265 to H.Y.), a grant from the Tianjin Natural Science Foundation of China (no. 21JCQNJC01870 to D.W.), and the Incubation Fund of Tianjin Third Central Hospital (no. 2019YNR6 to D.W.). References Ryan, D. P., Hong, T. S. & Bardeesy, N. Pancreatic Adenocarcinoma. New England Journal of Medicine 371 , 1039–1049 (2014). Hidalgo, M. Pancreatic Cancer. New England Journal of Medicine 362 , 1605–1617 (2010). Li, ( D et al. Pancreatic Cancer . THE LANCET • vol. 363 www.thelancet.com1049 (2004). Wang, S. et al. The molecular biology of pancreatic adenocarcinoma: translational challenges and clinical perspectives. Signal Transduction and Targeted Therapy vol. 6 Preprint at https://doi.org/10.1038/s41392-021-00659-4 (2021). Grossberg, A. J. et al. Multidisciplinary standards of care and recent progress in pancreatic ductal adenocarcinoma. CA Cancer J Clin 70 , 375–403 (2020). McGuigan, A. et al. Pancreatic cancer: A review of clinical diagnosis, epidemiology, treatment and outcomes. World Journal of Gastroenterology vol. 24 4846–4861 Preprint at https://doi.org/10.3748/wjg.v24.i43.4846 (2018). Halbrook, C. J., Lyssiotis, C. A., Pasca di Magliano, M. & Maitra, A. Pancreatic cancer: Advances and challenges. Cell vol. 186 1729–1754 Preprint at https://doi.org/10.1016/j.cell.2023.02.014 (2023). Park, W., Chawla, A. & O’Reilly, E. M. Pancreatic Cancer: A Review. JAMA - Journal of the American Medical Association vol. 326 851–862 Preprint at https://doi.org/10.1001/jama.2021.13027 (2021). Rahib, L. et al. Projecting cancer incidence and deaths to 2030: The unexpected burden of thyroid, liver, and pancreas cancers in the united states. Cancer Research vol. 74 2913–2921 Preprint at https://doi.org/10.1158/0008-5472.CAN-14-0155 (2014). Wood, L. D., Canto, M. I., Jaffee, E. M. & Simeone, D. M. Pancreatic Cancer: Pathogenesis, Screening, Diagnosis, and Treatment. Gastroenterology vol. 163 386-402.e1 Preprint at https://doi.org/10.1053/j.gastro.2022.03.056 (2022). Zhan, H. X. et al. Neoadjuvant therapy in pancreatic cancer: a systematic review and meta-analysis of prospective studies. Cancer Med 6 , (2017). Springfeld, C. et al. Neoadjuvant therapy for pancreatic cancer. Nature Reviews Clinical Oncology vol. 20 318–337 Preprint at https://doi.org/10.1038/s41571-023-00746-1 (2023). Falcomatà, C. et al. Context-Specific Determinants of the Immunosuppressive Tumor Microenvironment in Pancreatic Cancer. Cancer Discovery vol. 13 278–297 Preprint at https://doi.org/10.1158/2159-8290.CD-22-0876 (2023). Ho, W. J., Jaffee, E. M. & Zheng, L. The tumour microenvironment in pancreatic cancer — clinical challenges and opportunities. Nature Reviews Clinical Oncology vol. 17 527–540 Preprint at https://doi.org/10.1038/s41571-020-0363-5 (2020). Zhang, S. et al. Single cell transcriptomic analyses implicate an immunosuppressive tumor microenvironment in pancreatic cancer liver metastasis. Nat Commun 14 , (2023). Kzhyshkowska, J., Shen, J. & Larionova, I. Targeting of TAMs: can we be more clever than cancer cells? Cellular and Molecular Immunology Preprint at https://doi.org/10.1038/s41423-024-01232-z (2024). Mantovani, A., Marchesi, F., Jaillon, S., Garlanda, C. & Allavena, P. Tumor-associated myeloid cells: diversity and therapeutic targeting. Cellular and Molecular Immunology vol. 18 566–578 Preprint at https://doi.org/10.1038/s41423-020-00613-4 (2021). Zhu, Y. et al. Tissue-Resident Macrophages in Pancreatic Ductal Adenocarcinoma Originate from Embryonic Hematopoiesis and Promote Tumor Progression. Immunity 47 , 323-338.e6 (2017). Kloosterman, D. J. & Akkari, L. Macrophages at the interface of the co-evolving cancer ecosystem. Cell vol. 186 1627–1651 Preprint at https://doi.org/10.1016/j.cell.2023.02.020 (2023). Caronni, N. et al. IL-1β+ macrophages fuel pathogenic inflammation in pancreatic cancer. Nature 623 , 415–422 (2023). Garlanda, C. & Mantovani, A. Interleukin-1 in tumor progression, therapy, and prevention. Cancer Cell 39 , 1023–1027 (2021). Das, S., Shapiro, B., Vucic, E. A., Vogt, S. & Bar-Sagi, D. Tumor cell-derived IL1β promotes desmoplasia and immune suppression in pancreatic cancer. Cancer Res 80 , 1088–1101 (2020). Reis, M. B. et al. Interleukin-1 receptor-induced PGE2 production controls acetylcholine-mediated cardiac dysfunction and mortality during scorpion envenomation. Nat Commun 11 , (2020). Zasłona, Z. et al. The Induction of Pro–IL-1β by Lipopolysaccharide Requires Endogenous Prostaglandin E2 Production. The Journal of Immunology 198 , 3558–3564 (2017). Thumkeo, D. et al. PGE2-EP2/EP4 signaling elicits immunosuppression by driving the mregDC-Treg axis in inflammatory tumor microenvironment. Cell Rep 39 , (2022). Lacher, S. B. et al. PGE2 limits effector expansion of tumour-infiltrating stem-like CD8+ T cells. Nature 629 , 417–425 (2024). Morotti, M. et al. PGE2 inhibits TIL expansion by disrupting IL-2 signalling and mitochondrial function. Nature 629 , 426–434 (2024). Zitvogel, L., Galluzzi, L., Kepp, O., Smyth, M. J. & Kroemer, G. Type I interferons in anticancer immunity. Nature Reviews Immunology vol. 15 405–414 Preprint at https://doi.org/10.1038/nri3845 (2015). Boukhaled, G. M., Harding, S. & Brooks, D. G. Opposing Roles of Type i Interferons in Cancer Immunity. Annual Review of Pathology: Mechanisms of Disease vol. 16 167–198 Preprint at https://doi.org/10.1146/annurev-pathol-031920-093932 (2021). Yu, R., Zhu, B. & Chen, D. Type I interferon-mediated tumor immunity and its role in immunotherapy. Cellular and Molecular Life Sciences vol. 79 Preprint at https://doi.org/10.1007/s00018-022-04219-z (2022). Borden, E. C. Interferons α and β in cancer: therapeutic opportunities from new insights. Nature Reviews Drug Discovery vol. 18 219–234 Preprint at https://doi.org/10.1038/s41573-018-0011-2 (2019). Crouse, J., Kalinke, U. & Oxenius, A. Regulation of antiviral T cell responses by type i interferons. Nature Reviews Immunology vol. 15 231–242 Preprint at https://doi.org/10.1038/nri3806 (2015). Lu, C. et al. Type i interferon suppresses tumor growth through activating the STAT3-granzyme B pathway in tumor-infiltrating cytotoxic T lymphocytes. J Immunother Cancer 7 , (2019). Duong, E. et al. Type I interferon activates MHC class I-dressed CD11b+ conventional dendritic cells to promote protective anti-tumor CD8+ T cell immunity. Immunity 55 , 308-323.e9 (2022). Cao, X. et al. Next generation of tumor-activating type I IFN enhances anti-tumor immune responses to overcome therapy resistance. Nat Commun 12 , (2021). Thyrell, L. et al. Interferon α-induced apoptosis in tumor cells is mediated through the phosphoinositide 3-kinase/mammalian target of rapamycin signaling pathway. Journal of Biological Chemistry 279 , 24152–24162 (2004). Yu, R., Zhu, B. & Chen, D. Type I interferon-mediated tumor immunity and its role in immunotherapy. Cellular and Molecular Life Sciences vol. 79 Preprint at https://doi.org/10.1007/s00018-022-04219-z (2022). Elewaut, A. et al. Cancer cells impair monocyte-mediated T cell stimulation to evade immunity. Nature (2024) doi:10.1038/s41586-024-08257-4. Hangai, S. et al. PGE2 induced in and released by dying cells functions as an inhibitory DAMP. Proc Natl Acad Sci U S A 113 , 3844–3849 (2016). Fabricius, D. et al. Prostaglandin E2 Inhibits IFN-α Secretion and Th1 Costimulation by Human Plasmacytoid Dendritic Cells via E-Prostanoid 2 and E-Prostanoid 4 Receptor Engagement. The Journal of Immunology 184 , 677–684 (2010). Cheon, H. J., Wang, Y., Wightman, S. M., Jackson, M. W. & Stark, G. R. How cancer cells make and respond to interferon-I. Trends in Cancer vol. 9 83–92 Preprint at https://doi.org/10.1016/j.trecan.2022.09.003 (2023). Arashiba, K., Miyake, Y. & Nishibayashi, Y. A molybdenum complex bearing PNP-type pincer ligands leads to the catalytic reduction of dinitrogen into ammonia. Nat Chem 3 , 120–125 (2011). Merz, L. S., Ballmann, J. & Gade, L. H. Phosphines and N-Heterocycles Joining Forces: an Emerging Structural Motif in PNP-Pincer Chemistry. European Journal of Inorganic Chemistry vol. 2020 2023–2042 Preprint at https://doi.org/10.1002/ejic.202000206 (2020). Pandey, D. K., Khaskin, E. & Khusnutdinova, J. R. PNP-Pincer Ligands Armed With Methyls: New Tools To Control Sterics and Non-Innocence. ChemCatChem 15 , (2023). Both, N. F., Spannenberg, A., Junge, K. & Beller, M. Low-Valent Molybdenum PNP Pincer Complexes as Catalysts for the Semihydrogenation of Alkynes. Organometallics 41 , 1797–1805 (2022). Matsuda, Y., Nakajima, M. & Nemoto, T. Acridine PNP-Pincer Ligands Enabling Transition Metal-Catalyzed Photoreactions. ACS Catal 13 , 10224–10231 (2023). Mitsumoto, T. et al. Catalytic Activity of Molybdenum Complexes Bearing PNP-Type Pincer Ligand toward Ammonia Formation. Angewandte Chemie - International Edition 62 , (2023). Arora, V., Narjinari, H., Nandi, P. G. & Kumar, A. Recent advances in pincer-nickel catalyzed reactions. Dalton Transactions vol. 50 3394–3428 Preprint at https://doi.org/10.1039/d0dt03593a (2021). Ai, W., Zhong, R., Liu, X. & Liu, Q. Hydride Transfer Reactions Catalyzed by Cobalt Complexes. Chemical Reviews vol. 119 2876–2953 Preprint at https://doi.org/10.1021/acs.chemrev.8b00404 (2019). Manar, K. K. & Ren, P. Recent progress on group 10 metal complexes of pincer ligands: From synthesis to activities and catalysis. in Advances in Organometallic Chemistry vol. 76 185–259 (Academic Press Inc., 2021). Chakraborty, S., Bhattacharya, P., Dai, H. & Guan, H. Nickel and Iron Pincer Complexes as Catalysts for the Reduction of Carbonyl Compounds. Acc Chem Res 48 , 1995–2003 (2015). Liu, W., Sahoo, B., Junge, K. & Beller, M. Cobalt Complexes as an Emerging Class of Catalysts for Homogeneous Hydrogenations. Acc Chem Res 51 , 1858–1869 (2018). Kempf, H. A. et al. Synthesis of Ru‐PNPC Pincer Complexes and Applications in Catalytic Hydrogenation and Dehydrogenation Reactions. ChemCatChem (2024) doi:10.1002/cctc.202401481. Burbidge, R., Trotter, M., Buxton, B. & Holden, S. Drug design by machine learning: Support vector machines for pharmaceutical data analysis. Comput Chem 26 , (2001). Tang, Y. et al. Data-driven discovery of innate immunomodulators via machine learning-guided high throughput screening. Chem Sci 14 , 12747–12766 (2023). Tanoli, Z., Vähä-Koskela, M. & Aittokallio, T. Artificial intelligence, machine learning, and drug repurposing in cancer. Expert Opinion on Drug Discovery vol. 16 977–989 Preprint at https://doi.org/10.1080/17460441.2021.1883585 (2021). Vatansever, S. et al. Artificial intelligence and machine learning-aided drug discovery in central nervous system diseases: State-of-the-arts and future directions. Medicinal Research Reviews vol. 41 1427–1473 Preprint at https://doi.org/10.1002/med.21764 (2021). Koromina, M., Pandi, M. T. & Patrinos, G. P. Rethinking Drug Repositioning and Development with Artificial Intelligence, Machine Learning, and Omics. OMICS 23 , (2019). Issa, N. T., Stathias, V., Schürer, S. & Dakshanamurthy, S. Machine and deep learning approaches for cancer drug repurposing. Seminars in Cancer Biology vol. 68 132–142 Preprint at https://doi.org/10.1016/j.semcancer.2019.12.011 (2021). Mayr, A. et al. Large-scale comparison of machine learning methods for drug target prediction on ChEMBL. Chem Sci 9 , 5441–5451 (2018). Dragan, P., Merski, M., Wiśniewski, S., Sanmukh, S. G. & Latek, D. Chemokine Receptors—Structure-Based Virtual Screening Assisted by Machine Learning. Pharmaceutics 15 , (2023). Hie, B., Bryson, B. D. & Berger, B. Leveraging Uncertainty in Machine Learning Accelerates Biological Discovery and Design. Cell Syst 11 , 461-477.e9 (2020). Vamathevan, J. et al. Applications of machine learning in drug discovery and development. Nature Reviews Drug Discovery vol. 18 463–477 Preprint at https://doi.org/10.1038/s41573-019-0024-5 (2019). Butler, K. T., Davies, D. W., Cartwright, H., Isayev, O. & Walsh, A. Machine learning for molecular and materials science. Nature vol. 559 547–555 Preprint at https://doi.org/10.1038/s41586-018-0337-2 (2018). Yang, S., Varghese, P., Stephenson, E., Tu, K. & Gronsbell, J. Machine learning approaches for electronic health records phenotyping: a methodical review. J Am Med Inform Assoc 30 , 367–381 (2023). Issa, N. T., Stathias, V., Schürer, S. & Dakshanamurthy, S. Machine and deep learning approaches for cancer drug repurposing. Seminars in Cancer Biology vol. 68 132–142 Preprint at https://doi.org/10.1016/j.semcancer.2019.12.011 (2021). Grivennikov, S. I., Greten, F. R. & Karin, M. Immunity, Inflammation, and Cancer. Cell vol. 140 883–899 Preprint at https://doi.org/10.1016/j.cell.2010.01.025 (2010). Guo, X., Zhao, Y. & You, F. Identification and characterization of endogenous retroviruses upon SARS-CoV-2 infection. Front Immunol 15 , (2024). Neufeldt, C. J. et al. SARS-CoV-2 infection induces a pro-inflammatory cytokine response through cGAS-STING and NF-κB. Commun Biol 5 , (2022). Hadjadj, J. et al. Impaired Type I Interferon Activity and Inflammatory Responses in Severe COVID-19 Patients . Raghavan, B., Martin, S. F., Esser, P. R., Goebeler, M. & Schmidt, M. Metal allergens nickel and cobalt facilitate TLR4 homodimerization independently of MD2. EMBO Rep 13 , 1109–1115 (2012). Ullman, N. A., Burchard, P. R., Dunne, R. F. & Linehan, D. C. Immunologic Strategies in Pancreatic Cancer: Making Cold Tumors Hot. Journal of Clinical Oncology vol. 40 Preprint at https://doi.org/10.1200/JCO.21.02616 (2022). Chamma, H., Vila, I. K., Taffoni, C., Turtoi, A. & Laguette, N. Activation of STING in the pancreatic tumor microenvironment: A novel therapeutic opportunity. Cancer Letters vol. 538 Preprint at https://doi.org/10.1016/j.canlet.2022.215694 (2022). Cheng, S. et al. A pan-cancer single-cell transcriptional atlas of tumor infiltrating myeloid cells. Cell 184 , 792-809.e23 (2021). Bill, R. et al. CXCL9:SPP1 macrophage polarity identifies a network of cellular programs that control human cancers. Science (1979) 381 , (2023). Gungabeesoon, J. et al. A neutrophil response linked to tumor control in immunotherapy. Cell 186 , 1448-1464.e20 (2023). Hirschhorn, D. et al. T cell immunotherapies engage neutrophils to eliminate tumor antigen escape variants. Cell 186 , 1432-1447.e17 (2023). Benguigui, M. et al. Interferon-stimulated neutrophils as a predictor of immunotherapy response. Cancer Cell 42 , 253-265.e12 (2024). Park, B. S. et al. The structural basis of lipopolysaccharide recognition by the TLR4-MD-2 complex. Nature 458 , 1191–1195 (2009). Takeuchi, O. & Akira, S. Pattern Recognition Receptors and Inflammation. Cell vol. 140 805–820 Preprint at https://doi.org/10.1016/j.cell.2010.01.022 (2010). Guo, Q. et al. Induction of alarmin S100A8/A9 mediates activation of aberrant neutrophils in the pathogenesis of COVID-19. Cell Host Microbe 29 , 222-235.e4 (2021). Zhao, Y. et al. SARS-CoV-2 spike protein interacts with and activates TLR41. Cell Research vol. 31 818–820 Preprint at https://doi.org/10.1038/s41422-021-00495-9 (2021). Mantovani, A., Allavena, P., Sica, A. & Balkwill, F. Cancer-related inflammation. Nature vol. 454 Preprint at https://doi.org/10.1038/nature07205 (2008). Kawai, T., Ikegawa, M., Ori, D. & Akira, S. Decoding Toll-like receptors: Recent insights and perspectives in innate immunity. Immunity vol. 57 649–673 Preprint at https://doi.org/10.1016/j.immuni.2024.03.004 (2024). Ohto, U., Fukase, K., Miyake, K. & Shimizu, T. Structural basis of species-specific endotoxin sensing by innate immune receptor TLR4/MD-2. Proc Natl Acad Sci U S A 109 , 7421–7426 (2012). Mata-Haro, V. et al. The vaccine adjuvant monophosphoryl lipid A as a TRIF-biased agonist of TLR4. Science (1979) 316 , 1628–1632 (2007). Ciesielska, A., Matyjek, M. & Kwiatkowska, K. TLR4 and CD14 trafficking and its influence on LPS-induced pro-inflammatory signaling. Cellular and Molecular Life Sciences vol. 78 1233–1261 Preprint at https://doi.org/10.1007/s00018-020-03656-y (2021). Płóciennikowska, A., Hromada-Judycka, A., Borzęcka, K. & Kwiatkowska, K. Co-operation of TLR4 and raft proteins in LPS-induced pro-inflammatory signaling. Cellular and Molecular Life Sciences 72 , 557–581 (2015). Rajaiah, R., Perkins, D. J., Ireland, D. D. C., Vogel, S. N. & Kagan, J. C. CD14 dependence of TLR4 endocytosis and TRIF signaling displays ligand specificity and is dissociable in endotoxin tolerance. Proc Natl Acad Sci U S A 112 , 8391–8396 (2015). Perkins, D. J. et al. Autocrine–paracrine prostaglandin E 2 signaling restricts TLR4 internalization and TRIF signaling. Nat Immunol 19 , 1309–1318 (2018). Oblak, A., Pohar, J. & Jerala, R. MD-2 determinants of nickel and cobalt-mediated activation of human TLR4. PLoS One 10 , (2015). Domingo, I. K., Groenendyk, J., Michalak, M. & Bhavsar, A. P. Cisplatin toxicity is mediated by direct binding to TLR4 through a mechanism that is distinct from metal allergens Running Title: Cisplatin activates TLR4 by Direct Binding. doi:10.1101/2022.07.06.498982. Zhao, Y. et al. Publisher Correction: SARS-CoV-2 spike protein interacts with and activates TLR4 (Cell Research, (2021), 31, 7, (818-820), 10.1038/s41422-021-00495-9). Cell Research vol. 31 Preprint at https://doi.org/10.1038/s41422-021-00501-0 (2021). Volk-Draper, L. et al. Paclitaxel therapy promotes breast cancer metastasis in a TLR4-dependent manner. Cancer Res 74 , 5421–5434 (2014). Rajput, S., Volk-Draper, L. D. & Ran, S. TLR4 is a novel determinant of the response to paclitaxel in breast cancer. Mol Cancer Ther 12 , 1676–1687 (2013). Wanderley, C. W. et al. Paclitaxel reduces tumor growth by reprogramming tumor-associated macrophages to an M1 profile in a TLR4-dependent manner. Cancer Res 78 , 5891–5900 (2018). Liu, S. et al. Critical role of Syk-dependent STAT1 activation in innate antiviral immunity. Cell Rep 34 , (2021). Lin, Y.-C., Huang, D.-Y., Chu, C.-L., Lin, Y.-L. & Lin, W.-W. I M M U N O L O G Y The Tyrosine Kinase Syk Differentially Regulates Toll-like Receptor Signaling Downstream of the Adaptor Molecules TRAF6 and TRAF3 . www.SCIENCESIGNALING.org. Joshi, S. New insights into SYK targeting in solid tumors. Trends in Pharmacological Sciences Preprint at https://doi.org/10.1016/j.tips.2024.08.006 (2024). Balic, J. J. et al. STAT3 serine phosphorylation is required for TLR4 metabolic reprogramming and IL-1β expression. Nat Commun 11 , (2020). Ren, B. et al. Tumor microenvironment participates in metastasis of pancreatic cancer. Molecular Cancer vol. 17 Preprint at https://doi.org/10.1186/s12943-018-0858-1 (2018). Cilenti, F. et al. A PGE2-MEF2A axis enables context-dependent control of inflammatory gene expression. Immunity 54 , 1665-1682.e14 (2021). Perkins, D. J. et al. Autocrine–paracrine prostaglandin E 2 signaling restricts TLR4 internalization and TRIF signaling. Nat Immunol 19 , 1309–1318 (2018). Zheng, J. et al. Comprehensive elaboration of the cGAS-STING signaling axis in cancer development and immunotherapy. Molecular Cancer vol. 19 Preprint at https://doi.org/10.1186/s12943-020-01250-1 (2020). Nalio Ramos, R. et al. Tissue-resident FOLR2+ macrophages associate with CD8+ T cell infiltration in human breast cancer. Cell 185 , 1189-1207.e25 (2022). Knudsen, N. H. & Manguso, R. T. Tumor-Derived PGE2 Gives NK Cells a Headache. Immunity vol. 53 1131–1132 Preprint at https://doi.org/10.1016/j.immuni.2020.11.018 (2020). Ng, M., Cerezo-Wallis, D., Ng, L. G. & Hidalgo, A. Adaptations of neutrophils in cancer. Immunity vol. 58 40–58 Preprint at https://doi.org/10.1016/j.immuni.2024.12.009 (2025). Khan, N. et al. β-Glucan reprograms neutrophils to promote disease tolerance against influenza A virus. Nat Immunol (2025) doi:10.1038/s41590-024-02041-2. Huang, X. et al. Neutrophils in Cancer immunotherapy: friends or foes? Molecular Cancer vol. 23 Preprint at https://doi.org/10.1186/s12943-024-02004-z (2024). Gros, A. et al. PD-1 identifies the patient-specific CD8+ tumor-reactive repertoire infiltrating human tumors. Journal of Clinical Investigation 124 , 2246–2259 (2014). Guo, X., Zhao, Y. & You, F. MOI is a comprehensive database collecting processed multi-omics data associated with viral infection. Sci Rep 14 , (2024). Jumper, J. et al. Highly accurate protein structure prediction with AlphaFold. Nature 596 , (2021). Trott, O. & Olson, A. J. AutoDock Vina: Improving the speed and accuracy of docking with a new scoring function, efficient optimization, and multithreading. J Comput Chem 31 , 455–461 (2010). Pettersen, E. F. et al. UCSF Chimera - A visualization system for exploratory research and analysis. J Comput Chem 25 , (2004). Liao, Y., Smyth, G. K. & Shi, W. The Subread aligner: Fast, accurate and scalable read mapping by seed-and-vote. Nucleic Acids Res 41 , (2013). Liao, Y., Smyth, G. K. & Shi, W. FeatureCounts: An efficient general purpose program for assigning sequence reads to genomic features. Bioinformatics 30 , 923–930 (2014). Kanehisa, M. & Goto, S. KEGG: Kyoto Encyclopedia of Genes and Genomes . Nucleic Acids Research vol. 28 http://www.genome.ad.jp/kegg/ (2000). Yu, G., Wang, L. G., Han, Y. & He, Q. Y. ClusterProfiler: An R package for comparing biological themes among gene clusters. OMICS 16 , 284–287 (2012). Dennis, G. et al. DAVID: Database for Annotation, Visualization, and Integrated Discovery . Genome Biology vol. 4 http://dot.ped.med.umich.edu:2000/ (2003). Szklarczyk, D. et al. STRING v11: Protein-protein association networks with increased coverage, supporting functional discovery in genome-wide experimental datasets. Nucleic Acids Res 47 , (2019). Shannon, P. et al. Cytoscape: A software Environment for integrated models of biomolecular interaction networks. Genome Res 13 , (2003). Janky, R. et al. iRegulon: From a Gene List to a Gene Regulatory Network Using Large Motif and Track Collections. PLoS Comput Biol 10 , (2014). Buenrostro, J. D., Wu, B., Chang, H. Y. & Greenleaf, W. J. ATAC-seq: A method for assaying chromatin accessibility genome-wide. Curr Protoc Mol Biol 2015 , (2015). Langmead, B., Trapnell, C., Pop, M. & Salzberg, S. L. Ultrafast and memory-efficient alignment of short DNA sequences to the human genome. Genome Biol 10 , (2009). Li, H. et al. The Sequence Alignment/Map format and SAMtools. Bioinformatics 25 , 2078–2079 (2009). Zhang, Y. et al. Model-based analysis of ChIP-Seq (MACS). Genome Biol 9 , (2008). Quinlan, A. R. & Hall, I. M. BEDTools: A flexible suite of utilities for comparing genomic features. Bioinformatics 26 , 841–842 (2010). Ramírez, F., Dündar, F., Diehl, S., Grüning, B. A. & Manke, T. DeepTools: A flexible platform for exploring deep-sequencing data. Nucleic Acids Res 42 , (2014). Lun, A. T. L. & Smyth, G. K. Csaw: A Bioconductor package for differential binding analysis of ChIP-seq data using sliding windows. Nucleic Acids Res 44 , e45 (2015). Yu, G., Wang, L. G. & He, Q. Y. ChIP seeker: An R/Bioconductor package for ChIP peak annotation, comparison and visualization. Bioinformatics 31 , 2382–2383 (2015). Robinson, J. T. et al. Integrative genomics viewer. Nature Biotechnology vol. 29 24–26 Preprint at https://doi.org/10.1038/nbt.1754 (2011). Satija, R., Farrell, J. A., Gennert, D., Schier, A. F. & Regev, A. Spatial reconstruction of single-cell gene expression data. Nat Biotechnol 33 , 495–502 (2015). Jin, S. et al. Inference and analysis of cell-cell communication using CellChat. Nat Commun 12 , (2021). Trapnell, C. et al. The dynamics and regulators of cell fate decisions are revealed by pseudotemporal ordering of single cells. Nat Biotechnol 32 , 381–386 (2014). Subramanian, A. et al. A Next Generation Connectivity Map: L1000 Platform and the First 1,000,000 Profiles. Cell 171 , 1437-1452.e17 (2017). Wishart, D. S. et al. DrugBank: a comprehensive resource for in silico drug discovery and exploration. Nucleic Acids Res 34 , (2006). Mendez, D. et al. ChEMBL: Towards direct deposition of bioassay data. Nucleic Acids Res 47 , D930–D940 (2019). Additional Declarations (Not answered) Supplementary Files SupplementaryMaterials.zip Supplementary Table S1. Primer sequences used in this study. Supplementary Table S2. Information on shRNA targeting Stat1 and Syk , as well asgRNA for Syk CRISPR-Cas9 knockout. OriginalImages.pptx Original images GraphicalAbCo68.pdf Graphical Abstract SupplementaryFigure1.pdf Supplementary Figure Legends: Supplementary Figure 1. Effect of Co68 on cytokine and interferon expression in macrophages. A) Heatmap showing the differential expression of inflammatory cytokines and interferons in peripheral blood mononuclear cells (PBMCs) from COVID-19 patients compared to healthy controls. B) Volcano plot depicting the distribution of differentially expressed genes (DEGs) between lipopolysaccharide (LPS)- and dimethyl sulfoxide (DMSO)-treated groups in RAW 264.7 cells. The plot displays fold changes in gene expression, with upregulated DEGs in the LPS group shown in red and downregulated DEGs in blue. C) Gene Ontology (GO) enrichment analysis of DEGs in the LPS group compared to the DMSO group. D) The sticks representation of the structure of the PNP ligand used for Co68 synthesis and the synthesis reaction of Co68 from PNP ligand and CoCl₂. E) Cell viability of J774A.1 macrophages and peritoneal macrophages (PMs) treated with Co68 for 6 hours at the indicated concentrations. F-I) Ifnb1 expression levels in J774A.1 macrophages and PMs following Co68 treatment, measured in a dose- and time-dependent manner. J) Co68 elevated Ifnb1 expression across M0, M1, and M2 subtypes of THP-1 monocytes. K-L) Isg15 and Ifit3 were strongly upregulated in multiple organs and blood of C57BL/6J mice treated with Co68 compared to the PBS group. M) Co68 was unstable after dissolution in DMSO and almost completely failed to induce Ifnb1 expression by Day 3. N-P) Exposure to cobalt chloride (CoCl₂) resulted in the induction of Ifnb1 , Il1b , and Il6 in RAW264.7 macrophages at higher concentrations. Q-R) The expression of Ifna4 (Q) and Il1a (R) in RAW264.7 macrophages upon treatment with DMSO, Co68, CoCl₂, and LPS. RT-qPCR data were presented as means ± SEM from three independent experiments. Statistical significance was determined using one-way ANOVA with Bonferroni's multiple comparisons test or paired-sample t-test. ns, not significant; * P < 0.05, ** P < 0.01, *** P < 0.001. SupplementaryFigure2.pdf Supplementary Figure 2. Co68 activates innate immunity and reduces inflammation. A) Heatmap of innate immune response-related genes differentially expressed in Co68-treated vs. DMSO-treated groups. B-C) Protein-protein interaction (PPI) network analysis showing upregulated genes in Co68-treated cells are enriched in regulation of type I interferon production and negative regulation of viral genome replication pathways. D) Gene Ontology (GO) enrichment analysis of downregulated genes in Co68-treated cells compared to the DMSO group. E) Western blot analysis showing Co68 enhances phosphorylation of TBK1 and IRF3 at varying concentrations. F) Co68 reduces expression of pro-inflammatory cytokines and chemokines ( Il1a , Csf1 , Csf2 , Csf3 ) in LPS-stimulated cells. G) Co68 promotes expression of immune markers ( Tnf , Ifnb1 , Cxcl10 , Isg15 ) under LPS stimulation. H) Gene Set Enrichment Analysis (GSEA) showing that Co68 downregulates inflammation-related genes compared to LPS alone. I) Co68 reduces expression of Nos2 and Ptgs2 at protein level induced by LPS. RT-qPCR data were presented as means ± SEM from three independent experiments. Statistical significance was determined using one-way ANOVA with Bonferroni’s multiple comparisons test. * P < 0.05, ** P < 0.01, *** P < 0.001. SupplementaryFigure3.pdf Supplementary Figure 3. Co68 exerts a potent antitumor effect. A, B) Quantification of tumor weight in Pan02 tumor-bearing wild-type (WT) mice (A, n = 6 per group) or IFNAR1 knockout (KO) mice (B, n = 7 per group) treated with Co68 or PBS via intraperitoneal (i.p.) injection every other day for two weeks. C) Tumor weight quantification of Pan02 tumors in WT mice treated with isotype control antibody (200 µg/mouse, i.p.), Co68 (25 mg/kg, i.p.), anti-PD-1 antibody (200 µg/mouse, i.p.), or Co68 combined with anti-PD-1 antibody (n = 5 per group). D, E, F) Representative images (D) and quantification of tumor size (E) and weight (F) of Pan02 tumors in WT mice treated with DMXAA (25 mg/kg, i.p.) or PBS, administered every other day for two weeks (n = 6 per group). G, H, I) Representative images (G) and quantification of tumor size (H) and weight (I) of MC38 tumors in WT mice treated with Co68 (25 mg/kg, i.p.) or PBS via i.p. injection every other day for two weeks (n = 6 per group). J) t-SNE analysis illustrating immune cell clustering in Pan02 tumors, with clusters color-coded according to their identity. K) Comparison of immune cell clusters between Co68- and PBS-treated Pan02 tumors, indicating immune reprogramming following Co68 treatment. L) Dot plot displaying the expression of key marker genes across immune cell clusters. Data were presented as mean ± SEM from three independent experiments. Statistical significance was assessed using paired t-test. ns , not significant; P < 0.05; *P < 0.01. SupplementaryFigure4.pdf Supplementary Figure 4. Co68 reprograms the tumor microenvironment (TME) in Pan02 tumors. A) Interaction network of cell types across all cell clusters in Pan02 tumors treated with PBS or Co68. B) Interaction networks illustrating the communication between the Cancer cell1 cluster and other cell types within the TME. C-D) Cell interaction networks of tumor-associated macrophages (TAMs)2 (C) and monocytes (Mono) (D) with other cell clusters in the Pan02 TME. E) Chord diagrams depicting the SPP1 signaling pathway networks originating from TAMs and interacting with other cell populations within the pancreatic cancer TME. F) t-SNE plot illustrating subclusters of tumor-associated neutrophils (TANs) in Co68- and PBS-treated Pan02 tumors. G) Violin plots showing the expression of interferon-stimulated genes (ISGs) across TAN subpopulations, highlighting the upregulation of ISGs in Co68-induced TANs1. H) Chord diagram illustrating marker gene expression and biological processes associated with TANs1 in Co68-treated tumors. I) t-SNE plot showing cell clustering in Pan02 tumors treated with Co68 or DMXAA. J) Heatmap of gene expression profiles across various cell types from subclustering analysis of tumor-associated macrophages (TAMs) in Co68 and DMXAA treatment groups. K) Chord diagrams depicting differentially expressed genes (DEGs) and associated biological processes in IL1B+ TAMs from Co68-treated tumors compared to the DMXAA group. L) Violin plots comparing the expression of ISGs and pattern recognition receptors (PRRs) in IL1B+ TAMs from Co68- and DMXAA-treated tumors. SupplementaryFigure5.pdf Supplementary Figure 5. Co68 directly binds to TLR4. A) Co68 did not induce Ifnb1 expression in BMDMs from IRF3 knockout mice. B-D) Neutralizing antibodies against TLR4, CD14, or MD2 suppressed Co68-induced Isg15 expression in RAW 264.7 cells. E) Co-transfection of TLR4, CD14, and MD2 plasmids enhanced Co68-induced Isg15 expression in HEK293T cells. F) MST analysis revealed a K d of 4.37 µM for LPS binding to TLR4. G) SPR analysis showed Co68 binding to TLR4 with a K d of 31.71 µM. H) SPR analysis of LPS binding to TLR4, with sensorgram data showing RU at increasing LPS concentrations. I) SPR analysis of LPS-TLR4 dose-response curve with a K d of 147.64 µM. J) Electron microscopy structure of Co68, with key atoms in orange (phosphorus), blue (nitrogen), magenta (carbon), green (chloride), and pink (cobalt). K) TLR4-MD2 dimer with Co68 positioned in the MD2 binding pocket. L) Ribbon and surface representations of TLR4-MD2-Co68 complex. M-N) Detailed interactions between Co68, MD2 residues, and TLR4. O) Close-up molecular interface showing interactions between Co68 and MD2 and TLR4. P) Synthesis of Co65–Co74 and their chemical structures. RT-qPCR data were presented as means ± SEM from three independent experiments. Statistical significance was determined using one-way ANOVA with Bonferroni's multiple comparisons test or paired-sample t-test. ns, not significant; * P < 0.05, ** P < 0.01, *** P < 0.001. SupplementaryFigure6.pdf Supplementary Figure 6. Co68 induces the IFN-I signaling via TLR4. A-F) The dose-response curve of SPR analysis for six Ifnb1 -inducible analogs of Co68 with TLR4, including Co65 (A), Co69 (B), Co70 (C), Co71 (D), Co72 (E), and Co73 (F). G,I,J) Representative images (G), tumor size and weight quantification (I), and growth curves (J) of Pan02 tumors in TLR4 KO mice (n = 6) treated with Co68 or PBS via i.p. injection every other day for two weeks. H, K, L) Representative images (H), tumor size and weight quantification (K), and growth curves (L) of Pan02 tumors in IRF3 KO mice (n = 6) treated with Co68 or PBS via i.p. injection every other day for two weeks. SupplementaryFigure7.pdf Supplementary Figure 7. Co68 induces the non-classic activation of STAT1 via TLR4-SYK. A-C) Number and proportion of peaks identified in LPS (A), LPS-Co68 (B), and Co68 (C) groups, showing distinct chromatin accessibility profiles for Co68 compared to LPS and DMSO. D-E) Distribution of peaks in various genomic regions and transcription factor-binding loci relative to the TSS among LPS, LPS-Co68 and Co68 groups. F-H) Comparation of binding signal intensity of STAT1, EGR1, and REL between Co68 and LPS groups. I-K) IGV visualization of peak tracks for Usp18 , Ifit1 , Il1b , and Tnf . L) Heatmap showing early upregulation of ISGs following Co68 treatment for 1 hour. M) Co-immunoprecipitation (CoIP) of TLR4 from RAW 264.7 cells upon the DMSO, LPS and Co68 treatment. N) Knockdown efficiency of Stat1 and Syk in RAW 264.7 cells, confirmed by RT-qPCR. O) SYK knockdown abrogates the anti-inflammatory effect of Co68 in LPS-stimulated RAW 264.7 cells. P) Western blot confirming SYK knockout in RAW 264.7 cells, ensuring the observed effects were due to SYK deficiency. Q) Western blot showing that Co68 fails to inhibit LPS-induced phosphorylation of P65 and IκBα in SYK knockout cells, indicating SYK is essential for Co68’s anti-inflammatory effects. RT-qPCR data were presented as means ± SEM from three independent experiments. Statistical significance was analyzed using paired t-test. ns, not significant; * P < 0.05; ** P < 0.01. Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-7085332","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":485848914,"identity":"5ff7878b-2f50-4e28-a1b1-087bf35ad7fd","order_by":0,"name":"Xuefei Guo","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA0UlEQVRIiWNgGAWjYLCCigoJOfv2xsaHH4jWcuaMjbEBz+FmYwmitZxtS0vcIJHeJsBDjGr5iORjEgfYDjNul3zYxiDBYCen20BAi+GNtGSDAzyHmS1nJ7Y9KGBINjY7QEjLjBzDxx8kDrMx3E5sN5BgOJC4jQgtBgcOGBzmYbh5sE2Chxgt8hI5hg8OJKRJGNxgJFKLAc8zoF8O2BhI9iQCA9mACL/ItwND7OA/ifp+9uMPH36osJMjqMUAVYEBAeVgWxqIUDQKRsEoGAUjHAAAZ3lHYXMurKUAAAAASUVORK5CYII=","orcid":"","institution":"Peking University Health Science Center","correspondingAuthor":true,"prefix":"","firstName":"Xuefei","middleName":"","lastName":"Guo","suffix":""},{"id":485848915,"identity":"ce6a37ea-6d7e-4455-aed7-2beac5966f73","order_by":1,"name":"Yang Zhao","email":"","orcid":"","institution":"Peking University Health Science Center","correspondingAuthor":false,"prefix":"","firstName":"Yang","middleName":"","lastName":"Zhao","suffix":""},{"id":485848916,"identity":"88dd4b4d-92d2-4b96-b4be-7fcdfebb1745","order_by":2,"name":"Xianle Rong","email":"","orcid":"","institution":"Center of Basic Molecular Science (CBMS), Department of Chemistry, Tsinghua University","correspondingAuthor":false,"prefix":"","firstName":"Xianle","middleName":"","lastName":"Rong","suffix":""},{"id":485848917,"identity":"b22f3340-197f-417d-a31a-a1b1bfe8edbb","order_by":3,"name":"Xingyu Chen","email":"","orcid":"","institution":"Peking University Health Science Center","correspondingAuthor":false,"prefix":"","firstName":"Xingyu","middleName":"","lastName":"Chen","suffix":""},{"id":485848918,"identity":"af142723-866d-4f93-a226-361bb584c57e","order_by":4,"name":"Xiao Wang","email":"","orcid":"","institution":"Peking University Health Science Center","correspondingAuthor":false,"prefix":"","firstName":"Xiao","middleName":"","lastName":"Wang","suffix":""},{"id":485848919,"identity":"6ec0411d-4ffd-426c-b56b-6434ee2e52b4","order_by":5,"name":"Tian Liu","email":"","orcid":"","institution":"Peking University Health Science Center","correspondingAuthor":false,"prefix":"","firstName":"Tian","middleName":"","lastName":"Liu","suffix":""},{"id":485848920,"identity":"ad8e9119-2750-4411-a88e-fd016861eac6","order_by":6,"name":"Yunfei Xie","email":"","orcid":"","institution":"Peking University Health Science Center","correspondingAuthor":false,"prefix":"","firstName":"Yunfei","middleName":"","lastName":"Xie","suffix":""},{"id":485848921,"identity":"7bef5baa-f334-46e0-ac28-1e5c829be726","order_by":7,"name":"Yushu Zou","email":"","orcid":"","institution":"Peking University Health Science Center","correspondingAuthor":false,"prefix":"","firstName":"Yushu","middleName":"","lastName":"Zou","suffix":""},{"id":485848922,"identity":"3477b572-13f1-4763-980f-fe07722037ed","order_by":8,"name":"Tian Ming Chu","email":"","orcid":"","institution":"Key Laboratory of Infection and Immunity, National Laboratory of Macromolecules, Institute of Biophysics, Chinese Academy of Sciences, Beijing","correspondingAuthor":false,"prefix":"","firstName":"Tian","middleName":"Ming","lastName":"Chu","suffix":""},{"id":485848923,"identity":"01bba281-e4a0-411c-988e-c7dc3ac53157","order_by":9,"name":"Xiangxi Wang","email":"","orcid":"","institution":"","correspondingAuthor":false,"prefix":"","firstName":"Xiangxi","middleName":"","lastName":"Wang","suffix":""},{"id":485848924,"identity":"0ae94eab-ce4d-48d8-90db-d0355eca7291","order_by":10,"name":"Qiang Liu","email":"","orcid":"","institution":"Center of Basic Molecular Science (CBMS), Department of Chemistry, Tsinghua University","correspondingAuthor":false,"prefix":"","firstName":"Qiang","middleName":"","lastName":"Liu","suffix":""},{"id":485848925,"identity":"8761d00d-3990-406a-b6c8-3a84299f80b0","order_by":11,"name":"Fuping You","email":"","orcid":"https://orcid.org/0000-0002-7444-729X","institution":"Peking University Health Science Center","correspondingAuthor":false,"prefix":"","firstName":"Fuping","middleName":"","lastName":"You","suffix":""}],"badges":[],"createdAt":"2025-07-09 15:25:58","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-7085332/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-7085332/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":87024410,"identity":"0ad53b69-168b-45ca-92de-e032f7dffbc2","added_by":"auto","created_at":"2025-07-18 11:49:15","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":406252,"visible":true,"origin":"","legend":"\u003ch5\u003e\u003cstrong\u003eIdentification of immunomodulatory compounds using DLINP based on convolutional neural networks (CNN).\u003c/strong\u003e\u003c/h5\u003e\n\u003cp\u003e\u003cstrong\u003eA)\u003c/strong\u003e Overview of the experimental design and methodologies used in this study.\u003cbr\u003e\n\u003cstrong\u003eB)\u003c/strong\u003e Schematic illustrating the development of the CNN model, DLINP, which identifies compounds that both inhibit NF-κB-mediated inflammatory responses and promote innate immune activation through IRF3.\u003cbr\u003e\n\u003cstrong\u003eC, D)\u003c/strong\u003e Performance evaluation of DLINP using a confusion matrix (C) and a receiver operating characteristic (ROC) curve (D).\u003cbr\u003e\n\u003cstrong\u003eE)\u003c/strong\u003e Chemical structure of Co68.\u003cbr\u003e\n\u003cstrong\u003eF)\u003c/strong\u003e Cell viability of RAW264.7 macrophages and bone marrow-derived macrophages (BMDMs) treated with Co68 for 6 hours at varying concentrations.\u003cbr\u003e\n\u003cstrong\u003eG-J)\u003c/strong\u003e Dose- and time-dependent induction of \u003cem\u003eIfnb1\u003c/em\u003e expression in RAW264.7 macrophages and BMDMs following Co68 treatment.\u003cbr\u003e\n\u003cstrong\u003eK)\u003c/strong\u003e Elevated \u003cem\u003eIfnb1\u003c/em\u003e expression in multiple organs and blood of C57BL/6J mice after 24 hours of Co68 treatment, as measured by qRT-PCR.\u003cbr\u003e\n\u003cstrong\u003eL-N)\u003c/strong\u003e Expression levels of \u003cem\u003eIfnb1\u003c/em\u003e, \u003cem\u003eIl1b\u003c/em\u003e, and \u003cem\u003eIl6\u003c/em\u003e in RAW264.7 macrophages treated with Co68, cobalt chloride (CoCl2), or lipopolysaccharide (LPS). RT-qPCR data were expressed as mean ± SEM from three independent experiments. Statistical significance was assessed by one-way ANOVA followed by Bonferroni’s multiple comparisons test (G-N). * \u003cem\u003eP\u003c/em\u003e \u0026lt; 0.05, ** \u003cem\u003eP\u003c/em\u003e \u0026lt; 0.01, *** \u003cem\u003eP\u003c/em\u003e \u0026lt; 0.001.\u003c/p\u003e","description":"","filename":"fig1.png","url":"https://assets-eu.researchsquare.com/files/rs-7085332/v1/16c739b58f14a6b120912fd7.png"},{"id":87024375,"identity":"d19f899a-954b-461d-befc-d03248100edc","added_by":"auto","created_at":"2025-07-18 11:49:12","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":303132,"visible":true,"origin":"","legend":"\u003ch5\u003e\u003cstrong\u003eCo68 activates the innate immune response while suppressing inflammatory responses.\u003c/strong\u003e\u003c/h5\u003e\n\u003cp\u003e\u003cstrong\u003eA)\u003c/strong\u003e Volcano plot displaying log2-transformed fold changes of differentially expressed genes (DEGs) between Co68- and DMSO-treated RAW 264.7 cells. Red dots indicate upregulated DEGs in the Co68 group, and blue dots represent downregulated DEGs.\u003cbr\u003e\n\u003cstrong\u003eB)\u003c/strong\u003e Protein-protein interaction (PPI) network analysis of upregulated DEGs in Co68-treated RAW 264.7 cells, highlighting key genes involved in the innate immune response.\u003cbr\u003e\n\u003cstrong\u003eC)\u003c/strong\u003e Gene Ontology (GO) enrichment analysis of upregulated DEGs in Co68-treated RAW 264.7 cells, revealing significant biological processes related to the innate immune response.\u003cbr\u003e\n\u003cstrong\u003eD)\u003c/strong\u003e Co68 treatment reduces the expression of pro-inflammatory cytokines (\u003cem\u003eIl1b\u003c/em\u003e, \u003cem\u003eIl6\u003c/em\u003e) and inflammatory mediators (\u003cem\u003ePtgs2\u003c/em\u003e, \u003cem\u003eNos2\u003c/em\u003e) in RAW 264.7 cells following lipopolysaccharide (LPS) stimulation.\u003cbr\u003e\n\u003cstrong\u003eE)\u003c/strong\u003e Co68 suppresses \u003cem\u003eIl1b\u003c/em\u003e expression induced by vesicular stomatitis virus (VSV, multiplicity of infection [MOI] = 0.1), herpes simplex virus 1 (HSV-1, F strain, MOI = 0.5), and encephalomyocarditis virus (EMCV, MOI = 0.1) in RAW 264.7 cells.\u003cbr\u003e\n\u003cstrong\u003eF)\u003c/strong\u003e Co68 improves survival rates in LPS-challenged mice (n = 20, 20 mg/kg).\u003cbr\u003e\n\u003cstrong\u003eG)\u003c/strong\u003e Heatmap illustrating the differential gene expression patterns between LPS- and Co68-treated RAW 264.7 cells.\u003cbr\u003e\n\u003cstrong\u003eH)\u003c/strong\u003e Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis of downregulated DEGs in the LPS-Co68 group compared to the LPS group, highlighting enriched inflammation-related pathways.\u003cbr\u003e\n\u003cstrong\u003eI)\u003c/strong\u003e Transcription factor enrichment analysis (TFEA) of downregulated DEGs in the LPS-Co68 group compared to the LPS group identifies RelA as a potential key regulator. Yellow nodes represent transcription factors, and blue nodes represent downregulated genes by Co68.\u003cbr\u003e\n\u003cstrong\u003eJ)\u003c/strong\u003e Western blot analysis showing that Co68 inhibits LPS-induced phosphorylation of P65 and IκBα, key regulators of the NF-κB pathway. RT-qPCR data are presented as means ± SEM from three independent experiments. Statistical significance was determined by one-way ANOVA with Bonferroni’s multiple comparisons test (D, E), or the log-rank test (F). * \u003cem\u003eP\u003c/em\u003e \u0026lt; 0.05, ** \u003cem\u003eP\u003c/em\u003e \u0026lt; 0.01, and *** \u003cem\u003eP\u003c/em\u003e \u0026lt; 0.001.\u003c/p\u003e","description":"","filename":"fig2.png","url":"https://assets-eu.researchsquare.com/files/rs-7085332/v1/13c3e906828ae4c47828bd81.png"},{"id":87024378,"identity":"f15120ea-166a-4705-97fc-d758484d6ca2","added_by":"auto","created_at":"2025-07-18 11:49:12","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":401501,"visible":true,"origin":"","legend":"\u003ch5\u003e\u003cstrong\u003eCo68 treatment inhibits pancreatic cancer tumor (Pan02) growth in mice.\u003c/strong\u003e\u003c/h5\u003e\n\u003cp\u003e\u003cstrong\u003eA, B, C)\u003c/strong\u003e Representative images (A), quantification of tumor size (B), and growth curves (C) of Pan02 tumors in wild-type (WT) mice (n = 6 per group) treated with Co68 (25 mg/kg, i.p.) or PBS via intraperitoneal (i.p.) injection every other day for two weeks.\u003cbr\u003e\n\u003cstrong\u003eD, E, F)\u003c/strong\u003e Representative images (D), quantification of tumor size (E), and growth curves (F) of Pan02 tumors in IFNAR1 knockout (KO) mice (n = 7 per group) treated with Co68 (25 mg/kg, i.p.) or PBS via i.p. injection every other day for two weeks.\u003cbr\u003e\n\u003cstrong\u003eG)\u003c/strong\u003e Tumor images of subcutaneous Pan02 implants in mice treated with isotype control antibody (200 µg/mouse, i.p.), Co68 (25 mg/kg, i.p.), anti-PD-1 antibody (200 µg/mouse, i.p.), or Co68 combined with anti-PD-1 antibody (n = 5 per group).\u003cbr\u003e\n\u003cstrong\u003eH)\u003c/strong\u003e Quantification of tumor size of subcutaneous Pan02 implants in the same treatment groups as in panel G.\u003cbr\u003e\n\u003cstrong\u003eI)\u003c/strong\u003e Comparison of growth curves for subcutaneous Pan02 tumors between the Co68 and AntiPD1-Co68 treatment groups as in panel J.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eJ)\u003c/strong\u003e Tumor images of subcutaneous Pan02 implants in mice treated with DMXAA (25 mg/kg, i.p.), Co68 (25 mg/kg, i.p.), or solvent control (n = 5 per group).\u003cbr\u003e\n\u003cstrong\u003eK)\u003c/strong\u003e Representative growth curves of subcutaneous Pan02 tumors in the same treatment groups as in panel J.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eL)\u003c/strong\u003e Flow cytometry analysis of CD4 T cells, CD8 T cells, and NK cells in Pan02 tumors, showing increased cytotoxic immune cells in Co68-treated tumors compared to the PBS group.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eM)\u003c/strong\u003e t-SNE plot illustrating cell population clustering and annotation in Pan02 tumor samples treated with PBS or Co68, with clusters color-coded by cell type.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eN)\u003c/strong\u003e Bar chart comparing the proportions of major cell types between Co68- and PBS-treated groups in Pan02 tumors.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eO)\u003c/strong\u003e Gene Ontology (GO) enrichment analysis of upregulated differentially expressed genes (DEGs) in CD8\u003csup\u003e+\u003c/sup\u003e T cells following Co68 treatment compared to PBS. Data were presented as means ± SEM for the indicated number of mice per group. Data are representative of three independent experiments. Statistical significance was determined by one-way ANOVA with Bonferroni’s multiple comparisons test. \u003cem\u003ens\u003c/em\u003e, not significant, \u003cem\u003eP\u003c/em\u003e \u0026gt; 0.05; * \u003cem\u003eP\u003c/em\u003e \u0026lt; 0.05; ** \u003cem\u003eP\u003c/em\u003e \u0026lt; 0.01; *** \u003cem\u003eP\u003c/em\u003e \u0026lt; 0.001.\u003c/p\u003e","description":"","filename":"fig3.png","url":"https://assets-eu.researchsquare.com/files/rs-7085332/v1/cb683da82f526d160ba9490a.png"},{"id":87024377,"identity":"f0b57d41-a1ca-41b5-ab89-e43fac810d2a","added_by":"auto","created_at":"2025-07-18 11:49:12","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":231894,"visible":true,"origin":"","legend":"\u003ch5\u003e\u003cstrong\u003eCo68 reprograms tumor-associated macrophages (TAMs) within the tumor microenvironment (TME) in the Pan02 model.\u003c/strong\u003e\u003c/h5\u003e\n\u003cp\u003e\u003cstrong\u003eA)\u003c/strong\u003e Proportional distribution of tumor-associated cell types in Pan02 tumors.\u003cbr\u003e\n\u003cstrong\u003eB)\u003c/strong\u003e Gene set enrichment analysis (GSEA) of tumor-associated macrophages (TAMs), showing enriched innate immune pathways in Co68-treated tumors compared to PBS.\u003cbr\u003e\n\u003cstrong\u003eC)\u003c/strong\u003e UMAP visualization showing tumor-associated macrophages (TAMs) reclustering into three subpopulations in Co68- and PBS-treated groups for downstream pseudotime analysis.\u003cbr\u003e\n\u003cstrong\u003eD)\u003c/strong\u003e Pseudotime trajectory of TAMs in Co68- and PBS-treated groups visualized by UMAP, with differentiation states indicated by a purple-to-yellow gradient.\u003cbr\u003e\n\u003cstrong\u003eE)\u003c/strong\u003e Expression patterns of marker genes along the pseudotime trajectory of TAMs, with colors representing expression intensity.\u003cbr\u003e\n\u003cstrong\u003eF)\u003c/strong\u003e Chord diagram showing associations between marker genes and key biological processes in the yellow color cluster at the end of the pseudotime trajectory analysis. Line thickness represents the strength of the association.\u003cbr\u003e\n\u003cstrong\u003eG)\u003c/strong\u003e t-SNE plot of TAM subclusters (IL1B TAMs, S100A10 TAMs, HSP TAMs) in Co68- and PBS-treated groups..\u003cbr\u003e\n\u003cstrong\u003eH)\u003c/strong\u003e Violin plots showing the expression of \u003cem\u003eClec2d\u003c/em\u003e, \u003cem\u003eIfih1\u003c/em\u003e, \u003cem\u003eCxcl10\u003c/em\u003e, \u003cem\u003eLy96\u003c/em\u003e, \u003cem\u003eCd14\u003c/em\u003e, \u003cem\u003eIsg15\u003c/em\u003e, \u003cem\u003eOasl1\u003c/em\u003e, \u003cem\u003eRsad2\u003c/em\u003e, and \u003cem\u003eIfitm3\u003c/em\u003e in IL1B TAMs, which were significantly upregulated in Co68-treated groups compared to PBS.\u003cbr\u003e\n\u003cstrong\u003eI)\u003c/strong\u003e GSEA of IL1B TAMs comparing Co68 and PBS treatment groups, highlighting enriched pathways related to IFN-I signaling in the Co68 group.\u003cbr\u003e\n\u003cstrong\u003eJ)\u003c/strong\u003e The interaction network of TAMs with other cell types in the tumor microenvironment shows predicted interactions. Line thickness indicates the strength of the interactions.\u003cbr\u003e\n\u003cstrong\u003eK)\u003c/strong\u003e Bar plot comparing signaling pathways enriched in Co68- and PBS-treated groups in IL1B TAMs. Blue bars represent pathways enriched in PBS, while red bars represent pathways enriched in Co68.\u003cbr\u003e\n\u003cstrong\u003eL, M)\u003c/strong\u003e Chord diagrams representing IFN-I (M) and IL1B signaling pathway (N) networks in Co68-treated (M) and PBS-treated (N) groups, respectively. Chord width reflects interaction strength between cell populations.\u003cbr\u003e\n\u003cstrong\u003eN)\u003c/strong\u003e Comparison of distribution of TAN subpopulations (PMNs, TANs1, and TANs2) between Co68 and PBS groups.\u003cbr\u003e\n\u003cstrong\u003eO)\u003c/strong\u003e t-SNE plot illustrating cell type annotations in Pan02 tumors from Co68- and DMXAA-treated groups.\u003cbr\u003e\n\u003cstrong\u003eP)\u003c/strong\u003e t-SNE plot showing four TAM subclusters (IL1B TAMs, S100A10 TAMs, MKI67 TAMs, and HSP TAMs) in Pan02 tumors from Co68- and DMXAA-treated groups.\u003cbr\u003e\n\u003cstrong\u003eQ)\u003c/strong\u003e Volcano plot of DEGs in IL1B TAMs comparing Co68- and DMXAA-treated groups. Upregulated genes in Co68 group are shown in red, downregulated genes in Co68 group in blue, and non-significant genes in gray.\u003cbr\u003e\n\u003cstrong\u003eR)\u003c/strong\u003e Violin plots comparing the expression of inflammation-related genes (\u003cem\u003eIl1a\u003c/em\u003e, \u003cem\u003eIl1b\u003c/em\u003e, \u003cem\u003ePtgs2\u003c/em\u003e, and \u003cem\u003eIlrun\u003c/em\u003e) between Co68- and DMXAA-treated groups.\u003c/p\u003e","description":"","filename":"fig4.png","url":"https://assets-eu.researchsquare.com/files/rs-7085332/v1/6a9e5d2d5b74fdb1c22d632e.png"},{"id":87024376,"identity":"f440cfcb-170a-4dac-82b1-b9f8daebde9d","added_by":"auto","created_at":"2025-07-18 11:49:12","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":1163316,"visible":true,"origin":"","legend":"\u003ch5\u003e\u003cstrong\u003eCo68 activates IFN-I signaling by directly targeting TLR4.\u003c/strong\u003e\u003c/h5\u003e\n\u003cp\u003e\u003cstrong\u003eA-B)\u003c/strong\u003e Co68-induced \u003cem\u003eIfnb1\u003c/em\u003e expression was significantly reduced in RAW 264.7 and J774A.1 macrophages following inhibition of the Toll-like receptor 4 (TLR4) signaling pathway, when using inhibitors of TLR4, TRIF, and TBK1.\u003cbr\u003e\n\u003cstrong\u003eC-D)\u003c/strong\u003e Co68 failed to induce \u003cem\u003eIfnb1\u003c/em\u003e expression in bone marrow-derived macrophages (BMDMs) from TLR4 knockout (KO) or TRIF KO mice.\u003cbr\u003e\n\u003cstrong\u003eE-G)\u003c/strong\u003e Neutralizing antibodies targeting TLR4, CD14, or MD2 significantly suppressed Co68-induced \u003cem\u003eIfnb1\u003c/em\u003e expression in RAW 264.7 cells. Cells were pretreated with IgG isotype control or specific neutralizing antibodies for 48 hours, followed by Co68 treatment for 3 hours.\u003cbr\u003e\n\u003cstrong\u003eH)\u003c/strong\u003e Co-transfection of TLR4, CD14, and MD2 plasmids enabled Co68-induced \u003cem\u003eIfnb1\u003c/em\u003eexpression in HEK293T cells.\u003cbr\u003e\n\u003cstrong\u003eI)\u003c/strong\u003e Dose-response curve from microscale thermophoresis (MST) showing a dissociation constant (K\u003cem\u003ed\u003c/em\u003e) of 0.98 µM for the interaction between Co68 and TLR4.\u003cbr\u003e\n\u003cstrong\u003eJ)\u003c/strong\u003e Surface plasmon resonance (SPR) analysis of Co68 binding to TLR4, with a real-time binding curve and a calculated K\u003cem\u003ed\u003c/em\u003e of 33.19 µM.\u003cbr\u003e\n\u003cstrong\u003eK)\u003c/strong\u003e Molecular docking revealed that Co68 resides in the MD2 hydrophobic pocket of the TLR4-MD2 complex.\u003cbr\u003e\n\u003cstrong\u003eL)\u003c/strong\u003e Co68 was predicted to interact with both TLR4 and MD2 simultaneously (shown in red).\u003cbr\u003e\n\u003cstrong\u003eM)\u003c/strong\u003e Amino acid sites where Co68 may interact with TLR4 and MD2.\u003cbr\u003e\n\u003cstrong\u003eN)\u003c/strong\u003e Hydrogen bonds formed by Co68 atoms interacting with MD2 amino acids.\u003cbr\u003e\n\u003cstrong\u003eO)\u003c/strong\u003e AlphaFold3-predicted sites for direct interaction between Co ions and TLR4.\u003cbr\u003e\n\u003cstrong\u003eP)\u003c/strong\u003e TLR4 mutations alter Co68-induced IFNβ-Luciferase reporter activity, highlighting critical residues for functional interaction.\u003cbr\u003e\n\u003cstrong\u003eQ)\u003c/strong\u003e Mutations in TLR4 did not significantly impact NF-κB Luciferase reporter activity in response to CoCl2 or lipopolysaccharide (LPS) treatment.\u003cbr\u003e\n\u003cstrong\u003eR)\u003c/strong\u003e Comparison of Co68 and its nine analogs in RAW 264.7 macrophages demonstrates differential induction of \u003cem\u003eIfnb1\u003c/em\u003eexpression. RT-qPCR data and Luciferase assays were presented as means ± SEM from three independent experiments. Statistical significance was determined by one-way ANOVA with Bonferroni’s multiple comparisons test or paired-samples t-test. * \u003cem\u003eP\u003c/em\u003e \u0026lt; 0.05, ** \u003cem\u003eP\u003c/em\u003e \u0026lt; 0.01, and *** \u003cem\u003eP\u003c/em\u003e \u0026lt; 0.001.\u003c/p\u003e","description":"","filename":"fig5.png","url":"https://assets-eu.researchsquare.com/files/rs-7085332/v1/f1f57f7e0881bdab195f3ba8.png"},{"id":87025216,"identity":"00961f17-d836-47cf-84b0-ef009c94cfb6","added_by":"auto","created_at":"2025-07-18 11:57:13","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":794247,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eCo68 abolishes the inflammatory response via non-classical activation of STAT1.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eA)\u003c/strong\u003e Distribution of peaks around the transcription start site (TSS) in the LPS, LPS-Co68, and Co68 groups compared to the DMSO group.\u003cbr\u003e\n \u003cstrong\u003eB)\u003c/strong\u003e Gene Ontology (GO) annotation of differentially accessible regions (DARs) near genes involved in immune response-related pathways.\u003cbr\u003e\n \u003cstrong\u003eC)\u003c/strong\u003e Volcano plot showing differential enrichments of transcription factors (TFs) predicted by motif enrichment analysis between the Co68 and LPS treatment groups.\u003cbr\u003e\n \u003cstrong\u003eD)\u003c/strong\u003e STAT1 binding motifs derived from the JASPAR database.\u003cbr\u003e\n \u003cstrong\u003eE)\u003c/strong\u003e Mean aggregate binding signal of STAT1 across LPS, LPS-Co68, and Co68 groups compared to the DMSO group.\u003cbr\u003e\n \u003cstrong\u003eF, G)\u003c/strong\u003e Peak tracks of \u003cem\u003eRsad2\u003c/em\u003e, \u003cem\u003eIfi202b\u003c/em\u003e, and \u003cem\u003eIl6\u003c/em\u003evisualized in IGV software for the LPS, LPS-Co68, and Co68 groups compared to the DMSO group.\u003cbr\u003e\n \u003cstrong\u003eH)\u003c/strong\u003e Transcription factor enrichment analysis (TFEA) of upregulated differentially expressed genes (DEGs) in the Co68 group compared to the LPS group after 1 hour of treatment in RAW 264.7 cells, identifying STAT1 as a potential transcription factor. Yellow nodes represent enriched transcription factors, while red nodes represent upregulated DEGs in Co68-treated RAW 264.7 cells.\u003cbr\u003e\n \u003cstrong\u003eI)\u003c/strong\u003e Inhibition of STAT1 reduced Co68-induced anti-inflammatory effects in LPS-stimulated RAW 264.7 cells.\u003cbr\u003e\n \u003cstrong\u003eJ)\u003c/strong\u003e Knockdown of STAT1 abrogated the anti-inflammatory effect of Co68 in LPS-stimulated RAW 264.7 cells.\u003cbr\u003e\n \u003cstrong\u003eK)\u003c/strong\u003e Western blot analysis showing that Co68 inhibitedLPS-induced phosphorylation of P65 and IκBα after knockdown of STAT1.\u003cbr\u003e\n \u003cstrong\u003eL)\u003c/strong\u003e Inhibition of the TLR4 receptor significantly reducedCo68-induced STAT1 expression in RAW 264.7 cells.\u003cbr\u003e\n \u003cstrong\u003eM)\u003c/strong\u003e Venn diagram showing the overlap and unique kinases specific to LPS and Co68 treatments, identified by mass spectrometry after one hour of treatment with Co68 and LPS.\u003cbr\u003e\n \u003cstrong\u003eN)\u003c/strong\u003e Co-immunoprecipitation (CoIP) analysis of TLR4-SYK and STAT1-SYK interactions.\u003cbr\u003e\n \u003cstrong\u003eO)\u003c/strong\u003e SYK knockout abrogated Co68's anti-inflammatory effect in LPS-stimulated RAW 264.7 cells. RT-qPCR data were presented as means ± SEM from three independent experiments. Statistical significance was determined by one-way ANOVA with Bonferroni’s multiple comparisons test or paired-samples t-test. * \u003cem\u003eP\u003c/em\u003e \u0026lt; 0.05, ** \u003cem\u003eP\u003c/em\u003e \u0026lt; 0.01, *** \u003cem\u003eP\u003c/em\u003e \u0026lt; 0.001.\u003c/p\u003e","description":"","filename":"fig6.png","url":"https://assets-eu.researchsquare.com/files/rs-7085332/v1/eba110322d4201fa28d92b08.png"},{"id":89729207,"identity":"b7595d4a-801b-424a-9c81-f687e91da2dc","added_by":"auto","created_at":"2025-08-23 13:03:30","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":5170659,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7085332/v1/2e791c33-50ff-4efc-ab6b-30fb5e87b644.pdf"},{"id":87025209,"identity":"95ce8276-632c-49ce-808f-36d118b5f13f","added_by":"auto","created_at":"2025-07-18 11:57:12","extension":"zip","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":17131,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eSupplementary Table S1.\u003c/strong\u003e Primer sequences used in this study.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eSupplementary Table S2.\u003c/strong\u003e Information on shRNA targeting \u003cem\u003eStat1\u003c/em\u003e and \u003cem\u003eSyk\u003c/em\u003e, as well asgRNA for \u003cem\u003eSyk\u003c/em\u003e CRISPR-Cas9 knockout.\u003c/p\u003e","description":"","filename":"SupplementaryMaterials.zip","url":"https://assets-eu.researchsquare.com/files/rs-7085332/v1/3dd15dbbc4dd37b657a32f1e.zip"},{"id":87025215,"identity":"d3c0ac19-4f4c-41e1-a405-9595c144fb04","added_by":"auto","created_at":"2025-07-18 11:57:13","extension":"pptx","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":25283301,"visible":true,"origin":"","legend":"Original images","description":"","filename":"OriginalImages.pptx","url":"https://assets-eu.researchsquare.com/files/rs-7085332/v1/d35b86b4f4a4a05d8611c837.pptx"},{"id":87025210,"identity":"911c177e-a8eb-4486-90a6-a55d62679876","added_by":"auto","created_at":"2025-07-18 11:57:12","extension":"pdf","order_by":3,"title":"","display":"","copyAsset":false,"role":"supplement","size":1062069,"visible":true,"origin":"","legend":"Graphical Abstract","description":"","filename":"GraphicalAbCo68.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7085332/v1/7255edabb9069796abdb71d4.pdf"},{"id":87024382,"identity":"b62e9194-8cca-4c9f-a4b3-63928a0de19d","added_by":"auto","created_at":"2025-07-18 11:49:13","extension":"pdf","order_by":4,"title":"","display":"","copyAsset":false,"role":"supplement","size":5198322,"visible":true,"origin":"","legend":"\u003cp\u003eSupplementary Figure Legends:\u003c/p\u003e\n\u003ch5\u003eSupplementary Figure 1. Effect of Co68 on cytokine and interferon expression in macrophages.\u003c/h5\u003e\n\u003cp\u003e\u003cstrong\u003eA)\u003c/strong\u003e Heatmap showing the differential expression of inflammatory cytokines and interferons in peripheral blood mononuclear cells (PBMCs) from COVID-19 patients compared to healthy controls.\u003cbr\u003e\n\u003cstrong\u003eB)\u003c/strong\u003e Volcano plot depicting the distribution of differentially expressed genes (DEGs) between lipopolysaccharide (LPS)- and dimethyl sulfoxide (DMSO)-treated groups in RAW 264.7 cells. The plot displays fold changes in gene expression, with upregulated DEGs in the LPS group shown in red and downregulated DEGs in blue.\u003cbr\u003e\n\u003cstrong\u003eC)\u003c/strong\u003e Gene Ontology (GO) enrichment analysis of DEGs in the LPS group compared to the DMSO group.\u003cbr\u003e\n\u003cstrong\u003eD)\u003c/strong\u003e The sticks representation of the structure of the PNP ligand used for Co68 synthesis and the synthesis reaction of Co68 from PNP ligand and CoCl₂.\u003cbr\u003e\n\u003cstrong\u003eE)\u003c/strong\u003e Cell viability of J774A.1 macrophages and peritoneal macrophages (PMs) treated with Co68 for 6 hours at the indicated concentrations.\u003cbr\u003e\n\u003cstrong\u003eF-I)\u003c/strong\u003e \u003cem\u003eIfnb1\u003c/em\u003e expression levels in J774A.1 macrophages and PMs following Co68 treatment, measured in a dose- and time-dependent manner.\u003cbr\u003e\n\u003cstrong\u003eJ)\u003c/strong\u003e Co68 elevated \u003cem\u003eIfnb1\u003c/em\u003e expression across M0, M1, and M2 subtypes of THP-1 monocytes.\u003cbr\u003e\n\u003cstrong\u003eK-L)\u003c/strong\u003e \u003cem\u003eIsg15\u003c/em\u003e and \u003cem\u003eIfit3\u003c/em\u003e were strongly upregulated in multiple organs and blood of C57BL/6J mice treated with Co68 compared to the PBS group.\u003cbr\u003e\n\u003cstrong\u003eM)\u003c/strong\u003e Co68 was unstable after dissolution in DMSO and almost completely failed to induce \u003cem\u003eIfnb1\u003c/em\u003e expression by Day 3.\u003cbr\u003e\n\u003cstrong\u003eN-P)\u003c/strong\u003e Exposure to cobalt chloride (CoCl₂) resulted in the induction of \u003cem\u003eIfnb1\u003c/em\u003e, \u003cem\u003eIl1b\u003c/em\u003e, and \u003cem\u003eIl6\u003c/em\u003e in RAW264.7 macrophages at higher concentrations.\u003cbr\u003e\n\u003cstrong\u003eQ-R)\u003c/strong\u003e The expression of \u003cem\u003eIfna4\u003c/em\u003e (Q) and \u003cem\u003eIl1a\u003c/em\u003e (R) in RAW264.7 macrophages upon treatment with DMSO, Co68, CoCl₂, and LPS. RT-qPCR data were presented as means ± SEM from three independent experiments. Statistical significance was determined using one-way ANOVA with Bonferroni's multiple comparisons test or paired-sample t-test. \u003cstrong\u003ens\u003c/strong\u003e, not significant; * \u003cem\u003eP\u003c/em\u003e \u0026lt; 0.05, ** \u003cem\u003eP\u003c/em\u003e \u0026lt; 0.01, *** \u003cem\u003eP\u003c/em\u003e \u0026lt; 0.001.\u003c/p\u003e","description":"","filename":"SupplementaryFigure1.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7085332/v1/775ba03dd8753a7236cf492a.pdf"},{"id":87024407,"identity":"31c65506-796b-459b-9032-6273ea5f0b1d","added_by":"auto","created_at":"2025-07-18 11:49:14","extension":"pdf","order_by":5,"title":"","display":"","copyAsset":false,"role":"supplement","size":1875216,"visible":true,"origin":"","legend":"\u003cp\u003eSupplementary Figure 2. Co68 activates innate immunity and reduces inflammation.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eA)\u003c/strong\u003e Heatmap of innate immune response-related genes differentially expressed in Co68-treated vs. DMSO-treated groups.\u003cbr\u003e\n\u003cstrong\u003eB-C)\u003c/strong\u003e Protein-protein interaction (PPI) network analysis showing upregulated genes in Co68-treated cells are enriched in regulation of type I interferon production and negative regulation of viral genome replication pathways.\u003cbr\u003e\n\u003cstrong\u003eD)\u003c/strong\u003e Gene Ontology (GO) enrichment analysis of downregulated genes in Co68-treated cells compared to the DMSO group.\u003cbr\u003e\n\u003cstrong\u003eE)\u003c/strong\u003e Western blot analysis showing Co68 enhances phosphorylation of TBK1 and IRF3 at varying concentrations.\u003cbr\u003e\n\u003cstrong\u003eF)\u003c/strong\u003e Co68 reduces expression of pro-inflammatory cytokines and chemokines (\u003cem\u003eIl1a\u003c/em\u003e, \u003cem\u003eCsf1\u003c/em\u003e, \u003cem\u003eCsf2\u003c/em\u003e, \u003cem\u003eCsf3\u003c/em\u003e) in LPS-stimulated cells.\u003cbr\u003e\n\u003cstrong\u003eG)\u003c/strong\u003e Co68 promotes expression of immune markers (\u003cem\u003eTnf\u003c/em\u003e, \u003cem\u003eIfnb1\u003c/em\u003e, \u003cem\u003eCxcl10\u003c/em\u003e, \u003cem\u003eIsg15\u003c/em\u003e) under LPS stimulation.\u003cbr\u003e\n\u003cstrong\u003eH)\u003c/strong\u003e Gene Set Enrichment Analysis (GSEA) showing that Co68 downregulates inflammation-related genes compared to LPS alone.\u003cbr\u003e\n\u003cstrong\u003eI)\u003c/strong\u003e Co68 reduces expression of \u003cem\u003eNos2\u003c/em\u003e and \u003cem\u003ePtgs2\u003c/em\u003e at protein level induced by LPS. RT-qPCR data were presented as means ± SEM from three independent experiments. Statistical significance was determined using one-way ANOVA with Bonferroni’s multiple comparisons test. * \u003cem\u003eP\u003c/em\u003e \u0026lt; 0.05, ** \u003cem\u003eP\u003c/em\u003e \u0026lt; 0.01, *** \u003cem\u003eP\u003c/em\u003e \u0026lt; 0.001.\u003c/p\u003e","description":"","filename":"SupplementaryFigure2.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7085332/v1/6d48c4c8c6dce6990662de12.pdf"},{"id":87024386,"identity":"5f5587ca-1729-4619-a095-1461adc7686d","added_by":"auto","created_at":"2025-07-18 11:49:13","extension":"pdf","order_by":6,"title":"","display":"","copyAsset":false,"role":"supplement","size":1549364,"visible":true,"origin":"","legend":"\u003cp\u003eSupplementary Figure 3. Co68 exerts a potent antitumor effect.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eA, B)\u003c/strong\u003e Quantification of tumor weight in Pan02 tumor-bearing wild-type (WT) mice (A, n = 6 per group) or IFNAR1 knockout (KO) mice (B, n = 7 per group) treated with Co68 or PBS via intraperitoneal (i.p.) injection every other day for two weeks.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eC)\u003c/strong\u003e Tumor weight quantification of Pan02 tumors in WT mice treated with isotype control antibody (200 µg/mouse, i.p.), Co68 (25 mg/kg, i.p.), anti-PD-1 antibody (200 µg/mouse, i.p.), or Co68 combined with anti-PD-1 antibody (n = 5 per group).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eD, E, F)\u003c/strong\u003e Representative images (D) and quantification of tumor size (E) and weight (F) of Pan02 tumors in WT mice treated with DMXAA (25 mg/kg, i.p.) or PBS, administered every other day for two weeks (n = 6 per group).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eG, H, I)\u003c/strong\u003e Representative images (G) and quantification of tumor size (H) and weight (I) of MC38 tumors in WT mice treated with Co68 (25 mg/kg, i.p.) or PBS via i.p. injection every other day for two weeks (n = 6 per group).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eJ)\u003c/strong\u003e t-SNE analysis illustrating immune cell clustering in Pan02 tumors, with clusters color-coded according to their identity.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eK)\u003c/strong\u003e Comparison of immune cell clusters between Co68- and PBS-treated Pan02 tumors, indicating immune reprogramming following Co68 treatment.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eL)\u003c/strong\u003e Dot plot displaying the expression of key marker genes across immune cell clusters. Data were presented as mean ± SEM from three independent experiments. Statistical significance was assessed using paired t-test. \u003cem\u003ens\u003c/em\u003e, not significant; \u003cem\u003eP\u003c/em\u003e \u0026lt; 0.05; \u003cem\u003e*P\u003c/em\u003e \u0026lt; 0.01.\u003c/p\u003e","description":"","filename":"SupplementaryFigure3.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7085332/v1/55ef093158e3c10633535f0c.pdf"},{"id":87025217,"identity":"b6c98252-9533-4ffa-b2d6-5d96f629d927","added_by":"auto","created_at":"2025-07-18 11:57:13","extension":"pdf","order_by":7,"title":"","display":"","copyAsset":false,"role":"supplement","size":2742918,"visible":true,"origin":"","legend":"\u003cp\u003eSupplementary Figure 4. Co68 reprograms the tumor microenvironment (TME) in Pan02 tumors.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eA)\u003c/strong\u003e Interaction network of cell types across all cell clusters in Pan02 tumors treated with PBS or Co68.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eB)\u003c/strong\u003e Interaction networks illustrating the communication between the Cancer cell1 cluster and other cell types within the TME.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eC-D)\u003c/strong\u003e Cell interaction networks of tumor-associated macrophages (TAMs)2 (C) and monocytes (Mono) (D) with other cell clusters in the Pan02 TME.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eE)\u003c/strong\u003e Chord diagrams depicting the SPP1 signaling pathway networks originating from TAMs and interacting with other cell populations within the pancreatic cancer TME.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eF)\u003c/strong\u003e t-SNE plot illustrating subclusters of tumor-associated neutrophils (TANs) in Co68- and PBS-treated Pan02 tumors.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eG)\u003c/strong\u003e Violin plots showing the expression of interferon-stimulated genes (ISGs) across TAN subpopulations, highlighting the upregulation of ISGs in Co68-induced TANs1.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eH)\u003c/strong\u003e Chord diagram illustrating marker gene expression and biological processes associated with TANs1 in Co68-treated tumors.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eI)\u003c/strong\u003e t-SNE plot showing cell clustering in Pan02 tumors treated with Co68 or DMXAA.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eJ)\u003c/strong\u003e Heatmap of gene expression profiles across various cell types from subclustering analysis of tumor-associated macrophages (TAMs) in Co68 and DMXAA treatment groups.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eK)\u003c/strong\u003e Chord diagrams depicting differentially expressed genes (DEGs) and associated biological processes in IL1B+ TAMs from Co68-treated tumors compared to the DMXAA group.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eL)\u003c/strong\u003e Violin plots comparing the expression of ISGs and pattern recognition receptors (PRRs) in IL1B+ TAMs from Co68- and DMXAA-treated tumors.\u003c/p\u003e","description":"","filename":"SupplementaryFigure4.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7085332/v1/09905ff31737abc4809bca57.pdf"},{"id":87024380,"identity":"7493fead-51ae-4be6-89f6-0a10e92c57a9","added_by":"auto","created_at":"2025-07-18 11:49:13","extension":"pdf","order_by":8,"title":"","display":"","copyAsset":false,"role":"supplement","size":2829179,"visible":true,"origin":"","legend":"\u003cp\u003eSupplementary Figure 5. Co68 directly binds to TLR4.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eA)\u003c/strong\u003e Co68 did not induce \u003cem\u003eIfnb1\u003c/em\u003eexpression in BMDMs from IRF3 knockout mice.\u003cbr\u003e\n\u003cstrong\u003eB-D)\u003c/strong\u003e Neutralizing antibodies against TLR4, CD14, or MD2 suppressed Co68-induced \u003cem\u003eIsg15\u003c/em\u003eexpression in RAW 264.7 cells.\u003cbr\u003e\n\u003cstrong\u003eE)\u003c/strong\u003e Co-transfection of TLR4, CD14, and MD2 plasmids enhanced Co68-induced \u003cem\u003eIsg15\u003c/em\u003eexpression in HEK293T cells.\u003cbr\u003e\n\u003cstrong\u003eF)\u003c/strong\u003e MST analysis revealed a K\u003cem\u003ed\u003c/em\u003e of 4.37 µM for LPS binding to TLR4.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eG)\u003c/strong\u003e SPR analysis showed Co68 binding to TLR4 with a K\u003cem\u003ed\u003c/em\u003e of 31.71 µM.\u003cbr\u003e\n\u003cstrong\u003eH)\u003c/strong\u003e SPR analysis of LPS binding to TLR4, with sensorgram data showing RU at increasing LPS concentrations.\u003cbr\u003e\n\u003cstrong\u003eI)\u003c/strong\u003e SPR analysis of LPS-TLR4 dose-response curve with a K\u003cem\u003ed\u003c/em\u003e of 147.64 µM.\u003cbr\u003e\n\u003cstrong\u003eJ)\u003c/strong\u003e Electron microscopy structure of Co68, with key atoms in orange (phosphorus), blue (nitrogen), magenta (carbon), green (chloride), and pink (cobalt).\u003cbr\u003e\n\u003cstrong\u003eK)\u003c/strong\u003e TLR4-MD2 dimer with Co68 positioned in the MD2 binding pocket.\u003cbr\u003e\n\u003cstrong\u003eL)\u003c/strong\u003e Ribbon and surface representations of TLR4-MD2-Co68 complex.\u003cbr\u003e\n\u003cstrong\u003eM-N)\u003c/strong\u003e Detailed interactions between Co68, MD2 residues, and TLR4.\u003cbr\u003e\n\u003cstrong\u003eO)\u003c/strong\u003e Close-up molecular interface showing interactions between Co68 and MD2 and TLR4.\u003cbr\u003e\n\u003cstrong\u003eP)\u003c/strong\u003e Synthesis of Co65–Co74 and their chemical structures. RT-qPCR data were presented as means ± SEM from three independent experiments. Statistical significance was determined using one-way ANOVA with Bonferroni's multiple comparisons test or paired-sample t-test. \u003cstrong\u003ens\u003c/strong\u003e, not significant; * \u003cem\u003eP\u003c/em\u003e \u0026lt; 0.05, ** \u003cem\u003eP\u003c/em\u003e \u0026lt; 0.01, *** \u003cem\u003eP\u003c/em\u003e \u0026lt; 0.001.\u003c/p\u003e","description":"","filename":"SupplementaryFigure5.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7085332/v1/0ccf8728d04271a308eb8e1c.pdf"},{"id":87025211,"identity":"5eec3f61-2f38-41cd-8938-6ab484e7aa64","added_by":"auto","created_at":"2025-07-18 11:57:13","extension":"pdf","order_by":9,"title":"","display":"","copyAsset":false,"role":"supplement","size":874350,"visible":true,"origin":"","legend":"\u003cp\u003eSupplementary Figure 6. Co68 induces the IFN-I signaling via TLR4.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eA-F)\u003c/strong\u003e The dose-response curve of SPR analysis for six \u003cem\u003eIfnb1\u003c/em\u003e-inducible analogs of Co68 with TLR4, including Co65 (A), Co69 (B), Co70 (C), Co71 (D), Co72 (E), and Co73 (F).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eG,I,J)\u003c/strong\u003e Representative images (G), tumor size and weight quantification (I), and growth curves (J) of Pan02 tumors in TLR4 KO mice (n = 6) treated with Co68 or PBS via i.p. injection every other day for two weeks.\u003cbr\u003e\n\u003cstrong\u003eH, K, L)\u003c/strong\u003e Representative images (H), tumor size and weight quantification (K), and growth curves (L) of Pan02 tumors in IRF3 KO mice (n = 6) treated with Co68 or PBS via i.p. injection every other day for two weeks.\u003c/p\u003e","description":"","filename":"SupplementaryFigure6.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7085332/v1/31fe112692b14f2bbbf7a409.pdf"},{"id":87024403,"identity":"36bc6d6c-76e0-4b48-b837-3d201f57c33a","added_by":"auto","created_at":"2025-07-18 11:49:14","extension":"pdf","order_by":10,"title":"","display":"","copyAsset":false,"role":"supplement","size":1181532,"visible":true,"origin":"","legend":"\u003cp\u003eSupplementary Figure 7. Co68 induces the non-classic activation of STAT1 via TLR4-SYK.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eA-C)\u003c/strong\u003e Number and proportion of peaks identified in LPS (A), LPS-Co68 (B), and Co68 (C) groups, showing distinct chromatin accessibility profiles for Co68 compared to LPS and DMSO.\u003cbr\u003e\n\u003cstrong\u003eD-E)\u003c/strong\u003e Distribution of peaks in various genomic regions and transcription factor-binding loci relative to the TSS among LPS, LPS-Co68 and Co68 groups.\u003cbr\u003e\n\u003cstrong\u003eF-H)\u003c/strong\u003e Comparation of binding signal intensity of STAT1, EGR1, and REL between Co68 and LPS groups.\u003cbr\u003e\n\u003cstrong\u003eI-K)\u003c/strong\u003e IGV visualization of peak tracks for \u003cem\u003eUsp18\u003c/em\u003e, \u003cem\u003eIfit1\u003c/em\u003e, \u003cem\u003eIl1b\u003c/em\u003e, and \u003cem\u003eTnf\u003c/em\u003e.\u003cbr\u003e\n\u003cstrong\u003eL)\u003c/strong\u003e Heatmap showing early upregulation of ISGs following Co68 treatment for 1 hour.\u003cbr\u003e\n\u003cstrong\u003eM)\u003c/strong\u003e Co-immunoprecipitation (CoIP) of TLR4 from RAW 264.7 cells upon the DMSO, LPS and Co68 treatment.\u003cbr\u003e\n\u003cstrong\u003eN)\u003c/strong\u003e Knockdown efficiency of \u003cem\u003eStat1\u003c/em\u003e and \u003cem\u003eSyk\u003c/em\u003e in RAW 264.7 cells, confirmed by RT-qPCR.\u003cbr\u003e\n\u003cstrong\u003eO)\u003c/strong\u003e SYK knockdown abrogates the anti-inflammatory effect of Co68 in LPS-stimulated RAW 264.7 cells.\u003cbr\u003e\n\u003cstrong\u003eP)\u003c/strong\u003e Western blot confirming SYK knockout in RAW 264.7 cells, ensuring the observed effects were due to SYK deficiency.\u003cbr\u003e\n\u003cstrong\u003eQ)\u003c/strong\u003e Western blot showing that Co68 fails to inhibit LPS-induced phosphorylation of P65 and IκBα in SYK knockout cells, indicating SYK is essential for Co68’s anti-inflammatory effects. RT-qPCR data were presented as means ± SEM from three independent experiments. Statistical significance was analyzed using paired t-test. ns, not significant; * \u003cem\u003eP\u003c/em\u003e \u0026lt; 0.05; ** \u003cem\u003eP\u003c/em\u003e \u0026lt; 0.01.\u003c/p\u003e","description":"","filename":"SupplementaryFigure7.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7085332/v1/0076ce2a0d79e41f032e401f.pdf"}],"financialInterests":"(Not answered)","formattedTitle":"Phosphines-Nitrogen-Phosphines Chelated CoCl2 Exhibits Potent Antitumor Activity in Pancreatic Cancer","fulltext":[{"header":"Highlights","content":"\u003cul\u003e\n \u003cli\u003eCo68 features a classic PNP-pincer ligand acting as an immunomodifier identified by CNN model DLINP\u003c/li\u003e\n \u003cli\u003eCo68 reprograms tumor-associated macrophages within the pancreatic cancer microenvironment\u003c/li\u003e\n \u003cli\u003eCo68 induces type I interferon via a novel binding mode with TLR4\u003c/li\u003e\n \u003cli\u003eCo68 demonstrates significant anti-inflammatory effects via the TLR4-SYK-STAT1 axis\u003c/li\u003e\n\u003c/ul\u003e\n\n"},{"header":"In brief","content":"\u003cp\u003eA multiscale drug discovery platform integrating the convolutional neural network (CNN) model DLINP, multi-omics analysis, and animal models identified Co68—a compound featuring a classic Phosphines-Nitrogen-Phosphines (PNP) pincer ligand—as an effective agent with anti-inflammatory and anti-pancreatic cancer properties.\u003c/p\u003e"},{"header":"Introduction","content":"\u003cp\u003ePancreatic cancer is often referred to as the \"king of cancers\" due to its insidious onset, poor prognosis, and high resistance to treatment\u003csup\u003e\u003cspan additionalcitationids=\"CR2\" citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e\u003c/sup\u003e. Global statistics indicate that approximately 500,000 new cases of pancreatic cancer are diagnosed each year, with a dismal five-year survival rate of approximately 10%\u003csup\u003e4\u0026ndash;6\u003c/sup\u003e. This malignancy is characterized by a lack of early-stage symptoms, which often results in diagnosis at an advanced stage when surgical intervention is no longer an option\u003csup\u003e\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e\u003c/sup\u003e. The inherent biological traits of pancreatic cancer, such as rapid tumor cell proliferation and extensive metastatic potential, further contribute to its aggressive nature\u003csup\u003e\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e\u003c/sup\u003e. Projections suggest that pancreatic cancer may become the second leading cause of cancer-related deaths by 2030\u003csup\u003e9\u003c/sup\u003e. Currently, treatment options are limited to surgery, chemotherapy, radiotherapy, and targeted therapies, including those targeting the epidermal growth factor receptor (EGFR), angiogenesis, and hypoxia\u003csup\u003e\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e\u003c/sup\u003e. However, these therapeutic strategies have had limited success in improving long-term survival, and substantial challenges remain in effectively treating this disease.\u003c/p\u003e\u003cp\u003eThe limited progress in pancreatic cancer treatment has intensified the focus on immunotherapy, primarily driven by the challenges posed by the complex and immunosuppressive tumor microenvironment (TME) inherent to this disease\u003csup\u003e\u003cspan additionalcitationids=\"CR12\" citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e\u003c/sup\u003e. A hallmark of the TME is the intricate interplay among tumor-associated fibroblasts, immune cells, and tumor cells, which collectively facilitate immune evasion and complicate therapeutic approaches\u003csup\u003e\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e,\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e\u003c/sup\u003e. Tumor-associated macrophages (TAMs), the most abundant immune cell population within this microenvironment, play a central role in establishing a pro-inflammatory environment that promotes tumor progression\u003csup\u003e\u003cspan additionalcitationids=\"CR17 CR18\" citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e\u003c/sup\u003e. TAMs secrete interleukin-1 beta (IL-1β), a key cytokine that drives pancreatic cancer initiation, progression, and metastasis\u003csup\u003e\u003cspan additionalcitationids=\"CR21\" citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e\u003c/sup\u003e. Fueled by the positive feedback loop between IL-1β and prostaglandin E2 (PGE2), chronic inflammation has emerged as a critical driver of tumor development\u003csup\u003e\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e,\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e\u003c/sup\u003e. PGE2, secreted by tumor cells, enhances IL-1β secretion from TAMs, creating a vicious cycle that accelerates tumor progression and immune evasion, ultimately limiting the effectiveness of immune responses\u003csup\u003e\u003cspan additionalcitationids=\"CR25 CR26\" citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e\u003c/sup\u003e. The reciprocal relationship between TAMs and pancreatic cancer cells, mediated through the IL-1β-PGE2 axis, underscores the pivotal role of TAMs in the progression of pancreatic cancer\u003csup\u003e\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e,\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e\u003c/sup\u003e. Targeting IL-1β and PGE2 presents a promising strategy to disrupt this inflammatory axis, offering novel therapeutic avenues to combat this aggressive and challenging cancer.\u003c/p\u003e\u003cp\u003eType I interferons (IFN-I), primarily interferon-alpha (IFN-α) and interferon-beta (IFN-β), are critical players in promoting anti-tumor immunity through multiple mechanisms, positioning them as valuable therapeutic agents in cancer treatment\u003csup\u003e\u003cspan additionalcitationids=\"CR29\" citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e\u003c/sup\u003e. IFN-I enhances immune responses by activating natural killer (NK) cells and cytotoxic CD8\u003csup\u003e+\u003c/sup\u003e T lymphocytes, which are essential for recognizing and eliminating tumor cells\u003csup\u003e\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e,\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e\u003c/sup\u003e. By boosting the cytotoxic activities of these immune cells, IFN-I modifies the TME and suppresses tumor growth\u003csup\u003e\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e\u003c/sup\u003e. Additionally, IFN-I improves the presentation of tumor-associated antigens by antigen-presenting cells, such as dendritic cells, thus promoting the development of adaptive immune responses\u003csup\u003e\u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e\u003c/sup\u003e. Recent studies have further demonstrated IFN-I's ability to inhibit tumor cell proliferation and induce apoptosis, solidifying its potential as an effective cancer therapy\u003csup\u003e\u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e35\u003c/span\u003e\u003c/sup\u003e. Through the activation of specific signaling pathways, IFN-I can suppress tumor cell proliferation and initiate apoptotic mechanisms, making it a promising candidate for cancer treatment\u003csup\u003e\u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e36\u003c/span\u003e,\u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e37\u003c/span\u003e\u003c/sup\u003e. However, the TME is filled with factors that inhibit IFN-I signaling, including tumor-derived PGE2, which not only exacerbates IL-1β secretion in TAMs but also directly suppresses IFN-I signaling in TAMs\u003csup\u003e\u003cspan additionalcitationids=\"CR39\" citationid=\"CR38\" class=\"CitationRef\"\u003e38\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e40\u003c/span\u003e\u003c/sup\u003e. This suppression reduces the expression of interferon-stimulated genes (ISGs) and diminishes the overall anti-tumor immune response\u003csup\u003e\u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e41\u003c/span\u003e\u003c/sup\u003e. Given the immune-suppressive role of IL-1β-secreting TAMs and the critical importance of IFN-I signaling in cancer immunotherapy, there is an urgent need for novel therapeutic agents capable of modulating the balance between inflammation and IFN-I signaling within the pancreatic cancer TME.\u003c/p\u003e\u003cp\u003ePhosphines-Nitrogen-Phosphines (PNP) pincer chemistry, a subfield of coordination chemistry, focuses on the development and application of PNP pincer ligands\u0026mdash;tridentate ligands that feature a central nitrogen donor atom flanked by two phosphine groups\u003csup\u003e\u003cspan additionalcitationids=\"CR43\" citationid=\"CR42\" class=\"CitationRef\"\u003e42\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e44\u003c/span\u003e\u003c/sup\u003e. These ligands are known for their stability and reactivity, particularly in catalysis, and have been studied extensively in transition metal chemistry\u003csup\u003e\u003cspan additionalcitationids=\"CR46\" citationid=\"CR45\" class=\"CitationRef\"\u003e45\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e47\u003c/span\u003e\u003c/sup\u003e. Recent advances in PNP-pincer chemistry have incorporated N-heterocycles into the ligand framework, leveraging the combined electronic properties of phosphines and N-heterocycles to enhance metal-ligand interactions\u003csup\u003e\u003cspan additionalcitationids=\"CR49\" citationid=\"CR48\" class=\"CitationRef\"\u003e48\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR50\" class=\"CitationRef\"\u003e50\u003c/span\u003e\u003c/sup\u003e. This has led to the development of highly efficient catalysts for hydrogenation, C\u0026ndash;H activation, and cross-coupling reactions\u003csup\u003e\u003cspan additionalcitationids=\"CR52\" citationid=\"CR51\" class=\"CitationRef\"\u003e51\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR53\" class=\"CitationRef\"\u003e53\u003c/span\u003e\u003c/sup\u003e. Despite the growing interest in their catalytic potential, the biological activity of PNP-metal complexes remains largely unexplored, particularly in terms of their ability to modulate immune responses or serve as therapeutic agents. Machine learning (ML) techniques have revolutionized drug discovery by enabling the analysis of complex biological data and the prediction of drug efficacy and safety\u003csup\u003e\u003cspan additionalcitationids=\"CR55\" citationid=\"CR54\" class=\"CitationRef\"\u003e54\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR56\" class=\"CitationRef\"\u003e56\u003c/span\u003e\u003c/sup\u003e. Advanced ML models, including deep neural networks, convolutional neural networks (CNNs), and support vector machines (SVMs), excel at identifying patterns within large-scale biological datasets, accelerating the drug discovery process\u003csup\u003e\u003cspan additionalcitationids=\"CR58 CR59\" citationid=\"CR57\" class=\"CitationRef\"\u003e57\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR60\" class=\"CitationRef\"\u003e60\u003c/span\u003e\u003c/sup\u003e. In particular, deep learning frameworks such as TensorFlow and Keras have facilitated the development of sophisticated models that optimize various aspects of drug discovery, including molecular property prediction, de novo design, and retrosynthetic analysis\u003csup\u003e\u003cspan additionalcitationids=\"CR62 CR63 CR64\" citationid=\"CR61\" class=\"CitationRef\"\u003e61\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR65\" class=\"CitationRef\"\u003e65\u003c/span\u003e\u003c/sup\u003e. These advancements have significantly enhanced the speed and efficiency of identifying novel therapeutic agents, including drug repurposing applications\u003csup\u003e\u003cspan citationid=\"CR66\" class=\"CitationRef\"\u003e66\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e\u003cp\u003eIn this study, we developed a convolutional neural network model, DLINP, to screen small molecules that modulate both the IL-1β/PGE2 feedback loop and IFN-I signaling within the pancreatic cancer TME\u003csup\u003e\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e,\u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e38\u003c/span\u003e\u003c/sup\u003e. Our findings led to the unexpected identification of Co68, a PNP-pincer ligand complexed with CoCl₂, as a potent immune-modulatory agent. Co68 enhanced the IFN-I response and reprogrammed TAMs and tumor-associated neutrophils (TANs) to promote immune activation and reduce inflammation. Through single-cell RNA sequencing (scRNA-seq) and functional assays, we demonstrated that Co68 activates the TLR4-TRIF-IFN-I axis, which is essential for driving anti-tumor immunity. Moreover, we uncovered a novel TLR4-SYK-STAT1 signaling axis underlying the anti-inflammatory effects of Co68, distinguishing it from other immune-modulatory agents, such as DMXAA. Notably, Co68 is the first PNP-chelate identified to convert \"cold\" tumors into \"hot\" tumors by targeting TLR4, which has significant implications for overcoming immune evasion. This therapeutic strategy holds promise not only for pancreatic cancer but also for addressing immune evasion mechanisms in other malignancies. Our study demonstrates that Co68, a Phosphines-Nitrogen-Phosphines (PNP) chelated CoCl₂ compound, could serve as a novel therapeutic agent, offering new opportunities for cancer treatment by reshaping the immune landscape within the TME.\u003c/p\u003e"},{"header":"Results","content":"\u003cp\u003e\u003cem\u003eDevelop the DLINP model for screening potential immunomodulatory agents\u003c/em\u003e\u003c/p\u003e\u003cp\u003eExcessive inflammation is detrimental to health, which could result in damage to multiple organs and promote tumor development\u003csup\u003e\u003cspan citationid=\"CR67\" class=\"CitationRef\"\u003e67\u003c/span\u003e\u003c/sup\u003e. For example, COVID-19 is characterized by a \"cytokine storm,\" with elevated levels of pro-inflammatory cytokines such as IL-1β, IL-6, and TNF while concurrently suppressing the production of type I and III interferons (Figure \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003eA)\u003csup\u003e\u003cspan additionalcitationids=\"CR69\" citationid=\"CR68\" class=\"CitationRef\"\u003e68\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR70\" class=\"CitationRef\"\u003e70\u003c/span\u003e\u003c/sup\u003e. These findings suggest that targeting type I interferon (IFN-I) pathways may offer significant prognostic and therapeutic potential. Lipopolysaccharide (LPS) treatment in the RAW 264.7 macrophage cell line induced a robust upregulation of genes associated with both inflammatory and IFN-I responses, including \u003cem\u003eIl1a\u003c/em\u003e, \u003cem\u003eIl1b\u003c/em\u003e, \u003cem\u003eIl6\u003c/em\u003e, \u003cem\u003ePtgs2\u003c/em\u003e, \u003cem\u003eIfnb1\u003c/em\u003e, \u003cem\u003eIfit2\u003c/em\u003e, \u003cem\u003eMx1\u003c/em\u003e, and \u003cem\u003eOasl1\u003c/em\u003e (Figures \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003eB and S1C). These observations motivated us to identify compounds that not only promote IFN-I production but also suppress inflammatory responses. The approach workflow is summarized in Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eA. To achieve this, we developed the DLINP (Deep Learning for Innate immunity modulatory Potential) model, a convolutional neural network (CNN) designed to screen compounds from the CAS database, which contains over 127\u0026nbsp;million unique chemical entities. The screening revealed several compounds that both promoted IFN-I expression and suppressed inflammatory cytokine production. The top candidate compounds were further validated in vitro and in vivo, assessing their potential to inhibit NF-κB-mediated inflammatory pathways and enhance IRF3-mediated immune responses. \"Co68\" emerged as the most efficacious compound after comprehensive toxicological and functional evaluations. We subsequently investigated its molecular target and signaling pathways through which it activates IFN-I production. The antitumor efficacy of Co68 was further assessed in the Pan02 pancreatic cancer mouse model, employing single-cell RNA sequencing (scRNA-seq). Ultimately, we confirmed the broad anti-inflammatory signaling activity of Co68, positioning it as a promising candidate for therapeutic application in inflammatory diseases and cancers.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003eTo construct the DLINP model, we first extracted immune-modulating small molecules and their corresponding gene expression profiles from the L1000 and ChEMBL databases. As illustrated in Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eB, SMILES descriptors of each compound were converted into a one-hot encoded matrix, representing the atomic composition and chemical bonding features as binary values (0 and 1). Gene expression data associated with each compound were similarly one-hot encoded, with upregulated genes assigned a value of 1 and downregulated genes a value of 0, resulting in a compound-specific gene expression encoding matrix. The DLINP CNN model, consisting of one input layer, four hidden layers, and one output layer, was trained to predict the immune-regulatory potential of these compounds. The model\u0026rsquo;s performance was evaluated using confusion matrix analysis, Receiver Operating Characteristic (ROC) curves, and the area under the curve (AUC) metrics. As shown in Figs.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eC and \u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eD, the DLINP model demonstrated robust predictive capability, achieving an AUC of 0.91 and a true positive rate (TPR) of 93.11%, validating its accuracy and effectiveness in predicting immune-modulatory properties. Encouraged by these results, we applied the trained DLINP model to screen small molecules from the CAS database. Among the identified candidates, Co68 was predicted to have a particularly potent immune-modulatory effect. Its chemical structure, depicted in Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eE, combines the classical PNP structure commonly found in catalytic chemistry with cobalt dichloride, with the synthesis reaction outlined in Figure \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003eD. We first assessed the cytotoxicity of Co68 in primary macrophages and various cell lines, including the Bone Marrow-Derived Macrophages (BMDMs), peritoneal macrophages (PMs), RAW 264.7, and J774A.1 (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eF and Figure \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003eE). While Co68 at a high concentration of 500 \u0026micro;M caused cell death, lower concentrations (\u0026lt;\u0026thinsp;100 \u0026micro;M) showed minimal effects on cell viability.\u003c/p\u003e\u003cp\u003eCo68 was then found to induce the expression of \u003cem\u003eIfnb1\u003c/em\u003e, with peak levels observed at 200 \u0026micro;M in RAW 264.7, BMDMs, and PMs, and at 100 \u0026micro;M in J774A.1 cells (Figs.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eG\u0026ndash;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eH and Figure \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003eF\u0026ndash;S1G). These findings validated the predictive accuracy and efficacy of the DLINP model in identifying compounds capable of modulating immune responses. Time-course analysis revealed that Co68 induced an expression peak of \u003cem\u003eIfnb1\u003c/em\u003e at 3 hours in RAW 264.7, J774A.1, and PMs, while in BMDMs, the peak occurred at 5 hours (Figs.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eI\u0026ndash;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eJ and Figure \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003eH\u0026ndash;S1I). Based on these results, the optimal concentration of Co68 for in vitro experiments was determined to be 100 \u0026micro;M for a 3-hour treatment period. Additionally, Co68 significantly induced IFNB1 production in the human THP-1 macrophage cell line across different subtypes (Figure \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003eJ). In vivo, Co68 administration at 10 \u0026micro;M for 24 hours significantly increased the expression of \u003cem\u003eIfnb1\u003c/em\u003e and interferon-stimulated genes (ISGs), including \u003cem\u003eIsg15\u003c/em\u003e and \u003cem\u003eIfit3\u003c/em\u003e, in spleen, lung, and liver tissues, as well as in peripheral blood mononuclear cells (PBMCs) from mice (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eK, Figure \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003eK\u0026ndash;S1L), but did not affect heart and kidney tissues. Notably, Co68 showed instability when dissolved in DMSO. As shown in Figure \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003eM, the activation of Co68 in RAW 264.7 cells was significantly diminished 3 days post-dissolution. Consistent with previous studies demonstrating that Co\u0026sup2;⁺ directly binds to TLR4 to induce metal anaphylaxis\u003csup\u003e\u003cspan citationid=\"CR71\" class=\"CitationRef\"\u003e71\u003c/span\u003e\u003c/sup\u003e, we also observed that CoCl₂, a precursor of Co68, induced the expression of \u003cem\u003eIfnb1\u003c/em\u003e, \u003cem\u003eIl1b\u003c/em\u003e, and \u003cem\u003eIl6\u003c/em\u003e at a higher concentration (500 \u0026micro;M) following a 24-hour treatment. However, CoCl₂ did not trigger similar effects at the lower concentration (50 \u0026micro;M) for the shorter period (3 hours) as Co68 did (Figure \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003eN\u0026ndash;S1P). We further compared the gene expression profiles induced by Co68, CoCl₂, and LPS to investigate the differences in activation patterns. As shown in Figs.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eL\u0026ndash;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eN and Figure \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003eQ\u0026ndash;S1R, Co68 induced a robust expression of \u003cem\u003eIfnb1\u003c/em\u003e and \u003cem\u003eIfna4\u003c/em\u003e after just 3 hours of treatment, significantly surpassing both CoCl₂ and LPS in these marker genes. In contrast, the expression of \u003cem\u003eIl1b\u003c/em\u003e, \u003cem\u003eIl6\u003c/em\u003e, and \u003cem\u003eIl1a\u003c/em\u003e was markedly lower in the Co68 group than in the CoCl₂ and LPS groups, indicating that Co68 activates a distinct gene expression program compared to both CoCl₂ and LPS. This observation highlights the unique activation pattern of Co68, which contrasts with the more classical pro-inflammatory response driven by CoCl₂ and LPS. Collectively, these findings demonstrate the potential efficacy and safety of Co68 in modulating innate immune responses, thereby validating the high performance and precision of the DLINP model.\u003c/p\u003e\u003cp\u003e\u003cem\u003eCo68 induces type I interferon signaling and suppresses inflammatory responses\u003c/em\u003e\u003c/p\u003e\u003cp\u003eTo gain deeper insights into the effects of Co68 on macrophages, we conducted bulk RNA sequencing (RNA-seq) on RAW 264.7 cells treated with Co68 for 3 hours. As shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eA, a total of 472 genes were upregulated, while 153 genes were downregulated in the Co68-treated group compared to the DMSO control. As expected, Co68 treatment significantly upregulated the expression of interferons, ISGs, and major histocompatibility complex (MHC) molecules. Notably, genes such as \u003cem\u003eIfna2\u003c/em\u003e, \u003cem\u003eIfna4\u003c/em\u003e, \u003cem\u003eIfnb1\u003c/em\u003e, \u003cem\u003eOas1a\u003c/em\u003e, \u003cem\u003eRsad2\u003c/em\u003e, \u003cem\u003eIfit3\u003c/em\u003e, \u003cem\u003eCxcl10\u003c/em\u003e, \u003cem\u003eIsg15\u003c/em\u003e, \u003cem\u003eIsg20\u003c/em\u003e, \u003cem\u003eH2-D1\u003c/em\u003e, \u003cem\u003eH2-K1\u003c/em\u003e, and \u003cem\u003eH2-T23\u003c/em\u003e were prominently upregulated (Figure S2A). We performed a protein-protein interaction (PPI) network analysis of these differentially expressed genes (DEGs) using the STING database to elucidate the molecular mechanisms underlying these changes. The most enriched clusters of upregulated DEGs in the PPI network following Co68 treatment were predominantly related to innate immune responses, regulation of type I interferon production, and negative regulation of viral genome replication (Figs.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eB and S2B\u0026ndash;S2C), suggesting that Co68 acts as a modulator of innate immune responses. Gene Ontology (GO) analysis revealed that the upregulated genes in the Co68-treated group were enriched in pathways associated with defense responses to viruses, innate immune responses, cellular responses to interferon-beta, type I interferon-mediated signaling, and antiviral innate immune responses (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eC). In contrast, the downregulated genes were linked to processes such as the positive regulation of cell migration, negative regulation of dendritic cell differentiation, apoptotic processes, ubiquitin-dependent endocytosis, and cell proliferation (Figure S2D). Additionally, protein analysis demonstrated significant phosphorylation of TBK1 and IRF3, further supporting the activation of key signaling pathways involved in innate immune modulation (Figure S2E).\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003eTo evaluate the anti-inflammatory potential of Co68 in macrophages, we quantified the expression of genes encoding key inflammatory cytokines and enzymes, including \u003cem\u003eIl1b\u003c/em\u003e, \u003cem\u003eIl6\u003c/em\u003e, \u003cem\u003eNos2\u003c/em\u003e, and \u003cem\u003ePtgs2\u003c/em\u003e, which are crucial mediators of the inflammation observed in conditions such as sepsis and tumor progression\u003csup\u003e\u003cspan citationid=\"CR67\" class=\"CitationRef\"\u003e67\u003c/span\u003e\u003c/sup\u003e. As expected, Co68 significantly reduced the LPS-induced production of \u003cem\u003eIL-1α\u003c/em\u003e, \u003cem\u003eIL-1β\u003c/em\u003e, \u003cem\u003eIl6\u003c/em\u003e, \u003cem\u003eNos2\u003c/em\u003e, \u003cem\u003ePtgs2\u003c/em\u003e, \u003cem\u003eCsf1\u003c/em\u003e, \u003cem\u003eCsf2\u003c/em\u003e, and \u003cem\u003eCsf3\u003c/em\u003e in RAW 264.7 cells (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eD and Figure S2F). Interestingly, Co68 did not suppress TNF production; rather, it synergistically enhanced TNF levels in the presence of LPS (Figure S2G). A similar synergistic effect was observed for other immune mediators, including \u003cem\u003eIfnb1\u003c/em\u003e, \u003cem\u003eCxcl10\u003c/em\u003e, and \u003cem\u003eIsg15\u003c/em\u003e (Figure S2G). Co68 also exhibited broad-spectrum anti-inflammatory effects, significantly suppressing the expression of \u003cem\u003eIl1b\u003c/em\u003e induced by various viruses such as VSV, HSV, and EMCV in RAW 264.7 cells (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eE). In vivo, Co68 markedly reduced sepsis-related mortality following LPS injection into the peritoneal cavity (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eF), further supporting its anti-inflammatory potential. To further elucidate the molecular mechanisms underlying these effects, we performed bulk RNA-seq to analyze the anti-inflammatory phenotype and pathways modulated by Co68. As shown in the heatmap (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eG), Co68 prominently induced the expression of genes associated with IFN-I signaling, while LPS treatment primarily upregulated genes involved in the inflammatory response. Notably, Co68 induced the upregulation of several histone-coding genes, such as \u003cem\u003eH1f5\u003c/em\u003e, \u003cem\u003eH2bc4\u003c/em\u003e, and \u003cem\u003eH2bc18\u003c/em\u003e, a pattern absent in the LPS-treated group, suggesting that Co68 plays a specific role in the regulation of gene expression.\u003c/p\u003e\u003cp\u003eThe Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis of downregulated differentially expressed genes (DEGs) in the LPS-Co68 group, compared to the LPS group, revealed the suppression of several critical pathways, including the NF-κB signaling pathway, inflammatory bowel disease, rheumatoid arthritis, and the Toll-like receptor and NOD-like receptor signaling pathways (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eH). Gene Set Enrichment Analysis (GSEA) further confirmed these results, demonstrating significant enrichment of the \u003cem\u003eINFLAMMATORY_RESPONSE\u003c/em\u003e pathway within the downregulated gene set (Figure S2H). Transcription factor enrichment analysis, conducted using the iRegulon plugin in Cytoscape (v3.10.2), identified a significant enrichment of the NF-κB family transcription factor RELA (P65) among the downregulated genes (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eI). Consistent with this, Co68 effectively inhibited LPS-induced phosphorylation and degradation of IκB-α, along with the phosphorylation of P65 and the production of key downstream enzymes such as iNOS and Cox2 (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eJ and Figure S2I). These findings indicate that Co68 specifically targets and suppresses NF-κB-mediated inflammatory signaling. Collectively, these results demonstrate that Co68 not only robustly activates the type I interferon signaling pathway but also exerts broad anti-inflammatory effects, thereby positioning it as a promising therapeutic agent for the modulation of inflammation.\u003c/p\u003e\u003cp\u003e\u003cem\u003eCo68 demonstrates potent antitumor activity in the Pan02 model\u003c/em\u003e\u003c/p\u003e\u003cp\u003eGiven Co68\u0026rsquo;s ability to robustly activate IFN-I and its broad-spectrum anti-inflammatory properties in macrophages, we sought to investigate whether Co68 could suppress TAM-mediated inflammation and enhance antitumor immunity within the pancreatic cancer TME by promoting IFN-I signaling. To explore this hypothesis, we established a murine pancreatic cancer model using Pan02 cells. As shown in Figs.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eA-\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eC and Figure S3A, Co68 significantly inhibited tumor growth. This antitumor effect was found to be dependent on IFN-I signaling, as evidenced by the abrogation of Co68\u0026rsquo;s antitumor activity in Pan02 tumors implanted in IFNAR1 knockout (KO) mice (Figs.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eD-\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eF and Figure S3B). Consistent with previous studies\u003csup\u003e\u003cspan citationid=\"CR72\" class=\"CitationRef\"\u003e72\u003c/span\u003e\u003c/sup\u003e, Pan02 pancreatic tumors are classified as \"cold tumors\", showing limited sensitivity to PD-1 antibody treatment (Figs.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eG, \u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eH, and Figure S3C). However, we observed that the combination of recombinant PD-1 antibody and Co68 significantly inhibited tumor growth compared to either Co68 or PD-1 antibody treatment alone (Figs.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eG-\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eI). In contrast to PD-1 antibody treatment, DMXAA, a potent agonist of mouse Stimulator of Interferon Genes (STING), has been shown to inhibit Pan02 tumor growth significantly\u003csup\u003e\u003cspan citationid=\"CR73\" class=\"CitationRef\"\u003e73\u003c/span\u003e\u003c/sup\u003e. As depicted in Figures S3D-S3F, we confirmed the potent antitumor effect of DMXAA in the Pan02 tumor model in C57BL/6J mice. When comparing the antitumor effects of Co68 and DMXAA, we observed that Co68 exhibited a more pronounced antitumor effect than DMXAA (Figs.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eJ and \u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eK). This enhanced efficacy of Co68 may be attributed to its additional anti-inflammatory effects, which were absent in DMXAA treatment. Taken together, these results demonstrate that Co68 not only exhibits superior antitumor efficacy compared to DMXAA but also shows enhanced therapeutic potential when combined with PD-1 antibody treatment. Moreover, we observed that Co68 not only significantly inhibited the growth of Pan02 tumors but also demonstrated potent antitumor activity in the MC38 mouse colon tumor model (Figures S3G-S3I), indicating that Co68 may offer broad therapeutic potential across various tumor types.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003eFlow cytometry analysis further confirmed these results, showing a marked increase in the proportion of CD8⁺ T cells and NK cells, alongside a concurrent decrease in CD4⁺ T cells in the Co68-treated group compared to the PBS control (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eL) in the Pan02 tumor. To further investigate the antitumor mechanisms of Co68 and to compare its superior efficacy with DMXAA in the pancreatic cancer context, we performed single-cell RNA sequencing (scRNA-seq) on the Pan02 tumor model following treatment with Co68, DMXAA, and PBS. As shown in Figure S3J, unsupervised clustering using t-distributed Stochastic Neighbor Embedding (tSNE) of tumor cells from the PBS and Co68 treatment groups revealed 25 distinct cell subpopulations. Subsequent annotation of these clusters (Figure S3K), based on the expression of subpopulation-specific marker genes, identified various cell types within the tumor microenvironment, including cancer cells, cancer-associated fibroblasts (CAFs), TAMs, tumor-associated neutrophils (TANs), T cells, B cells, NK cells, dendritic cells (DCs), endothelial cells, and epithelial cells. Figure S3K also revealed that Co68 treatment led to significant alterations in the relative abundance of TAMs, TANs, T cells, and cancer cells compared to the PBS control group. To further investigate these observations, we performed sub-clustering and annotation of these cell populations, which enabled us to assess the impact of Co68 on gene expression and pathway alterations within each subpopulation. The gene markers used for cell annotation are shown in Figure S3L. Unsupervised clustering using t-SNE and subsequent annotation of cell populations revealed a tumor microenvironment predominantly enriched in CAFs, with TAMs being the most abundant immune cell population (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eM), consistent with previous reports\u003csup\u003e\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e\u003c/sup\u003e. We then compared the relative abundance of various immune cell subpopulations between the Co68 and PBS treatment groups. As shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eN, Co68 treatment led to a significant increase in the proportion of CD8⁺ T cells and NK cells compared to the PBS control group, consistent with the flow cytometry results presented above (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eL). In contrast, the proportion of the Cancer Cell 1 subpopulation was notably reduced. Differential gene expression analysis of the CD8⁺ T cell population in the Co68 group revealed upregulation of genes involved in T cell activation, lymphocyte differentiation and proliferation, response to interferon-gamma, and cytokine-mediated signaling pathways (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eO) compared to the PBS group. Collectively, these results suggest that Co68 treatment not only increases the abundance of CD8⁺ T cells within the pancreatic tumor microenvironment but also enhances their activation, thereby strengthening their antitumor ability.\u003c/p\u003e\u003cp\u003e\u003cem\u003eCo68 reprograms TAMs within the pancreatic tumor microenvironment\u003c/em\u003e\u003c/p\u003e\u003cp\u003eNotably, the proportion of monocytes, TAMs, and the TANs1 subpopulation was higher in the Co68-treated group compared to the PBS control (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eA), suggesting that these cell types play crucial roles in Co68-mediated modulation of the pancreatic tumor microenvironment. Given the established involvement of TAM-derived IL-1β and pancreatic cancer cell-derived PGE2 in the initiation, progression, and metastasis of pancreatic cancer\u003csup\u003e\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e,\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e,\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e,\u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e38\u003c/span\u003e\u003c/sup\u003e, we focused our subsequent analyses toward investigating the effects of Co68 on TAMs. Consistent with our hypothesis, gene set enrichment analysis revealed that Co68-treated TAMs exhibited significantly increased expression of genes associated with immune response activation, positive regulation of innate immune responses, cellular response to type I interferons, pattern recognition receptor signaling, and negative regulation of inflammatory responses (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eB). In contrast, PBS-treated TAMs were enriched in genes related to cellular responses to interleukin-1, regulation of mononuclear cell proliferation, and cholesterol metabolic processes (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eB).\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003ePseudotime analysis identified the TAMs1 subpopulation as the primary target of Co68-mediated reprogramming (Figs.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eC-\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eE). Differential gene expression and gene set enrichment analyses of the TAMs1, Mono, and TAMs2 subpopulations showed that TAMs1 exhibited significantly increased expression of genes involved in the positive regulation of T cell activation, antigen processing and presentation, and lymphocyte proliferation (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eF). Based on the expression of specific marker genes, we renamed the TAMs1, Mono, and TAMs2 subpopulations as IL-1β TAMs, HSP TAMs, and S100A10 TAMs, respectively, with IL-1β TAMs representing the Co68-responsive subpopulation (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eG). Further differential gene expression analysis of IL-1β TAMs revealed substantial upregulation of several PRRs and ISGs, including \u003cem\u003eClec2d, Ifih1, Ly96, Cd14, Ifitm3, Cxcl10, Isg15, Oasl1\u003c/em\u003e, and \u003cem\u003eRsad2\u003c/em\u003e, in the Co68-treated group compared to the PBS control (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eH). Gene set enrichment analysis of these differentially expressed genes demonstrated that Co68-upregulated genes in the IL-1β TAMs subcluster were significantly enriched in pathways associated with the activation of innate immune responses, while downregulated genes were associated with processes such as DNA replication, chromosome segregation, and cholesterol biosynthesis (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eI). Cell-cell communication analysis using the CellChat R package revealed significant interactions between the IL-1β TAM subpopulation and Cancer Cell 1, which showed a markedly reduced proportion in the Co68-treated group compared to the PBS control (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eJ, Figures S4A and S4B). These interactions were notably stronger than those observed between Cancer Cell 1 and the TAMs2 or Mono subpopulations (Figures S4C and S4D). Further analysis of intercellular communication pathways between the IL-1β TAMs in the Co68 and PBS groups (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eK) revealed that Co68 treatment significantly enhanced IFN-I-mediated communication between IL-1β TAMs and CD4\u003csup\u003e+\u003c/sup\u003e T cells, CD8\u003csup\u003e+\u003c/sup\u003e T cells, and endothelial cells (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eL). In contrast, PBS-treated IL-1β TAMs exhibited strong IL-1-mediated communication with CAFs, Cancer Cell 1, Cancer Cell 3, monocytes, and epithelial cells (Figs.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eK and \u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eM). Consistent with previous reports\u003csup\u003e\u003cspan citationid=\"CR74\" class=\"CitationRef\"\u003e74\u003c/span\u003e,\u003cspan citationid=\"CR75\" class=\"CitationRef\"\u003e75\u003c/span\u003e\u003c/sup\u003e, we also observed that TAMs communicate with other cell populations in the TME through the SPP1 signaling pathway (Figure S4E).\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003eGiven the significant alterations in neutrophil subpopulation proportions in the Co68-treated group (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eA) and the robust intercellular communication between IL-1β TAMs and neutrophils as identified by CellChat (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eJ), we further explored the effects of Co68 on neutrophil reprogramming within the pancreatic tumor microenvironment (Figure S4F). Co68 treatment significantly increased the proportion of TANs1 cells while decreasing the proportion of TANs2 cells compared to the PBS control (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eN). Differential gene expression analysis of the three neutrophil subpopulations revealed that Co68 treatment upregulated the expression of several ISGs in the TANs1 subpopulation (Figure S4G), including \u003cem\u003eIfit1, Ifit2, Ifit3, Rsad2, Parp14, Isg20, Irf7, Ifi204\u003c/em\u003e, and \u003cem\u003eIfi47\u003c/em\u003e. Gene set enrichment analysis demonstrated that the TANs1 subpopulation, with upregulated ISG expression, was significantly enriched in pathways related to the activation of innate immune responses, PRR signaling, regulation of innate immune responses, and response to interferon-beta (Figure S4H). These findings align with previous studies highlighting the enhanced antitumor activity of neutrophils with high ISG expression\u003csup\u003e\u003cspan additionalcitationids=\"CR77\" citationid=\"CR76\" class=\"CitationRef\"\u003e76\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR78\" class=\"CitationRef\"\u003e78\u003c/span\u003e\u003c/sup\u003e, suggesting that the induced ISG⁺ TANs1 may also play a crucial role in the antitumor effect of Co68 on pancreatic cancer.\u003c/p\u003e\u003cp\u003eTo further investigate the differential effects of Co68 and DMXAA at the single-cell level, we performed scRNA-seq analysis on Pan02 tumors treated with either compound. Unsupervised clustering and annotation of the resulting data revealed distinct differences in cell subpopulation ratios between the Co68 and DMXAA treatment groups (Figs.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eO and S4I) and compared to the Co68 and PBS groups (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eM). Subsequent subclustering of TAMs based on marker gene expression identified four distinct subpopulations: IL-1β TAMs, S100A10 TAMs, HSP TAMs, and MKI67 TAMs (Figs.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eP and S4J). Differential gene expression analysis of the IL-1β TAM subpopulation (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eQ) revealed that Co68 treatment, compared to DMXAA, significantly upregulated the expression of several IFN-I-related genes, including \u003cem\u003eIl1rn, Oasl1, Rsad2, Tnf, Cxcl10\u003c/em\u003e, and \u003cem\u003eIsg15\u003c/em\u003e (Figs.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eR and S4L). In parallel, Co68 treatment downregulated pro-inflammatory genes such as \u003cem\u003eIl1a, Il1b\u003c/em\u003e, and \u003cem\u003ePtgs2\u003c/em\u003e (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eR). Gene set enrichment analysis of the upregulated genes in Co68-treated IL-1β TAMs revealed significant enrichment in pathways associated with immune response activation, regulation of innate immune responses, and tumor necrosis factor production (Figure S4K). These results suggest that Co68, unlike DMXAA, not only elicits a stronger IFN-I response but also attenuates IL-1-driven inflammatory signaling. Collectively, these findings highlight that Co68 enhances antitumor immunity by reprogramming TAMs and TANs within the pancreatic tumor microenvironment, modulating the balance between IL-1 and IFN-I signaling, and exhibiting superior therapeutic potential over DMXAA.\u003c/p\u003e\u003cp\u003e\u003cem\u003eCo68 activates type I interferon via a novel binding mode with TLR4\u003c/em\u003e\u003c/p\u003e\u003cp\u003eTo identify the pattern recognition receptor (PRR) involved in Co68-mediated activation of interferon signaling, we initially employed pharmacological inhibitors targeting key innate immune pathways, followed by genetic knockout (KO) approaches. Inhibition of TLR4, TRIF, and TBK1/IKKe significantly reduced Co68-induced IFNB1 expression in both RAW 264.7 and J774A.1 cell lines (Figs.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eA-B). These results were further validated using BMDMs from TLR4, TRIF, and IRF3 knockout mice (Figs.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eC-D and Figure S5A), collectively confirming that Co68 induces IFNB1 expression via activation of the TLR4-TRIF-TBK1-IRF3 signaling axis. Previous studies have demonstrated that LPS-induced signaling through TLR4 requires the engagement of CD14 and MD2\u003csup\u003e79\u003c/sup\u003e. In contrast, Co\u003csup\u003e2+\u003c/sup\u003e directly binds to and activates TLR4 without the involvement of CD14 or MD2\u003csup\u003e71\u003c/sup\u003e. To investigate the roles of CD14 and MD2 in Co68-mediated TLR4 activation, we employed blocking antibodies specific to these molecules. As shown in Figs.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eE-\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eG and Figures S5B-S5D, blocking antibodies targeting TLR4, CD14, and MD2 significantly attenuated the production of \u003cem\u003eIfnb1\u003c/em\u003e and \u003cem\u003eIsg15\u003c/em\u003e induced by Co68. Notably, antibodies against TLR4 and MD2 nearly abolished Co68-induced expression of \u003cem\u003eIfnb1\u003c/em\u003e, suggesting that TLR4 and MD2 play a more dominant role than CD14 in the induction of IFN-I. Furthermore, the transfection of plasmids encoding TLR4, CD14, and MD2 into HEK293T cells significantly induced the expression of \u003cem\u003eIFNB1\u003c/em\u003e and \u003cem\u003eISG15\u003c/em\u003e upon Co68 treatment (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eH and Figure S5E), further corroborating the critical involvement of these molecules in Co68-mediated activation of the IFN-I signaling pathway.\u003c/p\u003e\u003cp\u003eTo explore the direct interaction between Co68 and TLR4, we employed both the microscale thermophoresis (MST) and surface plasmon resonance (SPR) techniques. MST analysis revealed a dissociation constant (Kd) of 0.98 \u0026micro;M for Co68 binding to TLR4, which is lower than the dissociation constant for LPS binding to TLR4 (Kd\u0026thinsp;=\u0026thinsp;4.37 \u0026micro;M) (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eI and Figure S5F). Consistent with these findings, SPR analysis showed a Kd of 31.71 \u0026micro;M for TLR4-Co68 binding (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eJ and Figure S5G), compared to 147.64 \u0026micro;M for TLR4-LPS binding (Figure S5H-S5I). These results collectively demonstrate that Co68 binds directly to TLR4 with a more stable interaction than LPS, suggesting a distinct binding mode. Given the observed differences in the expression patterns between Co68, LPS, and Co\u003csup\u003e2+\u003c/sup\u003e, we hypothesized that Co68 exhibits a unique binding mode to TLR4 compared to LPS and Co\u003csup\u003e2+\u003c/sup\u003e. To further explore these differences, we employed computational approaches, including AlphaFold3 and AutoDock Vina (V1.2.5) for molecular docking studies. The crystal structure of Co68 was visualized using PyMOL (Figure S5J). Molecular docking with AutoDock Vina (v1.2.5) identified the most stable binding mode of Co68 to TLR4, involving interaction with the hydrophobic pocket of MD2 (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eK and Figure S5K), with a binding energy of less than \u0026minus;\u0026thinsp;7 kcal/mol. As shown in Figs.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eL and S5L, Co68 interacted with both TLR4 and MD2, with a stronger interaction observed with MD2. Further analysis of the interacting amino acids using PyMOL revealed that Co68 interacted with eight amino acids in TLR4: ARG434, MET358, ARG380, ALA382, PHE406, ASN407, ILE411, and LYS433 (Figs.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eM and Figures S5M-S5N). These residues differ from those previously reported for TLR4-LPS interactions\u003csup\u003e\u003cspan citationid=\"CR79\" class=\"CitationRef\"\u003e79\u003c/span\u003e\u003c/sup\u003e, including ARG264, LYS341, LYS362, and LYS388. Additionally, chemical bonding analysis of Co68 and MD2 (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eN and Figure S5O) predicted hydrogen bonds involving residues such as ILE32, ILE46, ILE52, LEU54, VAL61, LEU78, CYS133, and ILE153, which were distinct from the bonding pattern seen in Co\u003csup\u003e2+\u003c/sup\u003e binding to TLR4 predicted by AlphaFold3 (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eO).\u003c/p\u003e\u003cp\u003eTo experimentally validate this distinct binding pattern, we introduced mutations in the eight predicted interacting amino acids of TLR4 with Co68. Luciferase assays revealed that mutations at F406A and K433R significantly reduced Co68-induced IFNB1 production (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eP). Notably, the simultaneous mutations at F406A and K433R led to a more significant reduction in IFNB1 expression. Strikingly, these mutations did not affect the activation of downstream NF-κB signaling induced by CoCl₂ and LPS (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eQ), further confirming that Co68 activates IFN-I signaling through a TLR4-dependent pathway that is distinct from the mechanisms of CoCl₂ and LPS. To further investigate the structure-activity relationship between Co68 and TLR4, we designed nine Co68 analogs with gradual substitutions in functional groups (Figure S5P). As shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eR, the PNP structure, tertiary butyl binding to the phosphate radical, chloride ion binding to the Co\u0026sup2;⁺ ion, and hydrogen atom binding to the nitrogen in the PNP structure were critical for Co68\u0026rsquo;s ability to induce \u003cem\u003eIfnb1\u003c/em\u003e expression. These structural features are crucial for Co68\u0026rsquo;s capacity to activate the IFN-I signaling pathway. Furthermore, analogs of Co68 that significantly induced \u003cem\u003eIfnb1\u003c/em\u003e expression in RAW 264.7 macrophages also demonstrated stable binding to TLR4, as validated by the SPR assay (Figure S6A-S6F). Consistent with these results, Co68 was unable to inhibit Pan02 tumor growth in TLR4 and IRF3 knockout mice (Figures S6G-S6L), further reinforcing the conclusion that Co68\u0026rsquo;s antitumor activity in vivo is mediated through the TLR4-TBK1-IRF3-IFN-I signaling axis. Taken together, these findings demonstrate that Co68 activates IFN-I signaling through TLR4, with a binding mode distinct from both LPS and Co\u0026sup2;⁺, underscoring its unique mechanism during immune modulation.\u003c/p\u003e\u003cp\u003e\u003cem\u003eCo68 inhibits inflammatory responses through the non-canonical activation of STAT1\u003c/em\u003e\u003c/p\u003e\u003cp\u003eAs demonstrated above, Co68 exhibits broad-spectrum anti-inflammatory effects and specifically induces the expression of several histone-modifying genes. Based on these observations, we hypothesized that a Co68-specific transcription factor mediates its anti-inflammatory activity. To identify potential Co68-responsive transcription factors, we conducted ATAC-seq analysis on samples treated with LPS, LPS-Co68, and Co68 for 1 hour, prior to detectable expression of \u003cem\u003eIfnb1\u003c/em\u003e. Remarkably, Co68 exhibited a potent anti-inflammatory effect on LPS even at this early time point. ATAC-seq identified 50165, 41324, and 33316 accessible peaks in the LPS, LPS-Co68, and Co68 treatment groups, respectively (Figures S7A-C). The annotation of these accessible peaks with the ChIPseeker R package revealed that both the Co68 and LPS-Co68 groups exhibited a more significant number of accessible peaks in promoter regions compared to the LPS-only group (Figure S7D). Additionally, there was a higher density of transcription factor-binding sites within 3 kb of transcription start sites (TSS) in the LPS-Co68 and Co68 groups compared to the LPS group, suggesting an increase in chromatin binding sites for transcription factors induced by Co68 (Figure S7E). In contrast, peaks identified in the LPS group were predominantly located in distal intergenic regions (Figure S7D). As illustrated in the heatmap (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003eA), the Co68 and LPS-Co68 groups demonstrated a marked increase in TSS-proximal peak accessibility compared to the LPS-only group. Using the ClueGO plug-in in Cytoscape, we annotated genes nearest to the differential accessible regions (DARs) between the LPS-Co68 and LPS groups (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003eB), revealing significant enrichment in pathways associated with myeloid cell activation, response to interferon-gamma, macrophage migration, dendritic cell differentiation, T cell receptor signaling, and activated T cell proliferation.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003eTo predict potential binding transcription factors, we performed a footprint analysis using TOBIAS software (v0.17.0), which revealed distinct transcription factors between the LPS and Co68 groups (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003eC). As expected, NFKB-REL exhibited significantly more potential binding sites in the LPS group compared to the Co68 group, while IRF3 and EGR1 were significantly enriched in the Co68 group. Surprisingly, STAT1 was also significantly enriched in the Co68 group, even at the 1-hour time point before detectable expression of the upstream signaling molecule \u003cem\u003eIfnb1\u003c/em\u003e (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003eC). The binding motif for STAT1, derived from the JASPAR database, is shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003eD. A heatmap of potential STAT1 binding sites revealed notably higher occupancy in the LPS-Co68 and Co68 groups compared to the LPS group (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003eE). Consistent with these findings, the binding motifs for STAT1 and EGR1, illustrated in Figures S7F and S7G, displayed markedly higher binding intensity in the Co68 group than in the LPS group. In contrast, the binding intensity of REL in the LPS group was stronger than in the Co68 group (Figure S7H). Track visualization using IGV software further confirmed these findings, showing stronger binding signals for genes such as \u003cem\u003eIl6\u003c/em\u003e, \u003cem\u003eIl1b\u003c/em\u003e, and \u003cem\u003eTnf\u003c/em\u003e in the LPS group compared to the LPS-Co68 and Co68 groups (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003eG and Figures S7J-S7K). In contrast, genes such as \u003cem\u003eRsad2\u003c/em\u003e, \u003cem\u003eIfi202b\u003c/em\u003e, \u003cem\u003eUsp18\u003c/em\u003e, and \u003cem\u003eIfit1\u003c/em\u003e exhibited stronger binding signals in the LPS-Co68 and Co68 groups compared to the LPS group (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003eF and Figure S7I). These results were further corroborated by RNA-seq analysis following Co68 treatment for 1 hour (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003eH and Figure S7L), suggesting that STAT1 activation occurs independently of, and prior to, IFN-I signaling via IFNAR.\u003c/p\u003e\u003cp\u003eTo investigate the functional role of early, IFN-I-independent STAT1 activation by Co68, we first employed the STAT1 inhibitor fludarabine. Notably, fludarabine treatment abrogated Co68\u0026rsquo;s anti-inflammatory effects (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003eI). This finding was further validated by Stat1 knockdown using shRNA (Figure S7N), which similarly abolished Co68\u0026rsquo;s anti-inflammatory activity (Figs.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003eJ and \u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003eK). To elucidate the mechanism through which Co68 activates STAT1, we examined the effects of inhibitors targeting TLR4, MyD88, and TRIF. Notably, only the TLR4 inhibitor significantly suppressed STAT1 activation, while MyD88 and TRIF inhibitors had no effect (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003eL). These results indicate that early STAT1 activation by Co68 is TLR4-dependent but MyD88- and TRIF-independent. To identify upstream regulators of STAT1 activation, we performed mass spectrometry analysis following immunoprecipitation (IP) of TLR4 (Figure S7M). This analysis revealed 20 kinases that were specifically upregulated 1-hour post-Co68 treatment, including SYK, a known direct upstream regulator of STAT1 phosphorylation (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003eM). Co-immunoprecipitation (Co-IP) experiments confirmed that TLR4 and STAT1 interacted with SYK only in the LPS-Co68 and Co68 treatment groups at 1-hour post-treatment (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003eN). Furthermore, Syk knockdown via shRNA (Figure S7N) and SYK knockout (KO) via CRISPR-Cas9 (Figure S7P) abolished Co68\u0026rsquo;s anti-inflammatory effects (Figure S7O, Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003eO, and Figure S7Q). Collectively, these data demonstrate that Co68\u0026rsquo;s broad-spectrum anti-inflammatory activity is dependent on the TLR4-SYK-STAT1 signaling pathway.\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eA growing body of evidence underscores the critical role of imbalances in innate immune signaling\u0026mdash;particularly between type I interferon (IFN-I) signaling and inflammatory responses\u0026mdash;in the pathogenesis and progression of a wide array of diseases\u003csup\u003e\u003cspan citationid=\"CR67\" class=\"CitationRef\"\u003e67\u003c/span\u003e,\u003cspan citationid=\"CR80\" class=\"CitationRef\"\u003e80\u003c/span\u003e\u003c/sup\u003e. The immune evasion strategies employed by pathogens such as SARS-CoV-2, coupled with the pro-tumor activity of tumor-associated macrophages (TAMs) within the pancreatic cancer microenvironment, highlight the centrality of this immune imbalance in both viral infections and cancer progression\u003csup\u003e\u003cspan additionalcitationids=\"CR82\" citationid=\"CR81\" class=\"CitationRef\"\u003e81\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR83\" class=\"CitationRef\"\u003e83\u003c/span\u003e\u003c/sup\u003e. Rectifying these immune dysregulations presents a promising therapeutic avenue to improve clinical outcomes across diverse disease settings. To this end, we developed DLINP, a convolutional neural network (CNN)-based deep learning model specifically designed to identify small molecules with dual immunomodulatory activities. Through this platform, we identified Co68 as a new immune modulator, a cobalt-based metal complex synthesized by the reaction of a PNP-pincer ligand with cobalt chloride in tetrahydrofuran. We demonstrated here that Co68 robustly activates IFN-I-dependent antiviral responses while simultaneously exhibiting broad-spectrum anti-inflammatory effects, both in vitro and in vivo. Mechanistically, Co68 enhances IFN-I signaling via the TLR4-TRIF pathway, a critical cascade that initiates antiviral and antitumor immunity, while concurrently exerting potent anti-inflammatory effects through the TLR4-SYK-STAT1 axis. Notably, Co68 demonstrated substantial antitumor activity in the Pan02 pancreatic cancer mouse model, which is resistant to PD-1 blockade, outperforming the established STING agonist DMXAA. Moreover, Co68 exhibited significant synergy with PD-1 antibody therapy, promoting enhanced tumor suppression and improved therapeutic outcomes. Further mechanistic insights derived from single-cell transcriptomic analysis revealed that Co68 disrupted the IL-1β-PGE2 axis between TAMs and cancer cells\u0026mdash;an essential pathway involved in tumor immune evasion. Co68 also substantially increased the abundance and functional activity of CD8\u003csup\u003e+\u003c/sup\u003e T cells and NK cells, likely through IFN-I signaling induction in TAMs, thereby bolstering antitumor immunity.\u003c/p\u003e\u003cp\u003eThis study emphasizes the transformative potential of deep learning in drug discovery. Co68, identified through the DLINP platform, represents the first PNP-pincer ligand to exhibit significant immunomodulatory activity. The PNP-pincer ligand structure of Co68 makes cobalt chloride exert dual effects on immune signaling via TLR4-mediated modulation of transcriptional landscapes, suppressing deleterious pro-inflammatory responses while enhancing IFN-I signaling. This positions Co68 as a next-generation immunomodulatory agent. Importantly, the PNP ligand (as shown in Figure \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003eD), used in the synthesis of Co68, does not activate IFN-I signaling in RAW 264.7 cells. Furthermore, cobalt chloride\u0026mdash;one of the substrates used to synthesize Co68\u0026mdash;only weakly activates the inflammatory response at high concentrations (500 \u0026micro;M) after 24 hours and does not induce significant transcriptional changes in macrophages within 3 hours of exposure. In contrast, Co68 activates IFN-I signaling at much lower concentrations (50 \u0026micro;M) within just 1 hour of treatment, with peak levels achieved by 3 hours. Co68 also exhibits potent anti-inflammatory activity against diverse stimuli, including LPS, VSV, HSV, and EMCV. Structure-activity relationship experiments demonstrated that the PNP ligand itself possessed broad-spectrum anti-inflammatory properties. Replacing the hydrogen atom bound to nitrogen in the PNP structure abrogates Co68's ability to activate IFN-I signaling. Additionally, substituting chloride ions in cobalt chloride with bromide or iodide significantly impairs Co68's capacity to activate IFN-I signaling. We hypothesize that the PNP ligand, when complexed with cobalt chloride, generated Co68 in a high-energy state, which endows the complex with unexpected biological activities, warranting further investigation to explore the underlying molecular mechanisms. As the research progresses, further studies on Co68\u0026rsquo;s pharmacokinetics, safety profile, and efficacy in additional disease models\u0026mdash;such as psoriasis and SARS-CoV-2 infection\u0026mdash;are necessary to assess its clinical potential. Future work will also focus on examining the role of histone-modifying genes, such as H1f5 and H2bc4, in Co68\u0026rsquo;s regulation of chromatin accessibility, providing deeper mechanistic insights into its molecular actions.\u003c/p\u003e\u003cp\u003eWe employed an array of pharmacological, genetic, and experimental techniques\u0026mdash;including Microscale Thermophoresis (MST) and Surface Plasmon Resonance (SPR)\u0026mdash;to confirm that Co68 directly interacts with TLR4, a member of the Toll-like receptor (TLR) family, which is key pattern recognition receptors (PRRs) involved in immune signaling\u003csup\u003e\u003cspan citationid=\"CR84\" class=\"CitationRef\"\u003e84\u003c/span\u003e\u003c/sup\u003e. TLR4 is particularly notable for its role in initiating immune responses. Previous studies have shown that TLR4 recognizes lipopolysaccharides (LPS) in conjunction with its co-receptors CD14 and MD2, triggering the activation of the NF-κB pathway through the MAL-MyD88 axis\u003csup\u003e\u003cspan citationid=\"CR79\" class=\"CitationRef\"\u003e79\u003c/span\u003e,\u003cspan additionalcitationids=\"CR86\" citationid=\"CR85\" class=\"CitationRef\"\u003e85\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR87\" class=\"CitationRef\"\u003e87\u003c/span\u003e\u003c/sup\u003e. This cascade leads to the production of pro-inflammatory cytokines central to the innate immune inflammatory response. TLR4 also activates a distinct signaling pathway upon internalization into endosomes, triggering the TRAM-TRIF axis to stimulate IRF3 and promote IFN-I production, thereby enhancing antiviral immunity\u003csup\u003e\u003cspan additionalcitationids=\"CR89\" citationid=\"CR88\" class=\"CitationRef\"\u003e88\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR90\" class=\"CitationRef\"\u003e90\u003c/span\u003e\u003c/sup\u003e. The dual roles of TLR4 in both inflammation and antiviral immunity make it a compelling target for therapeutic intervention, although its activation can also result in adverse effects, such as inflammation triggered by metal allergens like nickel and cobalt, or chemotherapeutic agents such as cisplatin, which activate TLR4 and exacerbate inflammation, potentially impairing patient outcomes\u003csup\u003e\u003cspan citationid=\"CR71\" class=\"CitationRef\"\u003e71\u003c/span\u003e,\u003cspan citationid=\"CR91\" class=\"CitationRef\"\u003e91\u003c/span\u003e,\u003cspan citationid=\"CR92\" class=\"CitationRef\"\u003e92\u003c/span\u003e\u003c/sup\u003e. Similarly, the spike (S) protein of SARS-CoV-2 has been shown to activate TLR4\u003csup\u003e93\u003c/sup\u003e, contributing to cytokine storms in COVID-19 patients, thus emphasizing TLR4's crucial role in immune regulation and its potential as a therapeutic target in a variety of diseases, ranging from infections to cancer and autoimmune disorders. Through AlphaFold predictions and molecular docking, we observed significant differences in the binding modes of Co68 compared to cobalt ions and LPS, with these findings further validated by point mutation experiments. Cryogenic electron microscopy (cryo-EM) is currently being employed to gain detailed structural insights into the Co68-TLR4 interaction. Notably, Co68 uniquely activates the TLR4-TRAM-TRIF-IRF3 pathway, bypassing the conventional MAL-MyD88-NF-κB cascade, thereby reprogramming TLR4-mediated immune responses. This unique mechanism may explain certain clinical observations in cancer therapies. For instance, paclitaxel, a widely used chemotherapeutic agent, targets TLR4, but its albumin-bound formulation (nab-paclitaxel) has been linked to enhanced tumorigenesis in specific clinical settings\u003csup\u003e\u003cspan additionalcitationids=\"CR95\" citationid=\"CR94\" class=\"CitationRef\"\u003e94\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR96\" class=\"CitationRef\"\u003e96\u003c/span\u003e\u003c/sup\u003e. This discrepancy may stem from structural differences in their binding modes to TLR4, resulting in divergent effects on immune signaling and inflammation.\u003c/p\u003e\u003cp\u003eA pivotal finding of this study is the identification of STAT1 as a critical transcription factor in Co68-induced early activation, which is independent of IFN-I signaling and essential for the compound's anti-inflammatory effects. ATAC-seq and footprint analysis revealed significant differences in chromatin accessibility between Co68- and LPS-treated cells, with STAT1 binding sites being notably enriched in the Co68 group. The early activation of STAT1, occurring prior to detectable \u003cem\u003eIfnb1\u003c/em\u003e expression, was found to be a key factor in Co68\u0026rsquo;s anti-inflammatory activity. Furthermore, our data confirmed that TLR4 is involved in the activation of STAT1, with MyD88 and TRIF being dispensable in this process. Mass spectrometry identified SYK as an upstream kinase in the TLR4-SYK-STAT1 signaling axis, and co-immunoprecipitation (Co-IP) experiments revealed that SYK interacts with both TLR4 and STAT1 in the presence of Co68. These findings provide novel mechanistic insights into the TLR4-SYK-STAT1 pathway, which had not been previously described\u003csup\u003e\u003cspan additionalcitationids=\"CR98\" citationid=\"CR97\" class=\"CitationRef\"\u003e97\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR99\" class=\"CitationRef\"\u003e99\u003c/span\u003e\u003c/sup\u003e. The TLR4-SYK-STAT1 axis emerges as a promising therapeutic target for inflammatory diseases. Although the precise mechanisms by which STAT1 mediates its anti-inflammatory effects remain to be fully understood, we propose that STAT1 activation may resemble the action of STAT3 in TLR4 signaling, where phosphorylation of serine 727 by downstream kinases such as TBK1 induces metabolic reprogramming and anti-inflammatory effects\u003csup\u003e\u003cspan citationid=\"CR100\" class=\"CitationRef\"\u003e100\u003c/span\u003e\u003c/sup\u003e. Further studies are needed to validate this hypothesis and deepen our understanding of the signaling dynamics at play. Co68 provides dual therapeutic benefits by concurrently activating the TLR4-TRIF and TLR4-SYK-STAT1 signaling pathways, thereby mitigating inflammation while enhancing antiviral and antitumor immunity.\u003c/p\u003e\u003cp\u003eIn the context of cancer, particularly pancreatic cancer, Co68\u0026rsquo;s ability to modulate the immune landscape by suppressing IL-1β signaling and enhancing IFN-I responses makes it a promising therapeutic candidate\u003csup\u003e\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e,\u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e38\u003c/span\u003e\u003c/sup\u003e. In this tumor type, TAMs are a major source of IL-1β, which drives tumor progression and inhibits IFN-I signaling through PGE2 produced by tumor cells, ultimately promoting immune evasion\u003csup\u003e\u003cspan additionalcitationids=\"CR102\" citationid=\"CR101\" class=\"CitationRef\"\u003e101\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR103\" class=\"CitationRef\"\u003e103\u003c/span\u003e\u003c/sup\u003e. Compared to other antitumor agents, such as DMXAA, which also significantly inhibits the growth of Pan02 pancreatic tumor in mice\u003csup\u003e\u003cspan citationid=\"CR104\" class=\"CitationRef\"\u003e104\u003c/span\u003e\u003c/sup\u003e, Co68 induces TLR4 activation more effectively than DMXAA activates STING. DMXAA\u0026rsquo;s activation of STING leads to potent NF-κB activation\u003csup\u003e\u003cspan citationid=\"CR73\" class=\"CitationRef\"\u003e73\u003c/span\u003e\u003c/sup\u003e, causing substantial production of IL-1β and IL-6\u0026mdash;cytokines that promote tumor growth and immune suppression. In contrast, Co68 attenuates IL-1β-mediated pro-inflammatory signaling, thereby reducing the tumor-promoting inflammation commonly observed in pancreatic cancer. Notably, Co68 treatment upregulated ISGs in specific subpopulations of TAMs, which facilitated the recruitment of CD8\u003csup\u003e+\u003c/sup\u003e T cells and NK cells\u0026mdash;key players in antitumor immunity\u003csup\u003e\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e,\u003cspan citationid=\"CR105\" class=\"CitationRef\"\u003e105\u003c/span\u003e,\u003cspan citationid=\"CR106\" class=\"CitationRef\"\u003e106\u003c/span\u003e\u003c/sup\u003e. A striking observation was the selective expansion of a neutrophil subpopulation enriched in ISGs. Given the growing evidence supporting the role of ISG-high neutrophils in antitumor immunity\u003csup\u003e\u003cspan additionalcitationids=\"CR77\" citationid=\"CR76\" class=\"CitationRef\"\u003e76\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR78\" class=\"CitationRef\"\u003e78\u003c/span\u003e,\u003cspan additionalcitationids=\"CR108\" citationid=\"CR107\" class=\"CitationRef\"\u003e107\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR109\" class=\"CitationRef\"\u003e109\u003c/span\u003e\u003c/sup\u003e, we hypothesize that this subset plays a crucial role in Co68\u0026rsquo;s enhanced therapeutic efficacy. However, further studies are required to fully elucidate the precise function of this neutrophil population within the TME. Additionally, Co68\u0026rsquo;s synergy with PD-1 blockade led to a more potent antitumor effect in the Pan02 model, suggesting that Co68 can overcome immunosuppressive barriers. This synergistic effect complements PD-1 antibody-mediated activation of CD8\u003csup\u003e+\u003c/sup\u003e T cells within the TME, restoring immune surveillance through dual mechanisms\u003csup\u003e\u003cspan citationid=\"CR110\" class=\"CitationRef\"\u003e110\u003c/span\u003e\u003c/sup\u003e. Co68\u0026rsquo;s dual activity is particularly beneficial for \u0026ldquo;cold\u0026rdquo; tumors, such as pancreatic cancer and triple-negative breast cancer, which are often characterized by immune evasion\u003csup\u003e\u003cspan citationid=\"CR95\" class=\"CitationRef\"\u003e95\u003c/span\u003e\u003c/sup\u003e. Future studies are essential to evaluate the broader applicability of Co68 in other tumor models and further explore its clinical potential as an immune-modulatory agent for cancer therapy.\u003c/p\u003e\u003cp\u003eIn conclusion, Co68's unique mechanism of action, mediated by its PNP-pincer ligand structure, positions it as a promising therapeutic agent for targeting immune dysregulation in diseases such as cancers and viral infections. As research progresses, Co68 could provide innovative strategies for overcoming immunosuppressive tumor microenvironments, ultimately enhancing immune responses and improving therapeutic outcomes in cancer treatment. Co68 represents a critical step forward in the development of PNP-pincer ligands, opening new avenues for immune system modulation and offering potential solutions for treating cancers that continue to claim millions of lives worldwide.\u003c/p\u003e"},{"header":"Materials and methods","content":"\u003ch4\u003eReagents and antibodies\u003c/h4\u003e\n\u003cp\u003eRabbit anti-COX2 antibodies were sourced from Wanleibio. Antibodies targeting LY96/MD2 (catalog no. 11784-1-AP), iNOS (catalog no. 80517-1-RR), TBK1 (catalog no. 28397-1-AP), NF-\u0026kappa;B p65 (catalog no. 10745-1-AP), and Beta Tubulin (catalog no. 10094-1-AP) were obtained from Proteintech. Mouse monoclonal antibodies against Beta Actin (catalog no. 66009-1-Ig), TLR4 (catalog no. 66350-1-Ig), CD14 (catalog no. 60253-1-Ig), and the HA tag (catalog no. 51064-2-AP) were also purchased from Proteintech. Phosphorylation-specific antibodies, including rabbit anti-phospho-NF-\u0026kappa;B p65 (Ser536) (93H1) (catalog no. 3033), anti-phospho-I\u0026kappa;B\u0026alpha; (Ser32) (14D4) (catalog no. 2859), anti-phospho-TBK1/NAK (Ser172) (D52C2) (catalog no. 5483), anti-phospho-IRF-3 (Ser396) (D6O1M) (catalog no. 29047), as well as antibodies for IRF-3 (D6I4C) (catalog no. 11904), Stat1 (D1K9Y) (catalog no. 14994), and Syk (D3Z1E) (catalog no. 13198), were obtained from Cell Signaling Technology. For secondary detection, horseradish peroxidase (HRP)-conjugated goat anti-rabbit IgG (H+L) (catalog no. SA00001-2) and HRP-conjugated goat anti-mouse IgG (H+L) (catalog no. SA00001-1) were both sourced from Proteintech. Flow cytometry antibodies included FITC-conjugated anti-mouse NK1.1 (catalog no. FITC-65138-25UG) and PE-conjugated anti-mouse CD8a (catalog no. PE-65069), which were procured from Proteintech, while APC-conjugated anti-mouse CD4 (catalog no. E-AB-F1097UE) was obtained from Elabscience. Recombinant mouse GM-CSF (catalog no. HY-P7361) and anti-mouse PD-1 antibody (catalog no. HY-P99144) were purchased from MedChemExpress. Recombinant human TLR4 protein (ECD, His Tag) was obtained from Sino Biological (catalog no. 10146-H08B).\u003c/p\u003e\n\u003ch4\u003eCell Culture\u003c/h4\u003e\n\u003cp\u003eThe cell lines RAW264.7, J774A.1, HEK293T, THP-1, and Vero were obtained from the American Type Culture Collection (ATCC). Bone marrow-derived macrophages (BMDMs) were isolated from the femurs and tibiae of 8-week-old C57BL/6J mice. The bone marrow was cultured in the presence of granulocyte-macrophage colony-stimulating factor (GM-CSF, RP01206, ABclonal) for 7 days to promote macrophage differentiation. Peritoneal macrophages (PMs) were similarly isolated from the peritoneal cavity and differentiated using GM-CSF for 7 days. HEK293T, RAW264.7, J774A.1, THP-1, Vero, and BMDMs were cultured in Dulbecco\u0026apos;s Modified Eagle\u0026apos;s Medium (DMEM; 03.1002C, EallBio), while PMs were maintained in RPMI 1640 (03.4001C, EallBio). All culture media were supplemented with 10% fetal bovine serum (FBS) and 1% penicillin-streptomycin to support optimal cell growth and maintenance.\u003c/p\u003e\n\u003ch3\u003e\u003cem\u003eVirus Infection and Propagation\u003c/em\u003e\u003c/h3\u003e\n\u003cp\u003eCells at 70\u0026ndash;80% confluence were infected with the following viruses at the indicated multiplicities of infection (MOI): Vesicular Stomatitis Virus (VSV, MOI = 0.1), Herpes Simplex Virus 1 (HSV-1, F strain, MOI = 0.5), and Encephalomyocarditis Virus (EMCV, MOI = 0.1). Virus stocks were prepared as follows: the VSV Indiana strain, kindly provided by Dr. J. Rose (Yale University), was propagated in Vero cells. The HSV-1 strain 17, a gift from Dr. Zhengfan Jiang (Peking University), was cultured in Vero cells for viral propagation. The EMCV strain (ATCC VR-129B) was acquired from ATCC and similarly propagated in Vero cells.\u003c/p\u003e\n\u003ch4\u003eMice and Construction of a Sepsis Model via LPS Injection\u003c/h4\u003e\n\u003cp\u003eWild-type (WT) C57BL/6J mice were obtained from the Department of Laboratory Animal Science, Peking University Health Science Center. \u003cem\u003eTlr4-/-\u003c/em\u003e, \u003cem\u003eIrf3-/-\u003c/em\u003e, and \u003cem\u003eTrif-/-\u003c/em\u003e mice, all on a C57BL/6J background, were acquired from the Institute of Experimental Animals, Chinese Academy of Medical Sciences. The \u003cem\u003eIfnar1-/-\u003c/em\u003e mice were generously provided by Prof. Erol Fikrig (Yale University). Genotyping of these mice was performed using the following primers: \u003cem\u003eTlr4\u003c/em\u003e KO (forward: 5\u0026prime;-TGCTCACACCATCATCAC-3\u0026prime;; reverse: 5\u0026prime;-CATGTACTAGGTTCGTCAGA-3\u0026prime;), \u003cem\u003eIrf3\u003c/em\u003e KO (forward: 5\u0026prime;-TCGTGCTTTACGCTATGCCGCTCCCGATT-3\u0026prime;; reverse: 5\u0026prime;-GAACCTCGGAGTTATCCCGAAGG-3\u0026prime;), and \u003cem\u003eIfnar1\u003c/em\u003e KO (forward: 5\u0026prime;-CGAGGCGAAGTGGTTAAAAG-3\u0026prime;; reverse: 5\u0026prime;-AATTCGCCAATGACAAGACG-3\u0026prime;). All animal experiments were carried out in accordance with the \u003cem\u003eGuide for the Care and Use of Laboratory Animals\u003c/em\u003e provided by the Chinese Association for Laboratory Animal Science. The protocols were approved by the Animal Care Committee of Peking University Health Science Center (permit number: LA 2016240). Mice were housed under specific pathogen-free (SPF) conditions at the Laboratory Animal Center of Peking University. Only male mice aged 6 to 8 weeks were used in the study. Lipopolysaccharide (LPS, HY-D1056, MedChemExpress) was dissolved in sterile PBS (P1020, Solarbio Life Science) to a final concentration of 20 mg/kg body weight. A total of 20 WT C57BL/6J mice were injected intraperitoneally with LPS using sterile insulin syringes. Following LPS injection, a subset of mice (n=10) received a 25 mg/kg dose of Co68 (200 \u0026mu;L in PBS) via intraperitoneal injection. The control group (n=10) was treated with an equivalent volume of PBS. Survival times were monitored and recorded following injection.\u003c/p\u003e\n\u003ch3\u003e\u003cem\u003eTumor Transplantation, Treatments, and Tissue Digestion\u003c/em\u003e\u003c/h3\u003e\n\u003cp\u003eMice were subcutaneously injected into the right flank with 5 \u0026times; 10⁵ Pan02 cells suspended in 100 \u0026mu;L PBS, unless otherwise specified. Tumor growth was monitored daily, with tumor size measured every 1\u0026ndash;2 days. Tumor volume was calculated using the formula: volume = length (mm) \u0026times; width\u0026sup2; (mm\u0026sup2;) \u0026times; 0.5. Tumor-bearing mice were treated with 25 mg/kg Co68 or DMXAA via intraperitoneal injection every 2 days. For immune checkpoint blockade, tumor-bearing mice were administered 200 \u0026mu;g of anti-mouse PD-1 antibody (200 \u0026mu;L saline) via intraperitoneal injection on days 3, 7, and 11 following tumor inoculation. Control mice received 200 \u0026mu;L of rat IgG2a isotype (Clone 2A3, BioXCell) on the same schedule. Following euthanasia, tumors and the surrounding skin were excised. Tumors were minced and digested in RPMI 1640 medium containing 0.5 mg/mL collagenase D (11088866001, Roche) and 0.1 mg/mL DNase I (DN25-100G, Sigma-Aldrich) at 37\u0026deg;C for 1 hour. The resulting cell mixture was then filtered through a 70-\u0026mu;m cell strainer to obtain a single-cell suspension, which was subsequently transferred to flow cytometry tubes for further analysis.\u003c/p\u003e\n\u003ch4\u003eRNA extraction, reverse transcription and real-time quantitative PCR\u003c/h4\u003e\n\u003cp\u003eRNA extraction from cells or tissues following various treatments or infections was performed using TRIzol reagent (TIANGEN, A0123A01). The purified RNA was then reverse transcribed into cDNA using HiScript II RT SuperMix (Vazyme, R223-01). Quantification of target gene expression was carried out using SYBR Green qMix (Vazyme, Q311) in a quantitative reverse transcription PCR (RT-qPCR) assay. The relative expression levels of target mRNAs were normalized to the housekeeping gene \u003cem\u003eGapdh\u003c/em\u003e. Detailed primer sequences used in this study are provided in Supplementary Table S1.\u003c/p\u003e\n\u003ch4\u003eLuciferase Assay\u003c/h4\u003e\n\u003cp\u003eCells were seeded into 6-well plates and cultured to 70\u0026ndash;80% confluence. Transient transfection was performed using Lipofectamine 2000 (Thermo Fisher) according to the manufacturer\u0026apos;s instructions. After 24 hours of transfection, the cells were treated with Co68, CoCl₂, or LPS at 37\u0026deg;C in a humidified incubator with 5% CO₂ for the indicated time points. Following treatment, cells were lysed directly in the wells using lysis buffer (FR203-01, Transgen), in accordance with the manufacturer\u0026rsquo;s guidelines. The resulting lysates were collected, and luciferase activity was measured using a TD20/20 Luminometer (Turner Designs).\u003c/p\u003e\n\u003ch4\u003eTotal Protein Extraction and Western Blot Analysis\u003c/h4\u003e\n\u003cp\u003eCells were lysed using RIPA Lysis Buffer (Strong) (HY-K1001, MedChemExpress), supplemented with a protease inhibitor cocktail (EDTA-Free Protease Inhibitor Cocktail) and phosphatase inhibitor cocktails (Phosphatase Inhibitor Cocktail I and III, both 100\u0026times; in DMSO). Clarified cell extracts (10 to 30 \u0026mu;g) were resolved on SDS-PAGE gels and subsequently transferred to nitrocellulose membranes (FFN08, Beyotime). Following blocking, membranes were incubated with primary antibodies specific to the target proteins. The bound secondary antibodies were detected using the enhanced chemiluminescence (ECL) method (07.10009-50, EallBio).\u003c/p\u003e\n\u003ch4\u003eCell Cytotoxicity Assay\u003c/h4\u003e\n\u003cp\u003eTarget cells were seeded into 96-well plates at a density of 2 \u0026times; 10\u0026sup3; cells per well in 100 \u0026mu;L of complete medium. Cells were treated with increasing concentrations of Co68 and incubated at 37\u0026deg;C in a humidified 5% CO₂ incubator for 48 hours. After the treatment period, 10 \u0026mu;L of CCK-8 reagent (40203ES60, YEASEN) was added to each well, and the plate was gently shaken to mix. The cells were then incubated at 37\u0026deg;C for an additional 1 hour. Absorbance was measured at 450 nm using a microplate reader to assess cell viability.\u003c/p\u003e\n\u003ch4\u003ePharmacological Inhibition of Proteins\u003c/h4\u003e\n\u003cp\u003eTo investigate the role of specific signaling pathways, cells were treated with a range of pharmacological inhibitors upon reaching 70-80% confluence. Each inhibitor was dissolved in anhydrous DMSO and diluted to the appropriate working concentration. The following inhibitors were used: C29 (10 \u0026micro;M, S6597, Selleck) for TLR2 inhibition, Procyanidin B1 (30 \u0026micro;M, HY-N0795, MedChemExpress) for TLR4 inhibition, MyD88-IN-1 (30 \u0026micro;M, HY-149992, MedChemExpress) for MyD88 inhibition, Pepinh-TRIF TFA (30 \u0026micro;M, HY-P2565, MedChemExpress) for TRIF inhibition, GSK8612 (5 \u0026micro;M, T5540, TargetMol) for TBK1/IKK\u0026epsilon; inhibition, MNS (10 \u0026micro;M, HY-78263, MedChemExpress) for SYK inhibition, and Fludarabine (10 \u0026micro;M, HY-B0069, MedChemExpress) for STAT1 inhibition. Following treatment, cells were incubated at 37\u0026deg;C for 3 hours. Control groups were treated with DMSO at a final concentration of less than 0.1%. This treatment scheme enabled the assessment of the contribution of each specific pathway to the cellular responses observed.\u003c/p\u003e\n\u003ch4\u003eCloning and Construction of Expression Plasmids\u003c/h4\u003e\n\u003cp\u003eMouse \u003cem\u003eTlr4\u003c/em\u003e, \u003cem\u003eCd14\u003c/em\u003e, and \u003cem\u003eMd2\u003c/em\u003e cDNAs were cloned into the pcDNA3.1-HA expression vector using a seamless cloning and assembly kit (Transgen, CU101-01). All expression constructs were generated using standard molecular biology techniques, and the coding sequences were fully verified by sequencing. Site-directed mutants were created using standard cloning procedures, with each mutant confirmed through sequencing. The Ifnb luciferase reporter and NF-\u0026kappa;B luciferase reporter plasmids were generously provided by Professor Zhengfan Jiang from Peking University. The pCMV-VSVG and psPAX2 plasmids were obtained from Addgene.\u003c/p\u003e\n\u003ch4\u003eGene Silencing of Stat1 and Syk using Short Hairpin RNA (shRNA)\u003c/h4\u003e\n\u003cp\u003eGene silencing of \u003cem\u003eStat1\u003c/em\u003e and \u003cem\u003eSyk\u003c/em\u003e was achieved using the pLKO.1 plasmid vector, which contains EcoRI and AgeI restriction enzyme cutting sites. The shRNA sequences designed to knock down \u003cem\u003eStat1\u003c/em\u003e and \u003cem\u003eSyk\u003c/em\u003e were sourced from the Sigma-Aldrich online database (https://www.sigmaaldrich.cn/CN/zh/product/sigma/shrna), converted by the online tools MOI (http://www.fynn-guo.cn/seq_tool.php)\u003csup\u003e111\u003c/sup\u003eand are listed in Supplementary Table S2. Initially, three pairs of shRNAs targeting \u003cem\u003eStat1\u003c/em\u003e and \u003cem\u003eSyk\u003c/em\u003e were designed, and following validation, the pairs with the highest knockdown efficiency were selected for further experiments.\u003c/p\u003e\n\u003cp\u003eThe final shRNA sequences for \u003cem\u003eStat1\u003c/em\u003e-shRNA-pair2 were:\u003c/p\u003e\n\u003cul\u003e\n \u003cli\u003e\u003cstrong\u003eSense strand\u003c/strong\u003e: CCGGCCCTGAAGTATCTGTATCCAACTCGAGTTGGATACAGATACTTCAGGGTTTTTG\u003c/li\u003e\n \u003cli\u003e\u003cstrong\u003eAntisense strand\u003c/strong\u003e: AATTCAAAAACCCTGAAGTATCTGTATCCAACTCGAGTTGGATACAGATACTTCAGGG\u003c/li\u003e\n\u003c/ul\u003e\n\u003cp\u003eThe final shRNA sequences for \u003cem\u003eSyk\u003c/em\u003e-shRNA-pair2 were:\u003c/p\u003e\n\u003cul\u003e\n \u003cli\u003e\u003cstrong\u003eSense strand\u003c/strong\u003e: CCGGGCAGGCCATCATCAGTCAGAACTCGAGTTCTGACTGATGATGGCCTGCTTTTTG\u003c/li\u003e\n \u003cli\u003e\u003cstrong\u003eAntisense strand\u003c/strong\u003e: AATTCAAAAAGCAGGCCATCATCAGTCAGAACTCGAGTTCTGACTGATGATGGCCTGC\u003c/li\u003e\n\u003c/ul\u003e\n\u003cp\u003eA nontargeting shRNA (scramble shRNA) from Sigma-Aldrich was used as a negative control. Transfections were performed at a final concentration of 10 nM using Lipofectamine\u0026reg; RNAiMAX Transfection Reagent (catalog number 13778030; Thermo Fisher Scientific), following the manufacturer\u0026rsquo;s instructions. After transfection, cells were split and selected with 5 \u0026micro;g/mL puromycin for two weeks. Stable clones were then isolated by limiting dilution and expanded for further analysis.\u003c/p\u003e\n\u003ch4\u003eCRISPR-Cas9 System for Syk Knockout\u003c/h4\u003e\n\u003cp\u003e\u003cem\u003eSyk\u003c/em\u003e knockout (KO) RAW264.7 cells were generated using the CRISPR-Cas9 system. Guide RNAs (gRNAs) with high efficiency and specificity were designed using the SYNTHEGO online CRISPR design tool (https://www.synthego.com/products/crispr-kits). The oligos were annealed and cloned into the lentiCRISPR v2 vector, which had been digested with the BsmBI enzyme (NEB). To generate lentivirus, 293T cells were transfected with the following plasmids: lentiCRISPR v2 (2400 ng), packaging plasmid psPAX2 (800 ng; Addgene 12260), envelope plasmid VSV-G (800 ng; Addgene 8454), and PEI (1600 ng). The transfection mixture was incubated at 37\u0026deg;C for 72 hours, after which viral supernatants were collected and used to infect RAW264.7 cells in the presence of polybrene (Beyotime Biotechnology, China). Forty-eight hours post-infection, cells were refreshed with fresh culture medium and selected with 10 \u0026micro;g/mL puromycin. Three pairs of sgRNAs targeting \u003cem\u003eSyk\u003c/em\u003e were initially designed (see Supplementary Table S2), and after validation, the pairs with the highest knockout efficiency were selected for further experiments. The successful knockout of \u003cem\u003eSyk\u003c/em\u003e was validated by Western blotting. The primer sequences used for the knockout validation are as follows:\u003c/p\u003e\n\u003cul\u003e\n \u003cli\u003e\u003cstrong\u003eSense strand\u003c/strong\u003e: CACCGUGAAGGGGUGCAGACAUGGC\u003c/li\u003e\n \u003cli\u003e\u003cstrong\u003eAntisense strand\u003c/strong\u003e: AAACGCCATGTCTGCACCCCTTCAC\u003c/li\u003e\n\u003c/ul\u003e\n\u003ch4\u003eImmunoprecipitation (IP) Assay\u003c/h4\u003e\n\u003cp\u003eImmunoprecipitation was performed using the ProteinIso\u0026reg; Protein A/G Resin (DP501, TransGen Biotech) according to the manufacturer\u0026rsquo;s instructions. Cells were lysed in NP-40 lysis buffer (HY-K1002, MedChemExpress), supplemented with protease inhibitors. The lysates were clarified by centrifugation at 12,000 rpm for 10 minutes at 4\u0026deg;C. Equal amounts of protein lysates were incubated with specific primary antibodies overnight at 4\u0026deg;C. Following incubation, ProteinIso\u0026reg; Protein A/G Resin was added and rotated gently for 4 hours at 4\u0026deg;C. The resin was washed three times with NP-40 lysis buffer to remove nonspecific binding. Immune complexes were eluted by heating the resin in SDS-PAGE loading buffer at 108\u0026deg;C for 5 minutes, and the resulting samples were analyzed by Western blotting.\u003c/p\u003e\n\u003ch4\u003eMicroscale Thermophoresis (MST) Assay\u003c/h4\u003e\n\u003cp\u003eThe interaction between TLR4 and Co68 was assessed using Microscale Thermophoresis (MST) at the State Key Laboratory, School of Pharmaceutical Sciences, Peking University Health Science Center. Recombinant human TLR4 protein was labeled with a fluorescent dye, and Co68 was prepared in a series of dilutions. The labeled TLR4 protein and Co68 were incubated together, and the mixture was loaded into capillaries for measurement. MST data were collected at 25\u0026deg;C, and binding affinity (K\u003cem\u003ed\u003c/em\u003e) was calculated using MO.Affinity Analysis software. All data analysis was independently conducted by our research team.\u003c/p\u003e\n\u003ch4\u003eSurface Plasmon Resonance (SPR) Assay\u003c/h4\u003e\n\u003cp\u003eSPR experiments were conducted using the S-Class label-free molecular interaction analysis system, in collaboration with Polariton Life Sciences, to evaluate the interaction between TLR4 protein and small molecules (Co65, Co66, Co67, Co68, Co69, Co70, Co71, Co72, Co73, and Co74). TLR4 protein was immobilized onto a C5 sensor chip via amine coupling, using a protein stock solution prepared at 10 \u0026micro;g/mL in 1X PBST (pH 4.5), with an immobilization time of 600 seconds. Small molecules were dissolved in 100% DMSO at 10 mM, then diluted in 1X PBST containing 5% DMSO before injection. Each analyte was flowed over the immobilized TLR4 at a rate of 30 \u0026micro;L/min, with association and dissociation phases set to 60 seconds and 120 seconds, respectively. Binding data were collected and analyzed to determine the kinetic parameters and affinities (K\u003cem\u003ed\u003c/em\u003e) of the interactions.\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eStructure Prediction with AlphaFold\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eThe binding interactions of the TLR4-MD2 complex with the small molecule Co68 and the ion Co\u003csup\u003e2+\u003c/sup\u003e were predicted using AlphaFold3\u003csup\u003e112\u003c/sup\u003e. The amino acid sequences of TLR4 and MD2 were retrieved from the UniProt database and used to predict the structure of the TLR4-MD2 complex. The predicted structure was evaluated for accuracy based on the confidence score (pLDDT) and aligned with known crystal structures for validation. For molecular docking, the TLR4-MD2 complex structure predicted by AlphaFold was used as the receptor. The structure of Co68 was obtained from the PubChem database, and its geometry was optimized prior to docking. Co\u003csup\u003e2+\u003c/sup\u003e binding sites were analyzed separately by including the ion in docking simulations. Molecular docking was performed using AutoDock Vina (V1.2.5)\u003csup\u003e113\u003c/sup\u003e, and binding interactions were visualized and analyzed using PyMOL (V2.1) and Chimera (V1.18)\u003csup\u003e114\u003c/sup\u003e. Key residues involved in the binding of Co68 and Co\u003csup\u003e2+\u003c/sup\u003e were identified and compared to evaluate their interaction patterns with the TLR4-MD2 complex.\u003c/p\u003e\n\u003ch4\u003eRNA-seq and data analysis\u003c/h4\u003e\n\u003cp\u003eTotal RNA was extracted using the high-throughput RNA extraction kit (TIANGEN, A0123A01). Quality control, library preparation, and sequencing were performed by Suzhou GENEWIZ Biotechnology company (https://www.genewiz.com.cn/), following established standard protocols. Data analysis was carried out in-house. Sequencing raw data in FASTQ format underwent quality control using FastQC (v0.11.9) and Trim-Galore (v0.6.4) software to generate clean data for downstream analysis. Trim-Galore was executed with the following parameters: trim_galore --gzip --trim-n --phred33 -j 7 --paired ${var}_1.fq.gz ${var}_2.fq.gz -o $wrk_dir/clean_result/. Clean reads were aligned to the mm10 reference genome using Subread (v2.0.0) with the following parameters\u003csup\u003e115\u003c/sup\u003e: subread-align -i $idx_dir -r $cle_dir/${var}_1_val_1.fq.gz -R $cle_dir/${var}_2_val_2.fq.gz -o $aln_dir/${var}.bam -T 30 -t 0. The gene count matrix was generated using featureCounts (v2.0.0) with the following parameters\u003csup\u003e116\u003c/sup\u003e: featureCounts -p -t exon -g gene_id -a $gtf_dir -o $cnt_dir/count_refGene $aln_dir/*.bam -T 29. The count data was normalized using the Fragments Per Kilobase Million (FPKM) formula. Differential expression analysis was conducted using the DESeq2 R package (v1.38.3), with the thresholds set at |log2(FoldChange)| \u0026gt; 2 and \u003cem\u003ep\u003c/em\u003e-value \u0026lt; 0.05 to identify significant differentially expressed genes (DEGs).\u003c/p\u003e\n\u003ch4\u003eGO annotation and KEGG pathway enrichment analysis\u003c/h4\u003e\n\u003cp\u003eIn this study, DEGs were subjected to enrichment analysis through Gene Ontology (GO) annotation and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis\u003csup\u003e117,118\u003c/sup\u003e. This analysis was performed using the clusterProfiler R package, specifically employing the enrichGO and enrichKEGG functions\u003csup\u003e118\u003c/sup\u003e. Alternatively, the online database DAVID (https://david.ncifcrf.gov/summary.jsp) (V6.8) was utilized for the same purpose\u003csup\u003e119\u003c/sup\u003e. Significance thresholds were established, with GO and KEGG terms having false discovery rates (FDR) less than 0.01 considered indicative of meaningful enrichments.\u003c/p\u003e\n\u003ch4\u003eGene set enrichment analysis (GSEA)\u003c/h4\u003e\n\u003cp\u003eGene Set Enrichment Analysis (GSEA) is a computational approach used to determine whether predefined gene sets exhibit statistically significant and coordinated differences between two biological states. Unlike traditional methods that focus solely on differentially expressed genes (DEGs), GSEA incorporates all genes in the analysis, regardless of their significance levels. In this study, GSEA was conducted using the clusterProfiler package, and the function gseaplot2 from the same package was employed to visualize the enrichment results\u003csup\u003e118\u003c/sup\u003e.\u003c/p\u003e\n\u003ch4\u003eProtein-protein interaction (PPI) network analysis\u003c/h4\u003e\n\u003cp\u003eProtein-Protein Interaction (PPI) network analysis is a computational approach used to investigate and map the interactions between proteins within a biological system. For this analysis, we utilized the online database STRING (https://string-db.org/) with default parameters\u003csup\u003e120\u003c/sup\u003e. The input consisted of differentially expressed genes (DEGs), and the output was a network representing the interactions between the proteins encoded by these DEGs. Visualization of the resulting PPI network was carried out using Cytoscape software (v3.10.2)\u003csup\u003e121\u003c/sup\u003e.\u003c/p\u003e\n\u003ch4\u003eTranscription factor enrichment analysis (TFEA)\u003c/h4\u003e\n\u003cp\u003eTranscription Factor Enrichment Analysis (TFEA) is a computational method used to identify transcription factors (TFs) that may regulate differentially expressed genes (DEGs). For this analysis, we utilized iRegulon (http://iregulon.aertslab.org/tutorial.html), a Cytoscape plug-in specifically designed for TFEA\u003csup\u003e122\u003c/sup\u003e. The input for iRegulon is a list of DEGs, and the output consists of a ranked list of predicted TFs based on enrichment scores. The results of TFEA can be directly visualized and further explored using Cytoscape\u0026apos;s integrated visualization tools (v3.10.2).\u003c/p\u003e\n\u003ch4\u003eATAC-seq\u003c/h4\u003e\n\u003cp\u003eWe employed a high-throughput sequencing methodology based on transposase-mediated chromatin accessibility profiling (ATAC-seq)\u003csup\u003e123\u003c/sup\u003e. This technique utilizes the specificity of Tn5 transposase to selectively cleave accessible regions of chromatin. The transposase, preloaded with DNA sequence adapters, was incubated with isolated cell nuclei, enabling the targeted insertion of adapters into open chromatin regions. Subsequently, indexed primers were utilized for PCR amplification to construct sequencing libraries. The resultant libraries, following sequencing, provided comprehensive insights into DNA regions associated with chromatin accessibility. This experiment was performed using the Hyperactive ATAC-Seq Library Prep Kit (Vazyme Biotech, TD711), with sequencing services conducted by GENEWIZ Biotechnology company. (Suzhou, China).\u003c/p\u003e\n\u003ch4\u003eData analysis of ATAC-seq\u003c/h4\u003e\n\u003cp\u003eFor ATAC-seq high-throughput data, sequenced by GENEWIZ Biotechnology, raw fastq data underwent quality control using Trim-Galore (v0.6.4). Filtered reads were then aligned and quantified against the mouse genome mm10 using the bowtie2 aligner (v2.3.5.1)\u003csup\u003e124\u003c/sup\u003e, generating SAM or BAM files. The bowtie2 alignment parameters were: bowtie2 --very-sensitive -X 2000 -x $Bowtie2Index -1 $cln_res/${sample}_1_val_1.fq.gz -2 $cln_res/${sample}_2_val_2.fq.gz -p $PPN | samtools view -buSh -@ $PPN | samtools sort -@ $PPN -O BAM -o $aln_res/${sample}.sorted.bam. Sorting and indexing of SAM/BAM files were completed using samtools (v1.10) with the command\u003csup\u003e125\u003c/sup\u003e: samtools index -@ $PPN $aln_res/${sample}.sorted.bam. Peak calling was performed using MACS3 (v3.0.0a5) with default parameters\u003csup\u003e126\u003c/sup\u003e, yielding peak files in BED format for each sample. For downstream analysis, different strategies were applied to ATAC-seq data. ATAC-seq peaks were merged across all samples using the bedtools merge (v2.31.1)\u003csup\u003e127\u003c/sup\u003e function, followed by normalization with bamCoverage (v3.3.2) in the deepTools suite\u003csup\u003e128\u003c/sup\u003e using the command: bamCoverage --bam $var -o ${var%.*}.bw --binSize 100 --normalizeUsing RPKM --effectiveGenomeSize 2864785220 --ignoreForNormalization chrM \u0026ndash;extendReads. Differential analysis for both ATAC-seq was conducted using csaw R package (v1.38.0)\u003csup\u003e129\u003c/sup\u003e, while peak annotation was performed using the ChIPseeker\u0026nbsp;R package (v1.34.1)\u003csup\u003e130\u003c/sup\u003e. Finally, peak visualization was achieved using Integrative Genomics Viewer (IGV) (v2.17.4)\u003csup\u003e131\u003c/sup\u003e.\u003c/p\u003e\n\u003ch4\u003eGeneration and Processing of Single-Cell RNA Sequencing (scRNA-seq) Data\u003c/h4\u003e\n\u003cp\u003eTumor tissues were harvested from mice following subcutaneous implantation and processed into single-cell suspensions. Single-cell suspensions, library construction, and sequencing were performed by Genewiz, Azenta Life Sciences, adhering to the 10\u0026times; Genomics Chromium Next GEM Single Cell 3\u0026prime; Reagent Kits v3.1 protocol. Sequencing was conducted on the Illumina NovaSeq 6000 platform in paired-end 150 bp (PE150) mode. Genewiz provided the binary FASTQ files for subsequent analysis. Raw FASTQ files were processed using the Cell Ranger pipeline (v3.1.0) for barcode demultiplexing, alignment, and unique molecular identifier (UMI) counting, generating feature-barcode matrices. Further analysis was performed in R using Seurat (v4.3.0)\u003csup\u003e132\u003c/sup\u003e. The data were normalized, reduced in dimensionality, and clustered for cellular classification. Cell types were manually annotated based on canonical marker gene expression, and the results were exported for downstream analyses. Cell-cell communication was analyzed using the CellChat package (v1.1.3)\u003csup\u003e133\u003c/sup\u003e. Gene set enrichment analysis (GSEA) and gene set variation analysis (GSVA) were conducted using the clusterProfiler package (v4.6.2)\u003csup\u003e118\u003c/sup\u003e. Pseudotime trajectory analysis was performed with Monocle3 (v1.0.0)\u003csup\u003e134\u003c/sup\u003e. Marker gene expression patterns were visualized using dot plots, violin plots, and heatmaps.\u003c/p\u003e\n\u003ch4\u003ePreprocessing and Encoding of Drug-Gene Expression Profiles Data\u003c/h4\u003e\n\u003cp\u003eThis study utilized a comprehensive dataset of gene expression profiles for small molecules, sourced from the Library of Integrated Network-Based Cellular Signatures (LINCS) project (https://clue.io/lincs)\u003csup\u003e135\u003c/sup\u003e, DrugBank\u003csup\u003e136\u003c/sup\u003e, and ChEMBL\u003csup\u003e137\u003c/sup\u003e, including chemical structure and bioactivity data. Strict data cleaning criteria were applied: molecules with fewer than seven replicates were excluded, and molecular SMILES were parsed using RDKit (v2024.03.5). The gene expression profiles for each molecule were averaged, disregarding variations in plate, dose, treatment time, and cell line. The focus was on landmark genes associated with NF-\u0026kappa;B and IRF3, which are induced by lipopolysaccharide (LPS). This resulted in a final dataset comprising 7,861 valid molecules, which was split into training (5,502) and test (2,359) sets. The chemical structures of small molecules were represented as SMILES strings, which were converted into grammar trees and then transformed into one-hot encoded arrays. These molecular features were processed into fixed-length vectors for input into a convolutional neural network (CNN). Dimensionality reduction was performed using a variational autoencoder (VAE), producing the input matrix (X). Simultaneously, the gene expression effects of each molecule were encoded into one-hot arrays, with \u0026lsquo;1\u0026rsquo; indicating gene upregulation and \u0026lsquo;0\u0026rsquo; indicating downregulation, creating the output matrix (Y).\u003c/p\u003e\n\u003ch4\u003eConstruction of the DLINP Model\u003c/h4\u003e\n\u003cp\u003eThe CNN-based DLINP (Deep Learning Inflammation and Immunity Prediction) model was designed to predict the ability of compounds to inhibit inflammatory responses and enhance innate immune responses. The model was implemented using TensorFlow (https://www.tensorflow.org/) and Keras (https://keras.io/) to predict small molecule-induced gene expression profiles. The architecture of the model was designed to capture interactions between molecular features and their corresponding gene expression patterns. Each small molecule was represented as a feature vector of dimensions (4717, 6042, 1). The model architecture began with a Conv1D layer containing 32 filters with a kernel size of 3, applying the ReLU activation function to extract relevant molecular features. This was followed by a MaxPooling1D layer to reduce dimensionality while preserving critical information. A second Conv1D layer with 64 filters was added, followed by another MaxPooling1D layer to refine the feature extraction process. The output from these convolutional layers was flattened into a one-dimensional vector and passed through a Dense layer containing 128 neurons with ReLU activation. To prevent overfitting, a dropout rate of 0.5 was applied. The final Dense output layer, using a sigmoid activation function, predicted the expression status (upregulated or downregulated) of landmark genes. The model was compiled with the Adam optimizer and binary cross-entropy loss for binary classification. The dataset was split into training, validation, and testing sets in a 70:30 ratio. The training was conducted over 1,000 epochs with a batch size of 32, with 10% of the training data reserved for validation to monitor performance during training. After model training, compounds from the CAS database (https://www.cas.org/) were screened using DLINP.\u003c/p\u003e\n\u003ch4\u003eEvaluation of the DLINP Model Performance\u003c/h4\u003e\n\u003cp\u003eThe performance of the DLINP model was evaluated using several key metrics, including the confusion matrix, precision-recall (PR) curve, receiver operating characteristic (ROC) curve, and the area under the ROC curve (AUC). These metrics offered a comprehensive assessment of the model\u0026apos;s ability to predict gene expression profiles induced by small molecules, providing insights into its precision, recall, overall classification accuracy, and discriminatory power. The confusion matrix revealed the distribution of true positive, true negative, false positive, and false negative predictions, providing detailed information on the model\u0026rsquo;s performance. The PR and ROC curves, along with the AUC, illustrated the trade-offs between sensitivity and specificity across various thresholds, further aiding in the evaluation of the model\u0026rsquo;s predictive capabilities.\u003c/p\u003e"},{"header":"Declarations","content":"\u003ch4\u003eData Availability\u003c/h4\u003e\n\u003cp\u003eThe RNA-seq, ATAC-seq, and scRNA-seq datasets generated in this study are available in the Gene Expression Omnibus (GEO) database under the following accession numbers: GSE288554, GSE288553, and GSE288555, respectively.\u003c/p\u003e\n\u003ch4\u003eCode availability\u003c/h4\u003e\n\u003cp\u003eThe R and Python code used to generate the figures in this study will be made available upon reasonable request. Interested parties are encouraged to contact the corresponding author for access to the code.\u003c/p\u003e\n\u003ch4\u003eStatistical analysis\u003c/h4\u003e\n\u003cp\u003eAll statistical analyses in this study were conducted using the Python package SciPy (https://pypi.org/project/scipy/). Results are presented as the mean \u0026plusmn; standard error of the mean (SEM). The \u003cem\u003ep\u003c/em\u003e values \u0026lt; 0.05 were considered statistically significant (*), \u003cem\u003ep\u003c/em\u003e values \u0026lt; 0.01, and \u003cem\u003ep\u003c/em\u003e values \u0026lt; 0.001 were regarded as highly statistically significant (** and ***).\u003c/p\u003e\n\u003ch2\u003eAcknowledgments\u003c/h2\u003e\n\u003cp\u003eWe thank the support from the Peking University High-performance Computing Platform for providing the computing clusters to facilitate the data analysis procedure in this research.\u003c/p\u003e\n\u003ch3\u003eAuthor contributions\u003c/h3\u003e\n\u003cp\u003eConceptualization, X.G.\u003cbr\u003e\u0026nbsp;Methodology, X.G.\u003cbr\u003e\u0026nbsp;Software, X.G.\u003cbr\u003e\u0026nbsp;Validation, X.G. and Y.Z. and S.H. and X.C.\u003cbr\u003e\u0026nbsp;Formal Analysis, X.G., and Y.Z. and F.Y.\u003cbr\u003e\u0026nbsp;Investigation, X.G. and Y.Z. and and Y.S.Z. and T.C.\u003cbr\u003e\u0026nbsp;Resources, F.Y. and Q.L. and X.R. and L.T. and X.W and X.W.\u003cbr\u003e\u0026nbsp;Data Curation, X.G.\u003cbr\u003e\u0026nbsp;Writing \u0026ndash; Original Draft, X.G.\u003cbr\u003e\u0026nbsp;Writing \u0026ndash; Review \u0026amp; Editing, F.Y. and Y.Z.\u003cbr\u003e\u0026nbsp;Visualization, X.G.\u003cbr\u003e\u0026nbsp;Supervision, X.G, and F.Y. and Q.L.\u003cbr\u003e\u0026nbsp;Project Administration, Y.Z. and T.C.\u003cbr\u003e\u0026nbsp;Funding Acquisition, F.Y.\u003c/p\u003e\n\u003ch3\u003eCompeting interests\u003c/h3\u003e\n\u003cp\u003eThe authors declare no competing interests.\u003c/p\u003e\n\u003ch3\u003eFunding\u003c/h3\u003e\n\u003cp\u003eThis work was supported by the Beijing Natural Science Foundation (Z210014), the National Key Research and Development Program of China (2021YFC2302602, 2020YFA0707800), the National Natural Science Foundation of China (31570891, 31872736, 32022028, 81991505, and 82201928), Peking University Clinical + X (PKU2020LCXQ009), the Peking University Medicine Fund (PKU2020LCXQ009), the Zhuhai Science and Technology Innovation Bureau (ZH22036302200063PWC to Z.Y.), the China Postdoctoral Science Foundation (2022M710265 to H.Y.), a grant from the Tianjin Natural Science Foundation of China (no. 21JCQNJC01870 to D.W.), and the Incubation Fund of Tianjin Third Central Hospital (no. 2019YNR6 to D.W.).\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eRyan, D. P., Hong, T. S. \u0026amp; Bardeesy, N. Pancreatic Adenocarcinoma. \u003cem\u003eNew England Journal of Medicine\u003c/em\u003e \u003cstrong\u003e371\u003c/strong\u003e, 1039\u0026ndash;1049 (2014).\u003c/li\u003e\n\u003cli\u003eHidalgo, M. Pancreatic Cancer. \u003cem\u003eNew England Journal of Medicine\u003c/em\u003e \u003cstrong\u003e362\u003c/strong\u003e, 1605\u0026ndash;1617 (2010).\u003c/li\u003e\n\u003cli\u003eLi, ( D \u003cem\u003eet al.\u003c/em\u003e \u003cem\u003ePancreatic Cancer\u003c/em\u003e. \u003cem\u003eTHE LANCET \u0026bull;\u003c/em\u003e vol. 363 www.thelancet.com1049 (2004).\u003c/li\u003e\n\u003cli\u003eWang, S. \u003cem\u003eet al.\u003c/em\u003e The molecular biology of pancreatic adenocarcinoma: translational challenges and clinical perspectives. \u003cem\u003eSignal Transduction and Targeted Therapy\u003c/em\u003e vol. 6 Preprint at https://doi.org/10.1038/s41392-021-00659-4 (2021).\u003c/li\u003e\n\u003cli\u003eGrossberg, A. J. \u003cem\u003eet al.\u003c/em\u003e Multidisciplinary standards of care and recent progress in pancreatic ductal adenocarcinoma. \u003cem\u003eCA Cancer J Clin\u003c/em\u003e \u003cstrong\u003e70\u003c/strong\u003e, 375\u0026ndash;403 (2020).\u003c/li\u003e\n\u003cli\u003eMcGuigan, A. \u003cem\u003eet al.\u003c/em\u003e Pancreatic cancer: A review of clinical diagnosis, epidemiology, treatment and outcomes. \u003cem\u003eWorld Journal of Gastroenterology\u003c/em\u003e vol. 24 4846\u0026ndash;4861 Preprint at https://doi.org/10.3748/wjg.v24.i43.4846 (2018).\u003c/li\u003e\n\u003cli\u003eHalbrook, C. J., Lyssiotis, C. A., Pasca di Magliano, M. \u0026amp; Maitra, A. Pancreatic cancer: Advances and challenges. \u003cem\u003eCell\u003c/em\u003e vol. 186 1729\u0026ndash;1754 Preprint at https://doi.org/10.1016/j.cell.2023.02.014 (2023).\u003c/li\u003e\n\u003cli\u003ePark, W., Chawla, A. \u0026amp; O\u0026rsquo;Reilly, E. M. Pancreatic Cancer: A Review. \u003cem\u003eJAMA - Journal of the American Medical Association\u003c/em\u003e vol. 326 851\u0026ndash;862 Preprint at https://doi.org/10.1001/jama.2021.13027 (2021).\u003c/li\u003e\n\u003cli\u003eRahib, L. \u003cem\u003eet al.\u003c/em\u003e Projecting cancer incidence and deaths to 2030: The unexpected burden of thyroid, liver, and pancreas cancers in the united states. \u003cem\u003eCancer Research\u003c/em\u003e vol. 74 2913\u0026ndash;2921 Preprint at https://doi.org/10.1158/0008-5472.CAN-14-0155 (2014).\u003c/li\u003e\n\u003cli\u003eWood, L. D., Canto, M. I., Jaffee, E. M. \u0026amp; Simeone, D. M. Pancreatic Cancer: Pathogenesis, Screening, Diagnosis, and Treatment. \u003cem\u003eGastroenterology\u003c/em\u003e vol. 163 386-402.e1 Preprint at https://doi.org/10.1053/j.gastro.2022.03.056 (2022).\u003c/li\u003e\n\u003cli\u003eZhan, H. X. \u003cem\u003eet al.\u003c/em\u003e Neoadjuvant therapy in pancreatic cancer: a systematic review and meta-analysis of prospective studies. \u003cem\u003eCancer Med\u003c/em\u003e \u003cstrong\u003e6\u003c/strong\u003e, (2017).\u003c/li\u003e\n\u003cli\u003eSpringfeld, C. \u003cem\u003eet al.\u003c/em\u003e Neoadjuvant therapy for pancreatic cancer. \u003cem\u003eNature Reviews Clinical Oncology\u003c/em\u003e vol. 20 318\u0026ndash;337 Preprint at https://doi.org/10.1038/s41571-023-00746-1 (2023).\u003c/li\u003e\n\u003cli\u003eFalcomat\u0026agrave;, C. \u003cem\u003eet al.\u003c/em\u003e Context-Specific Determinants of the Immunosuppressive Tumor Microenvironment in Pancreatic Cancer. \u003cem\u003eCancer Discovery\u003c/em\u003e vol. 13 278\u0026ndash;297 Preprint at https://doi.org/10.1158/2159-8290.CD-22-0876 (2023).\u003c/li\u003e\n\u003cli\u003eHo, W. J., Jaffee, E. M. \u0026amp; Zheng, L. The tumour microenvironment in pancreatic cancer \u0026mdash; clinical challenges and opportunities. \u003cem\u003eNature Reviews Clinical Oncology\u003c/em\u003e vol. 17 527\u0026ndash;540 Preprint at https://doi.org/10.1038/s41571-020-0363-5 (2020).\u003c/li\u003e\n\u003cli\u003eZhang, S. \u003cem\u003eet al.\u003c/em\u003e Single cell transcriptomic analyses implicate an immunosuppressive tumor microenvironment in pancreatic cancer liver metastasis. \u003cem\u003eNat Commun\u003c/em\u003e \u003cstrong\u003e14\u003c/strong\u003e, (2023).\u003c/li\u003e\n\u003cli\u003eKzhyshkowska, J., Shen, J. \u0026amp; Larionova, I. Targeting of TAMs: can we be more clever than cancer cells? \u003cem\u003eCellular and Molecular Immunology\u003c/em\u003e Preprint at https://doi.org/10.1038/s41423-024-01232-z (2024).\u003c/li\u003e\n\u003cli\u003eMantovani, A., Marchesi, F., Jaillon, S., Garlanda, C. \u0026amp; Allavena, P. Tumor-associated myeloid cells: diversity and therapeutic targeting. \u003cem\u003eCellular and Molecular Immunology\u003c/em\u003e vol. 18 566\u0026ndash;578 Preprint at https://doi.org/10.1038/s41423-020-00613-4 (2021).\u003c/li\u003e\n\u003cli\u003eZhu, Y. \u003cem\u003eet al.\u003c/em\u003e Tissue-Resident Macrophages in Pancreatic Ductal Adenocarcinoma Originate from Embryonic Hematopoiesis and Promote Tumor Progression. \u003cem\u003eImmunity\u003c/em\u003e \u003cstrong\u003e47\u003c/strong\u003e, 323-338.e6 (2017).\u003c/li\u003e\n\u003cli\u003eKloosterman, D. J. \u0026amp; Akkari, L. Macrophages at the interface of the co-evolving cancer ecosystem. \u003cem\u003eCell\u003c/em\u003e vol. 186 1627\u0026ndash;1651 Preprint at https://doi.org/10.1016/j.cell.2023.02.020 (2023).\u003c/li\u003e\n\u003cli\u003eCaronni, N. \u003cem\u003eet al.\u003c/em\u003e IL-1\u0026beta;+ macrophages fuel pathogenic inflammation in pancreatic cancer. \u003cem\u003eNature\u003c/em\u003e \u003cstrong\u003e623\u003c/strong\u003e, 415\u0026ndash;422 (2023).\u003c/li\u003e\n\u003cli\u003eGarlanda, C. \u0026amp; Mantovani, A. Interleukin-1 in tumor progression, therapy, and prevention. \u003cem\u003eCancer Cell\u003c/em\u003e \u003cstrong\u003e39\u003c/strong\u003e, 1023\u0026ndash;1027 (2021).\u003c/li\u003e\n\u003cli\u003eDas, S., Shapiro, B., Vucic, E. A., Vogt, S. \u0026amp; Bar-Sagi, D. Tumor cell-derived IL1\u0026beta; promotes desmoplasia and immune suppression in pancreatic cancer. \u003cem\u003eCancer Res\u003c/em\u003e \u003cstrong\u003e80\u003c/strong\u003e, 1088\u0026ndash;1101 (2020).\u003c/li\u003e\n\u003cli\u003eReis, M. B. \u003cem\u003eet al.\u003c/em\u003e Interleukin-1 receptor-induced PGE2 production controls acetylcholine-mediated cardiac dysfunction and mortality during scorpion envenomation. \u003cem\u003eNat Commun\u003c/em\u003e \u003cstrong\u003e11\u003c/strong\u003e, (2020).\u003c/li\u003e\n\u003cli\u003eZasłona, Z. \u003cem\u003eet al.\u003c/em\u003e The Induction of Pro\u0026ndash;IL-1\u0026beta; by Lipopolysaccharide Requires Endogenous Prostaglandin E2 Production. \u003cem\u003eThe Journal of Immunology\u003c/em\u003e \u003cstrong\u003e198\u003c/strong\u003e, 3558\u0026ndash;3564 (2017).\u003c/li\u003e\n\u003cli\u003eThumkeo, D. \u003cem\u003eet al.\u003c/em\u003e PGE2-EP2/EP4 signaling elicits immunosuppression by driving the mregDC-Treg axis in inflammatory tumor microenvironment. \u003cem\u003eCell Rep\u003c/em\u003e \u003cstrong\u003e39\u003c/strong\u003e, (2022).\u003c/li\u003e\n\u003cli\u003eLacher, S. B. \u003cem\u003eet al.\u003c/em\u003e PGE2 limits effector expansion of tumour-infiltrating stem-like CD8+ T cells. \u003cem\u003eNature\u003c/em\u003e \u003cstrong\u003e629\u003c/strong\u003e, 417\u0026ndash;425 (2024).\u003c/li\u003e\n\u003cli\u003eMorotti, M. \u003cem\u003eet al.\u003c/em\u003e PGE2 inhibits TIL expansion by disrupting IL-2 signalling and mitochondrial function. \u003cem\u003eNature\u003c/em\u003e \u003cstrong\u003e629\u003c/strong\u003e, 426\u0026ndash;434 (2024).\u003c/li\u003e\n\u003cli\u003eZitvogel, L., Galluzzi, L., Kepp, O., Smyth, M. J. \u0026amp; Kroemer, G. Type I interferons in anticancer immunity. \u003cem\u003eNature Reviews Immunology\u003c/em\u003e vol. 15 405\u0026ndash;414 Preprint at https://doi.org/10.1038/nri3845 (2015).\u003c/li\u003e\n\u003cli\u003eBoukhaled, G. M., Harding, S. \u0026amp; Brooks, D. G. Opposing Roles of Type i Interferons in Cancer Immunity. \u003cem\u003eAnnual Review of Pathology: Mechanisms of Disease\u003c/em\u003e vol. 16 167\u0026ndash;198 Preprint at https://doi.org/10.1146/annurev-pathol-031920-093932 (2021).\u003c/li\u003e\n\u003cli\u003eYu, R., Zhu, B. \u0026amp; Chen, D. Type I interferon-mediated tumor immunity and its role in immunotherapy. \u003cem\u003eCellular and Molecular Life Sciences\u003c/em\u003e vol. 79 Preprint at https://doi.org/10.1007/s00018-022-04219-z (2022).\u003c/li\u003e\n\u003cli\u003eBorden, E. C. Interferons \u0026alpha; and \u0026beta; in cancer: therapeutic opportunities from new insights. \u003cem\u003eNature Reviews Drug Discovery\u003c/em\u003e vol. 18 219\u0026ndash;234 Preprint at https://doi.org/10.1038/s41573-018-0011-2 (2019).\u003c/li\u003e\n\u003cli\u003eCrouse, J., Kalinke, U. \u0026amp; Oxenius, A. Regulation of antiviral T cell responses by type i interferons. \u003cem\u003eNature Reviews Immunology\u003c/em\u003e vol. 15 231\u0026ndash;242 Preprint at https://doi.org/10.1038/nri3806 (2015).\u003c/li\u003e\n\u003cli\u003eLu, C. \u003cem\u003eet al.\u003c/em\u003e Type i interferon suppresses tumor growth through activating the STAT3-granzyme B pathway in tumor-infiltrating cytotoxic T lymphocytes. \u003cem\u003eJ Immunother Cancer\u003c/em\u003e \u003cstrong\u003e7\u003c/strong\u003e, (2019).\u003c/li\u003e\n\u003cli\u003eDuong, E. \u003cem\u003eet al.\u003c/em\u003e Type I interferon activates MHC class I-dressed CD11b+ conventional dendritic cells to promote protective anti-tumor CD8+ T cell immunity. \u003cem\u003eImmunity\u003c/em\u003e \u003cstrong\u003e55\u003c/strong\u003e, 308-323.e9 (2022).\u003c/li\u003e\n\u003cli\u003eCao, X. \u003cem\u003eet al.\u003c/em\u003e Next generation of tumor-activating type I IFN enhances anti-tumor immune responses to overcome therapy resistance. \u003cem\u003eNat Commun\u003c/em\u003e \u003cstrong\u003e12\u003c/strong\u003e, (2021).\u003c/li\u003e\n\u003cli\u003eThyrell, L. \u003cem\u003eet al.\u003c/em\u003e Interferon \u0026alpha;-induced apoptosis in tumor cells is mediated through the phosphoinositide 3-kinase/mammalian target of rapamycin signaling pathway. \u003cem\u003eJournal of Biological Chemistry\u003c/em\u003e \u003cstrong\u003e279\u003c/strong\u003e, 24152\u0026ndash;24162 (2004).\u003c/li\u003e\n\u003cli\u003eYu, R., Zhu, B. \u0026amp; Chen, D. Type I interferon-mediated tumor immunity and its role in immunotherapy. \u003cem\u003eCellular and Molecular Life Sciences\u003c/em\u003e vol. 79 Preprint at https://doi.org/10.1007/s00018-022-04219-z (2022).\u003c/li\u003e\n\u003cli\u003eElewaut, A. \u003cem\u003eet al.\u003c/em\u003e Cancer cells impair monocyte-mediated T cell stimulation to evade immunity. \u003cem\u003eNature\u003c/em\u003e (2024) doi:10.1038/s41586-024-08257-4.\u003c/li\u003e\n\u003cli\u003eHangai, S. \u003cem\u003eet al.\u003c/em\u003e PGE2 induced in and released by dying cells functions as an inhibitory DAMP. \u003cem\u003eProc Natl Acad Sci U S A\u003c/em\u003e \u003cstrong\u003e113\u003c/strong\u003e, 3844\u0026ndash;3849 (2016).\u003c/li\u003e\n\u003cli\u003eFabricius, D. \u003cem\u003eet al.\u003c/em\u003e Prostaglandin E2 Inhibits IFN-\u0026alpha; Secretion and Th1 Costimulation by Human Plasmacytoid Dendritic Cells via E-Prostanoid 2 and E-Prostanoid 4 Receptor Engagement. \u003cem\u003eThe Journal of Immunology\u003c/em\u003e \u003cstrong\u003e184\u003c/strong\u003e, 677\u0026ndash;684 (2010).\u003c/li\u003e\n\u003cli\u003eCheon, H. J., Wang, Y., Wightman, S. M., Jackson, M. W. \u0026amp; Stark, G. R. How cancer cells make and respond to interferon-I. \u003cem\u003eTrends in Cancer\u003c/em\u003e vol. 9 83\u0026ndash;92 Preprint at https://doi.org/10.1016/j.trecan.2022.09.003 (2023).\u003c/li\u003e\n\u003cli\u003eArashiba, K., Miyake, Y. \u0026amp; Nishibayashi, Y. A molybdenum complex bearing PNP-type pincer ligands leads to the catalytic reduction of dinitrogen into ammonia. \u003cem\u003eNat Chem\u003c/em\u003e \u003cstrong\u003e3\u003c/strong\u003e, 120\u0026ndash;125 (2011).\u003c/li\u003e\n\u003cli\u003eMerz, L. S., Ballmann, J. \u0026amp; Gade, L. H. Phosphines and N-Heterocycles Joining Forces: an Emerging Structural Motif in PNP-Pincer Chemistry. \u003cem\u003eEuropean Journal of Inorganic Chemistry\u003c/em\u003e vol. 2020 2023\u0026ndash;2042 Preprint at https://doi.org/10.1002/ejic.202000206 (2020).\u003c/li\u003e\n\u003cli\u003ePandey, D. K., Khaskin, E. \u0026amp; Khusnutdinova, J. R. PNP-Pincer Ligands Armed With Methyls: New Tools To Control Sterics and Non-Innocence. \u003cem\u003eChemCatChem\u003c/em\u003e \u003cstrong\u003e15\u003c/strong\u003e, (2023).\u003c/li\u003e\n\u003cli\u003eBoth, N. F., Spannenberg, A., Junge, K. \u0026amp; Beller, M. Low-Valent Molybdenum PNP Pincer Complexes as Catalysts for the Semihydrogenation of Alkynes. \u003cem\u003eOrganometallics\u003c/em\u003e \u003cstrong\u003e41\u003c/strong\u003e, 1797\u0026ndash;1805 (2022).\u003c/li\u003e\n\u003cli\u003eMatsuda, Y., Nakajima, M. \u0026amp; Nemoto, T. Acridine PNP-Pincer Ligands Enabling Transition Metal-Catalyzed Photoreactions. \u003cem\u003eACS Catal\u003c/em\u003e \u003cstrong\u003e13\u003c/strong\u003e, 10224\u0026ndash;10231 (2023).\u003c/li\u003e\n\u003cli\u003eMitsumoto, T. \u003cem\u003eet al.\u003c/em\u003e Catalytic Activity of Molybdenum Complexes Bearing PNP-Type Pincer Ligand toward Ammonia Formation. \u003cem\u003eAngewandte Chemie - International Edition\u003c/em\u003e \u003cstrong\u003e62\u003c/strong\u003e, (2023).\u003c/li\u003e\n\u003cli\u003eArora, V., Narjinari, H., Nandi, P. G. \u0026amp; Kumar, A. Recent advances in pincer-nickel catalyzed reactions. \u003cem\u003eDalton Transactions\u003c/em\u003e vol. 50 3394\u0026ndash;3428 Preprint at https://doi.org/10.1039/d0dt03593a (2021).\u003c/li\u003e\n\u003cli\u003eAi, W., Zhong, R., Liu, X. \u0026amp; Liu, Q. Hydride Transfer Reactions Catalyzed by Cobalt Complexes. \u003cem\u003eChemical Reviews\u003c/em\u003e vol. 119 2876\u0026ndash;2953 Preprint at https://doi.org/10.1021/acs.chemrev.8b00404 (2019).\u003c/li\u003e\n\u003cli\u003eManar, K. K. \u0026amp; Ren, P. Recent progress on group 10 metal complexes of pincer ligands: From synthesis to activities and catalysis. in \u003cem\u003eAdvances in Organometallic Chemistry\u003c/em\u003e vol. 76 185\u0026ndash;259 (Academic Press Inc., 2021).\u003c/li\u003e\n\u003cli\u003eChakraborty, S., Bhattacharya, P., Dai, H. \u0026amp; Guan, H. Nickel and Iron Pincer Complexes as Catalysts for the Reduction of Carbonyl Compounds. \u003cem\u003eAcc Chem Res\u003c/em\u003e \u003cstrong\u003e48\u003c/strong\u003e, 1995\u0026ndash;2003 (2015).\u003c/li\u003e\n\u003cli\u003eLiu, W., Sahoo, B., Junge, K. \u0026amp; Beller, M. Cobalt Complexes as an Emerging Class of Catalysts for Homogeneous Hydrogenations. \u003cem\u003eAcc Chem Res\u003c/em\u003e \u003cstrong\u003e51\u003c/strong\u003e, 1858\u0026ndash;1869 (2018).\u003c/li\u003e\n\u003cli\u003eKempf, H. A. \u003cem\u003eet al.\u003c/em\u003e Synthesis of Ru‐PNPC Pincer Complexes and Applications in Catalytic Hydrogenation and Dehydrogenation Reactions. \u003cem\u003eChemCatChem\u003c/em\u003e (2024) doi:10.1002/cctc.202401481.\u003c/li\u003e\n\u003cli\u003eBurbidge, R., Trotter, M., Buxton, B. \u0026amp; Holden, S. Drug design by machine learning: Support vector machines for pharmaceutical data analysis. \u003cem\u003eComput Chem\u003c/em\u003e \u003cstrong\u003e26\u003c/strong\u003e, (2001).\u003c/li\u003e\n\u003cli\u003eTang, Y. \u003cem\u003eet al.\u003c/em\u003e Data-driven discovery of innate immunomodulators via machine learning-guided high throughput screening. \u003cem\u003eChem Sci\u003c/em\u003e \u003cstrong\u003e14\u003c/strong\u003e, 12747\u0026ndash;12766 (2023).\u003c/li\u003e\n\u003cli\u003eTanoli, Z., V\u0026auml;h\u0026auml;-Koskela, M. \u0026amp; Aittokallio, T. Artificial intelligence, machine learning, and drug repurposing in cancer. \u003cem\u003eExpert Opinion on Drug Discovery\u003c/em\u003e vol. 16 977\u0026ndash;989 Preprint at https://doi.org/10.1080/17460441.2021.1883585 (2021).\u003c/li\u003e\n\u003cli\u003eVatansever, S. \u003cem\u003eet al.\u003c/em\u003e Artificial intelligence and machine learning-aided drug discovery in central nervous system diseases: State-of-the-arts and future directions. \u003cem\u003eMedicinal Research Reviews\u003c/em\u003e vol. 41 1427\u0026ndash;1473 Preprint at https://doi.org/10.1002/med.21764 (2021).\u003c/li\u003e\n\u003cli\u003eKoromina, M., Pandi, M. T. \u0026amp; Patrinos, G. P. Rethinking Drug Repositioning and Development with Artificial Intelligence, Machine Learning, and Omics. \u003cem\u003eOMICS\u003c/em\u003e \u003cstrong\u003e23\u003c/strong\u003e, (2019).\u003c/li\u003e\n\u003cli\u003eIssa, N. T., Stathias, V., Sch\u0026uuml;rer, S. \u0026amp; Dakshanamurthy, S. Machine and deep learning approaches for cancer drug repurposing. \u003cem\u003eSeminars in Cancer Biology\u003c/em\u003e vol. 68 132\u0026ndash;142 Preprint at https://doi.org/10.1016/j.semcancer.2019.12.011 (2021).\u003c/li\u003e\n\u003cli\u003eMayr, A. \u003cem\u003eet al.\u003c/em\u003e Large-scale comparison of machine learning methods for drug target prediction on ChEMBL. \u003cem\u003eChem Sci\u003c/em\u003e \u003cstrong\u003e9\u003c/strong\u003e, 5441\u0026ndash;5451 (2018).\u003c/li\u003e\n\u003cli\u003eDragan, P., Merski, M., Wiśniewski, S., Sanmukh, S. G. \u0026amp; Latek, D. Chemokine Receptors\u0026mdash;Structure-Based Virtual Screening Assisted by Machine Learning. \u003cem\u003ePharmaceutics\u003c/em\u003e \u003cstrong\u003e15\u003c/strong\u003e, (2023).\u003c/li\u003e\n\u003cli\u003eHie, B., Bryson, B. D. \u0026amp; Berger, B. Leveraging Uncertainty in Machine Learning Accelerates Biological Discovery and Design. \u003cem\u003eCell Syst\u003c/em\u003e \u003cstrong\u003e11\u003c/strong\u003e, 461-477.e9 (2020).\u003c/li\u003e\n\u003cli\u003eVamathevan, J. \u003cem\u003eet al.\u003c/em\u003e Applications of machine learning in drug discovery and development. \u003cem\u003eNature Reviews Drug Discovery\u003c/em\u003e vol. 18 463\u0026ndash;477 Preprint at https://doi.org/10.1038/s41573-019-0024-5 (2019).\u003c/li\u003e\n\u003cli\u003eButler, K. T., Davies, D. W., Cartwright, H., Isayev, O. \u0026amp; Walsh, A. Machine learning for molecular and materials science. \u003cem\u003eNature\u003c/em\u003e vol. 559 547\u0026ndash;555 Preprint at https://doi.org/10.1038/s41586-018-0337-2 (2018).\u003c/li\u003e\n\u003cli\u003eYang, S., Varghese, P., Stephenson, E., Tu, K. \u0026amp; Gronsbell, J. Machine learning approaches for electronic health records phenotyping: a methodical review. \u003cem\u003eJ Am Med Inform Assoc\u003c/em\u003e \u003cstrong\u003e30\u003c/strong\u003e, 367\u0026ndash;381 (2023).\u003c/li\u003e\n\u003cli\u003eIssa, N. T., Stathias, V., Sch\u0026uuml;rer, S. \u0026amp; Dakshanamurthy, S. Machine and deep learning approaches for cancer drug repurposing. \u003cem\u003eSeminars in Cancer Biology\u003c/em\u003e vol. 68 132\u0026ndash;142 Preprint at https://doi.org/10.1016/j.semcancer.2019.12.011 (2021).\u003c/li\u003e\n\u003cli\u003eGrivennikov, S. I., Greten, F. R. \u0026amp; Karin, M. Immunity, Inflammation, and Cancer. \u003cem\u003eCell\u003c/em\u003e vol. 140 883\u0026ndash;899 Preprint at https://doi.org/10.1016/j.cell.2010.01.025 (2010).\u003c/li\u003e\n\u003cli\u003eGuo, X., Zhao, Y. \u0026amp; You, F. Identification and characterization of endogenous retroviruses upon SARS-CoV-2 infection. \u003cem\u003eFront Immunol\u003c/em\u003e \u003cstrong\u003e15\u003c/strong\u003e, (2024).\u003c/li\u003e\n\u003cli\u003eNeufeldt, C. J. \u003cem\u003eet al.\u003c/em\u003e SARS-CoV-2 infection induces a pro-inflammatory cytokine response through cGAS-STING and NF-\u0026kappa;B. \u003cem\u003eCommun Biol\u003c/em\u003e \u003cstrong\u003e5\u003c/strong\u003e, (2022).\u003c/li\u003e\n\u003cli\u003eHadjadj, J. \u003cem\u003eet al.\u003c/em\u003e \u003cem\u003eImpaired Type I Interferon Activity and Inflammatory Responses in Severe COVID-19 Patients\u003c/em\u003e.\u003c/li\u003e\n\u003cli\u003eRaghavan, B., Martin, S. F., Esser, P. R., Goebeler, M. \u0026amp; Schmidt, M. Metal allergens nickel and cobalt facilitate TLR4 homodimerization independently of MD2. \u003cem\u003eEMBO Rep\u003c/em\u003e \u003cstrong\u003e13\u003c/strong\u003e, 1109\u0026ndash;1115 (2012).\u003c/li\u003e\n\u003cli\u003eUllman, N. A., Burchard, P. R., Dunne, R. F. \u0026amp; Linehan, D. C. Immunologic Strategies in Pancreatic Cancer: Making Cold Tumors Hot. \u003cem\u003eJournal of Clinical Oncology\u003c/em\u003e vol. 40 Preprint at https://doi.org/10.1200/JCO.21.02616 (2022).\u003c/li\u003e\n\u003cli\u003eChamma, H., Vila, I. K., Taffoni, C., Turtoi, A. \u0026amp; Laguette, N. Activation of STING in the pancreatic tumor microenvironment: A novel therapeutic opportunity. \u003cem\u003eCancer Letters\u003c/em\u003e vol. 538 Preprint at https://doi.org/10.1016/j.canlet.2022.215694 (2022).\u003c/li\u003e\n\u003cli\u003eCheng, S. \u003cem\u003eet al.\u003c/em\u003e A pan-cancer single-cell transcriptional atlas of tumor infiltrating myeloid cells. \u003cem\u003eCell\u003c/em\u003e \u003cstrong\u003e184\u003c/strong\u003e, 792-809.e23 (2021).\u003c/li\u003e\n\u003cli\u003eBill, R. \u003cem\u003eet al.\u003c/em\u003e CXCL9:SPP1 macrophage polarity identifies a network of cellular programs that control human cancers. \u003cem\u003eScience (1979)\u003c/em\u003e \u003cstrong\u003e381\u003c/strong\u003e, (2023).\u003c/li\u003e\n\u003cli\u003eGungabeesoon, J. \u003cem\u003eet al.\u003c/em\u003e A neutrophil response linked to tumor control in immunotherapy. \u003cem\u003eCell\u003c/em\u003e \u003cstrong\u003e186\u003c/strong\u003e, 1448-1464.e20 (2023).\u003c/li\u003e\n\u003cli\u003eHirschhorn, D. \u003cem\u003eet al.\u003c/em\u003e T cell immunotherapies engage neutrophils to eliminate tumor antigen escape variants. \u003cem\u003eCell\u003c/em\u003e \u003cstrong\u003e186\u003c/strong\u003e, 1432-1447.e17 (2023).\u003c/li\u003e\n\u003cli\u003eBenguigui, M. \u003cem\u003eet al.\u003c/em\u003e Interferon-stimulated neutrophils as a predictor of immunotherapy response. \u003cem\u003eCancer Cell\u003c/em\u003e \u003cstrong\u003e42\u003c/strong\u003e, 253-265.e12 (2024).\u003c/li\u003e\n\u003cli\u003ePark, B. S. \u003cem\u003eet al.\u003c/em\u003e The structural basis of lipopolysaccharide recognition by the TLR4-MD-2 complex. \u003cem\u003eNature\u003c/em\u003e \u003cstrong\u003e458\u003c/strong\u003e, 1191\u0026ndash;1195 (2009).\u003c/li\u003e\n\u003cli\u003eTakeuchi, O. \u0026amp; Akira, S. Pattern Recognition Receptors and Inflammation. \u003cem\u003eCell\u003c/em\u003e vol. 140 805\u0026ndash;820 Preprint at https://doi.org/10.1016/j.cell.2010.01.022 (2010).\u003c/li\u003e\n\u003cli\u003eGuo, Q. \u003cem\u003eet al.\u003c/em\u003e Induction of alarmin S100A8/A9 mediates activation of aberrant neutrophils in the pathogenesis of COVID-19. \u003cem\u003eCell Host Microbe\u003c/em\u003e \u003cstrong\u003e29\u003c/strong\u003e, 222-235.e4 (2021).\u003c/li\u003e\n\u003cli\u003eZhao, Y. \u003cem\u003eet al.\u003c/em\u003e SARS-CoV-2 spike protein interacts with and activates TLR41. \u003cem\u003eCell Research\u003c/em\u003e vol. 31 818\u0026ndash;820 Preprint at https://doi.org/10.1038/s41422-021-00495-9 (2021).\u003c/li\u003e\n\u003cli\u003eMantovani, A., Allavena, P., Sica, A. \u0026amp; Balkwill, F. Cancer-related inflammation. \u003cem\u003eNature\u003c/em\u003e vol. 454 Preprint at https://doi.org/10.1038/nature07205 (2008).\u003c/li\u003e\n\u003cli\u003eKawai, T., Ikegawa, M., Ori, D. \u0026amp; Akira, S. Decoding Toll-like receptors: Recent insights and perspectives in innate immunity. \u003cem\u003eImmunity\u003c/em\u003e vol. 57 649\u0026ndash;673 Preprint at https://doi.org/10.1016/j.immuni.2024.03.004 (2024).\u003c/li\u003e\n\u003cli\u003eOhto, U., Fukase, K., Miyake, K. \u0026amp; Shimizu, T. Structural basis of species-specific endotoxin sensing by innate immune receptor TLR4/MD-2. \u003cem\u003eProc Natl Acad Sci U S A\u003c/em\u003e \u003cstrong\u003e109\u003c/strong\u003e, 7421\u0026ndash;7426 (2012).\u003c/li\u003e\n\u003cli\u003eMata-Haro, V. \u003cem\u003eet al.\u003c/em\u003e The vaccine adjuvant monophosphoryl lipid A as a TRIF-biased agonist of TLR4. \u003cem\u003eScience (1979)\u003c/em\u003e \u003cstrong\u003e316\u003c/strong\u003e, 1628\u0026ndash;1632 (2007).\u003c/li\u003e\n\u003cli\u003eCiesielska, A., Matyjek, M. \u0026amp; Kwiatkowska, K. TLR4 and CD14 trafficking and its influence on LPS-induced pro-inflammatory signaling. \u003cem\u003eCellular and Molecular Life Sciences\u003c/em\u003e vol. 78 1233\u0026ndash;1261 Preprint at https://doi.org/10.1007/s00018-020-03656-y (2021).\u003c/li\u003e\n\u003cli\u003ePł\u0026oacute;ciennikowska, A., Hromada-Judycka, A., Borzęcka, K. \u0026amp; Kwiatkowska, K. Co-operation of TLR4 and raft proteins in LPS-induced pro-inflammatory signaling. \u003cem\u003eCellular and Molecular Life Sciences\u003c/em\u003e \u003cstrong\u003e72\u003c/strong\u003e, 557\u0026ndash;581 (2015).\u003c/li\u003e\n\u003cli\u003eRajaiah, R., Perkins, D. J., Ireland, D. D. C., Vogel, S. N. \u0026amp; Kagan, J. C. CD14 dependence of TLR4 endocytosis and TRIF signaling displays ligand specificity and is dissociable in endotoxin tolerance. \u003cem\u003eProc Natl Acad Sci U S A\u003c/em\u003e \u003cstrong\u003e112\u003c/strong\u003e, 8391\u0026ndash;8396 (2015).\u003c/li\u003e\n\u003cli\u003ePerkins, D. J. \u003cem\u003eet al.\u003c/em\u003e Autocrine\u0026ndash;paracrine prostaglandin E 2 signaling restricts TLR4 internalization and TRIF signaling. \u003cem\u003eNat Immunol\u003c/em\u003e \u003cstrong\u003e19\u003c/strong\u003e, 1309\u0026ndash;1318 (2018).\u003c/li\u003e\n\u003cli\u003eOblak, A., Pohar, J. \u0026amp; Jerala, R. MD-2 determinants of nickel and cobalt-mediated activation of human TLR4. \u003cem\u003ePLoS One\u003c/em\u003e \u003cstrong\u003e10\u003c/strong\u003e, (2015).\u003c/li\u003e\n\u003cli\u003eDomingo, I. K., Groenendyk, J., Michalak, M. \u0026amp; Bhavsar, A. P. Cisplatin toxicity is mediated by direct binding to TLR4 through a mechanism that is distinct from metal allergens Running Title: Cisplatin activates TLR4 by Direct Binding. doi:10.1101/2022.07.06.498982.\u003c/li\u003e\n\u003cli\u003eZhao, Y. \u003cem\u003eet al.\u003c/em\u003e Publisher Correction: SARS-CoV-2 spike protein interacts with and activates TLR4 (Cell Research, (2021), 31, 7, (818-820), 10.1038/s41422-021-00495-9). \u003cem\u003eCell Research\u003c/em\u003e vol. 31 Preprint at https://doi.org/10.1038/s41422-021-00501-0 (2021).\u003c/li\u003e\n\u003cli\u003eVolk-Draper, L. \u003cem\u003eet al.\u003c/em\u003e Paclitaxel therapy promotes breast cancer metastasis in a TLR4-dependent manner. \u003cem\u003eCancer Res\u003c/em\u003e \u003cstrong\u003e74\u003c/strong\u003e, 5421\u0026ndash;5434 (2014).\u003c/li\u003e\n\u003cli\u003eRajput, S., Volk-Draper, L. D. \u0026amp; Ran, S. TLR4 is a novel determinant of the response to paclitaxel in breast cancer. \u003cem\u003eMol Cancer Ther\u003c/em\u003e \u003cstrong\u003e12\u003c/strong\u003e, 1676\u0026ndash;1687 (2013).\u003c/li\u003e\n\u003cli\u003eWanderley, C. W. \u003cem\u003eet al.\u003c/em\u003e Paclitaxel reduces tumor growth by reprogramming tumor-associated macrophages to an M1 profile in a TLR4-dependent manner. \u003cem\u003eCancer Res\u003c/em\u003e \u003cstrong\u003e78\u003c/strong\u003e, 5891\u0026ndash;5900 (2018).\u003c/li\u003e\n\u003cli\u003eLiu, S. \u003cem\u003eet al.\u003c/em\u003e Critical role of Syk-dependent STAT1 activation in innate antiviral immunity. \u003cem\u003eCell Rep\u003c/em\u003e \u003cstrong\u003e34\u003c/strong\u003e, (2021).\u003c/li\u003e\n\u003cli\u003eLin, Y.-C., Huang, D.-Y., Chu, C.-L., Lin, Y.-L. \u0026amp; Lin, W.-W. \u003cem\u003eI M M U N O L O G Y The Tyrosine Kinase Syk Differentially Regulates Toll-like Receptor Signaling Downstream of the Adaptor Molecules TRAF6 and TRAF3\u003c/em\u003e. www.SCIENCESIGNALING.org.\u003c/li\u003e\n\u003cli\u003eJoshi, S. New insights into SYK targeting in solid tumors. \u003cem\u003eTrends in Pharmacological Sciences\u003c/em\u003e Preprint at https://doi.org/10.1016/j.tips.2024.08.006 (2024).\u003c/li\u003e\n\u003cli\u003eBalic, J. J. \u003cem\u003eet al.\u003c/em\u003e STAT3 serine phosphorylation is required for TLR4 metabolic reprogramming and IL-1\u0026beta; expression. \u003cem\u003eNat Commun\u003c/em\u003e \u003cstrong\u003e11\u003c/strong\u003e, (2020).\u003c/li\u003e\n\u003cli\u003eRen, B. \u003cem\u003eet al.\u003c/em\u003e Tumor microenvironment participates in metastasis of pancreatic cancer. \u003cem\u003eMolecular Cancer\u003c/em\u003e vol. 17 Preprint at https://doi.org/10.1186/s12943-018-0858-1 (2018).\u003c/li\u003e\n\u003cli\u003eCilenti, F. \u003cem\u003eet al.\u003c/em\u003e A PGE2-MEF2A axis enables context-dependent control of inflammatory gene expression. \u003cem\u003eImmunity\u003c/em\u003e \u003cstrong\u003e54\u003c/strong\u003e, 1665-1682.e14 (2021).\u003c/li\u003e\n\u003cli\u003ePerkins, D. J. \u003cem\u003eet al.\u003c/em\u003e Autocrine\u0026ndash;paracrine prostaglandin E 2 signaling restricts TLR4 internalization and TRIF signaling. \u003cem\u003eNat Immunol\u003c/em\u003e \u003cstrong\u003e19\u003c/strong\u003e, 1309\u0026ndash;1318 (2018).\u003c/li\u003e\n\u003cli\u003eZheng, J. \u003cem\u003eet al.\u003c/em\u003e Comprehensive elaboration of the cGAS-STING signaling axis in cancer development and immunotherapy. \u003cem\u003eMolecular Cancer\u003c/em\u003e vol. 19 Preprint at https://doi.org/10.1186/s12943-020-01250-1 (2020).\u003c/li\u003e\n\u003cli\u003eNalio Ramos, R. \u003cem\u003eet al.\u003c/em\u003e Tissue-resident FOLR2+ macrophages associate with CD8+ T cell infiltration in human breast cancer. \u003cem\u003eCell\u003c/em\u003e \u003cstrong\u003e185\u003c/strong\u003e, 1189-1207.e25 (2022).\u003c/li\u003e\n\u003cli\u003eKnudsen, N. H. \u0026amp; Manguso, R. T. Tumor-Derived PGE2 Gives NK Cells a Headache. \u003cem\u003eImmunity\u003c/em\u003e vol. 53 1131\u0026ndash;1132 Preprint at https://doi.org/10.1016/j.immuni.2020.11.018 (2020).\u003c/li\u003e\n\u003cli\u003eNg, M., Cerezo-Wallis, D., Ng, L. G. \u0026amp; Hidalgo, A. Adaptations of neutrophils in cancer. \u003cem\u003eImmunity\u003c/em\u003e vol. 58 40\u0026ndash;58 Preprint at https://doi.org/10.1016/j.immuni.2024.12.009 (2025).\u003c/li\u003e\n\u003cli\u003eKhan, N. \u003cem\u003eet al.\u003c/em\u003e \u0026beta;-Glucan reprograms neutrophils to promote disease tolerance against influenza A virus. \u003cem\u003eNat Immunol\u003c/em\u003e (2025) doi:10.1038/s41590-024-02041-2.\u003c/li\u003e\n\u003cli\u003eHuang, X. \u003cem\u003eet al.\u003c/em\u003e Neutrophils in Cancer immunotherapy: friends or foes? \u003cem\u003eMolecular Cancer\u003c/em\u003e vol. 23 Preprint at https://doi.org/10.1186/s12943-024-02004-z (2024).\u003c/li\u003e\n\u003cli\u003eGros, A. \u003cem\u003eet al.\u003c/em\u003e PD-1 identifies the patient-specific CD8+ tumor-reactive repertoire infiltrating human tumors. \u003cem\u003eJournal of Clinical Investigation\u003c/em\u003e \u003cstrong\u003e124\u003c/strong\u003e, 2246\u0026ndash;2259 (2014).\u003c/li\u003e\n\u003cli\u003eGuo, X., Zhao, Y. \u0026amp; You, F. MOI is a comprehensive database collecting processed multi-omics data associated with viral infection. \u003cem\u003eSci Rep\u003c/em\u003e \u003cstrong\u003e14\u003c/strong\u003e, (2024).\u003c/li\u003e\n\u003cli\u003eJumper, J. \u003cem\u003eet al.\u003c/em\u003e Highly accurate protein structure prediction with AlphaFold. \u003cem\u003eNature\u003c/em\u003e \u003cstrong\u003e596\u003c/strong\u003e, (2021).\u003c/li\u003e\n\u003cli\u003eTrott, O. \u0026amp; Olson, A. J. AutoDock Vina: Improving the speed and accuracy of docking with a new scoring function, efficient optimization, and multithreading. \u003cem\u003eJ Comput Chem\u003c/em\u003e \u003cstrong\u003e31\u003c/strong\u003e, 455\u0026ndash;461 (2010).\u003c/li\u003e\n\u003cli\u003ePettersen, E. F. \u003cem\u003eet al.\u003c/em\u003e UCSF Chimera - A visualization system for exploratory research and analysis. \u003cem\u003eJ Comput Chem\u003c/em\u003e \u003cstrong\u003e25\u003c/strong\u003e, (2004).\u003c/li\u003e\n\u003cli\u003eLiao, Y., Smyth, G. K. \u0026amp; Shi, W. The Subread aligner: Fast, accurate and scalable read mapping by seed-and-vote. \u003cem\u003eNucleic Acids Res\u003c/em\u003e \u003cstrong\u003e41\u003c/strong\u003e, (2013).\u003c/li\u003e\n\u003cli\u003eLiao, Y., Smyth, G. K. \u0026amp; Shi, W. FeatureCounts: An efficient general purpose program for assigning sequence reads to genomic features. \u003cem\u003eBioinformatics\u003c/em\u003e \u003cstrong\u003e30\u003c/strong\u003e, 923\u0026ndash;930 (2014).\u003c/li\u003e\n\u003cli\u003eKanehisa, M. \u0026amp; Goto, S. \u003cem\u003eKEGG: Kyoto Encyclopedia of Genes and Genomes\u003c/em\u003e. \u003cem\u003eNucleic Acids Research\u003c/em\u003e vol. 28 http://www.genome.ad.jp/kegg/ (2000).\u003c/li\u003e\n\u003cli\u003eYu, G., Wang, L. G., Han, Y. \u0026amp; He, Q. Y. ClusterProfiler: An R package for comparing biological themes among gene clusters. \u003cem\u003eOMICS\u003c/em\u003e \u003cstrong\u003e16\u003c/strong\u003e, 284\u0026ndash;287 (2012).\u003c/li\u003e\n\u003cli\u003eDennis, G. \u003cem\u003eet al.\u003c/em\u003e \u003cem\u003eDAVID: Database for Annotation, Visualization, and Integrated Discovery\u003c/em\u003e. \u003cem\u003eGenome Biology\u003c/em\u003e vol. 4 http://dot.ped.med.umich.edu:2000/ (2003).\u003c/li\u003e\n\u003cli\u003eSzklarczyk, D. \u003cem\u003eet al.\u003c/em\u003e STRING v11: Protein-protein association networks with increased coverage, supporting functional discovery in genome-wide experimental datasets. \u003cem\u003eNucleic Acids Res\u003c/em\u003e \u003cstrong\u003e47\u003c/strong\u003e, (2019).\u003c/li\u003e\n\u003cli\u003eShannon, P. \u003cem\u003eet al.\u003c/em\u003e Cytoscape: A software Environment for integrated models of biomolecular interaction networks. \u003cem\u003eGenome Res\u003c/em\u003e \u003cstrong\u003e13\u003c/strong\u003e, (2003).\u003c/li\u003e\n\u003cli\u003eJanky, R. \u003cem\u003eet al.\u003c/em\u003e iRegulon: From a Gene List to a Gene Regulatory Network Using Large Motif and Track Collections. \u003cem\u003ePLoS Comput Biol\u003c/em\u003e \u003cstrong\u003e10\u003c/strong\u003e, (2014).\u003c/li\u003e\n\u003cli\u003eBuenrostro, J. D., Wu, B., Chang, H. Y. \u0026amp; Greenleaf, W. J. ATAC-seq: A method for assaying chromatin accessibility genome-wide. \u003cem\u003eCurr Protoc Mol Biol\u003c/em\u003e \u003cstrong\u003e2015\u003c/strong\u003e, (2015).\u003c/li\u003e\n\u003cli\u003eLangmead, B., Trapnell, C., Pop, M. \u0026amp; Salzberg, S. L. Ultrafast and memory-efficient alignment of short DNA sequences to the human genome. \u003cem\u003eGenome Biol\u003c/em\u003e \u003cstrong\u003e10\u003c/strong\u003e, (2009).\u003c/li\u003e\n\u003cli\u003eLi, H. \u003cem\u003eet al.\u003c/em\u003e The Sequence Alignment/Map format and SAMtools. \u003cem\u003eBioinformatics\u003c/em\u003e \u003cstrong\u003e25\u003c/strong\u003e, 2078\u0026ndash;2079 (2009).\u003c/li\u003e\n\u003cli\u003eZhang, Y. \u003cem\u003eet al.\u003c/em\u003e Model-based analysis of ChIP-Seq (MACS). \u003cem\u003eGenome Biol\u003c/em\u003e \u003cstrong\u003e9\u003c/strong\u003e, (2008).\u003c/li\u003e\n\u003cli\u003eQuinlan, A. R. \u0026amp; Hall, I. M. BEDTools: A flexible suite of utilities for comparing genomic features. \u003cem\u003eBioinformatics\u003c/em\u003e \u003cstrong\u003e26\u003c/strong\u003e, 841\u0026ndash;842 (2010).\u003c/li\u003e\n\u003cli\u003eRam\u0026iacute;rez, F., D\u0026uuml;ndar, F., Diehl, S., Gr\u0026uuml;ning, B. A. \u0026amp; Manke, T. DeepTools: A flexible platform for exploring deep-sequencing data. \u003cem\u003eNucleic Acids Res\u003c/em\u003e \u003cstrong\u003e42\u003c/strong\u003e, (2014).\u003c/li\u003e\n\u003cli\u003eLun, A. T. L. \u0026amp; Smyth, G. K. Csaw: A Bioconductor package for differential binding analysis of ChIP-seq data using sliding windows. \u003cem\u003eNucleic Acids Res\u003c/em\u003e \u003cstrong\u003e44\u003c/strong\u003e, e45 (2015).\u003c/li\u003e\n\u003cli\u003eYu, G., Wang, L. G. \u0026amp; He, Q. Y. ChIP seeker: An R/Bioconductor package for ChIP peak annotation, comparison and visualization. \u003cem\u003eBioinformatics\u003c/em\u003e \u003cstrong\u003e31\u003c/strong\u003e, 2382\u0026ndash;2383 (2015).\u003c/li\u003e\n\u003cli\u003eRobinson, J. T. \u003cem\u003eet al.\u003c/em\u003e Integrative genomics viewer. \u003cem\u003eNature Biotechnology\u003c/em\u003e vol. 29 24\u0026ndash;26 Preprint at https://doi.org/10.1038/nbt.1754 (2011).\u003c/li\u003e\n\u003cli\u003eSatija, R., Farrell, J. A., Gennert, D., Schier, A. F. \u0026amp; Regev, A. Spatial reconstruction of single-cell gene expression data. \u003cem\u003eNat Biotechnol\u003c/em\u003e \u003cstrong\u003e33\u003c/strong\u003e, 495\u0026ndash;502 (2015).\u003c/li\u003e\n\u003cli\u003eJin, S. \u003cem\u003eet al.\u003c/em\u003e Inference and analysis of cell-cell communication using CellChat. \u003cem\u003eNat Commun\u003c/em\u003e \u003cstrong\u003e12\u003c/strong\u003e, (2021).\u003c/li\u003e\n\u003cli\u003eTrapnell, C. \u003cem\u003eet al.\u003c/em\u003e The dynamics and regulators of cell fate decisions are revealed by pseudotemporal ordering of single cells. \u003cem\u003eNat Biotechnol\u003c/em\u003e \u003cstrong\u003e32\u003c/strong\u003e, 381\u0026ndash;386 (2014).\u003c/li\u003e\n\u003cli\u003eSubramanian, A. \u003cem\u003eet al.\u003c/em\u003e A Next Generation Connectivity Map: L1000 Platform and the First 1,000,000 Profiles. \u003cem\u003eCell\u003c/em\u003e \u003cstrong\u003e171\u003c/strong\u003e, 1437-1452.e17 (2017).\u003c/li\u003e\n\u003cli\u003eWishart, D. S. \u003cem\u003eet al.\u003c/em\u003e DrugBank: a comprehensive resource for in silico drug discovery and exploration. \u003cem\u003eNucleic Acids Res\u003c/em\u003e \u003cstrong\u003e34\u003c/strong\u003e, (2006).\u003c/li\u003e\n\u003cli\u003eMendez, D. \u003cem\u003eet al.\u003c/em\u003e ChEMBL: Towards direct deposition of bioassay data. \u003cem\u003eNucleic Acids Res\u003c/em\u003e \u003cstrong\u003e47\u003c/strong\u003e, D930\u0026ndash;D940 (2019).\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Pancreatic ductal adenocarcinoma (PDAC), Inflammatory response, Innate immunity response, Convolutional neural network, Phosphines-Nitrogen-Phosphines (PNP)-pincer ligands, TLR4, STAT1, NF-κB","lastPublishedDoi":"10.21203/rs.3.rs-7085332/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-7085332/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003ePancreatic cancer is projected to become the second leading cause of cancer-related deaths globally by 2030, yet effective therapeutic options remain limited. Within the pancreatic cancer tumor microenvironment (TME), tumor-associated macrophages (TAMs) secrete interleukin-1 beta (IL-1β), promoting cancer progression while suppressing type I interferons (IFN-I), which is critical for tumor killing. Utilizing the convolutional neural network (CNN)-based DLINP model developed in our laboratory, we identified Co68\u0026mdash;an effective metal catalyst featuring a Phosphines-Nitrogen-Phosphines (PNP)-chelated CoCl₂ complex\u0026mdash;as a promising candidate to modulate innate immune responses. In animal models of pancreatic cancer, Co68 demonstrated superior antitumor efficacy compared to the STING agonist DMXAA and showed enhanced therapeutic effects when combined with PD-1 blockade. Single-cell RNA sequencing (scRNA-seq) revealed that Co68 reprogrammed TAMs to express interferon-stimulated genes (ISGs), attenuated pro-inflammatory cytokine secretion, and disrupted the IL-1β-PGE2 feedback loop, thereby facilitating the recruitment of NK and cytotoxic CD8\u003csup\u003e+\u003c/sup\u003e T cells into the TME. Mechanistically, Co68 activated the IFN-I signaling pathway through the TLR4-TRIF-IFN-I axis and inhibited inflammation via the TLR4-SYK-STAT1 pathway. Collectively, these findings highlight the therapeutic potential of Co68, derived from PNP-pincer chemistry, to reshape immune dynamics within the pancreatic cancer TME, positioning it as a promising candidate for innovative immunotherapy strategies.\u003c/p\u003e","manuscriptTitle":"Phosphines-Nitrogen-Phosphines Chelated CoCl2 Exhibits Potent Antitumor Activity in Pancreatic Cancer","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-07-18 11:49:07","doi":"10.21203/rs.3.rs-7085332/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"5a2404fe-c661-4e96-9402-44dd92fb297b","owner":[],"postedDate":"July 18th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":51578017,"name":"Biological sciences/Immunology/Innate immunity/Pattern recognition receptors/Toll-like receptors"},{"id":51578018,"name":"Biological sciences/Cancer/Cancer microenvironment"},{"id":51578019,"name":"Biological sciences/Drug discovery/Drug screening/Virtual screening"},{"id":51578020,"name":"Biological sciences/Drug discovery/Target identification"}],"tags":[],"updatedAt":"2025-08-23T12:55:20+00:00","versionOfRecord":[],"versionCreatedAt":"2025-07-18 11:49:07","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-7085332","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-7085332","identity":"rs-7085332","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}