SLC39-driven zinc influx orchestrates pleiotropic tumor–immune crosstalk to establish an immune-suppressive microenvironment in colorectal cancer

SLC39-driven zinc influx orchestrates pleiotropic tumor–immune crosstalk to establish an immune-suppressive microenvironment in colorectal 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 Research Article SLC39-driven zinc influx orchestrates pleiotropic tumor–immune crosstalk to establish an immune-suppressive microenvironment in colorectal cancer Seok June Hong, Seheum Park, Sunghoon Kim, Young Il Park, Mikyung Kang, and 4 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8267774/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 Tumor metabolic reprogramming profoundly influences immune regulation, yet the mechanisms linking solute carrier (SLC) transporter activity to immune suppression remain elusive. Through integrative multi-omics, spatial, and single-cell analyses of colorectal cancer (CRC), we uncover a zinc influx–driven signaling axis mediated by the SLC39 family that establishes an immune-suppressive tumor ecosystem. Multi-omics clustering of 258 CRC patients identified three SLC-centered archetypes, among which an SLC39-enriched subtype displayed zinc pathway activation and correspondence to the “immune-desert” CMS2 subtype. Mechanistically, SLC39-mediated zinc influx activated the transcription factor CDX2, promoting enhancer-driven transcription of CD24, an anti-phagocytic “don’t eat me” signal. Zinc-dependent CD24 upregulation occurred independently of CD47 and was restricted to malignant epithelial cells. Single-cell and spatial transcriptomics revealed that CD24-expressing tumor cells interact with SIGLEC10 + monocyte-derived macrophages, dendritic cells, and resident macrophages, triggering pleiotropic immunoregulatory programs that suppress phagocytosis and remodel adhesion networks. This SLC39–CD24–SIGLEC10 axis defined spatially recurrent immune-suppressive niches and was associated with poor survival and resistance to immune checkpoint blockade. Functional assays confirmed that zinc-induced CD24–SIGLEC10 engagement attenuates macrophage phagocytosis, reversible by CD24 blockade. Furthermore, integrative modeling across five ICI-treated CRC cohorts demonstrated that a seven-gene signature encompassing SLC39 transporters, CDX2, CD24, and SIGLEC10 robustly predicted clinical response to immunotherapy, outperforming established biomarkers. These findings identify SLC39-mediated zinc influx as a regulator of tumor–immune crosstalk in CMS2-like CRC and highlight the SLC39–zinc–CD24–SIGLEC10 axis as a promising therapeutic target to overcome immune exclusion and immunotherapy resistance. Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Introduction Tumor cells exhibit heightened metabolic demands and reprogram nutrient acquisition through solute carrier (SLC) transporters to sustain uncontrolled proliferation, survival, and metastasis. Members of the SLC family have emerged as pivotal mediators of metabolic reprogramming, facilitating the import of essential nutrients and the export of metabolic waste, thereby coupling metabolic flux to oncogenic signaling 1 , 2 . SLC2A1 (GLUT1), a key glucose transporter, is upregulated in many cancers to support aerobic glycolysis, while SLC7A5 (LAT1) enables uptake of large neutral amino acids such as leucine, sustaining protein synthesis and mTORC1 activation 2 . Therapeutic inhibition of LAT1 has shown efficacy in preclinical models by restricting amino acid availability and suppressing tumor progression. Similarly, SLC7A11 (xCT) imports cystine to fuel glutathione biosynthesis, enhancing antioxidant defense and protecting cancer cells from ferroptosis 3 . These findings underscore SLC transporters as central mediators of metabolic reprogramming and potential therapeutic targets in oncology 4 . Although metabolic rewiring is a hallmark of cancer, the extent to which SLC transporters orchestrate interactions between tumor cells and the tumor microenvironment (TME) remains poorly defined. Accumulating evidence suggests that metabolic byproducts, such as lactate generated through the Warburg effect, can acidify the TME and impair immune cell activity, thereby facilitating immune escape 5 . However, most studies have emphasized tumor-intrinsic metabolic programs while neglecting the reciprocal influence of stromal and immune components 6 . As such, current frameworks inadequately capture the interplay between metabolic adaptation and immune modulation. This disconnect is particularly critical when considering that metabolic competition and nutrient availability in the TME can shape both tumor behavior and therapeutic response. The absence of integrated, systems-level studies that dissect these interactions has limited our capacity to identify context-specific metabolic vulnerabilities and design effective metabolism-targeted therapies. Colorectal cancer (CRC), with its well-characterized consensus molecular subtypes (CMS), presents a robust model to interrogate the intersection of metabolic regulation and immune microenvironment. CMS1 tumors exhibit hypermutation and immune infiltration, CMS2 tumors feature canonical WNT signaling and an immune-desert phenotype, CMS3 tumors display metabolic dysregulation, and CMS4 7 tumors are marked by stromal activation and poor prognosis, occurring at frequencies of 16%, 32%, 17%, and 35%, respectively 8 . Despite these subtype-specific landscapes, the contribution of SLC transporters to CRC subtype identity and TME modulation remains unclear. We hypothesize that specific SLC transporters define metabolically regulated immunosuppressive states in CRC. In particular, we propose that SLC39-mediated zinc influx drives subtype-specific transcriptional reprogramming, culminating in immune checkpoint activation and exclusion of anti-tumoral immune response. By integrating multi-omics, spatial, and single-cell data, our study aims to elucidate the SLC–TME axis as a mechanistic bridge linking metabolic inputs to immune modulation across CRC subtypes. Materials and Methods Patient sample collection and multi-omics profiling The primary dataset used for this study was obtained from The Cancer Genome Atlas (TCGA), specifically focusing on colorectal cancer (CRC) samples. A total of 255 CRC samples were selected, encompassing multi-omics data including transcriptomics, mutations, copy number variations (CNV), and DNA methylation. These multi-omics profiles were retrieved from Xena Browser (https://xenabrowser.net/), which provides access to publicly available TCGA datasets. These multi-omics data were utilized to investigate the molecular landscape of CRC. For data validation and additional analyses, we used data from the Clinical Proteomic Tumor Analysis Consortium (CPTAC), which contains proteomic and phosphoproteomic profiles of CRC samples. A total of 109 CRC samples were included in this validation cohort. The CPTAC data was retrieved from Linkedomics (https://linkedomics.org/), which provides integrative analyses of omics datasets. These data were used to cross-validate findings from the TCGA dataset and further refine the molecular characterization of CRC. Clinical information for the TCGA cohort is provided in Table 1, and corresponding information for the CPTAC cohort is available in Table 2. Data integration and gene selection To investigate metabolic dependencies in CRC, we focused on solute carrier (SLC) transporters, which regulate the import and export of metabolites, as well as cancer-specific oncogenic driver genes. We retained only the genes annotated as part of the SLC family or as known oncogenes in CRC. The list of SLC genes was curated by excluding pseudogenes and antisense transcripts to ensure biological relevance and transcriptional activity. For oncogenic driver genes, we utilized the OncoVar database, which provides curated lists of cancer-type-specific driver genes along with a scoring system reflecting their pathogenic potential. Genes annotated for colorectal adenocarcinoma (COAD) with a Driver level of 4 (i.e., “pathogenic,” score ≥ 20) were selected for analysis. Additionally, to increase the robustness of the selected oncogenes, we further filtered for genes with a Total score ≥ 3, representing the number of supporting databases in which the gene was identified as a driver. This filtering strategy ensured that only metabolically and oncogenically relevant genes were retained for subsequent integrative multi-omics analyses. Dimensionality reduction and multi-omics integration To integrate the multiple layers of omics data and identify latent molecular features associated with colorectal cancer, we applied Multi-Omics Factor Analysis (MOFA), a statistical framework designed for unsupervised integration of multi-omics datasets. MOFA enables extraction of low-dimensional latent factors that capture sources of variation shared across or specific to individual omics layers. We constructed a MOFA model using five omics layers: transcriptomics, somatic mutations, copy number variations, DNA promoter methylation, and DNA gene body methylation. Given the nature of the data, we specified ‘gaussian’ likelihoods for continuous data types (transcriptomics, CNVs, methylation), and ‘bernoulli’ likelihood for the mutation matrix, which was binarized to represent the presence or absence of nonsynonymous variants in each gene. The model was trained using the following parameters: the number of latent factors was set to 10, the convergence mode was set to ‘slow’ to ensure a thorough exploration of the posterior space, and maxiter was set to 10,000 to allow sufficient iterations for convergence. This dimensionality reduction provided a shared latent representation of the dataset, which was used for downstream clustering, module discovery, and association analyses with clinical or immunological features. Archetypal analysis using ParetoTI To identify representative molecular states, archetypal analysis was performed using the ParetoTI R package, which geometrically infers extremal points (archetypes) from input dimensional space. As input, the top four latent factors derived from MOFA were selected based on the proportion of explained variance. All six pairwise combinations of these four factors were evaluated using the t-ratio metric, which quantifies the fit between the polytope and convex hull of the data. The factor pair with the highest t-ratio (Factor 1 and Factor 4) was selected for downstream modeling. To functionally annotate the selected factors, gene set variation analysis (GSVA) was performed using metabolic pathway gene sets curated from Gene Ontology (GO: biological process category, filtered for terms containing “metabolic process”) and KEGG (metabolism category). Enrichment scores were computed for each sample and correlated with factor values to support interpretation of the molecular programs captured by the selected latent dimensions. Archetype assignment and multi-omics characterization Samples were assigned to one of three archetypes (Arc1, Arc2, Arc3) based on proximity to polytope vertices inferred by ParetoTI using MOFA-derived latent factors. To annotate biological relevance, consensus molecular subtypes (CMS) were inferred using the CMScaller R package. Molecular characteristics of each archetype were evaluated by analyzing enrichment patterns across multi-omics layers. Specifically, transcriptomic, promoter methylation, and gene body methylation profiles of SLC genes were compared across archetypes. GSVA was performed using curated pathway collections from GO, KEGG, WikiPathways, Reactome, and MSigDB Hallmark gene sets, focusing on the identification of pathways enriched in each archetype. To assess tumor microenvironment (TME) differences, xCell was used to infer immune and stromal cell-type enrichment scores from bulk transcriptomic data. Archetype-specific correlations with TME composition were calculated using Pearson correlation to evaluate immune landscape variation across subtypes. Identification of CD24 enhancer region and transcription factor binding motifs Putative enhancer elements regulating CD24 were identified using the GeneHancer database (GH06J106781). Open chromatin accessibility at this region was confirmed using ATAC-seq data from 25 colorectal adenocarcinoma (COAD) samples 9 . To investigate enhancer activity specific to the archetype 1, chromatin accessibility signals were compared across archetype-classified samples. Topological organization of the enhancer–promoter interaction was assessed using Hi-C data via the 3D Genome Browser (http://3dgenome.org), which revealed that both CD24 and GH06J106781 reside within the same topologically associating domain (TAD), supporting the potential regulatory interaction. To identify transcription factors that potentially bind to the enhancer, motif enrichment analysis was performed using the TFBSTools R package in conjunction with the JASPAR2022 motif database 10 . The enhancer sequence was scanned for transcription factor binding motifs using position weight matrices (PWMs) from JASPAR, and matches were identified based on a minimum score threshold of 80% relative to the minimal PWM score. For each candidate binding site, associated p-values and matched binding sequences were computed using searchSeq(), and high-confidence hits were retained for downstream structural modeling. Clinical information for scATAC-seq cohort is summarized in Table 3. Single-cell RNA-seq data processing and clustering Tumor-derived single-cell RNA-seq data were obtained from the Synapse dataset (accession: syn26844071). Only cells annotated as originating from tumor tissue were included in the analysis. Quality control filtering was applied to retain cells with more than 200 detected genes, fewer than 100,000 total counts, and less than 3% mitochondrial gene content. Genes expressed in fewer than three cells were removed. The processed data were normalized and scaled, and 2,000 highly variable genes were selected for downstream analysis. Principal component analysis (PCA) was performed, and the first 10 dimensions were used for clustering (resolution = 0.5) and UMAP-based visualization. All analyses were conducted using Seurat v5.2.1. Clinical information for this cohort is summarized in Table 4. Single-cell RNA-seq dataset from ICI-treated colorectal cancer Single-cell transcriptomic data from immune checkpoint inhibitor (ICI)-treated colorectal cancer was obtained from the publicly available dataset SCP2079, which corresponds to the study "Combined PD-1, BRAF and MEK inhibition in BRAFV600E colorectal cancer." For this study, we specifically used the pre-treatment samples to analyze the baseline immune and tumor microenvironmental states prior to therapy. Clinical information for this cohort is summarized in Table 5. Spatial transcriptomics data acquisition and preprocessing Spatial transcriptomics data were obtained from the publicly available dataset GSE225857, which includes FFPE tissue sections from four colorectal cancer (CRC) patient samples. The dataset was generated using the 10x Genomics Visium Spatial Gene Expression platform for FFPE tissues, and sequencing was performed on the Illumina NovaSeq 6000 system. For this study, we used the preprocessed UMI count matrices and accompanying H&E-stained tissue images provided with the dataset. Spatial transcriptomic analyses were carried out using Seurat (v5.2.1) and the SPATA2 11 package in R. Analyses included quality control filtering, normalization, dimensional reduction, clustering, and spatial feature extraction. Clinical information for this cohort is summarized in Table 6. Cell lines and cell culture LS174T cells were a gift from Dr Young-Hee Lim. THP-1 (40202) cells were obtained from the Korean Cell Line Bank. LS174T and THP-1 cells were cultured in Dulbecco’s modified Eagle medium (DMEM; Welgene) or Roswell Park Memorial Institute (RPMI) medium (Welgene), respectively, supplemented with 10% fetal bovine serum (FBS, Gibco, 12483020) and 1% penicillin–streptomycin (Hyclone, SV30010) in a humidified incubator at 37 °C with 5% CO₂. Lentiviral constructs and transduction The CRISPR–dCas9 Synergistic Activation Mediator (SAM) system was employed as a CRISPR activation (CRISPRa) tool to enhance gene expression. This system comprised two plasmids: lentiMPHv2 (Addgene, 89308) and lentiSAMv2 (Addgene, 75112). Single guide RNAs (sgRNAs) were designed using the CRISPICK and CHOPCHOP platforms and synthesized by Bionics (Korea). sgRNA sequences were cloned into the lentiSAMv2 backbone according to the protocol provided by Addgene. Briefly, lentiSAMv2 was digested with Esp3I (Enzynomics, R116S), and annealed oligonucleotides were ligated into the digested vector. Positive constructs were verified by Sanger sequencing. For lentiviral production, HEK293T cells were seeded in 6-well plates and cultured overnight to reach 70–80% confluence. Cells were transfected with a plasmid mixture of pMD2.G (Addgene, 12259), psPAX2 (Addgene, 12260), and either lentiMPHv2, lentiSAMv2-sgRNA, or lentiSAMv2-scramble gRNA using polyethylenimine in serum-free DMEM. After 6 h, the medium was replaced with DMEM supplemented with 10% FBS and 1% penicillin–streptomycin. The viral supernatant was harvested 48 h later, filtered through a 0.45-µm filter, and used for transduction. For gene activation in LS174T cells, lentiMPHv2 infection was performed for 48 h, followed by hygromycin selection, and subsequently, cells were infected with lentiSAMv2-sgRNA for 48 h and selected with blasticidin. The sgRNA sequences (5′–3′) were as follows: sgCDX2, GAGGTTAAAGTGCACCAGGT; sgSLC39A4, GGTTGTCCAGGGCCAGACTG; sgSLC39A5, GGTGGGGTGTCCCTAGAAGG. Real-time qPCR Total RNA was isolated from cultured cells using RiboEx™ (GeneAll, 301) and quantified with a NanoDrop spectrophotometer (Thermo Fisher Scientific). Complementary DNA (cDNA) was synthesized from 2.5 µg of total RNA using the RevertAid First Strand cDNA Synthesis Kit (Thermo Fisher Scientific, K1621) in a total volume of 10 µL. Quantitative PCR was performed on either a StepOnePlus or QuantStudio 6 Flex Real-Time PCR System (Applied Biosystems) using SYBR Green reagents (Enzynomics, RT500 or RT501). Gene expression levels were normalized to β-actin, and relative expression was calculated using the ΔΔCt method. All primer sequences used in this study are listed in Table 7. Western blot and antibodies Cells were lysed in RIPA buffer supplemented with protease and phosphatase inhibitor cocktails. Protein concentrations were determined using a Bradford assay. Equal amounts of protein (30–40 µg) were separated by SDS–PAGE, transferred to PVDF membranes, and blocked with 3% BSA in TBST. Membranes were incubated with primary antibodies overnight at 4 °C, followed by HRP-conjugated secondary antibodies for 1 h at room temperature. Protein bands were detected using an ECL detection kit (Thermo Fisher Scientific, 32106 or GLPBio, GK10008) and imaged with a UVP ChemStudio system (Analytik Jena). Primary antibodies were as follows: anti-CD24 (clone SN3, Santa Cruz Biotechnology, SC-19585; Western blot (WB) 1:1,000; flow cytometry and phagocytosis assay 10 μg ml⁻¹) and anti-CDX2 (clone D11D10, Cell Signaling Technology, 12306; WB 1:1,000; immunofluorescence (IF) 1:400). Secondary antibodies included Goat anti-Mouse IgG (H+L) Cross-Adsorbed Secondary Antibody, Alexa Fluor™ 488 (Invitrogen, A-11001; flow cytometry 1:500) and Goat anti-Rabbit IgG (H+L) Cross-Adsorbed Secondary Antibody, Alexa Fluor™ 594 (Invitrogen, A-11012; IF 1:500). Immunofluorescence Cells were seeded on coverslips and fixed with 4% paraformaldehyde for 15 min at room temperature. Fixed cells were washed three times with DPBS and permeabilized with 0.1% Triton X-100 for 10 min at room temperature, followed by blocking in 3% BSA in DPBS for 1 h at room temperature. Cells were then incubated with primary antibodies diluted in 1% BSA in DPBS at 4 °C overnight. After washing with DPBS, cells were incubated with fluorescently labeled secondary antibodies for 1 h at room temperature. Nuclei were counterstained with DAPI, and coverslips were mounted using Fluoromount Aqueous Mounting Medium (Sigma). Images were acquired using an ECLIPSE Ts2-FL microscope (Nikon). Flow cytometry Flow cytometric analyses were performed to measure intracellular Zn²⁺ levels and membrane CD24 expression. For intracellular Zn²⁺ quantification, cells were harvested by trypsinization and incubated with 1 μM FluoZin™-3 (Invitrogen, F24195) for 30 min at 37 °C. Cells were then washed twice with DPBS and further incubated for 30 min at 37 °C to allow de-esterification. For analysis of membrane CD24 expression, cells were detached using Accutase (Millipore, SCR005), incubated with 10 μg ml⁻¹ anti-CD24 antibody in FACS buffer (0.5% BSA in DPBS) for 1 h on ice, washed twice with ice-cold FACS buffer, and then incubated with Alexa Fluor™ 488-conjugated secondary antibody for 30 min on ice prior to flow cytometric analysis. All samples were analyzed using an Attune™ NxT flow cytometer (Thermo Fisher Scientific) with excitation/emission, 488/520 nm. Data were analyzed using FlowJo software (v10, BD Biosciences), and MFI values were calculated from the fluorescence signals obtained. Phagocytosis assay THP-1 cells were differentiated into macrophages by incubation with 100 ng ml⁻¹ phorbol 12-myristate 13-acetate (PMA; GLPBio, GN10444) for 48 h. Macrophages were polarized to the moMC phenotype with 20 ng ml⁻¹ interferon-γ (IFN-γ; Enzynomics, C006) and 100 ng ml⁻¹ lipopolysaccharide (LPS; Sigma-Aldrich, L2880), or to the RTMC phenotype with 20 ng ml⁻¹ interleukin-4 (IL-4; Enzynomics, C008) and 20 ng ml⁻¹ interleukin-13 (IL-13; Enzynomics, C009) for 48 h. A total of GFP⁺ THP-1 cells (2 × 10⁵) were differentiated into moMC-like macrophages on 18-mm coverslips in 12-well plates. LS174T cells were detached using Accutase, labeled with pHrodo™ Red succinimidyl ester (Thermo Fisher Scientific, P36600) at a 1:30,000 dilution in PBS for 1 h at 37 °C, and washed twice with DMEM supplemented with 10% FBS and 1% penicillin–streptomycin. Labeled LS174T cells (1 × 10⁶ per well) were added to macrophages in serum-free RPMI with or without 10 μg ml⁻¹ anti-CD24 antibody (clone SN3; Santa Cruz Biotechnology); an isotype control antibody (clone MOPC-21, BioXcell) was included at the same concentration. After 2 h of co-culture at 37 °C, cells were washed five times with serum-free RPMI and imaged using a fluorescence microscope (ECLIPSE Ts2-FL, Nikon). Phagocytosis was quantified as the ratio of pHrodo Red+ GFP⁺ THP-1 macrophages. Statistical analysis in experimental validations Statistical analyses were performed using GraphPad Prism version 6 (GraphPad Software). All experiments were performed in at least three independent biological replicates. Data are presented as mean ± SEM. For comparisons between two groups, exact P values were calculated using unpaired two-tailed t-tests. For comparisons among more than two groups, P values were determined using two-way ANOVA followed by Tukey’s multiple comparisons test. Data distribution was assumed to be normal, but this was not formally tested. Statistical significance was defined as P < 0.05. Constructing a predictive model for ICI response in CRC To develop a predictive model of ICI response in CRC, we integrated five publicly available ICI-treated CRC transcriptomic cohorts (MSS1: n = 9, MSS2: n = 34, MSS3: n = 12, MSI1: n = 70, MSI2: n = 41), totaling 166 samples. Batch effects were removed using the removeBatchEffect function from the limma R package. The integrated dataset was randomly partitioned into a modeling set (n = 138) and an independent testing set (n = 28). The modeling set was further split into a training set (n = 110) and validation set (n = 28). Model construction was performed using the PyCaret machine learning framework in Python. PyCaret was used to automate the comparison of multiple classification algorithms under standardized settings (fold = 5, fold_strategy = StratifiedKFold, session_id = 2025). Among the tested models, the Gradient Boosting Classifier demonstrated the highest predictive performance (AUC = 0.7262) and was selected for final deployment. Other top-performing models included AdaBoost Classifier (AUC = 0.6735), Extra Trees Classifier (AUC = 0.6515), Extreme Gradient Boosting (AUC = 0.6314), and Random Forest Classifier (AUC = 0.6263). The final model was trained using seven biologically informed features—SLC39A2, SLC39A4, SLC39A5, SLC39A10, CDX2, CD24, and SIGLEC10—identified in this study as key regulators of zinc transport and immune evasion in colorectal cancer. Model performance was evaluated using the area under the receiver operating characteristic curve (AUC). To benchmark our model, we trained and evaluated additional Gradient Boosting Classifier using previously published gene sets known to be associated with ICI response in either pancancer or CRC-specific contexts 12–19 . Each comparator model was trained using the same 138-sample modeling set and evaluated on the 28-sample testing set to assess generalizability. To ensure balanced representation of both clinical cohorts and response classes, each of the five ICI-treated CRC datasets (MSI1, MSI2, MSS1, MSS2, MSS3) was independently split into modeling (85%) and external testing (15%) sets. Splitting was performed separately per cohort to maintain cohort-wise sample proportions in both subsets. As a result, the modeling set (n = 138) and external testing set (n = 28) preserved the original composition of the five cohorts: MSI1 (42.75% vs. 39.29%), MSI2 (24.64% vs. 25.00%), MSS1 (5.07% vs. 7.14%), MSS2 (20.29% vs. 21.43%), MSS3 (7.25% vs. 7.14%). Response class distribution was also maintained, with responders (R) comprising 62.32% of the modeling set and 53.57% of the testing set, and non-responders (NR) comprising 37.68% and 46.43%, respectively. Clinical information for this cohort is summarized in Table 8 and corresponding biomarker gene sets for benchmarking are summarized in Table 9. Results Multiomics clustering identifies three SLC transporter-centered archetypes in CRC patients SLC transporters are pivotal mediators of metabolite flux that sculpt the tumor microenvironment 20 . To elucidate the regulatory architecture of SLC transporters in CRC, we conducted integrative multiomics profiling encompassing more than 300 SLC transporter genes and 500 cancer driver genes across 258 TCGA-COAD patient samples (Fig. 1A). The multiomics dataset included transcriptomic alterations, somatic mutations, copy number variations, and DNA methylation profiles, enabling a comprehensive assessment of SLC transporter-associated regulatory programs. To resolve distinct patient subtypes orchestrated by SLC transporter-centric programs, we applied a dimensionality reduction-based clustering strategy using the Multi-Omics Factor Analysis (MOFA) algorithm 21 . The clustering model inferred latent factors that integrated multiple feature modalities, capturing variation across patient samples. The optimal factor combination was determined by maximizing the t-ratio, ensuring clear separation among clusters (Fig. S1A). Functional annotation of these factors revealed strong associations with metabolic and immune-related pathways (Fig. 1B). Mapping patients onto the best-explained factor space identified three distinct archetypes (Fig. 1C), which were subsequently characterized based on their association with specific SLC transporters, biological pathways, microenvironmental signatures, and molecular subtypes (Fig. 1D). Among these, archetype 1 exhibited a notable enrichment of SLC39 family members (SLC39A2, SLC39A4, SLC39A5, and SLC39A10 were ranked in top 5%, Fig. S1B), which correlated with the activation of zinc-related pathways. In line with prior studies implicating the SLC39 family in zinc import 22 , zinc influx transporters were more highly expressed than SLC30 family-mediated efflux counterparts (Fig. S1C). Patients within this archetype predominantly aligned with the CMS2 molecular subtype, characterized by WNT activation, microsatellite stability (MSS), and an immune-desert phenotype 23 . In line with these features, this archetype showed elevated activity in WNT- and MSS-associated pathways while displaying a depletion of key infiltrating immune cells, as inferred from transcriptomic enrichment analysis. These findings suggest that archetype 1 represents a CMS2-like subtype, distinguished by SLC39 overexpression and enhanced zinc transport. SLC39-mediated zinc transport induces CD24 overexpression via CDX2 enhancer binding in tumor cells To elucidate the molecular mechanisms underlying SLC39 family overexpression and zinc transport activation in archetype 1, we identified key signaling molecules associated with this pathway (Fig. 2A, top). Cancer driver genes relevant to archetype 1 were analyzed, revealing CDX2 as a top-ranked transcription factor (TF), along with the SLC39 family (Fig. S1D, E). CDX2 plays a critical role in gut development and homeostasis 24 . Notably, both CDX2 expression and inferred protein activity were positively correlated with SLC39-mediated zinc influx and archetype 1 status (Fig. S2A). Comprehensive multi-omics correlation analyses further identified CD24 as a downstream effector of CDX2, consistent with its reported function in mediating immune-repressive signaling 25 and aligning with the immune-desert characteristics of archetype 1. Across the TCGA and CPTAC CRC cohorts, expression of SLC39 transporters, CDX2, and CD24 demonstrated broad positive correlations with zinc pathway activity (Fig. 2A), supporting a zinc-driven transcriptional cascade linking SLC39 influx to CDX2-dependent CD24 activation. This regulatory cascade was further supported by COMPASS-based metabolic flux modeling 26 , which quantifies zinc-associated reaction-level flux states beyond conventional gene set scoring, and independently recapitulated the SLC39–CDX2–CD24 signaling axis (Fig. S2B). To investigate the genomic basis of SLC39 overexpression, we analyzed copy number variation (CNV) and identified a positive correlation between SLC39 expression and CNV in both TCGA and CPTAC cohorts (Fig. 2B). Additionally, CNV inference from single-cell transcriptomic data of 21,698 CRC cells 27 using the inferCNV algorithm 28 confirmed genomic amplification associated with SLC39 family overexpression. Given that CDX2 functions as transcriptional regulators, we estimated their regulatory activity on target genes using the VIPER algorithm 29 , based on a CRC regulon network. The TF activities of CDX2 exhibited strong positive correlations with upstream SLC39-mediated zinc influx and downstream CD24 expression in both TCGA and CPTAC cohorts (Fig. 2C). To further validate the regulatory link, we analyzed epigenomic features of the CD24 locus (Fig. 2D). Hi-C chromatin interaction data from the HCT116 CRC cell line 30 confirmed that the CD24 locus and a putative enhancer region 31 were co-localized within the same topologically associated domain (Fig. 2D, top). In the Caco-2 CRC cell line, active histone marks 32 such as H3K4me1, H3K36me3, and H3K27ac were enriched at the putative enhancer (Fig. 2D, middle). scATAC-seq data from 18,275 cells across three CRC patients 33 further revealed high chromatin accessibility at this enhancer region, specifically in cancer cells compared with other cell types (Fig. 2D, bottom). To assess enhancer activity in patient samples, we analyzed ATAC-seq data from 25 TCGA-COAD patients 9 and observed significant positive correlations between CD24 enhancer accessibility and upstream regulatory signals, including SLC39-mediated zinc influx, CDX2 TF activity, and CD24 expression (Fig. 2E). Furthermore, motif enrichment analysis confirmed significant binding of CDX2 at the putative enhancer region, supporting their direct role in CD24 activation (Fig. 2F). Given reports of zinc-induced phosphorylation of CDX2 modulating its transcriptional activity, we examined phosphoproteome datasets and observed positive trends linking zinc influx, CDX2 activity, CD24 expression, and phosphorylation at CDX2 S176, albeit without statistical significance due to limited sample size (Fig. 2G). Candidate regulatory kinases potentially mediating this phosphorylation were also enriched in zinc-associated networks (Fig. S2C). To corroborate these findings, we performed 3D protein–DNA docking using AlphaFold3-predicted complex structures, a deep learning model capable of high-accuracy biomolecular complex prediction 34 , and the HDOCK server, a hybrid docking platform supporting protein–DNA interactions with integrated scoring and modeling features 35 . Simulations confirmed high-confidence CDX2 binding to the CD24 enhancer, with S176 phosphorylation enhancing CDX2 binding affinity (Fig. 2H, I). Collectively, these findings define a zinc-dependent transcriptional circuit in which SLC39-driven zinc influx promotes CDX2 phosphorylation and enhancer-mediated activation of CD24 in CRC. CD24 overexpression driven by zinc influx via the SLC39 family is restricted to cancer cells in the tumor microenvironment Most signal cascades identified in this study were derived from bulk tissue data, which consist of heterogeneous cell populations. To resolve cell type-specific signaling, we analyzed the impact of zinc influx mediated by SLC39 overexpression at the single-cell level using a transcriptomic dataset comprising 174,547 single cells from 26 CRC patients 27 . Canonical cell types were annotated based on established marker genes (Fig. 3A and Fig. S3A), and the expression patterns of zinc influx-mediated SLC39 signaling leading to CD24 upregulation were examined across cell types (Fig. 3B, top). Correlation analyses among the signaling factors were then performed in a cell type-specific manner (Fig. 3B, bottom). These analyses revealed a strong and selective association of SLC39-driven zinc influx with CDX2 activity and CD24 expression in malignant epithelial cells, whereas non-malignant compartments exhibited minimal or marginal correlations, consistent with the limited resolution of scRNA-seq data. To address this limitation, we conducted metacell analysis using SEACells 36 , which mitigates scRNA-seq data sparsity while preserving cellular heterogeneity by defining compact and well-separated metacells (Fig. 3C and Fig. S3B). This approach enabled the detection of robust CD24 overexpression driven by zinc influx via SLC39, exclusively in cancer cells (Fig. 3D), accompanied by the most significant positive correlation in cancer cells compared to other cell types (Fig. 3E). To functionally validate these findings, we overexpressed SLC39A4 or SLC39A5, the two SLC39 family members most strongly associated with zinc influx in our multiomics analysis (Fig. 2A), in the CRC cell line LS174T (Fig. S3C). Both transporters enhanced intracellular zinc levels, as evidenced by FluoZin-3 fluorescence (Fig. 3F). Concomitantly, overexpression of either gene upregulated CD24 at transcript and protein levels (Fig. 3G, H) and increased membrane-localized CD24, confirmed by flow cytometry (Fig. 3I and Fig. S3D). These cells further displayed enhanced CDX2 transcriptional activity and increased nuclear localization (Fig. 3G and Fig. S3E–G). CD24 has been established as a potent “don’t eat me” signal analogous to CD47 and is broadly overexpressed across multiple tumor types 25 . However, comparative analysis demonstrated that CD24 but not CD47 was selectively enriched in cancer cells and exhibited significant positive correlations with SLC39-mediated zinc influx (Fig. S3H–J). Consistently, overexpression of SLC39A4 or SLC39A5 induced CD24 but not CD47 transcripts in CRC cells (Fig. 3J). Together, these findings delineate a zinc-driven regulatory mechanism in which SLC39 activity selectively couples to CD24 expression in cancer cells, establishing a distinct immune evasion pathway independent of the canonical CD47 axis. SLC39-mediated zinc influx directs cell state trajectory toward CMS2-like cancer cells associated with poor survival and immunotherapy resistance To investigate the influence of SLC39-mediated zinc influx on cancer cell state dynamics, we applied the Palantir algorithm 37 to single-cell transcriptomic data (Fig. S4A). CMS-based subtyping was performed to annotate the inferred cell state trajectories (Fig. 4A, left). Three terminal states were identified based on pseudotime and entropy (Fig. 4A, right), with terminal state 1 converging into a subset of CMS2-like cancer cells, while terminal states 2 and 3 remained heterogeneous or unclassified (Fig. S4B). Analysis of SLC39-driven zinc influx components revealed their selective overexpression or TF activation in terminal state 1 (Fig. 4B). Gene trend analysis further confirmed that signaling cascades downstream of SLC39-mediated zinc influx were specifically upregulated along the trajectory toward terminal state 1 (Fig. 4C). Notably, CD24 expression exhibited a delayed onset relative to upstream regulators, consistent with its function as a downstream effector of zinc-induced CDX2 activation. Pathway enrichment analysis of terminal state-specific expression signatures confirmed that terminal state 1 exhibited significant activation or repression of CMS2 subtype-associated pathways (Fig. 4D). Given the immunosuppressive nature of CMS2-like tumors identified in our archetype analysis (Fig. 1E), we next examined the immune-modulatory potential of cancer cells in terminal state 1. Notably, CD24 emerged as the top-ranked immunomodulator overexpressed in this state, reinforcing its role as a key effector of the SLC39-driven zinc influx axis (Fig. 4E and Fig. S4C). The canonical ‘don’t eat me’ signal CD47 was not significantly upregulated in terminal state 1. Extension of the single-cell–derived findings to patient cohorts was achieved by constructing a terminal state signature matrix using the CellRank algorithm 38 , which enabled calculation of terminal state probability scores from bulk transcriptomic datasets (Fig. 4F, left). Application of this model to 288 TCGA-COAD patient samples revealed that terminal state 1 probability scores were significantly enriched in archetype 1 patients, characterized by CMS2-like features and activated SLC39-mediated zinc influx signaling (Fig. 4F, right). In contrast, terminal states 2 and 3 probabilities were heterogeneously distributed across archetype 2 and 3 patients. Survival analysis demonstrated that patients with higher terminal state 1 probability scores exhibited significantly shorter progression-free and disease-free survival (Fig. 4G). This association was specific to terminal state 1, as no significant survival differences were observed for terminal states 2 or 3 (Fig. S4D). To investigate the relevance of this state to immunotherapy response, we analyzed single-cell transcriptomes from 70,718 cancer cells derived from CRC patients treated with PD-1 inhibitors 39 (Fig. 4H, I). Intriguingly, overexpression or activation of components within the SLC39-driven zinc influx pathway was consistently associated with non-response to therapy and shorter progression-free survival. To elucidate the regulatory landscape underlying these clinical associations, we constructed patient group-specific cancer cell networks using scHumanNet 40 stratified by clinical response or progression-free survival status. Mapping of terminal state marker genes and core components of the SLC39-driven zinc influx pathway onto these networks revealed significant enrichment of terminal state 1 markers and zinc influx components within non-responder and short-survivor networks (Fig. 4J, K). Conversely, terminal state 2 and 3 markers were predominantly enriched in responder and long-survivor networks. This distribution pattern remained consistent regardless of the number of marker genes included in the analysis (Fig. S4E). Systematic importance evaluation using centrality metrics further demonstrated that terminal state 1 markers and SLC39-driven zinc influx components exhibited higher importance scores specifically in non-responder and short-survivor networks compared to other marker genes (Fig. 4L). This pattern was robust to variations in marker gene selection criteria (Fig. S4F). Collectively, these findings demonstrate that the SLC39-driven zinc influx axis drives a CMS2-like cancer cell state marked by CD24 overexpression, which is clinically associated with poor prognosis and resistance to immunotherapy. CD24 mediates pleiotropic immune evasion signaling in the tumor microenvironment through SIGLEC10 interactions CMS2-like cancer cells of archetype 1 exhibit pronounced immune-desert characteristics within the tumor microenvironment (TME) (Fig. 1E), prompting investigation of how CD24-overexpressing tumor cells communicate with surrounding immune populations. We sought to define the cellular interface of CD24-overexpressing tumor cells by systematically mapping receptor–ligand interactions within the TME. Leveraging a curated ligand–receptor database 41 and applying the NICHES framework 42 , we prioritized the top five candidate interactions, with cancer cells designated as the sender population (Fig. 5A and Fig. S5A). Among these, CD24 emerged as a tumor-selective ligand predicted to interact exclusively with SIGLEC10 receptors expressed on myeloid cells, suggesting a specialized axis of immune suppression mediated through CD24–SIGLEC10 engagement. Chromatin accessibility analysis further revealed that SIGLEC10 promoter were epigenetically active exclusively in myeloid populations within CRC patient samples 33 (Fig. S5B). To pinpoint the specific myeloid subpopulations expressing SIGLEC10, we performed subclustering using lineage-specific markers (Fig. S5C). This analysis revealed significant overexpression of SIGLEC10 in monocytes, resident tissue macrophages (RTMCs), monocyte-derived macrophages (moMCs), and monocyte-derived dendritic cells (moDCs) compared to other myeloid subsets (Fig. 5B). These myeloid-lineage cells originate from distinct progenitors: RTMCs derive from hematopoietic stem cells, while monocytes, moMCs, and moDCs derive from erythro-myeloid progenitors (Fig. 5C). Cell state trajectory analysis using the Palantir algorithm 37 revealed bifurcating developmental paths from monocytes toward moMC and moDC lineages (Fig. 5D and Fig. S5D). Gene trend profiling along these trajectories showed progressive upregulation of SIGLEC10, confirming their lineage development through stage-specific marker expression (Fig. 5E). To elucidate the downstream signaling cascades triggered by CD24-SIGLEC10 interactions, we employed the NicheNet algorithm 43 , integrating ligand-receptor interactions, signaling pathways, and transcriptional regulatory networks. In moMCs, CD24 exhibited high ligand activity and selectively engaged SIGLEC10 (Fig. 5F, left), leading to the activation of downstream transcriptional programs exclusively expressed in moMCs (Fig. 5F, right). Functional enrichment analysis of these target genes revealed strong associations with negative regulation of macrophage activation, phagocytosis, endocytosis, and immune responses—hallmarks of the ‘don’t eat me’ signal (Fig. 5G). Similarly, moDCs exhibited a parallel immune-repressive signaling trend, though the specific downstream gene sets differed from those in moMCs (Fig. S5E, F). Beyond moMCs and moDCs, RTMCs also interacted with cancer cells via CD24-SIGLEC10, but their downstream signaling profile diverged, predominantly influencing cell adhesion and chemotaxis pathways (Fig. 5H, I). These cell–cell interaction profiles collectively reveal CD24 overexpression in cancer cells elicits pleiotropic responses via SIGLEC10 binding across moDCs, moMCs, and RTMCs, orchestrating immunosuppressive programs involving immune evasion and adhesion remodeling. Recurrent CD24–SIGLEC10-enriched niches define immunorepressive microenvironments in colorectal cancer To evaluate whether the pleiotropic cellular interactions delineated in our analysis are spatially organized within tumor tissues, we analyzed spatial transcriptomes from four CRC patient samples 44 . Cell type deconvolution was performed using SPOTlight 45 with canonical marker genes (Fig. 6A). Spatial domains enriched for CD24–SIGLEC10 interactions were identified using SPATA2 11 , enabling the delineation of pleiotropic signaling regions (Fig. 6B). Notably, cancer cells, moMCs, moDCs, and RTMCs—key cell types engaged in CD24–SIGLEC10 signaling—were consistently enriched at the tumor periphery near these interaction hotspots (Fig. 6A, B, bottom). To investigate the functional relevance of this cellular ecosystem, we quantified the activity of core components in the SLC39-mediated zinc influx axis relative to spatial proximity to CD24–SIGLEC10-enriched areas. All signaling components demonstrated peak activity adjacent to these regions, with progressive attenuation observed at increasing distances (Fig. S6A, B). Consistent spatial gradients were also observed for immune evasion and cell adhesion pathways, with maximal activation near the interaction zones and gradual decline further away (Fig. 6C, D), providing functional validation of these spatially defined immunoregulatory microenvironments. Based on these observations, we characterized a cellular niche representing an immunosuppressive microenvironmental ecosystem mediated by CD24-SIGLEC10 pleiotropic interactions, encompassing signaling cascades of CD24 overexpression in cancer cells (Fig. 6E, bottom, red background box), CD24-SIGLEC10 interactions (Fig. 6E, bottom, blue background box), and downstream pleiotropic signals regulating immune evasion and cell adhesion (Fig. 6E, bottom, green background box). Using the FindClusters function of Seurat 46 , we identified between 10 and 14 distinct cellular niches across the four CRC tissue samples (Fig. 6E and Fig. S6C). Importantly, two of the four tissue samples exhibited the immunosuppressive cellular niches we characterized, with concordant activation patterns of signaling cascades mediated by CD24-SIGLEC10 interactions in tissue peripheral areas (Fig. 6E, top). In contrast to these niches, the CD47-SIRPA interaction—a putative major ‘don't eat me’ signal—showed minimal or only partial activation in these regions (Fig. S6D). Collectively, these spatial transcriptomic analyses confirmed the recurrent presence of cellular niches characterized by immune-suppressive pleiotropic signaling via CD24-SIGLEC10 interactions across CRC patient tissues, providing spatial context to our molecular findings. Conserved pleiotropic CD24–SIGLEC10 signaling defines immune-evasive myeloid networks associated with immunotherapy resistance We investigated whether pleiotropic immune-evasive signaling mediated by CD24–SIGLEC10 interactions is preserved at the patient level. To this end, we derived cell type-specific gene signatures from the top 50 marker genes of moMCs, moDCs, and RTMCs identified in the single-cell transcriptomic datasets 47 . These gene sets were mutually exclusive and independent of the downstream CD24–SIGLEC10 signaling modules (Fig. S7A). Using these signatures, we computed enrichment scores for moMCs and RTMCs across TCGA-COAD samples and stratified patients into subgroups based on combinatorial high (+) or low (–) enrichment status (Fig. 7A). The potential for intercellular CD24–SIGLEC10 interactions was assessed using the LIANA module from the DecoupleR framework 48 , which quantifies ligand-receptor co-expression relative to background distributions. Patients exhibiting concurrent high enrichment of both moMCs and RTMCs demonstrated the strongest predicted CD24–SIGLEC10 interaction potential (Fig. 7B). To assess functional consequences of these interactions, we analyzed gene modules associated with CD24–SIGLEC10 activity, encompassing immune-evasive ‘don’t eat me’ signals and adhesion-regulatory pathways. These modules showed robust positive correlations with moMC and RTMC enrichment scores, despite minimal gene overlap with the respective marker sets (Fig. 7C and Fig. S7A). This association remained consistent when moDCs were substituted for moMCs (Fig. S7B–D). Additionally, the immune-suppressive cellular landscape identified in TCGA patients was independently recapitulated in the CPTAC CRC cohort (Fig. S7E), underscoring the robustness and translational significance of these findings. To mechanistically recapitulate these observations, we employed THP-1 cells differentiated into moMC-like or RTMC-like lineages using distinct polarization protocols (Fig. 7D). Each protocol selectively enhanced marker expression of the corresponding myeloid subtype. To test whether CD24–SIGLEC10 engagement suppresses phagocytic activity, LS174T CRC cells were co-cultured with the differentiated moMC-like cells, and phagocytosis was quantified (Fig. 7E). SLC39A4- or SLC39A5-overexpressing CRC cells exhibited significantly reduced phagocytosis compared with control cells, an effect reversed by anti-CD24 antibody treatment, confirming that the suppression occurred through CD24–SIGLEC10 interaction. Motivated by these mechanistic insights, we hypothesized that expression of key components within the CD24–SIGLEC10 axis could stratify immunotherapy response. To test this, we integrated five independent ICI-treated CRC cohorts (n = 166) encompassing both MSS and MSI cases and applied batch correction to harmonize expression profiles (Fig. 7F and Fig. S7F). A Gradient Boosting Classifier was trained on a biologically informed seven-gene signature—SLC39A2, SLC39A4, SLC39A5, SLC39A10, CDX2, CD24, and SIGLEC10—selected for their central roles in zinc-driven immunosuppression. This model achieved the highest predictive accuracy among all benchmarked classifiers, with an AUC of 0.728 in an independent testing set, outperforming established biomarkers of adaptive immunity, stromal resistance, and tumor-intrinsic evasion 12–19 (Fig. 7G). Together, these findings define a conserved, pleiotropic immunosuppressive circuit orchestrated by SLC39-driven zinc influx and CD24–SIGLEC10 engagement. This pathway shapes an immune-evasive tumor microenvironment, suppresses macrophage phagocytosis, and serves as a clinically actionable predictor of ICI response in colorectal cancer. Discussion Our study delineates a previously uncharacterized pleiotropic signaling axis initiated by SLC39-mediated zinc influx, which orchestrates immune-suppressive reprogramming in CRC. Through SLC transporter-centered multiomics clustering, we identified a distinct patient subgroup aligned with the CMS2 molecular subtype, marked by WNT activation, microsatellite stability, and an immune-desert phenotype. This CMS2-like archetype was characterized by overexpression of the SLC39 family, particularly SLC39A4 and 5, which drives zinc influx and triggers a transcriptional cascade culminating in epigenetic activation of CD24—a recently recognized “don’t eat me” signal analogous to CD47. We uncovered that zinc functions as a cofactor enhancing the binding of the CDX2 transcription factor to a CD24 regulatory enhancer, promoting its overexpression, consistent with prior evidence that zinc activates CDX2 phosphorylation and transcriptional activity via the PI3K–CDX2 pathway to regulate downstream targets in intestinal epithelial cells 49 . Multi-omics correlation, chromatin accessibility profiling, enhancer activity assessment, and 3D protein–DNA docking collectively established the direct transcriptional control of CD24 by this zinc-responsive TF. Functionally, CRC cells overexpressing SLC39A4 or SLC39A5 exhibited increased intracellular zinc levels, confirming their zinc transport capacity, and upregulation of CDX2 target genes including CD24. Elevated transcript levels corresponded to increased protein abundance and membrane localization of CD24, indicating that zinc-mediated transcriptional regulation extends to cell–cell communication through extracellular binding partners. Notably, this zinc-dependent overexpression of CD24 occurred independently of CD47, defining a discrete immunoevasive pathway within CMS2-like tumors. At the patient level, the SLC39-driven zinc influx state was robustly recapitulated and associated with significantly reduced progression-free and disease-free survival across over 250 TCGA-COAD patients, as well as resistance to ICIs in 26 CRC patients, where elevated expression of “don’t eat me” signals such as CD24 or CD47 has been reported to correlate with diminished ICI efficacy and primary resistance in colorectal and head and neck cancer 50 , 51 . Extending our analysis to the tumor microenvironment, we found that CD24 overexpression in cancer cells facilitates immune-evasive communication via SIGLEC10-expressing myeloid cells. Interestingly, this interaction exhibited lineage-specific pleiotropy: engagement of SIGLEC10 on moMCs and moDCs led to repression of phagocytosis-related genes, while signaling in RTMCs favored adhesion-related pathways. Spatial and single-cell transcriptomic analyses validated that these pleiotropic CD24–SIGLEC10 interactions define recurrent immunosuppressive niches at tumor borders, where cancer–myeloid cell contact is spatially enriched. Furthermore, CD24–SIGLEC10 interaction strength and downstream signaling modules correlated with non-responsiveness to ICI therapy. Consistently, phagocytosis assays employing moMC-like macrophages and CRC cells demonstrated that overexpression of SLC39A4 or SLC39A5 suppressed macrophage-mediated engulfment, an effect reversed by CD24–SIGLEC10 blockade. To translate these mechanistic findings into clinically applicable framework, we developed an ICI response prediction model integrating multi-cohort transcriptomic data. The model, trained on a seven-gene SLC39–CD24–SIGLEC10 signature, achieved the highest predictive performance (AUC = 0.728) in the integrated ICI-treated CRC cohort, establishing the zinc-driven immunosuppressive program as a robust molecular predictor of therapeutic response. Although clinical translation of this axis is still emerging, anti-SIGLEC10 antibodies such as ONC-841 have recently entered early-phase clinical trials 52 , highlighting the druggability of this checkpoint pathway, whereas CD24-targeting antibodies (e.g., IMM47) have demonstrated preclinical efficacy but remain at the investigational stage 53 . These results confirm that SLC39-mediated zinc influx establishes a therapeutically actionable axis linking metabolic reprogramming to innate immune suppression. Together, our findings identify SLC39-driven zinc influx as a central regulator of immune-suppressive plasticity in CMS2-like CRC and uncover a multimodal immune evasion mechanism orchestrated through pleiotropic CD24–SIGLEC10 signaling. SLC transporters have been broadly implicated in tumor metabolism and microenvironmental remodeling 54 , yet their roles in CRC remain incompletely defined, particularly in the context of immune-desert or cold tumor states. Here, we delineate a mechanistic axis in which SLC39-driven zinc influx orchestrates an immunosuppressive program characteristic of the CMS2-like molecular subtype. This signaling cascade links metabolic reprogramming to immune evasion, establishing zinc-dependent transcriptional control as a critical determinant of tumor–immune interactions. Within this framework, we identify CD24–SIGLEC10 as the dominant anti-phagocytic checkpoint in CMS2-like CRC. Although CD47–SIRPα signaling is a canonical “don’t eat me” pathway frequently upregulated in CRC and known to potentiate adaptive immune priming upon blockade 55 , 56 , our data show that SLC39-mediated zinc influx preferentially promotes CDX2-dependent CD24 transcription without concomitant CD47 upregulation. This nominates CD24–SIGLEC10 as the principal innate immune checkpoint in this setting, reinforcing its relevance as a therapeutic target 57 . Importantly, our integrative analyses reveal that CD24–SIGLEC10 interactions extend beyond tumor-associated macrophages (TAMs), engaging multiple myeloid subsets to enforce pleiotropic immune suppression. Among these, SPP1 + moMCs sculpt hypoxic, fibroblast-enriched niches associated with poor prognosis 58 , 59 , while CCR7 + LAMP3 + moDCs acquire mregDC programs that dampen T-cell priming 60 . RTMCs further contribute by activating adhesion and hypoxia-adaptive pathways. Collectively, these lineage-specific programs demonstrate that zinc-dependent CD24 upregulation in tumor cells enables SIGLEC10 engagement across multiple myeloid compartments, thereby extending the canonical CD24–SIGLEC10 immune checkpoint into a multicellular suppressive circuit. While our study delineates the mechanistic underpinnings of SLC39-driven zinc influx and CD24-mediated immune evasion in CMS2-like CRC, it also reveals distinct regulatory architectures underpinning other archetypes. Archetypes 2 and 3, associated with the SLC22 and SLC35 families, respectively, exhibit immune and proliferative traits that are not attributable to zinc signaling, yet converge on unique CMS subtypes. As shown in Fig. 1E, archetype 2 aligns with the CMS4 subtype, characterized by epithelial–mesenchymal transition (EMT), angiogenesis, and relative repression of cell-cycle pathways. Notably, this subtype is marked by epigenetic silencing of SLC22 transporters such as SLC22A18 and SLC22A5, whose promoter hypermethylation enhances tumor cell invasion and survival 61 , 62 . In contrast, archetype 3 recapitulates features of CMS1 tumors, which harbor microsatellite instability (MSI) and mismatch repair (MMR) deficiency. These tumors paradoxically couple defective DNA repair with proliferative advantage due to impaired TGF-β signaling 63 , 64 . Within this context, aberrant methylation and suppression of SLC35A3—a nucleotide-sugar transporter—has been implicated in compromised DNA repair fidelity and attenuated immune infiltration 65 . Together, these findings suggest that CRC subtypes deploy divergent transporter circuits: SLC22 silencing supports mesenchymal transition in CMS4, SLC35A3 repression augments immune dysfunction in CMS1, and SLC39-mediated zinc signaling orchestrates immune escape in CMS2. This underscores the metabolic plasticity that stratifies consensus molecular subtypes and reveals context-dependent vulnerabilities across CRC archetypes 66 . In acknowledging the limitation that our study did not employ mass spectrometry–based metabolic flux assays to directly quantify metabolite exchange, we sought to overcome this gap by leveraging multi-omics integration, pathway-level analyses, and reaction-level flux modeling. Recent high-impact studies have demonstrated that transcriptomic and proteogenomic profiling can robustly infer tumor metabolic rewiring, such as glycolytic activation in MSI-high colorectal tumors 67 and oxidative phosphorylation imbalance in clear cell renal cell carcinoma 68 . Similarly, integrative omics of thyroid cancers revealed subtype-specific TCA cycle and one-carbon metabolism programs with elevated transporters and enzymes 69 , while proteogenomic stratification of NSCLC uncovered a distinct metabolic subtype with heightened glycolytic and bioenergetic activity 70 . Moreover, spatial multi-omics in gastric cancer has shown that correlating metabolite distributions with gene expression can resolve cell-type–specific metabolic crosstalk in the tumor microenvironment 71 . Beyond conventional enrichment-based methods, the metabolic flux modeling 26 applied in this study has been validated by its ability to recapitulate lipid-induced shifts toward serine–methionine metabolism in breast epithelial subpopulations, revealing redox-coupled metabolic plasticity that mirrors cancer-associated reprogramming dynamics 72 . Together, these references underscore that although direct metabolite tracing was not feasible, our use of multi-omics data and pathway inference represents a rigorous and widely validated strategy to dissect tumor metabolic states. In bulk transcriptomic analyses of CRC, the predominance of malignant epithelial cells often masks immune-derived signals, limiting the sensitivity of projecting moDC, moMC, or RTMC signatures for prognostic stratification or subtype classification. The high tumor purity observed in TCGA and similar cohorts skew expression profiles toward tumor-intrinsic programs, which tend to show stronger correlations with clinical outcomes than immune-related signatures 73 , 74 . Notably, molecular subtypes initially interpreted as mesenchymal tumor cell states were later attributed predominantly to stromal and immune cell expression, revealing how shifts in cellular composition can confound transcriptomic interpretation 75 . These observations underscore a central limitation of our approach—namely, that the weaker associations observed for immune cell signatures in bulk CRC datasets likely stem from tumor cell dominance rather than a lack of biological relevance. This study provides the most comprehensive evidence to date linking metabolic reprogramming via SLC transporters to tumor microenvironmental phenotypes, including canonical molecular subtypes in colorectal cancer. Specifically, we identify a CMS2-like patient subgroup in which SLC39-mediated zinc influx drives CD24 overexpression through epigenetic enhancer activation in cancer cells. This zinc-dependent signaling cascade establishes pleiotropic cell–cell interactions with myeloid-lineage cells—moDCs, moMCs, and RTMCs—leading to immune-evasive ‘don’t eat me’ signaling and adhesion remodeling. Notably, this immunosuppressive mechanism is independent of the canonical CD47 axis and is enriched in CMS2-like tumors, frequently associated with SLC39 copy number gain. To extend these findings toward clinical translation, we developed an ICI response prediction model integrating multi-cohort transcriptomic data. The model, trained on a seven-gene zinc–CD24–SIGLEC10 signature, achieved superior accuracy in predicting therapeutic response, highlighting the clinical relevance of this zinc-driven immunosuppressive program. Our findings suggest that therapeutic strategies targeting the SLC39–zinc influx–CD24 axis, including blockade of CD24–SIGLEC10 interactions, may reverse immune exclusion in this CRC subtype. Given the lineage-specific engagement of SIGLEC10 + myeloid populations in CMS2-like tumors, engineering CAR-based interventions targeting these myeloid subsets represents a promising direction for precision immunotherapy. Conclussion In conclusion, we identify SLC39 driven zinc influx as a central metabolic trigger that establishes an immune suppressive state in colorectal cancer. Zinc import enhances CDX2 dependent activation of a CD24 enhancer, driving cancer cell specific CD24 overexpression and engagement of SIGLEC10 positive myeloid cells. This CD24–SIGLEC10 axis suppresses phagocytosis, forms recurrent immune evasive niches and defines a CMS2 like tumor state associated with poor prognosis and resistance to immune checkpoint blockade. Functionally, zinc induced CD24 signaling reduces macrophage phagocytosis and can be reversed by CD24 blockade. Clinically, a seven gene SLC39–CD24–SIGLEC10 signature outperforms established biomarkers in predicting immunotherapy response, highlighting this zinc driven pathway as a therapeutically actionable target in CRC. Abbreviations AUC: Area under the receiver operating characteristic curve AUPR: Area under the precision-recall curve cDNA: Complementary DNA CMS: Consensus Molecular Subtypes CNV: Copy number variation COAD: Colorectal adenocarcinoma CPTAC: Clinical Proteomic Tumor Analysis Consortium CRC: Colorectal cancer CRISPRa: CRISPR activation DFI: Disease-free interval DMEM: Dulbecco’s Modified Eagle Medium EMT: Epithelial–mesenchymal transition FBS: Fetal bovine serum GO: Gene Ontology GSVA: Gene Set Variation Analysis ICI: Immune checkpoint inhibitor IF: Immunofluorescence IFN-γ: Interferon gamma KEGG: Kyoto Encyclopedia of Genes and Genomes LPS: Lipopolysaccharide MFI: Mean fluorescence intensity MMR: Mismatch repair moDCs: Monocyte-derived dendritic cells MOFA: Multi-Omics Factor Analysis moMCs: Monocyte-derived macrophages mregDC: Myeloid regulatory dendritic cell MSI: Microsatellite instability MSigDB: Molecular Signatures Database MSS: Microsatellite stable PCA: Principal component analysis PFI: Progression-free interval PFS: Progression-free survival PMA: Phorbol 12-myristate 13-acetate PWMs: Position weight matrices RPMI: Roswell Park Memorial Institute RTMCs: Resident tissue macrophages SAM: Synergistic Activation Mediator sgRNAs: Single-guide RNAs SLC: Solute carrier TAD: Topologically associating domain TAMs: Tumor-associated macrophages TCGA: The Cancer Genome Atlas TME: Tumor microenvironment UMAP: Uniform Manifold Approximation and Projection WB: Western blot Declarations Data availability Previously published datasets analyzed in this study are available under their respective repositories and accession codes as follows: TCGA colorectal cancer multi-omics (UCSC Xena, “TCGA Pan-Cancer (PANCAN)” cohort); CPTAC colorectal proteogenomics (LinkedOmics CPTAC-COAD); colorectal scRNA-seq (Synapse, accession syn26844071); ICI-treated CRC scRNA-seq (Broad Single Cell Portal, study SCP2079); ICI-treated CRC bulk RNA-seq (NCBI GEO, accession GSE179351, GSE235919, GSE302922); and CRC spatial transcriptomics (NCBI GEO, accession GSE225857). All processed data generated in this study that support the findings are deposited in Zenodo (https://doi.org/10.5281/zenodo.17292085). Additional information is available from the corresponding author upon reasonable request. Acknowledgements Not applicable. Funding This work was supported by the National Research Foundation of Korea (NRF) grant funded by the Korea government (MSIT) (RS-2022-NR072092, RS-2023-00214527, RS-2025-23525624, and RS-2025-02304837), Hyundai Motor Chung Mong-Koo Foundation. Author contributions S.J.H. and S.P performed data analyses and wrote the manuscript. S.K. and Y.I.P. participated in data analyses. J.L., M.K., and M.H. reviewed the manuscript. S.E.K. and K.K. conceived and supervised the study. Ethics declarations Not applicable. References Schlessinger, A., Zatorski, N., Hutchinson, K. & Colas, C. Targeting SLC transporters: small molecules as modulators and therapeutic opportunities. Trends in Biochemical Sciences vol. 48 801–814 Preprint at https://doi.org/10.1016/j.tibs.2023.05.011 (2023). Nwosu, Z. C., Song, M. G., di Magliano, M. P., Lyssiotis, C. A. & Kim, S. E. Nutrient transporters: connecting cancer metabolism to therapeutic opportunities. Oncogene vol. 42 711–724 Preprint at https://doi.org/10.1038/s41388-023-02593-x (2023). Yan, Y. et al. SLC7A11 expression level dictates differential responses to oxidative stress in cancer cells. Nat Commun 14 , (2023). Papalazarou, V. et al. Phenotypic profiling of solute carriers characterizes serine transport in cancer. Nat Metab 5 , 2148–2168 (2023). Zheng, Y. et al. Metabolic gatekeepers: harnessing tumor-derived metabolites to optimize T cell-based immunotherapy efficacy in the tumor microenvironment. Cell Death and Disease vol. 15 Preprint at https://doi.org/10.1038/s41419-024-07122-6 (2024). Stine, Z. E., Schug, Z. T., Salvino, J. M. & Dang, C. V. Targeting cancer metabolism in the era of precision oncology. Nature Reviews Drug Discovery vol. 21 141–162 Preprint at https://doi.org/10.1038/s41573-021-00339-6 (2022). Chowdhury, S. et al. Consensus molecular subtyping of metastatic colorectal cancer expands biomarker-directed therapeutic benefit for patients with CMS1 and CMS2 tumors. Br J Cancer 131 , 1328–1339 (2024). Green, C., Roccia, P. & Rufini, A. Making sense of human colorectal cancer molecular subtypes: mice are stepping in. Cell Death Discovery vol. 11 Preprint at https://doi.org/10.1038/s41420-025-02594-7 (2025). Corces, M. R. et al. The chromatin accessibility landscape of primary human cancers. Science (1979) 362 , (2018). Castro-Mondragon, J. A. et al. JASPAR 2022: The 9th release of the open-access database of transcription factor binding profiles. Nucleic Acids Res 50 , D165–D173 (2022). Kueckelhaus, J. et al. Inferring histology-associated gene expression gradients in spatial transcriptomic studies. Nat Commun 15 , (2024). Chen, D. et al. Pan-cancer analysis implicates novel insights of lactate metabolism into immunotherapy response prediction and survival prognostication. Journal of Experimental and Clinical Cancer Research 43 , (2024). Bulen, B. J. et al. Validation of Immunotherapy Response Score as Predictive of Pan-solid Tumor Anti-PD-1/PD-L1 Benefit. Cancer Research Communications 3 , 1335–1349 (2023). Zhang, W. et al. Pan-cancer evaluation of regulated cell death to predict overall survival and immune checkpoint inhibitor response. NPJ Precis Oncol 8 , (2024). Chida, K. et al. Transcriptomic Profiling of MSI-H/dMMR Gastrointestinal Tumors to Identify Determinants of Responsiveness to Anti–PD-1 Therapy. Clinical Cancer Research 28 , 2110–2117 (2022). Antoniotti, C. et al. An Immune-Related Gene Expression Signature Predicts Benefit from Adding Atezolizumab to FOLFOXIRI plus Bevacizumab in Metastatic Colorectal Cancer. Clinical Cancer Research 29 , 2291–2298 (2023). Ayers, M. et al. IFN-γ-related mRNA profile predicts clinical response to PD-1 blockade. Journal of Clinical Investigation 127 , 2930–2940 (2017). Hugo, W. et al. Genomic and Transcriptomic Features of Response to Anti-PD-1 Therapy in Metastatic Melanoma. Cell 165 , 35–44 (2016). Cristescu, R. et al. Pan-tumor genomic biomarkers for PD-1 checkpoint blockade-based immunotherapy. Science (1979) 362 , (2018). Chen, R. & Chen, L. Solute carrier transporters: emerging central players in tumour immunotherapy. Trends in Cell Biology vol. 32 186–201 Preprint at https://doi.org/10.1016/j.tcb.2021.08.002 (2022). Argelaguet, R. et al. MOFA+: A statistical framework for comprehensive integration of multi-modal single-cell data. Genome Biol 21 , (2020). Jeong, J. & Eide, D. J. The SLC39 family of zinc transporters. Molecular Aspects of Medicine vol. 34 612–619 Preprint at https://doi.org/10.1016/j.mam.2012.05.011 (2013). Ciardiello, D. et al. Immunotherapy of colorectal cancer: Challenges for therapeutic efficacy. Cancer Treatment Reviews vol. 76 22–32 Preprint at https://doi.org/10.1016/j.ctrv.2019.04.003 (2019). Gao, N., White, P. & Kaestner, K. H. Establishment of Intestinal Identity and Epithelial-Mesenchymal Signaling by Cdx2. Dev Cell 16 , 588–599 (2009). Barkal, A. A. et al. CD24 signalling through macrophage Siglec-10 is a target for cancer immunotherapy. Nature 572 , 392–396 (2019). Wagner, A. et al. Metabolic modeling of single Th17 cells reveals regulators of autoimmunity. Cell 184 , 4168-4185.e21 (2021). Joanito, I. et al. Single-cell and bulk transcriptome sequencing identifies two epithelial tumor cell states and refines the consensus molecular classification of colorectal cancer. Nat Genet 54 , 963–975 (2022). Patel, A. P. et al. Single-cell RNA-seq highlights intratumoral heterogeneity in primary glioblastoma. Science (1979) 344 , 1396–1401 (2014). Alvarez, M. J. et al. Functional characterization of somatic mutations in cancer using network-based inference of protein activity. Nat Genet 48 , 838–847 (2016). Wang, Y. et al. The 3D Genome Browser: A web-based browser for visualizing 3D genome organization and long-range chromatin interactions. Genome Biol 19 , (2018). Fishilevich, S. et al. GeneHancer: Genome-wide integration of enhancers and target genes in GeneCards. Database 2017 , (2017). Abascal, F. et al. Expanded encyclopaedias of DNA elements in the human and mouse genomes. Nature 583 , 699–710 (2020). Becker, W. R. et al. Single-cell analyses define a continuum of cell state and composition changes in the malignant transformation of polyps to colorectal cancer. Nat Genet 54 , 985–995 (2022). Abramson, J. et al. Accurate structure prediction of biomolecular interactions with AlphaFold 3. Nature 630 , 493–500 (2024). Yan, Y., Tao, H., He, J. & Huang, S. Y. The HDOCK server for integrated protein–protein docking. Nat Protoc 15 , 1829–1852 (2020). Persad, S. et al. SEACells infers transcriptional and epigenomic cellular states from single-cell genomics data. Nat Biotechnol 41 , 1746–1757 (2023). Setty, M. et al. Characterization of cell fate probabilities in single-cell data with Palantir. Nat Biotechnol 37 , 451–460 (2019). Lange, M. et al. CellRank for directed single-cell fate mapping. Nat Methods 19 , 159–170 (2022). Tian, J. et al. Combined PD-1, BRAF and MEK inhibition in BRAFV600E colorectal cancer: a phase 2 trial. Nat Med 29 , 458–466 (2023). Cha, J., Yu, J., Cho, J. W., Hemberg, M. & Lee, I. scHumanNet: a single-cell network analysis platform for the study of cell-type specificity of disease genes. Nucleic Acids Res 51 , E8 (2023). Dimitrov, D. et al. LIANA+ provides an all-in-one framework for cell–cell communication inference. Nat Cell Biol 26 , 1613–1622 (2024). Raredon, M. S. B. et al. Comprehensive visualization of cell–cell interactions in single-cell and spatial transcriptomics with NICHES. Bioinformatics 39 , (2023). Browaeys, R., Saelens, W. & Saeys, Y. NicheNet: modeling intercellular communication by linking ligands to target genes. Nat Methods 17 , 159–162 (2020). Wang, F. et al. Single-Cell and Spatial Transcriptome Analysis Reveals the Cellular Heterogeneity of Liver Metastatic Colorectal Cancer . (2023). Elosua-Bayes, M., Nieto, P., Mereu, E., Gut, I. & Heyn, H. SPOTlight: Seeded NMF regression to deconvolute spatial transcriptomics spots with single-cell transcriptomes. Nucleic Acids Res 49 , E50 (2021). Butler, A., Hoffman, P., Smibert, P., Papalexi, E. & Satija, R. Integrating single-cell transcriptomic data across different conditions, technologies, and species. Nat Biotechnol 36 , 411–420 (2018). Dunsmore, G. et al. C A N C E R I M M U N O L O G Y Timing and Location Dictate Monocyte Fate and Their Transition to Tumor-Associated Macrophages . https://www.science.org (2024). Badia-I-Mompel, P. et al. decoupleR: ensemble of computational methods to infer biological activities from omics data. Bioinformatics Advances 2 , (2022). Ramavath, H. N. et al. Zinc induces hephaestin expression via a PI3K-CDX2 dependent mechanism to regulate iron transport in intestinal Caco-2 cells. Biochem Biophys Res Commun 626 , 1–7 (2022). Yang, W. et al. IL-4-STAT6-induced high Siglec-G/10 expression aggravates the severe immune suppressive tumor microenvironment and impedes the efficacy of immunotherapy in head and neck squamous cell carcinoma. J Immunother Cancer 13 , (2025). Chen, S. H. et al. Dual checkpoint blockade of CD47 and PD-L1 using an affinity-tuned bispecific antibody maximizes antitumor immunity. J Immunother Cancer 9 , (2021). Sobczak, K. et al. The unique molecular recognition features of Siglec-10: structural insights into sialoglycan and antibody interactions. Preprint at https://doi.org/10.1101/2025.06.10.658867 (2025). Li, S. et al. IMM47, a humanized monoclonal antibody that targets CD24, exhibits exceptional anti-Tumor efficacy by blocking the CD24/Siglec-10 interaction and can be used as monotherapy or in combination with anti-PD1 antibodies for cancer immunotherapy. Antib Ther 6 , 240–252 (2023). Papalazarou, V. et al. Phenotypic profiling of solute carriers characterizes serine transport in cancer. Nat Metab 5 , 2148–2168 (2023). Tseng, D. et al. Anti-CD47 antibody-mediated phagocytosis of cancer by macrophages primes an effective antitumor T-cell response. Proc Natl Acad Sci U S A 110 , 11103–11108 (2013). Hu, T. et al. Tumor-intrinsic CD47 signal regulates glycolysis and promotes colorectal cancer cell growth and metastasis. Theranostics 10 , 4056–4072 (2020). Barkal, A. A. et al. CD24 signalling through macrophage Siglec-10 is a target for cancer immunotherapy. Nature 572 , 392–396 (2019). Sathe, A. et al. Colorectal Cancer Metastases in the Liver Establish Immunosuppressive Spatial Networking between Tumor-Associated SPP1þ Macrophages and Fibroblasts. Clinical Cancer Research 29 , 244–260 (2023). Matusiak, M. et al. Spatially Segregated Macrophage Populations Predict Distinct Outcomes in Colon Cancer. Cancer Discov 14 , 1418–1439 (2024). Maier, B. et al. A conserved dendritic-cell regulatory program limits antitumour immunity. Nature 580 , 257–262 (2020). Song, H. S., Ha, S. Y., Kim, J. Y., Kim, M. & Choi, J. H. The effect of genetic variants of SLC22A18 on proliferation, migration, and invasion of colon cancer cells. Sci Rep 14 , (2024). Scalise, M. et al. Human OCTN2 (SLC22A5) is down-regulated in virus- and nonvirus-mediated cancer. Cell Biochem Funct 30 , 419–425 (2012). Biswas, S. et al. Mutational inactivation of TGFBR2 in microsatellite unstable colon cancer arises from the cooperation of genomic instability and the clonal outgrowth of transforming growth factor β resistant cells. Genes Chromosomes Cancer 47 , 95–106 (2008). Grady, W. M. et al. Proliferation and Cdk4 expression in microsatellite unstable colon cancers with TGFBR2 mutations. Int J Cancer 118 , 600–608 (2006). Lu, S. et al. Colorectal cancer with low SLC35A3 is associated with immune infiltrates and poor prognosis. Sci Rep 14 , (2024). Guinney, J. et al. The consensus molecular subtypes of colorectal cancer. Nat Med 21 , 1350–1356 (2015). Vasaikar, S. et al. Proteogenomic Analysis of Human Colon Cancer Reveals New Therapeutic Opportunities. Cell 177 , 1035-1049.e19 (2019). Clark, D. J. et al. Integrated Proteogenomic Characterization of Clear Cell Renal Cell Carcinoma. Cell 179 , 964-983.e31 (2019). Kim, Y. H. et al. Integrative Multi-omics Analysis Reveals Different Metabolic Phenotypes Based on Molecular Characteristics in Thyroid Cancer. Clinical Cancer Research 30 , 883–894 (2024). Lehtiö, J. et al. Proteogenomics of non-small cell lung cancer reveals molecular subtypes associated with specific therapeutic targets and immune-evasion mechanisms. Nat Cancer 2 , 1224–1242 (2021). Sun, C. et al. Spatially resolved multi-omics highlights cell-specific metabolic remodeling and interactions in gastric cancer. Nat Commun 14 , (2023). Bustamante Eduardo, M. et al. A Metabolic Shift to the Serine Pathway Induced by Lipids Fosters Epigenetic Reprogramming in Nontransformed Breast Cells . Sci. Adv vol. 11 https://maayanlab.cloud/Enrichr/ (2025). Rhee, J. K. et al. Impact of tumor purity on immune gene expression and clustering analyses across multiple cancer types. Cancer Immunol Res 6 , 87–97 (2018). Aran, D., Sirota, M. & Butte, A. J. Systematic pan-cancer analysis of tumour purity. Nat Commun 6 , (2015). Isella, C. et al. Stromal contribution to the colorectal cancer transcriptome. Nat Genet 47 , 312–319 (2015). Tables Tables 1 to 9 are available in the Supplementary Files section. Additional Declarations No competing interests reported. Supplementary Files Tables.xlsx SupplementaryMaterials.pdf BlotSupplementaryfile.pdf 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-8267774","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":560595967,"identity":"4f7ba75d-ac1e-450f-addc-674b4c360522","order_by":0,"name":"Seok June Hong","email":"","orcid":"","institution":"Kyung Hee University","correspondingAuthor":false,"prefix":"","firstName":"Seok","middleName":"June","lastName":"Hong","suffix":""},{"id":560595968,"identity":"70e34c76-1c13-4486-beb2-d2358139faab","order_by":1,"name":"Seheum Park","email":"","orcid":"","institution":"Korea University","correspondingAuthor":false,"prefix":"","firstName":"Seheum","middleName":"","lastName":"Park","suffix":""},{"id":560595969,"identity":"68e75d6b-dc8d-4652-a9cd-4f8e635da3a6","order_by":2,"name":"Sunghoon Kim","email":"","orcid":"","institution":"Korea University","correspondingAuthor":false,"prefix":"","firstName":"Sunghoon","middleName":"","lastName":"Kim","suffix":""},{"id":560595970,"identity":"6d148c7a-f811-4331-9a93-e2e9143140c7","order_by":3,"name":"Young Il Park","email":"","orcid":"","institution":"Korea University","correspondingAuthor":false,"prefix":"","firstName":"Young","middleName":"Il","lastName":"Park","suffix":""},{"id":560595971,"identity":"7e4661d1-cbf8-40e4-8a86-2c59730789a8","order_by":4,"name":"Mikyung Kang","email":"","orcid":"","institution":"Korea University","correspondingAuthor":false,"prefix":"","firstName":"Mikyung","middleName":"","lastName":"Kang","suffix":""},{"id":560595972,"identity":"e8a60062-aa89-43d8-ba7d-d0445cb79212","order_by":5,"name":"Jaewoong Lee","email":"","orcid":"","institution":"Korea University","correspondingAuthor":false,"prefix":"","firstName":"Jaewoong","middleName":"","lastName":"Lee","suffix":""},{"id":560595973,"identity":"5548cfd7-f369-4596-a2b6-0112c7c2213d","order_by":6,"name":"Martin Hemberg","email":"","orcid":"","institution":"Massachusetts General Hospital","correspondingAuthor":false,"prefix":"","firstName":"Martin","middleName":"","lastName":"Hemberg","suffix":""},{"id":560595975,"identity":"b466dff1-40ad-4c9d-a6e1-5d5d680880c4","order_by":7,"name":"Sung Eun Kim","email":"","orcid":"","institution":"Korea University","correspondingAuthor":false,"prefix":"","firstName":"Sung","middleName":"Eun","lastName":"Kim","suffix":""},{"id":560595977,"identity":"ef75b71f-c705-4d4f-8784-0952c88c7d0a","order_by":8,"name":"Kwoneel Kim","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA2UlEQVRIiWNgGAWjYBACCQhlYcDGwHwAxJchVosEUAtbAojBQ7wWBgYeAxCLsBbJGcnHHnyokDDmEzvz+dWNGgseBvbDRzfg0yItkZZuOOOMhBmbdO4265xjQIfxpKXdwKdFTjrHTJq3TcIGpMU4hw2oRYLHjLCWv/9AWnKeGef8I0KLNEgLYwPIYTnMj3PbiNAiOf9ZmmTPMQljNuk0M+bcPgkeNkJ+kThz+JjEjxobw/mzkx9/zvlWJ8fPfvgYXi3IgA0cSWzEKgcB5g+kqB4Fo2AUjIKRAwA6PTyIYmxB9wAAAABJRU5ErkJggg==","orcid":"","institution":"Kyung Hee University","correspondingAuthor":true,"prefix":"","firstName":"Kwoneel","middleName":"","lastName":"Kim","suffix":""}],"badges":[],"createdAt":"2025-12-03 08:24:03","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-8267774/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-8267774/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":98751784,"identity":"7a76d783-0488-48e5-9921-93beb6c8ccfb","added_by":"auto","created_at":"2025-12-22 09:12:52","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":19411208,"visible":true,"origin":"","legend":"","description":"","filename":"Manuscript.docx","url":"https://assets-eu.researchsquare.com/files/rs-8267774/v1/debccb7f7cc8e2d044d2b31c.docx"},{"id":98751778,"identity":"1ddba1ca-1d45-4f54-90ea-1f4691930736","added_by":"auto","created_at":"2025-12-22 09:12:52","extension":"xlsx","order_by":3,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":209674,"visible":true,"origin":"","legend":"","description":"","filename":"Tables.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-8267774/v1/6c3f0f45ec064f882de72ad1.xlsx"},{"id":98751787,"identity":"194b14e8-14d7-42d7-a482-027cc4be0bf9","added_by":"auto","created_at":"2025-12-22 09:12:52","extension":"json","order_by":4,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":10847,"visible":true,"origin":"","legend":"","description":"","filename":"cbaf42eea6b84748aca49e47cbef5679.json","url":"https://assets-eu.researchsquare.com/files/rs-8267774/v1/60c17f8945419bd1b4b7d19b.json"},{"id":98778472,"identity":"a459f172-4e76-444f-b0cd-30b8501dcf56","added_by":"auto","created_at":"2025-12-22 12:29:16","extension":"pdf","order_by":5,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":893953,"visible":true,"origin":"","legend":"","description":"","filename":"BlotSupplementaryfile.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8267774/v1/2a436a61700125a05b022207.pdf"},{"id":98751791,"identity":"ab0d3386-0bb8-410a-863a-9c96efebb0a4","added_by":"auto","created_at":"2025-12-22 09:12:52","extension":"pdf","order_by":6,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":2991110,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryMaterials.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8267774/v1/1b3487636ecd581e96c2a366.pdf"},{"id":98751793,"identity":"fc732925-0a57-45e9-832e-7c12a5823456","added_by":"auto","created_at":"2025-12-22 09:12:53","extension":"xml","order_by":7,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":193687,"visible":true,"origin":"","legend":"","description":"","filename":"cbaf42eea6b84748aca49e47cbef56791enriched.xml","url":"https://assets-eu.researchsquare.com/files/rs-8267774/v1/9d64d0f0b59e68e6ca5d81a4.xml"},{"id":98751807,"identity":"d698218f-121f-416f-8250-6de08e703f2f","added_by":"auto","created_at":"2025-12-22 09:12:53","extension":"pdf","order_by":8,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":37533711,"visible":true,"origin":"","legend":"","description":"","filename":"MainFigures.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8267774/v1/050dcad95a8259d95d9830b7.pdf"},{"id":98777681,"identity":"8dd5cbae-af83-431a-82df-6217f819a8ff","added_by":"auto","created_at":"2025-12-22 12:28:20","extension":"pdf","order_by":9,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":16139778,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryFigures.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8267774/v1/24f6c7c754d1cf78fac8504e.pdf"},{"id":98777780,"identity":"cefdacda-ce06-4c2c-a197-b2838f12019f","added_by":"auto","created_at":"2025-12-22 12:28:27","extension":"jpeg","order_by":10,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":3185605,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage1.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-8267774/v1/a3fdfcf4589cf6a88e2c2b31.jpeg"},{"id":98777553,"identity":"41cc0f03-3b98-40c3-a36d-4b77b315eb95","added_by":"auto","created_at":"2025-12-22 12:28:02","extension":"jpeg","order_by":11,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":2459770,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage2.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-8267774/v1/598606d8a2eb9029fd1b52cd.jpeg"},{"id":98751801,"identity":"277aa021-5880-410e-8bec-5bc8595d9789","added_by":"auto","created_at":"2025-12-22 09:12:53","extension":"jpeg","order_by":12,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":2243364,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage3.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-8267774/v1/a058b1c72c10498f525dad27.jpeg"},{"id":98777498,"identity":"1200e29a-bd31-497e-856b-45ee14747a98","added_by":"auto","created_at":"2025-12-22 12:27:43","extension":"jpeg","order_by":13,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":2401932,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage4.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-8267774/v1/3109b3b996010e0cd3fd803d.jpeg"},{"id":98777887,"identity":"34434c09-df73-4f25-967d-edc1daec29b5","added_by":"auto","created_at":"2025-12-22 12:28:37","extension":"jpeg","order_by":14,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":2478694,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage5.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-8267774/v1/90b9e71a0076e896aed4ddc8.jpeg"},{"id":98780301,"identity":"b3aa7a22-3b77-4729-9381-826831dbe1c7","added_by":"auto","created_at":"2025-12-22 12:31:12","extension":"jpeg","order_by":15,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":5283296,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage6.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-8267774/v1/b3ff6fcb034cd2e2f10c0364.jpeg"},{"id":98751802,"identity":"ea258ce5-cb10-4d9c-b3cd-d6cb4fc4067c","added_by":"auto","created_at":"2025-12-22 09:12:53","extension":"jpeg","order_by":16,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":2229150,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage7.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-8267774/v1/35fcda87d2a21848c5d10c4e.jpeg"},{"id":98778835,"identity":"b338db65-f7cc-494e-a7e9-0bdc6d76115c","added_by":"auto","created_at":"2025-12-22 12:29:44","extension":"png","order_by":17,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":643778,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-8267774/v1/6f126dc8e948003adb676f3d.png"},{"id":98751799,"identity":"40038ddc-59b5-43da-af65-267fc65f2561","added_by":"auto","created_at":"2025-12-22 09:12:53","extension":"png","order_by":18,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":328013,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-8267774/v1/cb70b119ae8acaa84ae9bd4e.png"},{"id":98777813,"identity":"9f8e53d6-6ac9-4047-8e45-74279b6a6535","added_by":"auto","created_at":"2025-12-22 12:28:30","extension":"png","order_by":19,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":439226,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-8267774/v1/e7145e0a6792856541c5f0e7.png"},{"id":98751804,"identity":"833d7bd0-46ff-41be-bf4a-31fff87fd013","added_by":"auto","created_at":"2025-12-22 09:12:53","extension":"png","order_by":20,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":491321,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-8267774/v1/5f7f15e98cdc17fdd0715381.png"},{"id":98751796,"identity":"02fd7b9d-0618-4735-ae21-62249245d57c","added_by":"auto","created_at":"2025-12-22 09:12:53","extension":"png","order_by":21,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":440387,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage5.png","url":"https://assets-eu.researchsquare.com/files/rs-8267774/v1/6d639972b99065b570ae5d4a.png"},{"id":98751794,"identity":"4674f20e-7f25-47d9-a2b3-290d3e9e73b3","added_by":"auto","created_at":"2025-12-22 09:12:53","extension":"png","order_by":22,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":1443168,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage6.png","url":"https://assets-eu.researchsquare.com/files/rs-8267774/v1/6cc5e70e3968c2426a7c7bac.png"},{"id":98777704,"identity":"5a80c338-70a6-4100-bb79-d01a507aa01a","added_by":"auto","created_at":"2025-12-22 12:28:22","extension":"png","order_by":23,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":332118,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage7.png","url":"https://assets-eu.researchsquare.com/files/rs-8267774/v1/28d14caf15aac2bf69db5cc2.png"},{"id":98751805,"identity":"5147632e-024d-4e61-9512-1067a7dc40f2","added_by":"auto","created_at":"2025-12-22 09:12:53","extension":"xml","order_by":24,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":192850,"visible":true,"origin":"","legend":"","description":"","filename":"cbaf42eea6b84748aca49e47cbef56791structuring.xml","url":"https://assets-eu.researchsquare.com/files/rs-8267774/v1/246052eb82e6786a3f0630bf.xml"},{"id":98777915,"identity":"69bbf0c5-d97d-4a31-94b9-6450999492fa","added_by":"auto","created_at":"2025-12-22 12:28:40","extension":"html","order_by":25,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":212278,"visible":true,"origin":"","legend":"","description":"","filename":"earlyproof.html","url":"https://assets-eu.researchsquare.com/files/rs-8267774/v1/e02fa80d3a11b1e990f16bdb.html"},{"id":98751776,"identity":"dc0e1161-2883-4f9d-be24-fe673a32e3f2","added_by":"auto","created_at":"2025-12-22 09:12:52","extension":"jpeg","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":3185605,"visible":true,"origin":"","legend":"\u003cp\u003eStudy design and SLC-centered archetypes in colorectal cancer.\u003cstrong\u003e A, \u003c/strong\u003eSchematic representation of the study design. Multi-omics data from 255 colorectal cancer patients in the TCGA cohort were collected and filtered to only include solute carrier (SLC) genes and known cancer driver genes. \u003cstrong\u003eB,\u003c/strong\u003e Interpretation of MOFA latent factors. Latent factor 1 was predominantly associated with metabolic pathways and latent factor 4 was enriched for immune-related and RNA processing pathways. Each dot represents the correlation coefficient for a pathway obtained from Pearson correlation analysis between latent factor values and pathway activity scores following sample-based pathway analysis. \u003cstrong\u003eC,\u003c/strong\u003e Sample projection in latent factor space. Samples were plotted along metabolic (factor 1) and immune-RNA (factor 3) axes and mapped onto a ternary plot indicating relative distances to the three inferred archetypes. Points are colored by consensus molecular subtypes (CMSs), with background shading reflecting immune infiltration scores. \u003cstrong\u003eD,\u003c/strong\u003e Overview of phenotypes, pathways and tumor immune microenvironment identified in each archetype. Left, Samples with specific CMSs were mapped onto the latent factor space and arrows indicate representative SLC gene families enriched in each archetype. Middle, expression heatmaps of representative pathways. Rows indicate the pathways that are either positively or negatively enriched, while columns indicate the samples ordered by increasing archetype scores. Colored annotations on the right highlight representative pathways previously reported to be associated with specific CMSs. Right, correlation between archetype scores and immune cell enrichments inferred by xCell analysis.\u003c/p\u003e","description":"","filename":"floatimage1.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-8267774/v1/5c688b99d0b9efad4f16771d.jpeg"},{"id":98751781,"identity":"37da018d-fa98-4752-b18f-9a964c1cb81a","added_by":"auto","created_at":"2025-12-22 09:12:52","extension":"jpeg","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":2459770,"visible":true,"origin":"","legend":"\u003cp\u003eIdentification of the zinc-SLC39–CDX2 axis and CD24 enhancer regulated by CDX2 transcriptional activity.\u003cstrong\u003e A,\u003c/strong\u003e Integrative correlation analysis of the SLC39–CDX2–CD24 axis across transcriptome and proteome datasets. Correlation matrices were generated using gene expression data from 288 TCGA samples and proteomic data from 90 CPTAC samples to assess the relationships between zinc pathway scores and the expression of SLC39 family genes, CDX2 activity, and downstream targets. \u003cstrong\u003eB,\u003c/strong\u003e Correlation between CNV and gene expression for SLC39 family members using bulk-level TCGA/CPTAC data and single-cell inferred CNV data by inferCNV from colorectal cancer (CRC) samples. \u003cstrong\u003eC,\u003c/strong\u003e To focus on membrane-localized zinc import processes, the zinc influx by SLC39 score was computed by averaging normalized expression of nine SLC39 family genes known to mediate extracellular zinc uptake through plasma membrane localization (SLC39A1, A2, A3, A4, A5, A6, A8, A10, and A14). Correlation matrix demonstrating pairwise associations among zinc influx by SLC39 scores, CDX2 expression, CDX2 activity, and CD24 expression across both TCGA and CPTAC cohorts. \u003cstrong\u003eD,\u003c/strong\u003e Epigenomic characterization of a putative CD24 enhancer. Hi-C data from HCT116 cells confirmed that the predicted enhancer resides within the same TAD as CD24. ChIP–seq data from Caco-2 cells revealed enrichment of active enhancer marks (H3K4me1, H3K36me3, H3K27ac) in the predicted region. Single-cell ATAC-seq from three CRC patients further demonstrated cancer cell-specific chromatin accessibility of the CD24 enhancer, compared to other cell types including B cells, endothelial cells, fibroblasts, myeloid cells, and T cells. \u003cstrong\u003eE,\u003c/strong\u003e Correlation between chromatin accessibility of the CD24 enhancer (TCGA-ATAC dataset) and zinc influx, CDX2 expression and activity, and CD24 expression, highlighting a coordinated regulatory relationship. \u003cstrong\u003eF,\u003c/strong\u003e Motif analysis within the CD24 enhancer region identified canonical binding sequences for CDX2. \u003cstrong\u003eG,\u003c/strong\u003e Correlation analysis across CPTAC CRC patients (n = 9) revealed that phosphorylation levels of CDX2 at serine 176 (CDX2-pS176), as measured by phosphoproteomics, were positively associated with zinc influx by SLC39 scores (P = 0.098), CDX2 activity scores (P = 0.34), and CD24 expression scores (P = 0.21), suggesting a coordinated activation of the zinc–CDX2–CD24 axis. \u003cstrong\u003eH,\u003c/strong\u003e Structural modeling of CDX2 and phosphorylated CDX2 at serine 176 (CDX2-pS176) using AlphaFold3, illustrating DNA-binding configurations. \u003cstrong\u003eI,\u003c/strong\u003e Protein–DNA docking analysis using HDOCK revealed that CDX2 exhibits strong binding affinity to the identified CD24 enhancer sequence, with all top 8 docking models showing confidence scores \u0026gt; 0.7. Confidence scores were computed based on an empirical logistic function of docking scores where scores \u0026gt; 0.7 indicate high likelihood of molecular interaction. Notably, the phosphorylated form (CDX2-pS176) demonstrated significantly higher docking confidence scores than unmodified CDX2 (P = 0.0078), suggesting enhanced enhancer-binding potential upon phosphorylation.\u003c/p\u003e","description":"","filename":"floatimage2.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-8267774/v1/36c9cb76980e8e686e6252f3.jpeg"},{"id":98778504,"identity":"a868f189-b1d9-48fe-8b0a-0f0d2888c19d","added_by":"auto","created_at":"2025-12-22 12:29:22","extension":"jpeg","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":2243364,"visible":true,"origin":"","legend":"\u003cp\u003eSingle-cell analyses identify cancer cell-specific activation of the SLC39–CDX2–CD24 axis with functional validation.\u003cstrong\u003e A,\u003c/strong\u003e UMAP visualization of 174,547 cells from 88 tumor samples across 26 CRC patients, annotated into major cell types including cancer, T cells, B cells, fibroblasts, myeloid cells, and endothelial cells. \u003cstrong\u003eB,\u003c/strong\u003e Violin plots showing the expression and activity scores of zinc influx by SLC39, CDX2 expression, CDX2 activity and CD24 expression across annotated cell types. Corresponding pairwise correlation analyses were performed within each cell type, with significance indicated by dots only when P \u0026lt; 0.05 and absolute correlation coefficient ≥ 0.1, highlighting differential coordination of this axis across lineages. \u003cstrong\u003eC,\u003c/strong\u003e Metacells inferred from the single-cell dataset using SEACells (100 cells per metacell), with UMAP visualization of metacell structure and projection from the original single-cell embedding. Cancer cell–enriched metacells are indicated by a dotted boundary. \u003cstrong\u003eD,\u003c/strong\u003e Spatial distribution of zinc influx by SLC39, CDX2 expression, CDX2 activity, and CD24 expression across metacells, revealing a coherent gradient enriched in cancer-associated metacell clusters. \u003cstrong\u003eE,\u003c/strong\u003e Correlation plots between zinc influx by SLC39 and downstream transcriptional components in cancer-derived metacells, with linear regression lines overlaid. Corresponding single-cell level expression data are shown in the bottom right of each panel, illustrating the denoising effect of metacell aggregation. Bar plots below summarize correlation coefficients and significance across all major cell types, highlighting the cancer cell-specific coordination of the SLC39–CDX2–CD24 axis. \u003cstrong\u003eF,\u003c/strong\u003e Schematic diagram and quantification of intracellular zinc levels through FluoZin-3 fluorescence in SLC39A4- and SLC39A5-overexpressing cells. Fluorescence was quantified as MFI (mean fluorescence intensity) per cell. \u003cstrong\u003eG,\u003c/strong\u003e Quantitative RT-PCR analysis of CDX2 target gene transcripts confirmed by CDX2 overexpression in Fig. S3E reveal enhanced CDX2 activity in SLC39A4- and SLC39A5-overexpressing cells. \u003cstrong\u003eH-I,\u003c/strong\u003e Western blot analysis (H) and flow cytometry using anti-CD24 antibody (I) indicate enhanced CD24 protein and membrane localization levels. \u003cstrong\u003eJ,\u003c/strong\u003e Quantitative RT-PCR analysis of CD24 and CD47 in SLC39A4- and SLC39A5-overexpressing cells.\u003c/p\u003e","description":"","filename":"floatimage3.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-8267774/v1/037148c3fdee8894d14c32e0.jpeg"},{"id":98751783,"identity":"75c4fadb-2af8-40e0-8ebd-51950945fc49","added_by":"auto","created_at":"2025-12-22 09:12:52","extension":"jpeg","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":2401932,"visible":true,"origin":"","legend":"\u003cp\u003eTrajectory-informed characterization of zinc signaling–driven terminal states reveals a CD24⁺ immune-evasive fate associated with poor prognosis and immunotherapy resistance.\u003cstrong\u003e A,\u003c/strong\u003e UMAP visualization of 21,066 cancer cells from 26 CRC patients, annotated by CMS classification. Fate probabilities toward three terminal states were inferred using Palantir, highlighting transcriptionally distinct differentiation endpoints. \u003cstrong\u003eB,\u003c/strong\u003e Violin plots showing expression and activity levels of zinc influx by SLC39, CDX2, CDX2 activity, and CD24 across cells assigned to each terminal fate. \u003cstrong\u003eC,\u003c/strong\u003e Smoothed gene expression trends along the trajectory toward each terminal state, demonstrating coordinated upregulation of the zinc signaling axis and CD24 expression specifically in Terminal 1. \u003cstrong\u003eD,\u003c/strong\u003e Terminal-specific enrichment analysis of hallmark and colorectal cancer–relevant pathways revealed distinct transcriptional programs across terminal states. Terminal 1 was preferentially enriched for CMS2-associated molecular features, including activation of WNT and MSS programs, consistent with its immune-cold identity. \u003cstrong\u003eE,\u003c/strong\u003e Terminal 1–specific differential expression analysis revealed CD24 as the top upregulated immune modulatory gene compared to other terminal states (log₂ fold change \u0026gt; 0.5), suggesting that CD24 marks an immune-evasive trajectory endpoint. \u003cstrong\u003eF,\u003c/strong\u003e Projection of terminal states onto CMS subtypes (left), and gene signature–based deconvolution of TCGA-COAD bulk tumors using Terminal 1–3 marker genes (right), confirming that A1 archetype corresponds to the Terminal 1 program. \u003cstrong\u003eG,\u003c/strong\u003e Kaplan–Meier curves showing that patients with high Terminal 1 scores exhibit significantly worse progression-free interval (PFI) and disease-free interval (DFI) in TCGA-COAD. \u003cstrong\u003eH,\u003c/strong\u003e UMAP and density plots from the SCP2079 CRC single-cell dataset profiled prior to immune checkpoint inhibitor (ICI) treatment. Cells were annotated by clinical response, revealing that NR (non-responder) tumors exhibited elevated zinc influx, CDX2 expression, CDX2 activity, and CD24 expression compared to R (responder) tumors. \u003cstrong\u003eI,\u003c/strong\u003e The same features analyzed with respect to progression-free survival (PFS), showing stronger activation of the zinc–CD24 axis in tumors with PFS \u0026lt; 6 months. Statistical comparison of the distributions in \u003cstrong\u003eH\u003c/strong\u003e and \u003cstrong\u003eI\u003c/strong\u003e using a two-sided t-test confirmed significant differences (P \u0026lt; 2.2 × 10⁻¹²). \u003cstrong\u003eJ,\u003c/strong\u003e Cell type–specific gene networks constructed using scHumannet, comparing NR and R samples. Terminal 1–specific marker genes (top 300) were overlaid, revealing selective integration into the NR-specific network. \u003cstrong\u003eK,\u003c/strong\u003e Analogous gene network analysis for PFS \u0026lt; 6 months versus longer responders. Terminal 1–specific genes were preferentially embedded in the PFS\u0026lt;6 network. \u003cstrong\u003eL,\u003c/strong\u003e Quantitative network analysis of Terminal 1–3 marker genes across the two conditions, showing that Terminal 1 genes exhibit the highest connectivity, degree centrality, eigenvector centrality, and page rank in networks derived from non-responders and early progressors.\u003c/p\u003e","description":"","filename":"floatimage4.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-8267774/v1/741999395b5947a36a9c33b9.jpeg"},{"id":98751789,"identity":"2fab8509-c910-42d6-ba89-edd28d77b417","added_by":"auto","created_at":"2025-12-22 09:12:52","extension":"jpeg","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":2478694,"visible":true,"origin":"","legend":"\u003cp\u003ePleiotropic CD24–SIGLEC10 signaling bifurcates into immunosuppressive and adhesive programs in distinct macrophage lineages.\u003cstrong\u003e A,\u003c/strong\u003e Cell–cell interaction profiling using NICHES across CRC single-cell transcriptomes was performed by setting cancer cells as the sender and evaluating interactions with all other major cell types. For each cancer-to-receiver pair, top-ranking ligand–receptor interactions were identified, and the top five interactions were selected based on specificity and expression strength. Among these, CD24–SIGLEC10 emerged as a dominant interaction uniquely enriched in cancer–myeloid signaling, as highlighted by a red box. \u003cstrong\u003eB,\u003c/strong\u003e UMAP-based classification of myeloid cells into fine-grained subtypes, including monocyte-derived macrophages (moMC), tissue-resident macrophages (RTMC), monocytes, and dendritic cells (cDC, pDC, moDC), reveals distinct lineage states. SIGLEC10 expression was selectively enriched in moMC, RTMC, moDC, and monocyte populations, as marked in red. \u003cstrong\u003eC,\u003c/strong\u003e Schematic of myeloid developmental trajectories illustrating that both moMC and RTMC arise from distinct ontogenies: moMC from circulating monocytes, and RTMC from embryonic precursors, establishing the basis for functional divergence. Cell populations known to exhibit high SIGLEC10 expression (as shown in panel B) are highlighted with red boxes. \u003cstrong\u003eD,\u003c/strong\u003e Diffusion-based pseudotime and fate probability analysis of monocytes, moMC, and moDC revealed bifurcating differentiation paths, each with distinct lineage commitment. \u003cstrong\u003eE,\u003c/strong\u003e Gene expression trends along both moMC and moDC pseudotime trajectories demonstrated progressive upregulation of SIGLEC10, indicating that SIGLEC10 activation occurs along both monocyte derived macrophage and dendritic lineages. Lineage-specific markers CD81 (moMC) and CD1A (moDC) were shown as references. \u003cstrong\u003eF–G,\u003c/strong\u003e Ligand–receptor interaction modeling using NichNet for cancer–moMC pairs identified the top five ligands with the highest regulatory potential over downstream gene programs based on area under the precision–recall curve (AUPR). CD24 was among these top-ranked ligands, supported by its strong expression in cancer cells and high prior interaction potential with SIGLEC10. Predicted downstream target genes showed robust expression, indicating active signaling in moMCs. Functional enrichment of these genes revealed immunosuppressive pathways, including negative regulation of phagocytosis, innate immune response, and antigen presentation. \u003cstrong\u003eH–I,\u003c/strong\u003e A parallel NichNet analysis was performed for cancer–RTMC pairs, where CD24 was also included among the top five ligands ranked by AUPR. Predicted downstream target genes in RTMCs showed robust expression and were functionally enriched for adhesion and chemotaxis related pathways\u003c/p\u003e","description":"","filename":"floatimage5.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-8267774/v1/1e1744d75294201f94716cdc.jpeg"},{"id":98751792,"identity":"c22cccef-5693-4d70-b249-2afe5a5d9eb7","added_by":"auto","created_at":"2025-12-22 09:12:52","extension":"jpeg","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":5283296,"visible":true,"origin":"","legend":"\u003cp\u003eSpatial mapping of CD24–SIGLEC10 interaction and associated signaling pathways in CRC tissue sections.\u003cstrong\u003e A,\u003c/strong\u003e Spatial transcriptomic data from a representative CRC tissue section (1 out of 4 patients) profiled using the 10x Genomics Visium platform. The top panel shows the matched H\u0026amp;E image. Major and minor cell type annotations were obtained using the SPOTlight deconvolution algorithm, with previously annotated single-cell RNA-seq profiles (from Fig. 3) used as the reference. Regions enriched for CD24–SIGLEC10 interaction, as defined by NICHES and SPATA2 segmentation, are outlined with black contour lines on the minor cell type map. \u003cstrong\u003eB,\u003c/strong\u003e Spatial distribution of CD24 and SIGLEC10 expression levels across spatial spots, quantified from normalized expression matrices. CD24–SIGLEC10 interaction scores were computed using the NICHES framework. Black contour lines indicate the same interaction-enriched region as in a. \u003cstrong\u003eC,\u003c/strong\u003e Spatial projection of inferred pathway activities based on known gene sets associated with (i) CD24-mediated \"Don’t eat me\" signaling, (ii) negative regulation of phagocytosis, and (iii) regulation of cell–cell adhesion. Black contour lines indicate the same interaction-enriched region as in a. \u003cstrong\u003eD,\u003c/strong\u003e Enrichment scores for the gene sets shown in (c) were plotted relative to their distance from spatial annotations enriched for CD24–SIGLEC10 interaction, using the plotSasLineplot() function from SPATA2. Line plots represent smoothed expression trends estimated by local polynomial regression (loess), with distance calculated from the annotated border of the interaction-enriched region. \u003cstrong\u003eE,\u003c/strong\u003e Identification of spatial niches based on unsupervised clustering of spatial transcriptomes. The left panel shows schematic representations of the molecular features mapped in the dataset, including zinc influx by SLC39, CDX2 expression, CDX2 activity, CD24 and SIGLEC10 expression, CD24–SIGLEC10 interaction scores, downstream pathway signatures and enrichment scores of relevant cell types. The right panel shows spatial domains annotated as distinct niches across two representative samples (S2 and S3), where each domain was scored for enrichment of the aforementioned molecular features. Enrichment scores were visualized using dot plots, with dot size representing –log₁₀(P) and color indicating effect size (normalized enrichment value). Niches enclosed by black dotted boxes represent regions where all axes of the zinc–CDX2–CD24–SIGLEC10 signaling cascade are concurrently enriched.\u003c/p\u003e","description":"","filename":"floatimage6.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-8267774/v1/28fc6a76cfd12c8ebb5273a0.jpeg"},{"id":98779601,"identity":"f92d3f0b-22fe-43bb-b205-fa55b71ef207","added_by":"auto","created_at":"2025-12-22 12:30:30","extension":"jpeg","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":2229150,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cbr\u003e\n\u003c/p\u003e\n\u003cp\u003ePleiotropic CD24–SIGLEC10 signaling promotes divergent macrophage programs and correlates with poor immune outcome, functionally validated through targeted assays.\u003cstrong\u003e A,\u003c/strong\u003e Marker gene sets (n = 50 per population) from single-cell–defined moMC and RTMC subtypes were used to score bulk RNA-seq profiles from the TCGA-COAD cohort (n = 288). Samples were projected into a two-dimensional space based on moMC and RTMC enrichment scores, and unsupervised clustering was performed using the \u003cem\u003efviz_cluster() \u003c/em\u003efunction from the \u003cem\u003efactoextra\u003c/em\u003e package, identifying four distinct subgroups along the moMC and RTMC axes. \u003cstrong\u003eB,\u003c/strong\u003e CD24–SIGLEC10 interaction scores were inferred from TCGA bulk expression data using DecoupleR, applying ligand–receptor pair sets from the LIANA database. Each pair was treated as a gene set, and enrichment scores were computed using weighted mean-based scoring. Interaction scores were compared across the four moMC/RTMC-defined subtypes. \u003cstrong\u003eC,\u003c/strong\u003e Bulk-level correlation analysis between moMC and RTMC axis scores and pathway activities. moMC scores were positively associated with CD24-mediated \"don't eat me\" signaling, negative regulation of phagocytosis, and negative regulation of immune responses. RTMC scores were positively correlated with the regulation of cell–cell adhesion. \u003cstrong\u003eD,\u003c/strong\u003e Schematic diagram of lineage-selective differentiation of THP-1 cells into moMC-like or RTMC-like macrophages and quantitative RT-PCR analysis of representative moMC and RTMC markers. \u003cstrong\u003eE,\u003c/strong\u003e Schematic diagram of phagocytosis assay using moMC-like differentiated THP-1 cells expressing GFP and pHrodo Red-treated LS174T CRC cells (left top) and cellular images of GFP\u003csup\u003e+\u003c/sup\u003e THP-1 cells that have phagocytosed pHrodo Red-treated LS174T cells (left bottom). Graph shows quantification of the ratio of pHrodo Red-positive GFP\u003csup\u003e+\u003c/sup\u003e THP-1 cells cocultured with SLC39A4- or SLC39A5-overexpressing LS174T cells in the presence of either IgG or anti-CD24 antibodies (right). \u003cstrong\u003eF,\u003c/strong\u003e Five independent immune checkpoint inhibitor (ICI)-treated colorectal cancer (CRC) cohorts—MSS1 (n = 9), MSS2 (n = 34), MSS3 (n = 12), MSI1 (n = 70), and MSI2 (n = 41)—were integrated (n = 166 total), followed by batch effect removal using the limma package. The integrated dataset was randomly partitioned into a modeling set (n = 138) and an independent testing set (n = 28). The modeling set was further split into training (n = 110) and validation (n = 28) subsets. A Gradient Boosting Classifier was trained using seven biologically informed features—SLC39A2, SLC39A4, SLC39A5, SLC39A10, CDX2, CD24, and SIGLEC10— that were identified in this study as key regulators of immune evasion by zinc transport in colorectal cancer. \u003cstrong\u003eG,\u003c/strong\u003e Receiver operating characteristic (ROC) curves generated using the independent testing set (n = 28) to evaluate the reproducibility and generalizability of our model compared to previously reported CRC-specific or pancancer biomarkers. The model trained using our seven-gene signature (SLC39A2, SLC39A4, SLC39A5, SLC39A10, CDX2, CD24, SIGLEC10) consistently achieved the highest AUC (0.728) across all benchmarked models.\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;\u003c/p\u003e","description":"","filename":"floatimage7.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-8267774/v1/12197d8d454d77389a7d54e4.jpeg"},{"id":100910876,"identity":"493d2e23-cc17-4b7c-a5ee-50c85a4f1b45","added_by":"auto","created_at":"2026-01-22 16:54:26","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":21781303,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8267774/v1/a3111e6c-2daf-47c1-bf62-d803156a8236.pdf"},{"id":98751777,"identity":"7af25cbe-a00d-41a8-a91f-eac50702b534","added_by":"auto","created_at":"2025-12-22 09:12:52","extension":"xlsx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":209674,"visible":true,"origin":"","legend":"","description":"","filename":"Tables.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-8267774/v1/a3cc04a3cd6daf666ef01a64.xlsx"},{"id":98777508,"identity":"8dce5488-3bce-4f0f-ae1d-53d12972e960","added_by":"auto","created_at":"2025-12-22 12:27:51","extension":"pdf","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":2991110,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryMaterials.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8267774/v1/059256cdb45349e4f486ea93.pdf"},{"id":98778723,"identity":"5b259b6c-6af0-4ddc-9a26-ac03dd206c11","added_by":"auto","created_at":"2025-12-22 12:29:34","extension":"pdf","order_by":3,"title":"","display":"","copyAsset":false,"role":"supplement","size":893953,"visible":true,"origin":"","legend":"","description":"","filename":"BlotSupplementaryfile.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8267774/v1/faabaf787cb72878b8e027da.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"SLC39-driven zinc influx orchestrates pleiotropic tumor–immune crosstalk to establish an immune-suppressive microenvironment in colorectal cancer","fulltext":[{"header":"Introduction","content":"\u003cp\u003eTumor cells exhibit heightened metabolic demands and reprogram nutrient acquisition through solute carrier (SLC) transporters to sustain uncontrolled proliferation, survival, and metastasis. Members of the SLC family have emerged as pivotal mediators of metabolic reprogramming, facilitating the import of essential nutrients and the export of metabolic waste, thereby coupling metabolic flux to oncogenic signaling\u003csup\u003e\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e,\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e\u003c/sup\u003e. SLC2A1 (GLUT1), a key glucose transporter, is upregulated in many cancers to support aerobic glycolysis, while SLC7A5 (LAT1) enables uptake of large neutral amino acids such as leucine, sustaining protein synthesis and mTORC1 activation\u003csup\u003e\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e\u003c/sup\u003e. Therapeutic inhibition of LAT1 has shown efficacy in preclinical models by restricting amino acid availability and suppressing tumor progression. Similarly, SLC7A11 (xCT) imports cystine to fuel glutathione biosynthesis, enhancing antioxidant defense and protecting cancer cells from ferroptosis\u003csup\u003e\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e\u003c/sup\u003e. These findings underscore SLC transporters as central mediators of metabolic reprogramming and potential therapeutic targets in oncology\u003csup\u003e\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eAlthough metabolic rewiring is a hallmark of cancer, the extent to which SLC transporters orchestrate interactions between tumor cells and the tumor microenvironment (TME) remains poorly defined. Accumulating evidence suggests that metabolic byproducts, such as lactate generated through the Warburg effect, can acidify the TME and impair immune cell activity, thereby facilitating immune escape\u003csup\u003e\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e\u003c/sup\u003e. However, most studies have emphasized tumor-intrinsic metabolic programs while neglecting the reciprocal influence of stromal and immune components\u003csup\u003e\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e\u003c/sup\u003e. As such, current frameworks inadequately capture the interplay between metabolic adaptation and immune modulation. This disconnect is particularly critical when considering that metabolic competition and nutrient availability in the TME can shape both tumor behavior and therapeutic response. The absence of integrated, systems-level studies that dissect these interactions has limited our capacity to identify context-specific metabolic vulnerabilities and design effective metabolism-targeted therapies.\u003c/p\u003e \u003cp\u003eColorectal cancer (CRC), with its well-characterized consensus molecular subtypes (CMS), presents a robust model to interrogate the intersection of metabolic regulation and immune microenvironment. CMS1 tumors exhibit hypermutation and immune infiltration, CMS2 tumors feature canonical WNT signaling and an immune-desert phenotype, CMS3 tumors display metabolic dysregulation, and CMS4\u003csup\u003e7\u003c/sup\u003e tumors are marked by stromal activation and poor prognosis, occurring at frequencies of 16%, 32%, 17%, and 35%, respectively\u003csup\u003e\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e\u003c/sup\u003e. Despite these subtype-specific landscapes, the contribution of SLC transporters to CRC subtype identity and TME modulation remains unclear. We hypothesize that specific SLC transporters define metabolically regulated immunosuppressive states in CRC. In particular, we propose that SLC39-mediated zinc influx drives subtype-specific transcriptional reprogramming, culminating in immune checkpoint activation and exclusion of anti-tumoral immune response. By integrating multi-omics, spatial, and single-cell data, our study aims to elucidate the SLC\u0026ndash;TME axis as a mechanistic bridge linking metabolic inputs to immune modulation across CRC subtypes.\u003c/p\u003e"},{"header":"Materials and Methods","content":"\u003cp\u003e\u003cstrong\u003ePatient sample collection and multi-omics profiling\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe primary dataset used for this study was obtained from The Cancer Genome Atlas (TCGA), specifically focusing on colorectal cancer (CRC) samples. A total of 255 CRC samples were selected, encompassing multi-omics data including transcriptomics, mutations, copy number variations (CNV), and DNA methylation. These multi-omics profiles were retrieved from Xena Browser (https://xenabrowser.net/), which provides access to publicly available TCGA datasets. These multi-omics data were utilized to investigate the molecular landscape of CRC. For data validation and additional analyses, we used data from the Clinical Proteomic Tumor Analysis Consortium (CPTAC), which contains proteomic and phosphoproteomic profiles of CRC samples. A total of 109 CRC samples were included in this validation cohort. The CPTAC data was retrieved from Linkedomics (https://linkedomics.org/), which provides integrative analyses of omics datasets. These data were used to cross-validate findings from the TCGA dataset and further refine the molecular characterization of CRC.\u0026nbsp;Clinical information for the TCGA cohort is provided in\u0026nbsp;Table 1, and corresponding information for the CPTAC cohort is available in Table 2.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eData integration and gene selection\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTo investigate metabolic dependencies in CRC, we focused on solute carrier (SLC) transporters, which regulate the import and export of metabolites, as well as cancer-specific oncogenic driver genes. We retained only the genes annotated as part of the SLC family or as known oncogenes in CRC. The list of SLC genes was curated by excluding pseudogenes and antisense transcripts to ensure biological relevance and transcriptional activity. For oncogenic driver genes, we utilized the OncoVar database, which provides curated lists of cancer-type-specific driver genes along with a scoring system reflecting their pathogenic potential. Genes annotated for colorectal adenocarcinoma (COAD) with a Driver level of 4 (i.e., \u0026ldquo;pathogenic,\u0026rdquo; score \u0026ge; 20) were selected for analysis. Additionally, to increase the robustness of the selected oncogenes, we further filtered for genes with a Total score \u0026ge; 3, representing the number of supporting databases in which the gene was identified as a driver. This filtering strategy ensured that only metabolically and oncogenically relevant genes were retained for subsequent integrative multi-omics analyses.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDimensionality reduction and multi-omics integration\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTo integrate the multiple layers of omics data and identify latent molecular features associated with colorectal cancer, we applied Multi-Omics Factor Analysis (MOFA), a statistical framework designed for unsupervised integration of multi-omics datasets. MOFA enables extraction of low-dimensional latent factors that capture sources of variation shared across or specific to individual omics layers. We constructed a MOFA model using five omics layers: transcriptomics, somatic mutations, copy number variations, DNA promoter methylation, and DNA gene body methylation. Given the nature of the data, we specified \u0026lsquo;gaussian\u0026rsquo; likelihoods for continuous data types (transcriptomics, CNVs, methylation), and \u0026lsquo;bernoulli\u0026rsquo; likelihood for the mutation matrix, which was binarized to represent the presence or absence of nonsynonymous variants in each gene. The model was trained using the following parameters: the number of latent factors was set to 10, the convergence mode was set to \u0026lsquo;slow\u0026rsquo; to ensure a thorough exploration of the posterior space, and maxiter was set to 10,000 to allow sufficient iterations for convergence. This dimensionality reduction provided a shared latent representation of the dataset, which was used for downstream clustering, module discovery, and association analyses with clinical or immunological features.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eArchetypal analysis using ParetoTI\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTo identify representative molecular states, archetypal analysis was performed using the ParetoTI R package, which geometrically infers extremal points (archetypes) from input dimensional space. As input, the top four latent factors derived from MOFA were selected based on the proportion of explained variance. All six pairwise combinations of these four factors were evaluated using the t-ratio metric, which quantifies the fit between the polytope and convex hull of the data. The factor pair with the highest t-ratio (Factor 1 and Factor 4) was selected for downstream modeling. To functionally annotate the selected factors, gene set variation analysis (GSVA) was performed using metabolic pathway gene sets curated from Gene Ontology (GO: biological process category, filtered for terms containing \u0026ldquo;metabolic process\u0026rdquo;) and KEGG (metabolism category). Enrichment scores were computed for each sample and correlated with factor values to support interpretation of the molecular programs captured by the selected latent dimensions.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eArchetype assignment and multi-omics characterization\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eSamples were assigned to one of three archetypes (Arc1, Arc2, Arc3) based on proximity to polytope vertices inferred by ParetoTI using MOFA-derived latent factors. To annotate biological relevance, consensus molecular subtypes (CMS) were inferred using the CMScaller R package. Molecular characteristics of each archetype were evaluated by analyzing enrichment patterns across multi-omics layers. Specifically, transcriptomic, promoter methylation, and gene body methylation profiles of SLC genes were compared across archetypes. GSVA was performed using curated pathway collections from GO, KEGG, WikiPathways, Reactome, and MSigDB Hallmark gene sets, focusing on the identification of pathways enriched in each archetype. To assess tumor microenvironment (TME) differences, xCell was used to infer immune and stromal cell-type enrichment scores from bulk transcriptomic data. Archetype-specific correlations with TME composition were calculated\u0026nbsp;using Pearson correlation\u0026nbsp;to evaluate immune landscape variation across subtypes.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eIdentification of CD24 enhancer region and transcription factor binding motifs\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003ePutative enhancer elements regulating CD24 were identified using the GeneHancer database (GH06J106781). Open chromatin accessibility at this region was confirmed using ATAC-seq data from 25 colorectal adenocarcinoma (COAD) samples\u003csup\u003e9\u003c/sup\u003e. To investigate enhancer activity specific to the archetype 1, chromatin accessibility signals were compared across archetype-classified samples. Topological organization of the enhancer\u0026ndash;promoter interaction was assessed using Hi-C data via the 3D Genome Browser (http://3dgenome.org), which revealed that both CD24 and GH06J106781 reside within the same topologically associating domain (TAD), supporting the potential regulatory interaction. To identify transcription factors that potentially bind to the enhancer, motif enrichment analysis was performed using the TFBSTools R package in conjunction with the JASPAR2022 motif database\u003csup\u003e10\u003c/sup\u003e. The enhancer sequence was scanned for transcription factor binding motifs using position weight matrices (PWMs) from JASPAR, and matches were identified based on a minimum score threshold of 80% relative to the minimal PWM score. For each candidate binding site, associated p-values and matched binding sequences were computed using searchSeq(), and high-confidence hits were retained for downstream structural modeling.\u0026nbsp;Clinical information for\u0026nbsp;scATAC-seq\u0026nbsp;cohort is summarized in\u0026nbsp;Table 3.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eSingle-cell RNA-seq data processing and clustering\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTumor-derived single-cell RNA-seq data were obtained from the Synapse dataset (accession: syn26844071). Only cells annotated as originating from tumor tissue were included in the analysis. Quality control filtering was applied to retain cells with more than 200 detected genes, fewer than 100,000 total counts, and less than 3% mitochondrial gene content.\u0026nbsp;Genes expressed in fewer than three cells were removed. The processed data were normalized and scaled, and 2,000 highly variable genes were selected for downstream analysis. Principal component analysis (PCA) was performed, and the first 10 dimensions were used for clustering (resolution = 0.5) and UMAP-based visualization. All analyses were conducted using Seurat v5.2.1.\u0026nbsp;Clinical information for this cohort is summarized in Table 4.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eSingle-cell RNA-seq dataset from ICI-treated colorectal cancer\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eSingle-cell transcriptomic data from immune checkpoint inhibitor (ICI)-treated colorectal cancer was obtained from the publicly available dataset SCP2079, which corresponds to the study \u0026quot;Combined PD-1, BRAF and MEK inhibition in BRAFV600E colorectal cancer.\u0026quot; For this study, we specifically used the pre-treatment samples to analyze the baseline immune and tumor microenvironmental states prior to therapy. Clinical information for this cohort is summarized in\u0026nbsp;Table 5.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eSpatial transcriptomics data acquisition and preprocessing\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eSpatial transcriptomics data were obtained from the publicly available dataset GSE225857, which includes FFPE tissue sections from four colorectal cancer (CRC) patient samples. The dataset was generated using the 10x Genomics Visium Spatial Gene Expression platform for FFPE tissues, and sequencing was performed on the Illumina NovaSeq 6000 system. For this study, we used the preprocessed UMI count matrices and accompanying H\u0026amp;E-stained tissue images provided with the dataset. Spatial transcriptomic analyses were carried out using Seurat (v5.2.1) and the SPATA2\u003csup\u003e11\u003c/sup\u003e package in R. Analyses included quality control filtering, normalization, dimensional reduction, clustering, and spatial feature extraction. Clinical information for this cohort is summarized in Table 6.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCell lines and cell culture\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eLS174T cells were a gift from Dr Young-Hee Lim. THP-1 (40202) cells were obtained from the Korean Cell Line Bank. LS174T and THP-1 cells were cultured in Dulbecco\u0026rsquo;s modified Eagle medium (DMEM; Welgene) or Roswell Park Memorial Institute (RPMI) medium (Welgene), respectively, supplemented with 10% fetal bovine serum (FBS, Gibco, 12483020) and 1% penicillin\u0026ndash;streptomycin (Hyclone, SV30010) in a humidified incubator at 37\u0026thinsp;\u0026deg;C with 5% CO₂.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eLentiviral constructs and transduction\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe CRISPR\u0026ndash;dCas9 Synergistic Activation Mediator (SAM) system was employed as a CRISPR activation (CRISPRa) tool to enhance gene expression. This system comprised two plasmids: lentiMPHv2 (Addgene, 89308) and lentiSAMv2 (Addgene, 75112). Single guide RNAs (sgRNAs) were designed using the CRISPICK and CHOPCHOP platforms and synthesized by Bionics (Korea). sgRNA sequences were cloned into the lentiSAMv2 backbone according to the protocol provided by Addgene. Briefly, lentiSAMv2 was digested with Esp3I (Enzynomics, R116S), and annealed oligonucleotides were ligated into the digested vector. Positive constructs were verified by Sanger sequencing. For lentiviral production, HEK293T cells were seeded in 6-well plates and cultured overnight to reach 70\u0026ndash;80% confluence. Cells were transfected with a plasmid mixture of pMD2.G (Addgene, 12259), psPAX2 (Addgene, 12260), and either lentiMPHv2, lentiSAMv2-sgRNA, or lentiSAMv2-scramble gRNA using polyethylenimine in serum-free DMEM. After 6 h, the medium was replaced with DMEM supplemented with 10% FBS and 1% penicillin\u0026ndash;streptomycin. The viral supernatant was harvested 48 h later, filtered through a 0.45-\u0026micro;m filter, and used for transduction. For gene activation in LS174T cells, lentiMPHv2 infection was performed for 48 h, followed by hygromycin selection, and subsequently, cells were infected with lentiSAMv2-sgRNA for 48 h and selected with blasticidin. The sgRNA sequences (5\u0026prime;\u0026ndash;3\u0026prime;) were as follows: sgCDX2, GAGGTTAAAGTGCACCAGGT; sgSLC39A4, GGTTGTCCAGGGCCAGACTG; sgSLC39A5, GGTGGGGTGTCCCTAGAAGG.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eReal-time qPCR\u0026nbsp;\u003c/strong\u003e\u003cbr\u003e\u0026nbsp;Total RNA was isolated from cultured cells using RiboEx\u0026trade; (GeneAll, 301) and quantified with a NanoDrop spectrophotometer (Thermo Fisher Scientific). Complementary DNA (cDNA) was synthesized from 2.5\u0026thinsp;\u0026micro;g of total RNA using the RevertAid First Strand cDNA Synthesis Kit (Thermo Fisher Scientific, K1621) in a total volume of 10\u0026thinsp;\u0026micro;L. Quantitative PCR was performed on either a StepOnePlus or QuantStudio 6 Flex Real-Time PCR System (Applied Biosystems) using SYBR Green reagents (Enzynomics, RT500 or RT501). Gene expression levels were normalized to \u0026beta;-actin, and relative expression was calculated using the \u0026Delta;\u0026Delta;Ct method. All primer sequences used in this study are listed in Table 7.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eWestern blot and antibodies\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eCells were lysed in RIPA buffer supplemented with protease and phosphatase inhibitor cocktails. Protein concentrations were determined using a Bradford assay. Equal amounts of protein (30\u0026ndash;40\u0026thinsp;\u0026micro;g) were separated by SDS\u0026ndash;PAGE, transferred to PVDF membranes, and blocked with 3% BSA in TBST. Membranes were incubated with primary antibodies overnight at 4\u0026thinsp;\u0026deg;C, followed by HRP-conjugated secondary antibodies for 1\u0026thinsp;h at room temperature. Protein bands were detected using an ECL detection kit (Thermo Fisher Scientific, 32106 or GLPBio, GK10008) and imaged with a UVP ChemStudio system (Analytik Jena). Primary antibodies were as follows: anti-CD24 (clone SN3, Santa Cruz Biotechnology, SC-19585; Western blot (WB) 1:1,000; flow cytometry and phagocytosis assay 10 \u0026mu;g\u0026thinsp;ml⁻\u0026sup1;) and anti-CDX2 (clone D11D10, Cell Signaling Technology, 12306; WB 1:1,000; immunofluorescence (IF) 1:400). Secondary antibodies included Goat anti-Mouse IgG (H+L) Cross-Adsorbed Secondary Antibody, Alexa Fluor\u0026trade; 488 (Invitrogen, A-11001; flow cytometry 1:500) and Goat anti-Rabbit IgG (H+L) Cross-Adsorbed Secondary Antibody, Alexa Fluor\u0026trade; 594 (Invitrogen, A-11012; IF 1:500).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eImmunofluorescence\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eCells were seeded on coverslips and fixed with 4% paraformaldehyde for 15\u0026thinsp;min at room temperature. Fixed cells were washed three times with DPBS and permeabilized with 0.1% Triton X-100 for 10\u0026thinsp;min at room temperature, followed by blocking in 3% BSA in DPBS for 1\u0026thinsp;h at room temperature. Cells were then incubated with primary antibodies diluted in 1% BSA in DPBS at 4\u0026thinsp;\u0026deg;C overnight. After washing with DPBS, cells were incubated with fluorescently labeled secondary antibodies for 1\u0026thinsp;h at room temperature. Nuclei were counterstained with DAPI, and coverslips were mounted using Fluoromount Aqueous Mounting Medium (Sigma). Images were acquired using an ECLIPSE Ts2-FL microscope (Nikon).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFlow cytometry\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eFlow cytometric analyses were performed to measure intracellular Zn\u0026sup2;⁺ levels and membrane CD24 expression. For intracellular Zn\u0026sup2;⁺ quantification, cells were harvested by trypsinization and incubated with 1\u0026thinsp;\u0026mu;M FluoZin\u0026trade;-3 (Invitrogen, F24195) for 30\u0026thinsp;min at 37 \u0026deg;C. Cells were then washed twice with DPBS and further incubated for 30\u0026thinsp;min at 37\u0026thinsp;\u0026deg;C to allow de-esterification. For analysis of membrane CD24 expression, cells were detached using Accutase (Millipore, SCR005), incubated with 10\u0026thinsp;\u0026mu;g\u0026thinsp;ml⁻\u0026sup1; anti-CD24 antibody in FACS buffer (0.5% BSA in DPBS) for 1\u0026thinsp;h on ice, washed twice with ice-cold FACS buffer, and then incubated with Alexa Fluor\u0026trade; 488-conjugated secondary antibody for 30\u0026thinsp;min on ice prior to flow cytometric analysis. All samples were analyzed using an Attune\u0026trade; NxT flow cytometer (Thermo Fisher Scientific) with excitation/emission, 488/520 nm. Data were analyzed using FlowJo software (v10, BD Biosciences), and MFI values were calculated from the fluorescence signals obtained.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ePhagocytosis assay\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTHP-1 cells were differentiated into macrophages by incubation with 100\u0026thinsp;ng\u0026thinsp;ml⁻\u0026sup1; phorbol 12-myristate 13-acetate (PMA; GLPBio, GN10444) for 48\u0026thinsp;h. Macrophages were polarized to the moMC phenotype with 20\u0026thinsp;ng\u0026thinsp;ml⁻\u0026sup1; interferon-\u0026gamma; (IFN-\u0026gamma;; Enzynomics, C006) and 100\u0026thinsp;ng\u0026thinsp;ml⁻\u0026sup1; lipopolysaccharide (LPS; Sigma-Aldrich, L2880), or to the RTMC phenotype with 20\u0026thinsp;ng\u0026thinsp;ml⁻\u0026sup1; interleukin-4 (IL-4; Enzynomics, C008) and 20\u0026thinsp;ng\u0026thinsp;ml⁻\u0026sup1; interleukin-13 (IL-13; Enzynomics, C009) for 48\u0026thinsp;h. A total of GFP⁺ THP-1 cells (2\u0026thinsp;\u0026times;\u0026thinsp;10⁵) were differentiated into moMC-like macrophages on 18-mm coverslips in 12-well plates. LS174T cells were detached using Accutase, labeled with pHrodo\u0026trade; Red succinimidyl ester (Thermo Fisher Scientific, P36600) at a 1:30,000 dilution in PBS for 1\u0026thinsp;h at 37\u0026thinsp;\u0026deg;C, and washed twice with DMEM supplemented with 10% FBS and 1% penicillin\u0026ndash;streptomycin. Labeled LS174T cells (1\u0026thinsp;\u0026times;\u0026thinsp;10⁶ per well) were added to macrophages in serum-free RPMI with or without 10\u0026thinsp;\u0026mu;g\u0026thinsp;ml⁻\u0026sup1; anti-CD24 antibody (clone SN3; Santa Cruz Biotechnology); an isotype control antibody (clone MOPC-21, BioXcell) was included at the same concentration. After 2\u0026thinsp;h of co-culture at 37\u0026thinsp;\u0026deg;C, cells were washed five times with serum-free RPMI and imaged using a fluorescence microscope (ECLIPSE Ts2-FL, Nikon). Phagocytosis was quantified as the ratio of pHrodo Red+ GFP⁺ THP-1 macrophages.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eStatistical analysis in experimental validations\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eStatistical analyses were performed using GraphPad Prism version 6 (GraphPad Software). All experiments were performed in at least three independent biological replicates. Data are presented as mean\u0026thinsp;\u0026plusmn;\u0026thinsp;SEM. For comparisons between two groups, exact P values were calculated using unpaired two-tailed t-tests. For comparisons among more than two groups, P values were determined using two-way ANOVA followed by Tukey\u0026rsquo;s multiple comparisons test. Data distribution was assumed to be normal, but this was not formally tested. Statistical significance was defined as P\u0026thinsp;\u0026lt;\u0026thinsp;0.05.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConstructing a predictive model for ICI response in CRC\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTo develop a predictive model of ICI response in CRC, we integrated five publicly available ICI-treated CRC transcriptomic cohorts (MSS1: n = 9, MSS2: n = 34, MSS3: n = 12, MSI1: n = 70, MSI2: n = 41), totaling 166 samples. Batch effects were removed using the removeBatchEffect function from the limma R package. The integrated dataset was randomly partitioned into a modeling set (n = 138) and an independent testing set (n = 28). The modeling set was further split into a training set (n = 110) and validation set (n = 28). Model construction was performed using the PyCaret machine learning framework in Python. PyCaret was used to automate the comparison of multiple classification algorithms under standardized settings (fold = 5, fold_strategy = StratifiedKFold, session_id = 2025). Among the tested models, the Gradient Boosting Classifier demonstrated the highest predictive performance (AUC = 0.7262) and was selected for final deployment. Other top-performing models included AdaBoost Classifier (AUC = 0.6735), Extra Trees Classifier (AUC = 0.6515), Extreme Gradient Boosting (AUC = 0.6314), and Random Forest Classifier (AUC = 0.6263). The final model was trained using seven biologically informed features\u0026mdash;SLC39A2, SLC39A4, SLC39A5, SLC39A10, CDX2, CD24, and SIGLEC10\u0026mdash;identified in this study as key regulators of zinc transport and immune evasion in colorectal cancer. Model performance was evaluated using the area under the receiver operating characteristic curve (AUC). To benchmark our model, we trained and evaluated additional Gradient Boosting Classifier using previously published gene sets known to be associated with ICI response in either pancancer or CRC-specific contexts\u003csup\u003e12\u0026ndash;19\u003c/sup\u003e. Each comparator model was trained using the same 138-sample modeling set and evaluated on the 28-sample testing set to assess generalizability. To ensure balanced representation of both clinical cohorts and response classes, each of the five ICI-treated CRC datasets (MSI1, MSI2, MSS1, MSS2, MSS3) was independently split into modeling (85%) and external testing (15%) sets. Splitting was performed separately per cohort to maintain cohort-wise sample proportions in both subsets. As a result, the modeling set (n = 138) and external testing set (n = 28) preserved the original composition of the five cohorts: MSI1 (42.75% vs. 39.29%), MSI2 (24.64% vs. 25.00%), MSS1 (5.07% vs. 7.14%), MSS2 (20.29% vs. 21.43%), MSS3 (7.25% vs. 7.14%). Response class distribution was also maintained, with responders (R) comprising 62.32% of the modeling set and 53.57% of the testing set, and non-responders (NR) comprising 37.68% and 46.43%, respectively. Clinical information for this cohort is summarized in Table 8 and corresponding biomarker gene sets for benchmarking are summarized in Table 9.\u003c/p\u003e"},{"header":"Results","content":"\u003cp\u003e\u003cstrong\u003eMultiomics clustering identifies three SLC transporter-centered archetypes in CRC patients \u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eSLC transporters are pivotal mediators of metabolite flux that sculpt the tumor microenvironment\u003csup\u003e20\u003c/sup\u003e. To elucidate the regulatory architecture of SLC transporters in CRC, we conducted integrative multiomics profiling encompassing more than 300 SLC transporter genes and 500 cancer driver genes across 258 TCGA-COAD patient samples (Fig. 1A). The multiomics dataset included transcriptomic alterations, somatic mutations, copy number variations, and DNA methylation profiles, enabling a comprehensive assessment of SLC transporter-associated regulatory programs. To resolve distinct patient subtypes orchestrated by SLC transporter-centric programs, we applied a dimensionality reduction-based clustering strategy using the Multi-Omics Factor Analysis (MOFA) algorithm\u003csup\u003e21\u003c/sup\u003e.\u0026nbsp;The clustering model inferred latent factors that integrated multiple feature modalities, capturing variation across patient samples. The optimal factor combination was determined by maximizing the t-ratio, ensuring clear separation among clusters (Fig. S1A). Functional annotation of these factors revealed strong associations with metabolic and immune-related pathways (Fig. 1B). Mapping patients onto the best-explained factor space identified three distinct archetypes (Fig. 1C), which were subsequently characterized based on their association with specific SLC transporters, biological pathways, microenvironmental signatures, and molecular subtypes (Fig. 1D). Among these, archetype 1 exhibited a notable enrichment of SLC39 family members (SLC39A2, SLC39A4, SLC39A5, and SLC39A10 were ranked in top 5%, Fig. S1B), which correlated with the activation of zinc-related pathways. In line with prior studies implicating the SLC39 family in zinc import\u003csup\u003e22\u003c/sup\u003e, zinc influx transporters were more highly expressed than SLC30 family-mediated efflux counterparts (Fig. S1C). Patients within this archetype predominantly aligned with the CMS2 molecular subtype, characterized by WNT activation, microsatellite stability (MSS), and an immune-desert phenotype\u003csup\u003e23\u003c/sup\u003e. In line with these features, this archetype showed elevated activity in WNT- and MSS-associated pathways while displaying a depletion of key infiltrating immune cells, as inferred from transcriptomic enrichment analysis. These findings suggest that archetype 1 represents a CMS2-like subtype, distinguished by SLC39 overexpression and enhanced zinc transport.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eSLC39-mediated zinc transport induces CD24 overexpression via CDX2 enhancer binding in tumor cells\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTo elucidate the molecular mechanisms underlying SLC39 family overexpression and zinc transport activation in archetype 1, we identified key signaling molecules associated with this pathway (Fig. 2A, top). Cancer driver genes relevant to archetype 1 were analyzed, revealing CDX2 as a top-ranked transcription factor (TF), along with the SLC39 family (Fig. S1D, E). CDX2 plays a critical role in gut development and homeostasis\u003csup\u003e24\u003c/sup\u003e. Notably, both CDX2 expression and inferred protein activity were positively correlated with SLC39-mediated zinc influx and archetype 1 status (Fig. S2A). Comprehensive multi-omics correlation analyses further identified CD24 as a downstream effector of CDX2, consistent with its reported function in mediating immune-repressive signaling\u003csup\u003e25\u003c/sup\u003e and aligning with the immune-desert characteristics of archetype 1. Across the TCGA and CPTAC CRC cohorts, expression of SLC39 transporters, CDX2, and CD24 demonstrated broad positive correlations with zinc pathway activity (Fig. 2A), supporting a zinc-driven transcriptional cascade linking SLC39 influx to CDX2-dependent CD24 activation. This regulatory cascade was further supported by COMPASS-based metabolic flux modeling\u003csup\u003e26\u003c/sup\u003e, which quantifies zinc-associated reaction-level flux states beyond conventional gene set scoring, and independently recapitulated the SLC39\u0026ndash;CDX2\u0026ndash;CD24 signaling axis (Fig.\u0026nbsp;S2B).\u0026nbsp;To investigate the genomic basis of SLC39 overexpression, we analyzed copy number variation (CNV) and\u0026nbsp;identified\u0026nbsp;a positive correlation between SLC39 expression and CNV in both TCGA and CPTAC cohorts (Fig. 2B). Additionally, CNV inference from single-cell transcriptomic data of 21,698 CRC cells\u003csup\u003e27\u003c/sup\u003e using the inferCNV algorithm\u003csup\u003e28\u003c/sup\u003e confirmed genomic amplification associated with SLC39 family overexpression. Given that CDX2 functions as transcriptional regulators, we estimated their regulatory activity on target genes using the VIPER algorithm\u003csup\u003e29\u003c/sup\u003e, based on a CRC regulon network. The TF activities of CDX2 exhibited strong positive correlations with upstream SLC39-mediated zinc influx and downstream CD24 expression in both TCGA and CPTAC cohorts (Fig. 2C). To further validate the regulatory link, we analyzed epigenomic features of the CD24 locus (Fig. 2D). Hi-C chromatin interaction data from the HCT116 CRC cell line\u003csup\u003e30\u003c/sup\u003e confirmed that the CD24 locus and a putative enhancer region\u003csup\u003e31\u003c/sup\u003e were co-localized within the same topologically associated domain (Fig. 2D, top). In the Caco-2 CRC cell line, active histone marks\u003csup\u003e32\u003c/sup\u003e such as H3K4me1, H3K36me3, and H3K27ac were enriched at the putative enhancer (Fig. 2D, middle). scATAC-seq data from 18,275 cells across three CRC patients\u003csup\u003e33\u003c/sup\u003e further revealed high chromatin accessibility at this enhancer region, specifically in cancer cells compared with other cell types (Fig. 2D, bottom). To assess enhancer activity in patient samples, we analyzed ATAC-seq data from 25 TCGA-COAD patients\u003csup\u003e9\u003c/sup\u003e and observed significant positive correlations between CD24 enhancer accessibility and upstream regulatory signals, including SLC39-mediated zinc influx, CDX2 TF activity, and CD24 expression (Fig. 2E). Furthermore, motif enrichment analysis confirmed significant binding of CDX2 at the putative enhancer region, supporting their direct role in CD24 activation (Fig. 2F). Given reports of zinc-induced phosphorylation of CDX2 modulating its transcriptional activity, we examined phosphoproteome datasets and observed positive trends linking zinc influx, CDX2 activity, CD24 expression, and phosphorylation at CDX2 S176, albeit without statistical significance due to limited sample size (Fig. 2G). Candidate regulatory kinases potentially mediating this phosphorylation were also enriched in zinc-associated networks (Fig. S2C). To corroborate these findings, we performed 3D protein\u0026ndash;DNA docking using AlphaFold3-predicted complex structures, a deep learning model capable of high-accuracy biomolecular complex prediction\u003csup\u003e34\u003c/sup\u003e, and the HDOCK server, a hybrid docking platform supporting protein\u0026ndash;DNA interactions with integrated scoring and modeling features\u003csup\u003e35\u003c/sup\u003e. Simulations confirmed high-confidence CDX2 binding to the CD24 enhancer, with S176 phosphorylation enhancing CDX2 binding affinity (Fig. 2H, I). Collectively, these findings define a zinc-dependent transcriptional circuit in which SLC39-driven zinc influx promotes CDX2 phosphorylation and enhancer-mediated activation of CD24 in CRC.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCD24 overexpression driven by zinc influx via the SLC39 family is restricted to cancer cells in the tumor microenvironment \u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eMost signal cascades identified in this study were derived from bulk tissue data, which consist of heterogeneous cell populations. To resolve cell type-specific signaling, we analyzed the impact of zinc influx mediated by SLC39 overexpression at the single-cell level using a transcriptomic dataset comprising 174,547 single cells from 26 CRC patients\u003csup\u003e27\u003c/sup\u003e. Canonical cell types were annotated based on established marker genes (Fig. 3A and Fig. S3A), and the expression patterns of zinc influx-mediated SLC39 signaling leading to CD24 upregulation were examined across cell types (Fig. 3B, top). Correlation analyses among the signaling factors were then performed in a cell type-specific manner (Fig. 3B, bottom). These analyses revealed a strong and selective association of SLC39-driven zinc influx with CDX2 activity and CD24 expression in malignant epithelial cells, whereas non-malignant compartments exhibited minimal or marginal correlations, consistent with the limited resolution of scRNA-seq data. To address this limitation, we conducted metacell analysis using SEACells\u003csup\u003e36\u003c/sup\u003e, which mitigates scRNA-seq data sparsity while preserving cellular heterogeneity by defining compact and well-separated metacells (Fig. 3C and Fig. S3B). This approach enabled the detection of robust CD24 overexpression driven by zinc influx via SLC39, exclusively in cancer cells (Fig. 3D), accompanied by the most significant positive correlation in cancer cells compared to other cell types (Fig. 3E).\u0026nbsp;To functionally validate these findings, we overexpressed SLC39A4 or SLC39A5, the two SLC39 family members most strongly associated with zinc influx in our multiomics analysis (Fig. 2A), in the CRC cell line LS174T (Fig. S3C). Both transporters enhanced intracellular zinc levels, as evidenced by FluoZin-3 fluorescence (Fig. 3F). Concomitantly, overexpression of either gene upregulated CD24 at transcript and protein levels (Fig. 3G, H) and increased membrane-localized CD24, confirmed by flow cytometry (Fig. 3I and Fig. S3D). These cells further displayed enhanced CDX2 transcriptional activity and increased nuclear localization (Fig. 3G and Fig. S3E\u0026ndash;G).\u0026nbsp;CD24 has been established as a potent \u0026ldquo;don\u0026rsquo;t eat me\u0026rdquo; signal analogous to CD47 and is broadly overexpressed across multiple tumor types\u003csup\u003e25\u003c/sup\u003e. However, comparative analysis demonstrated that CD24 but not CD47 was selectively enriched in cancer cells and exhibited significant positive correlations with SLC39-mediated zinc influx (Fig. S3H\u0026ndash;J). Consistently, overexpression of SLC39A4 or SLC39A5 induced CD24 but not CD47 transcripts in CRC cells (Fig. 3J). Together, these findings delineate a zinc-driven regulatory mechanism in which SLC39 activity selectively couples to CD24 expression in cancer cells, establishing a distinct immune evasion pathway independent of the canonical CD47 axis.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eSLC39-mediated zinc influx directs cell state trajectory toward CMS2-like cancer cells associated with poor survival and immunotherapy resistance\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTo investigate the influence of SLC39-mediated zinc influx on cancer cell state dynamics, we applied the Palantir algorithm\u003csup\u003e37\u003c/sup\u003e to single-cell transcriptomic data (Fig. S4A). CMS-based subtyping was performed to annotate the inferred cell state trajectories (Fig. 4A, left). Three terminal states were identified based on pseudotime and entropy (Fig. 4A, right), with terminal state 1 converging into a subset of CMS2-like cancer cells, while terminal states 2 and 3 remained heterogeneous or unclassified (Fig. S4B). Analysis of SLC39-driven zinc influx components revealed their selective overexpression or TF activation in terminal state 1 (Fig. 4B). Gene trend analysis further confirmed that signaling cascades downstream of SLC39-mediated zinc influx were specifically upregulated along the trajectory toward terminal state 1 (Fig. 4C). Notably, \u003cem\u003eCD24\u003c/em\u003e expression exhibited a delayed onset relative to upstream regulators, consistent with its function as a downstream effector of zinc-induced \u003cem\u003eCDX2\u003c/em\u003e activation. Pathway enrichment analysis of terminal state-specific expression signatures confirmed that terminal state 1 exhibited significant activation or repression of CMS2 subtype-associated pathways (Fig. 4D). Given the immunosuppressive nature of CMS2-like tumors identified in our archetype analysis (Fig. 1E), we next examined the immune-modulatory potential of cancer cells in terminal state 1. Notably, CD24 emerged as the top-ranked immunomodulator overexpressed in this state, reinforcing its role as a key effector of the SLC39-driven zinc influx axis (Fig. 4E and Fig. S4C). The canonical \u0026lsquo;don\u0026rsquo;t eat me\u0026rsquo; signal CD47 was not significantly upregulated in terminal state 1.\u003c/p\u003e\n\u003cp\u003eExtension of the single-cell\u0026ndash;derived findings to patient cohorts was achieved by constructing a terminal state signature matrix using the CellRank algorithm\u003csup\u003e38\u003c/sup\u003e,\u0026nbsp;which enabled calculation of terminal state probability scores from bulk transcriptomic datasets\u0026nbsp;(Fig. 4F, left). Application of this model to 288 TCGA-COAD patient samples revealed that terminal state 1 probability scores were significantly enriched in archetype 1 patients, characterized by CMS2-like features and activated SLC39-mediated zinc influx signaling (Fig. 4F, right). In contrast, terminal states 2 and 3 probabilities were heterogeneously distributed across archetype 2 and 3 patients. Survival analysis demonstrated that patients with higher terminal state 1 probability scores exhibited significantly shorter progression-free and disease-free survival (Fig. 4G). This association was specific to terminal state 1, as no significant survival differences were observed for terminal states 2 or 3 (Fig. S4D). To investigate the relevance of this state to immunotherapy response, we analyzed single-cell transcriptomes from 70,718 cancer cells derived from CRC patients treated with PD-1 inhibitors\u003csup\u003e39\u003c/sup\u003e (Fig. 4H, I). Intriguingly, overexpression or activation of components within the SLC39-driven zinc influx pathway was consistently associated with non-response to therapy and shorter progression-free survival. To elucidate the regulatory landscape underlying these clinical associations, we constructed patient group-specific cancer cell networks using scHumanNet\u003csup\u003e40\u003c/sup\u003e stratified by clinical response or progression-free survival status. Mapping of terminal state marker genes and core components of the SLC39-driven zinc influx pathway onto these networks revealed significant enrichment of terminal state 1 markers and zinc influx components within non-responder and short-survivor networks (Fig. 4J, K). Conversely, terminal state 2 and 3 markers were predominantly enriched in responder and long-survivor networks. This distribution pattern remained consistent regardless of the number of marker genes included in the analysis (Fig. S4E). Systematic importance evaluation using centrality metrics further demonstrated that terminal state 1 markers and SLC39-driven zinc influx components exhibited higher importance scores specifically in non-responder and short-survivor networks compared to other marker genes (Fig. 4L). This pattern was robust to variations in marker gene selection criteria (Fig. S4F). Collectively, these findings demonstrate that the SLC39-driven zinc influx axis drives a CMS2-like cancer cell state marked by CD24 overexpression, which is clinically associated with poor prognosis and resistance to immunotherapy.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCD24 mediates pleiotropic immune evasion signaling in the tumor microenvironment through SIGLEC10 interactions\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eCMS2-like cancer cells of archetype 1 exhibit pronounced immune-desert characteristics within the tumor microenvironment (TME) (Fig. 1E), prompting investigation of how CD24-overexpressing tumor cells communicate with surrounding immune populations.\u0026nbsp;We sought to define the cellular interface of CD24-overexpressing tumor cells by systematically mapping receptor\u0026ndash;ligand interactions within the TME. Leveraging a curated ligand\u0026ndash;receptor database\u003csup\u003e41\u003c/sup\u003e and applying the NICHES framework\u003csup\u003e42\u003c/sup\u003e, we prioritized the top five candidate interactions, with cancer cells designated as the sender population (Fig. 5A and Fig. S5A). Among these, CD24 emerged as a tumor-selective ligand predicted to interact exclusively with SIGLEC10 receptors expressed on myeloid cells, suggesting a specialized axis of immune suppression mediated through CD24\u0026ndash;SIGLEC10 engagement. Chromatin accessibility analysis further revealed that SIGLEC10 promoter were epigenetically active exclusively in myeloid populations within CRC patient samples\u003csup\u003e33\u003c/sup\u003e (Fig. S5B). To pinpoint the specific myeloid subpopulations expressing SIGLEC10, we performed subclustering using lineage-specific markers (Fig. S5C). This analysis revealed significant overexpression of SIGLEC10 in monocytes, resident tissue macrophages (RTMCs), monocyte-derived macrophages (moMCs), and monocyte-derived dendritic cells (moDCs) compared to other myeloid subsets (Fig. 5B). These myeloid-lineage cells originate from distinct progenitors: RTMCs derive from hematopoietic stem cells, while monocytes, moMCs, and moDCs derive from erythro-myeloid progenitors (Fig. 5C). Cell state trajectory analysis using the Palantir algorithm\u003csup\u003e37\u003c/sup\u003e revealed bifurcating developmental paths from monocytes toward moMC and moDC lineages (Fig. 5D and Fig. S5D). Gene trend profiling along these trajectories showed progressive upregulation of SIGLEC10, confirming their lineage development through stage-specific marker expression (Fig. 5E). To elucidate the downstream signaling cascades triggered by CD24-SIGLEC10 interactions, we employed the NicheNet algorithm\u003csup\u003e43\u003c/sup\u003e, integrating ligand-receptor interactions, signaling pathways, and transcriptional regulatory networks. In moMCs, CD24 exhibited high ligand activity and selectively engaged SIGLEC10 (Fig. 5F, left), leading to the activation of downstream transcriptional programs exclusively expressed in moMCs (Fig. 5F, right). Functional enrichment analysis of these target genes revealed strong associations with negative regulation of macrophage activation, phagocytosis, endocytosis, and immune responses\u0026mdash;hallmarks of the \u0026lsquo;don\u0026rsquo;t eat me\u0026rsquo; signal (Fig. 5G). Similarly, moDCs exhibited a parallel immune-repressive signaling trend, though the specific downstream gene sets differed from those in moMCs (Fig. S5E, F). Beyond moMCs and moDCs, RTMCs also interacted with cancer cells via CD24-SIGLEC10, but their downstream signaling profile diverged, predominantly influencing cell adhesion and chemotaxis pathways (Fig. 5H, I). These cell\u0026ndash;cell interaction profiles collectively reveal CD24 overexpression in cancer cells elicits pleiotropic responses via SIGLEC10 binding across moDCs, moMCs, and RTMCs, orchestrating immunosuppressive programs involving immune evasion and adhesion remodeling.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eRecurrent CD24\u0026ndash;SIGLEC10-enriched niches define immunorepressive microenvironments in colorectal cancer\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTo evaluate whether the pleiotropic cellular interactions delineated in our analysis are spatially organized within tumor tissues, we analyzed spatial transcriptomes from four CRC patient samples\u003csup\u003e44\u003c/sup\u003e. Cell type deconvolution was performed using SPOTlight\u003csup\u003e45\u003c/sup\u003e with canonical marker genes (Fig. 6A). Spatial domains enriched for CD24\u0026ndash;SIGLEC10 interactions were identified using SPATA2\u003csup\u003e11\u003c/sup\u003e, enabling the delineation of pleiotropic signaling regions (Fig. 6B). Notably, cancer cells, moMCs, moDCs, and RTMCs\u0026mdash;key cell types engaged in CD24\u0026ndash;SIGLEC10 signaling\u0026mdash;were consistently enriched at the tumor periphery near these interaction hotspots (Fig. 6A, B, bottom). To investigate the functional relevance of this cellular ecosystem, we quantified the activity of core components in the SLC39-mediated zinc influx axis relative to spatial proximity to CD24\u0026ndash;SIGLEC10-enriched areas. All signaling components demonstrated peak activity adjacent to these regions, with progressive attenuation observed at increasing distances (Fig. S6A, B). Consistent spatial gradients were also observed for immune evasion and cell adhesion pathways, with maximal activation near the interaction zones and gradual decline further away (Fig. 6C, D), providing functional validation of these spatially defined immunoregulatory microenvironments. Based on these observations, we characterized a cellular niche representing an immunosuppressive microenvironmental ecosystem mediated by CD24-SIGLEC10 pleiotropic interactions, encompassing signaling cascades of CD24 overexpression in cancer cells (Fig. 6E, bottom, red background box), CD24-SIGLEC10 interactions (Fig. 6E, bottom, blue background box), and downstream pleiotropic signals regulating immune evasion and cell adhesion (Fig. 6E, bottom, green background box). Using the FindClusters function of Seurat\u003csup\u003e46\u003c/sup\u003e, we identified between 10 and 14 distinct cellular niches across the four CRC tissue samples (Fig. 6E and Fig. S6C). Importantly, two of the four tissue samples exhibited the immunosuppressive cellular niches we characterized, with concordant activation patterns of signaling cascades mediated by CD24-SIGLEC10 interactions in tissue peripheral areas (Fig. 6E, top). In contrast to these niches, the CD47-SIRPA interaction\u0026mdash;a putative major \u0026lsquo;don\u0026apos;t eat me\u0026rsquo; signal\u0026mdash;showed minimal or only partial activation in these regions (Fig. S6D). Collectively, these spatial transcriptomic analyses confirmed the recurrent presence of cellular niches characterized by immune-suppressive pleiotropic signaling via CD24-SIGLEC10 interactions across CRC patient tissues, providing spatial context to our molecular findings.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConserved pleiotropic CD24\u0026ndash;SIGLEC10 signaling defines immune-evasive myeloid networks associated with immunotherapy resistance\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWe investigated whether pleiotropic immune-evasive signaling mediated by CD24\u0026ndash;SIGLEC10 interactions is preserved at the patient level. To this end,\u0026nbsp;we derived cell type-specific gene signatures from the top 50 marker genes of moMCs, moDCs, and RTMCs identified in the single-cell transcriptomic datasets\u003csup\u003e47\u003c/sup\u003e. These gene sets were mutually exclusive and independent of the downstream CD24\u0026ndash;SIGLEC10 signaling modules (Fig.\u0026nbsp;S7A). Using these signatures, we computed enrichment scores for moMCs and RTMCs across TCGA-COAD samples and stratified patients into subgroups based on combinatorial high (+) or low (\u0026ndash;) enrichment status (Fig. 7A). The potential for intercellular CD24\u0026ndash;SIGLEC10 interactions was assessed using the LIANA module from the DecoupleR framework\u003csup\u003e48\u003c/sup\u003e, which quantifies ligand-receptor co-expression relative to background distributions. Patients exhibiting concurrent high enrichment of both moMCs and RTMCs demonstrated the strongest predicted CD24\u0026ndash;SIGLEC10 interaction potential (Fig. 7B). To assess functional consequences of these interactions, we analyzed gene modules associated with CD24\u0026ndash;SIGLEC10 activity, encompassing immune-evasive \u0026lsquo;don\u0026rsquo;t eat me\u0026rsquo; signals and adhesion-regulatory pathways. These modules showed robust positive correlations with moMC and RTMC enrichment scores, despite minimal gene overlap with the respective marker sets (Fig. 7C\u0026nbsp;and Fig.\u0026nbsp;S7A). This association remained consistent when moDCs were substituted for moMCs (Fig.\u0026nbsp;S7B\u0026ndash;D). Additionally, the immune-suppressive cellular landscape identified in TCGA patients was independently recapitulated in the CPTAC CRC cohort (Fig.\u0026nbsp;S7E), underscoring the robustness and translational significance of these findings. To mechanistically recapitulate these observations, we employed THP-1 cells differentiated into moMC-like or RTMC-like lineages using distinct polarization protocols (Fig. 7D). Each protocol selectively enhanced marker expression of the corresponding myeloid subtype. To test whether CD24\u0026ndash;SIGLEC10 engagement suppresses phagocytic activity, LS174T CRC cells were co-cultured with the differentiated moMC-like cells, and phagocytosis was quantified (Fig. 7E). SLC39A4- or SLC39A5-overexpressing CRC cells exhibited significantly reduced phagocytosis compared with control cells, an effect reversed by anti-CD24 antibody treatment, confirming that the suppression occurred through CD24\u0026ndash;SIGLEC10 interaction. Motivated by these mechanistic insights, we hypothesized that expression of key components within the CD24\u0026ndash;SIGLEC10 axis could stratify immunotherapy response. To test this, we integrated five independent ICI-treated CRC cohorts (n = 166) encompassing both MSS and MSI cases and applied batch correction to harmonize expression profiles (Fig. 7F and\u0026nbsp;Fig.\u0026nbsp;S7F). A Gradient Boosting Classifier was trained on a biologically informed seven-gene signature\u0026mdash;SLC39A2, SLC39A4, SLC39A5, SLC39A10, CDX2, CD24, and SIGLEC10\u0026mdash;selected for their central roles in zinc-driven immunosuppression. This model achieved the highest predictive accuracy among all benchmarked classifiers, with an AUC of 0.728 in an independent testing set, outperforming established biomarkers of adaptive immunity, stromal resistance, and tumor-intrinsic evasion\u003csup\u003e12\u0026ndash;19\u003c/sup\u003e (Fig. 7G). Together, these findings define a conserved, pleiotropic immunosuppressive circuit orchestrated by SLC39-driven zinc influx and CD24\u0026ndash;SIGLEC10 engagement. This pathway shapes an immune-evasive tumor microenvironment, suppresses macrophage phagocytosis, and serves as a clinically actionable predictor of ICI response in colorectal cancer.\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eOur study delineates a previously uncharacterized pleiotropic signaling axis initiated by SLC39-mediated zinc influx, which orchestrates immune-suppressive reprogramming in CRC. Through SLC transporter-centered multiomics clustering, we identified a distinct patient subgroup aligned with the CMS2 molecular subtype, marked by WNT activation, microsatellite stability, and an immune-desert phenotype. This CMS2-like archetype was characterized by overexpression of the SLC39 family, particularly SLC39A4 and 5, which drives zinc influx and triggers a transcriptional cascade culminating in epigenetic activation of CD24\u0026mdash;a recently recognized \u0026ldquo;don\u0026rsquo;t eat me\u0026rdquo; signal analogous to CD47. We uncovered that zinc functions as a cofactor enhancing the binding of the CDX2 transcription factor to a CD24 regulatory enhancer, promoting its overexpression, consistent with prior evidence that zinc activates CDX2 phosphorylation and transcriptional activity via the PI3K\u0026ndash;CDX2 pathway to regulate downstream targets in intestinal epithelial cells\u003csup\u003e\u003cspan citationid=\"CR49\" class=\"CitationRef\"\u003e49\u003c/span\u003e\u003c/sup\u003e. Multi-omics correlation, chromatin accessibility profiling, enhancer activity assessment, and 3D protein\u0026ndash;DNA docking collectively established the direct transcriptional control of CD24 by this zinc-responsive TF. Functionally, CRC cells overexpressing SLC39A4 or SLC39A5 exhibited increased intracellular zinc levels, confirming their zinc transport capacity, and upregulation of CDX2 target genes including CD24. Elevated transcript levels corresponded to increased protein abundance and membrane localization of CD24, indicating that zinc-mediated transcriptional regulation extends to cell\u0026ndash;cell communication through extracellular binding partners. Notably, this zinc-dependent overexpression of CD24 occurred independently of CD47, defining a discrete immunoevasive pathway within CMS2-like tumors. At the patient level, the SLC39-driven zinc influx state was robustly recapitulated and associated with significantly reduced progression-free and disease-free survival across over 250 TCGA-COAD patients, as well as resistance to ICIs in 26 CRC patients, where elevated expression of \u0026ldquo;don\u0026rsquo;t eat me\u0026rdquo; signals such as CD24 or CD47 has been reported to correlate with diminished ICI efficacy and primary resistance in colorectal and head and neck cancer\u003csup\u003e\u003cspan citationid=\"CR50\" class=\"CitationRef\"\u003e50\u003c/span\u003e,\u003cspan citationid=\"CR51\" class=\"CitationRef\"\u003e51\u003c/span\u003e\u003c/sup\u003e. Extending our analysis to the tumor microenvironment, we found that CD24 overexpression in cancer cells facilitates immune-evasive communication via SIGLEC10-expressing myeloid cells. Interestingly, this interaction exhibited lineage-specific pleiotropy: engagement of SIGLEC10 on moMCs and moDCs led to repression of phagocytosis-related genes, while signaling in RTMCs favored adhesion-related pathways. Spatial and single-cell transcriptomic analyses validated that these pleiotropic CD24\u0026ndash;SIGLEC10 interactions define recurrent immunosuppressive niches at tumor borders, where cancer\u0026ndash;myeloid cell contact is spatially enriched. Furthermore, CD24\u0026ndash;SIGLEC10 interaction strength and downstream signaling modules correlated with non-responsiveness to ICI therapy. Consistently, phagocytosis assays employing moMC-like macrophages and CRC cells demonstrated that overexpression of SLC39A4 or SLC39A5 suppressed macrophage-mediated engulfment, an effect reversed by CD24\u0026ndash;SIGLEC10 blockade. To translate these mechanistic findings into clinically applicable framework, we developed an ICI response prediction model integrating multi-cohort transcriptomic data. The model, trained on a seven-gene SLC39\u0026ndash;CD24\u0026ndash;SIGLEC10 signature, achieved the highest predictive performance (AUC\u0026thinsp;=\u0026thinsp;0.728) in the integrated ICI-treated CRC cohort, establishing the zinc-driven immunosuppressive program as a robust molecular predictor of therapeutic response. Although clinical translation of this axis is still emerging, anti-SIGLEC10 antibodies such as ONC-841 have recently entered early-phase clinical trials\u003csup\u003e\u003cspan citationid=\"CR52\" class=\"CitationRef\"\u003e52\u003c/span\u003e\u003c/sup\u003e, highlighting the druggability of this checkpoint pathway, whereas CD24-targeting antibodies (e.g., IMM47) have demonstrated preclinical efficacy but remain at the investigational stage\u003csup\u003e\u003cspan citationid=\"CR53\" class=\"CitationRef\"\u003e53\u003c/span\u003e\u003c/sup\u003e. These results confirm that SLC39-mediated zinc influx establishes a therapeutically actionable axis linking metabolic reprogramming to innate immune suppression. Together, our findings identify SLC39-driven zinc influx as a central regulator of immune-suppressive plasticity in CMS2-like CRC and uncover a multimodal immune evasion mechanism orchestrated through pleiotropic CD24\u0026ndash;SIGLEC10 signaling.\u003c/p\u003e \u003cp\u003eSLC transporters have been broadly implicated in tumor metabolism and microenvironmental remodeling\u003csup\u003e\u003cspan citationid=\"CR54\" class=\"CitationRef\"\u003e54\u003c/span\u003e\u003c/sup\u003e, yet their roles in CRC remain incompletely defined, particularly in the context of immune-desert or cold tumor states. Here, we delineate a mechanistic axis in which SLC39-driven zinc influx orchestrates an immunosuppressive program characteristic of the CMS2-like molecular subtype. This signaling cascade links metabolic reprogramming to immune evasion, establishing zinc-dependent transcriptional control as a critical determinant of tumor\u0026ndash;immune interactions. Within this framework, we identify CD24\u0026ndash;SIGLEC10 as the dominant anti-phagocytic checkpoint in CMS2-like CRC. Although CD47\u0026ndash;SIRPα signaling is a canonical \u0026ldquo;don\u0026rsquo;t eat me\u0026rdquo; pathway frequently upregulated in CRC and known to potentiate adaptive immune priming upon blockade\u003csup\u003e\u003cspan citationid=\"CR55\" class=\"CitationRef\"\u003e55\u003c/span\u003e,\u003cspan citationid=\"CR56\" class=\"CitationRef\"\u003e56\u003c/span\u003e\u003c/sup\u003e, our data show that SLC39-mediated zinc influx preferentially promotes CDX2-dependent CD24 transcription without concomitant CD47 upregulation. This nominates CD24\u0026ndash;SIGLEC10 as the principal innate immune checkpoint in this setting, reinforcing its relevance as a therapeutic target\u003csup\u003e\u003cspan citationid=\"CR57\" class=\"CitationRef\"\u003e57\u003c/span\u003e\u003c/sup\u003e. Importantly, our integrative analyses reveal that CD24\u0026ndash;SIGLEC10 interactions extend beyond tumor-associated macrophages (TAMs), engaging multiple myeloid subsets to enforce pleiotropic immune suppression. Among these, SPP1\u003csup\u003e+\u003c/sup\u003e moMCs sculpt hypoxic, fibroblast-enriched niches associated with poor prognosis\u003csup\u003e\u003cspan citationid=\"CR58\" class=\"CitationRef\"\u003e58\u003c/span\u003e,\u003cspan citationid=\"CR59\" class=\"CitationRef\"\u003e59\u003c/span\u003e\u003c/sup\u003e, while CCR7\u003csup\u003e+\u003c/sup\u003eLAMP3\u003csup\u003e+\u003c/sup\u003e moDCs acquire mregDC programs that dampen T-cell priming\u003csup\u003e\u003cspan citationid=\"CR60\" class=\"CitationRef\"\u003e60\u003c/span\u003e\u003c/sup\u003e. RTMCs further contribute by activating adhesion and hypoxia-adaptive pathways. Collectively, these lineage-specific programs demonstrate that zinc-dependent CD24 upregulation in tumor cells enables SIGLEC10 engagement across multiple myeloid compartments, thereby extending the canonical CD24\u0026ndash;SIGLEC10 immune checkpoint into a multicellular suppressive circuit.\u003c/p\u003e \u003cp\u003eWhile our study delineates the mechanistic underpinnings of SLC39-driven zinc influx and CD24-mediated immune evasion in CMS2-like CRC, it also reveals distinct regulatory architectures underpinning other archetypes. Archetypes 2 and 3, associated with the SLC22 and SLC35 families, respectively, exhibit immune and proliferative traits that are not attributable to zinc signaling, yet converge on unique CMS subtypes. As shown in Fig.\u0026nbsp;1E, archetype 2 aligns with the CMS4 subtype, characterized by epithelial\u0026ndash;mesenchymal transition (EMT), angiogenesis, and relative repression of cell-cycle pathways. Notably, this subtype is marked by epigenetic silencing of SLC22 transporters such as SLC22A18 and SLC22A5, whose promoter hypermethylation enhances tumor cell invasion and survival\u003csup\u003e\u003cspan citationid=\"CR61\" class=\"CitationRef\"\u003e61\u003c/span\u003e,\u003cspan citationid=\"CR62\" class=\"CitationRef\"\u003e62\u003c/span\u003e\u003c/sup\u003e. In contrast, archetype 3 recapitulates features of CMS1 tumors, which harbor microsatellite instability (MSI) and mismatch repair (MMR) deficiency. These tumors paradoxically couple defective DNA repair with proliferative advantage due to impaired TGF-β signaling\u003csup\u003e\u003cspan citationid=\"CR63\" class=\"CitationRef\"\u003e63\u003c/span\u003e,\u003cspan citationid=\"CR64\" class=\"CitationRef\"\u003e64\u003c/span\u003e\u003c/sup\u003e. Within this context, aberrant methylation and suppression of SLC35A3\u0026mdash;a nucleotide-sugar transporter\u0026mdash;has been implicated in compromised DNA repair fidelity and attenuated immune infiltration\u003csup\u003e\u003cspan citationid=\"CR65\" class=\"CitationRef\"\u003e65\u003c/span\u003e\u003c/sup\u003e. Together, these findings suggest that CRC subtypes deploy divergent transporter circuits: SLC22 silencing supports mesenchymal transition in CMS4, SLC35A3 repression augments immune dysfunction in CMS1, and SLC39-mediated zinc signaling orchestrates immune escape in CMS2. This underscores the metabolic plasticity that stratifies consensus molecular subtypes and reveals context-dependent vulnerabilities across CRC archetypes\u003csup\u003e\u003cspan citationid=\"CR66\" class=\"CitationRef\"\u003e66\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eIn acknowledging the limitation that our study did not employ mass spectrometry\u0026ndash;based metabolic flux assays to directly quantify metabolite exchange, we sought to overcome this gap by leveraging multi-omics integration, pathway-level analyses, and reaction-level flux modeling. Recent high-impact studies have demonstrated that transcriptomic and proteogenomic profiling can robustly infer tumor metabolic rewiring, such as glycolytic activation in MSI-high colorectal tumors\u003csup\u003e\u003cspan citationid=\"CR67\" class=\"CitationRef\"\u003e67\u003c/span\u003e\u003c/sup\u003e and oxidative phosphorylation imbalance in clear cell renal cell carcinoma\u003csup\u003e\u003cspan citationid=\"CR68\" class=\"CitationRef\"\u003e68\u003c/span\u003e\u003c/sup\u003e. Similarly, integrative omics of thyroid cancers revealed subtype-specific TCA cycle and one-carbon metabolism programs with elevated transporters and enzymes\u003csup\u003e\u003cspan citationid=\"CR69\" class=\"CitationRef\"\u003e69\u003c/span\u003e\u003c/sup\u003e, while proteogenomic stratification of NSCLC uncovered a distinct metabolic subtype with heightened glycolytic and bioenergetic activity\u003csup\u003e\u003cspan citationid=\"CR70\" class=\"CitationRef\"\u003e70\u003c/span\u003e\u003c/sup\u003e. Moreover, spatial multi-omics in gastric cancer has shown that correlating metabolite distributions with gene expression can resolve cell-type\u0026ndash;specific metabolic crosstalk in the tumor microenvironment\u003csup\u003e\u003cspan citationid=\"CR71\" class=\"CitationRef\"\u003e71\u003c/span\u003e\u003c/sup\u003e. Beyond conventional enrichment-based methods, the metabolic flux modeling\u003csup\u003e\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e\u003c/sup\u003e applied in this study has been validated by its ability to recapitulate lipid-induced shifts toward serine\u0026ndash;methionine metabolism in breast epithelial subpopulations, revealing redox-coupled metabolic plasticity that mirrors cancer-associated reprogramming dynamics\u003csup\u003e\u003cspan citationid=\"CR72\" class=\"CitationRef\"\u003e72\u003c/span\u003e\u003c/sup\u003e. Together, these references underscore that although direct metabolite tracing was not feasible, our use of multi-omics data and pathway inference represents a rigorous and widely validated strategy to dissect tumor metabolic states. In bulk transcriptomic analyses of CRC, the predominance of malignant epithelial cells often masks immune-derived signals, limiting the sensitivity of projecting moDC, moMC, or RTMC signatures for prognostic stratification or subtype classification. The high tumor purity observed in TCGA and similar cohorts skew expression profiles toward tumor-intrinsic programs, which tend to show stronger correlations with clinical outcomes than immune-related signatures\u003csup\u003e\u003cspan citationid=\"CR73\" class=\"CitationRef\"\u003e73\u003c/span\u003e,\u003cspan citationid=\"CR74\" class=\"CitationRef\"\u003e74\u003c/span\u003e\u003c/sup\u003e. Notably, molecular subtypes initially interpreted as mesenchymal tumor cell states were later attributed predominantly to stromal and immune cell expression, revealing how shifts in cellular composition can confound transcriptomic interpretation\u003csup\u003e\u003cspan citationid=\"CR75\" class=\"CitationRef\"\u003e75\u003c/span\u003e\u003c/sup\u003e. These observations underscore a central limitation of our approach\u0026mdash;namely, that the weaker associations observed for immune cell signatures in bulk CRC datasets likely stem from tumor cell dominance rather than a lack of biological relevance.\u003c/p\u003e \u003cp\u003eThis study provides the most comprehensive evidence to date linking metabolic reprogramming via SLC transporters to tumor microenvironmental phenotypes, including canonical molecular subtypes in colorectal cancer. Specifically, we identify a CMS2-like patient subgroup in which SLC39-mediated zinc influx drives CD24 overexpression through epigenetic enhancer activation in cancer cells. This zinc-dependent signaling cascade establishes pleiotropic cell\u0026ndash;cell interactions with myeloid-lineage cells\u0026mdash;moDCs, moMCs, and RTMCs\u0026mdash;leading to immune-evasive \u0026lsquo;don\u0026rsquo;t eat me\u0026rsquo; signaling and adhesion remodeling. Notably, this immunosuppressive mechanism is independent of the canonical CD47 axis and is enriched in CMS2-like tumors, frequently associated with SLC39 copy number gain. To extend these findings toward clinical translation, we developed an ICI response prediction model integrating multi-cohort transcriptomic data. The model, trained on a seven-gene zinc\u0026ndash;CD24\u0026ndash;SIGLEC10 signature, achieved superior accuracy in predicting therapeutic response, highlighting the clinical relevance of this zinc-driven immunosuppressive program. Our findings suggest that therapeutic strategies targeting the SLC39\u0026ndash;zinc influx\u0026ndash;CD24 axis, including blockade of CD24\u0026ndash;SIGLEC10 interactions, may reverse immune exclusion in this CRC subtype. Given the lineage-specific engagement of SIGLEC10\u003csup\u003e+\u003c/sup\u003e myeloid populations in CMS2-like tumors, engineering CAR-based interventions targeting these myeloid subsets represents a promising direction for precision immunotherapy.\u003c/p\u003e"},{"header":"Conclussion","content":"\u003cp\u003eIn conclusion, we identify SLC39 driven zinc influx as a central metabolic trigger that establishes an immune suppressive state in colorectal cancer. Zinc import enhances CDX2 dependent activation of a CD24 enhancer, driving cancer cell specific CD24 overexpression and engagement of SIGLEC10 positive myeloid cells. This CD24\u0026ndash;SIGLEC10 axis suppresses phagocytosis, forms recurrent immune evasive niches and defines a CMS2 like tumor state associated with poor prognosis and resistance to immune checkpoint blockade. Functionally, zinc induced CD24 signaling reduces macrophage phagocytosis and can be reversed by CD24 blockade. Clinically, a seven gene SLC39\u0026ndash;CD24\u0026ndash;SIGLEC10 signature outperforms established biomarkers in predicting immunotherapy response, highlighting this zinc driven pathway as a therapeutically actionable target in CRC.\u003c/p\u003e "},{"header":"Abbreviations","content":"\u003cp\u003eAUC: Area under the receiver operating characteristic curve \u0026nbsp;\u003c/p\u003e\n\u003cp\u003eAUPR: Area under the precision-recall curve \u0026nbsp;\u003c/p\u003e\n\u003cp\u003ecDNA: Complementary DNA \u0026nbsp;\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eCMS: Consensus Molecular Subtypes \u0026nbsp;\u003c/p\u003e\n\u003cp\u003eCNV: Copy number variation \u0026nbsp;\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eCOAD: Colorectal adenocarcinoma \u0026nbsp;\u003c/p\u003e\n\u003cp\u003eCPTAC: Clinical Proteomic Tumor Analysis Consortium \u0026nbsp;\u003c/p\u003e\n\u003cp\u003eCRC: Colorectal cancer \u0026nbsp;\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eCRISPRa: CRISPR activation \u0026nbsp;\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eDFI: Disease-free interval \u0026nbsp;\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eDMEM: Dulbecco\u0026rsquo;s Modified Eagle Medium \u0026nbsp;\u003c/p\u003e\n\u003cp\u003eEMT: Epithelial\u0026ndash;mesenchymal transition \u0026nbsp;\u003c/p\u003e\n\u003cp\u003eFBS: Fetal bovine serum \u0026nbsp;\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eGO: Gene Ontology \u0026nbsp;\u003c/p\u003e\n\u003cp\u003eGSVA: Gene Set Variation Analysis \u0026nbsp;\u003c/p\u003e\n\u003cp\u003eICI: Immune checkpoint inhibitor \u0026nbsp;\u003c/p\u003e\n\u003cp\u003eIF: Immunofluorescence \u0026nbsp;\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eIFN-\u0026gamma;: Interferon gamma \u0026nbsp;\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eKEGG: Kyoto Encyclopedia of Genes and Genomes \u0026nbsp;\u003c/p\u003e\n\u003cp\u003eLPS: Lipopolysaccharide \u0026nbsp;\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eMFI: Mean fluorescence intensity \u0026nbsp;\u003c/p\u003e\n\u003cp\u003eMMR: Mismatch repair \u0026nbsp;\u003c/p\u003e\n\u003cp\u003emoDCs: Monocyte-derived dendritic cells \u0026nbsp;\u003c/p\u003e\n\u003cp\u003eMOFA: Multi-Omics Factor Analysis \u0026nbsp;\u003c/p\u003e\n\u003cp\u003emoMCs: Monocyte-derived macrophages \u0026nbsp;\u003c/p\u003e\n\u003cp\u003emregDC: Myeloid regulatory dendritic cell\u003c/p\u003e\n\u003cp\u003eMSI: Microsatellite instability \u0026nbsp;\u003c/p\u003e\n\u003cp\u003eMSigDB: Molecular Signatures Database \u0026nbsp;\u003c/p\u003e\n\u003cp\u003eMSS: Microsatellite stable \u0026nbsp;\u0026nbsp;\u003c/p\u003e\n\u003cp\u003ePCA: Principal component analysis \u0026nbsp;\u003c/p\u003e\n\u003cp\u003ePFI: Progression-free interval \u0026nbsp;\u003c/p\u003e\n\u003cp\u003ePFS: Progression-free survival \u0026nbsp;\u003c/p\u003e\n\u003cp\u003ePMA: Phorbol 12-myristate 13-acetate \u0026nbsp;\u003c/p\u003e\n\u003cp\u003ePWMs: Position weight matrices \u0026nbsp;\u003c/p\u003e\n\u003cp\u003eRPMI: Roswell Park Memorial Institute \u0026nbsp;\u003c/p\u003e\n\u003cp\u003eRTMCs: Resident tissue macrophages \u0026nbsp;\u003c/p\u003e\n\u003cp\u003eSAM: Synergistic Activation Mediator \u0026nbsp;\u003c/p\u003e\n\u003cp\u003esgRNAs: Single-guide RNAs \u0026nbsp;\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eSLC: Solute carrier \u0026nbsp;\u003c/p\u003e\n\u003cp\u003eTAD: Topologically associating domain \u0026nbsp;\u003c/p\u003e\n\u003cp\u003eTAMs: Tumor-associated macrophages \u0026nbsp;\u003c/p\u003e\n\u003cp\u003eTCGA: The Cancer Genome Atlas \u0026nbsp;\u003c/p\u003e\n\u003cp\u003eTME: Tumor microenvironment \u0026nbsp;\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eUMAP: Uniform Manifold Approximation and Projection \u0026nbsp;\u003c/p\u003e\n\u003cp\u003eWB: Western blot \u0026nbsp;\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eData availability\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003ePreviously published datasets analyzed in this study are available under their respective repositories and accession codes as follows: TCGA colorectal cancer multi-omics (UCSC Xena, \u0026ldquo;TCGA Pan-Cancer (PANCAN)\u0026rdquo; cohort); CPTAC colorectal proteogenomics (LinkedOmics CPTAC-COAD); colorectal scRNA-seq (Synapse, accession syn26844071); ICI-treated CRC scRNA-seq (Broad Single Cell Portal, study SCP2079); ICI-treated CRC bulk RNA-seq (NCBI GEO, accession GSE179351, GSE235919, GSE302922); and CRC spatial transcriptomics (NCBI GEO, accession GSE225857). All processed data generated in this study that support the findings are deposited in Zenodo (https://doi.org/10.5281/zenodo.17292085). Additional information is available from the corresponding author upon reasonable request.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAcknowledgements\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNot applicable.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis work was supported by the National Research Foundation of Korea (NRF) grant funded by the Korea government (MSIT) (RS-2022-NR072092, RS-2023-00214527, RS-2025-23525624, and RS-2025-02304837), Hyundai Motor Chung Mong-Koo Foundation.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthor contributions\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eS.J.H. and S.P performed data analyses and wrote the manuscript. S.K. and Y.I.P. participated in data analyses. J.L., M.K., and M.H. reviewed the manuscript. S.E.K. and K.K. conceived and supervised the study.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eEthics declarations\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNot applicable.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eSchlessinger, A., Zatorski, N., Hutchinson, K. \u0026amp; Colas, C. Targeting SLC transporters: small molecules as modulators and therapeutic opportunities. \u003cem\u003eTrends in Biochemical Sciences\u003c/em\u003e vol. 48 801\u0026ndash;814 Preprint at https://doi.org/10.1016/j.tibs.2023.05.011 (2023).\u003c/li\u003e\n\u003cli\u003eNwosu, Z. C., Song, M. G., di Magliano, M. P., Lyssiotis, C. A. \u0026amp; Kim, S. E. Nutrient transporters: connecting cancer metabolism to therapeutic opportunities. \u003cem\u003eOncogene\u003c/em\u003e vol. 42 711\u0026ndash;724 Preprint at https://doi.org/10.1038/s41388-023-02593-x (2023).\u003c/li\u003e\n\u003cli\u003eYan, Y. \u003cem\u003eet al.\u003c/em\u003e SLC7A11 expression level dictates differential responses to oxidative stress in cancer cells. \u003cem\u003eNat Commun\u003c/em\u003e \u003cstrong\u003e14\u003c/strong\u003e, (2023).\u003c/li\u003e\n\u003cli\u003ePapalazarou, V. \u003cem\u003eet al.\u003c/em\u003e Phenotypic profiling of solute carriers characterizes serine transport in cancer. \u003cem\u003eNat Metab\u003c/em\u003e \u003cstrong\u003e5\u003c/strong\u003e, 2148\u0026ndash;2168 (2023).\u003c/li\u003e\n\u003cli\u003eZheng, Y. \u003cem\u003eet al.\u003c/em\u003e Metabolic gatekeepers: harnessing tumor-derived metabolites to optimize T cell-based immunotherapy efficacy in the tumor microenvironment. \u003cem\u003eCell Death and Disease\u003c/em\u003e vol. 15 Preprint at https://doi.org/10.1038/s41419-024-07122-6 (2024).\u003c/li\u003e\n\u003cli\u003eStine, Z. E., Schug, Z. T., Salvino, J. M. \u0026amp; Dang, C. V. Targeting cancer metabolism in the era of precision oncology. \u003cem\u003eNature Reviews Drug Discovery\u003c/em\u003e vol. 21 141\u0026ndash;162 Preprint at https://doi.org/10.1038/s41573-021-00339-6 (2022).\u003c/li\u003e\n\u003cli\u003eChowdhury, S. \u003cem\u003eet al.\u003c/em\u003e Consensus molecular subtyping of metastatic colorectal cancer expands biomarker-directed therapeutic benefit for patients with CMS1 and CMS2 tumors. \u003cem\u003eBr J Cancer\u003c/em\u003e \u003cstrong\u003e131\u003c/strong\u003e, 1328\u0026ndash;1339 (2024).\u003c/li\u003e\n\u003cli\u003eGreen, C., Roccia, P. \u0026amp; Rufini, A. Making sense of human colorectal cancer molecular subtypes: mice are stepping in. \u003cem\u003eCell Death Discovery\u003c/em\u003e vol. 11 Preprint at https://doi.org/10.1038/s41420-025-02594-7 (2025).\u003c/li\u003e\n\u003cli\u003eCorces, M. R. \u003cem\u003eet al.\u003c/em\u003e The chromatin accessibility landscape of primary human cancers. \u003cem\u003eScience (1979)\u003c/em\u003e \u003cstrong\u003e362\u003c/strong\u003e, (2018).\u003c/li\u003e\n\u003cli\u003eCastro-Mondragon, J. A. \u003cem\u003eet al.\u003c/em\u003e JASPAR 2022: The 9th release of the open-access database of transcription factor binding profiles. \u003cem\u003eNucleic Acids Res\u003c/em\u003e \u003cstrong\u003e50\u003c/strong\u003e, D165\u0026ndash;D173 (2022).\u003c/li\u003e\n\u003cli\u003eKueckelhaus, J. \u003cem\u003eet al.\u003c/em\u003e Inferring histology-associated gene expression gradients in spatial transcriptomic studies. \u003cem\u003eNat Commun\u003c/em\u003e \u003cstrong\u003e15\u003c/strong\u003e, (2024).\u003c/li\u003e\n\u003cli\u003eChen, D. \u003cem\u003eet al.\u003c/em\u003e Pan-cancer analysis implicates novel insights of lactate metabolism into immunotherapy response prediction and survival prognostication. \u003cem\u003eJournal of Experimental and Clinical Cancer Research\u003c/em\u003e \u003cstrong\u003e43\u003c/strong\u003e, (2024).\u003c/li\u003e\n\u003cli\u003eBulen, B. J. \u003cem\u003eet al.\u003c/em\u003e Validation of Immunotherapy Response Score as Predictive of Pan-solid Tumor Anti-PD-1/PD-L1 Benefit. \u003cem\u003eCancer Research Communications\u003c/em\u003e \u003cstrong\u003e3\u003c/strong\u003e, 1335\u0026ndash;1349 (2023).\u003c/li\u003e\n\u003cli\u003eZhang, W. \u003cem\u003eet al.\u003c/em\u003e Pan-cancer evaluation of regulated cell death to predict overall survival and immune checkpoint inhibitor response. \u003cem\u003eNPJ Precis Oncol\u003c/em\u003e \u003cstrong\u003e8\u003c/strong\u003e, (2024).\u003c/li\u003e\n\u003cli\u003eChida, K. \u003cem\u003eet al.\u003c/em\u003e Transcriptomic Profiling of MSI-H/dMMR Gastrointestinal Tumors to Identify Determinants of Responsiveness to Anti\u0026ndash;PD-1 Therapy. \u003cem\u003eClinical Cancer Research\u003c/em\u003e \u003cstrong\u003e28\u003c/strong\u003e, 2110\u0026ndash;2117 (2022).\u003c/li\u003e\n\u003cli\u003eAntoniotti, C. \u003cem\u003eet al.\u003c/em\u003e An Immune-Related Gene Expression Signature Predicts Benefit from Adding Atezolizumab to FOLFOXIRI plus Bevacizumab in Metastatic Colorectal Cancer. \u003cem\u003eClinical Cancer Research\u003c/em\u003e \u003cstrong\u003e29\u003c/strong\u003e, 2291\u0026ndash;2298 (2023).\u003c/li\u003e\n\u003cli\u003eAyers, M. \u003cem\u003eet al.\u003c/em\u003e IFN-\u0026gamma;-related mRNA profile predicts clinical response to PD-1 blockade. \u003cem\u003eJournal of Clinical Investigation\u003c/em\u003e \u003cstrong\u003e127\u003c/strong\u003e, 2930\u0026ndash;2940 (2017).\u003c/li\u003e\n\u003cli\u003eHugo, W. \u003cem\u003eet al.\u003c/em\u003e Genomic and Transcriptomic Features of Response to Anti-PD-1 Therapy in Metastatic Melanoma. \u003cem\u003eCell\u003c/em\u003e \u003cstrong\u003e165\u003c/strong\u003e, 35\u0026ndash;44 (2016).\u003c/li\u003e\n\u003cli\u003eCristescu, R. \u003cem\u003eet al.\u003c/em\u003e Pan-tumor genomic biomarkers for PD-1 checkpoint blockade-based immunotherapy. \u003cem\u003eScience (1979)\u003c/em\u003e \u003cstrong\u003e362\u003c/strong\u003e, (2018).\u003c/li\u003e\n\u003cli\u003eChen, R. \u0026amp; Chen, L. Solute carrier transporters: emerging central players in tumour immunotherapy. \u003cem\u003eTrends in Cell Biology\u003c/em\u003e vol. 32 186\u0026ndash;201 Preprint at https://doi.org/10.1016/j.tcb.2021.08.002 (2022).\u003c/li\u003e\n\u003cli\u003eArgelaguet, R. \u003cem\u003eet al.\u003c/em\u003e MOFA+: A statistical framework for comprehensive integration of multi-modal single-cell data. \u003cem\u003eGenome Biol\u003c/em\u003e \u003cstrong\u003e21\u003c/strong\u003e, (2020).\u003c/li\u003e\n\u003cli\u003eJeong, J. \u0026amp; Eide, D. J. The SLC39 family of zinc transporters. \u003cem\u003eMolecular Aspects of Medicine\u003c/em\u003e vol. 34 612\u0026ndash;619 Preprint at https://doi.org/10.1016/j.mam.2012.05.011 (2013).\u003c/li\u003e\n\u003cli\u003eCiardiello, D. \u003cem\u003eet al.\u003c/em\u003e Immunotherapy of colorectal cancer: Challenges for therapeutic efficacy. \u003cem\u003eCancer Treatment Reviews\u003c/em\u003e vol. 76 22\u0026ndash;32 Preprint at https://doi.org/10.1016/j.ctrv.2019.04.003 (2019).\u003c/li\u003e\n\u003cli\u003eGao, N., White, P. \u0026amp; Kaestner, K. H. Establishment of Intestinal Identity and Epithelial-Mesenchymal Signaling by Cdx2. \u003cem\u003eDev Cell\u003c/em\u003e \u003cstrong\u003e16\u003c/strong\u003e, 588\u0026ndash;599 (2009).\u003c/li\u003e\n\u003cli\u003eBarkal, A. A. \u003cem\u003eet al.\u003c/em\u003e CD24 signalling through macrophage Siglec-10 is a target for cancer immunotherapy. \u003cem\u003eNature\u003c/em\u003e \u003cstrong\u003e572\u003c/strong\u003e, 392\u0026ndash;396 (2019).\u003c/li\u003e\n\u003cli\u003eWagner, A. \u003cem\u003eet al.\u003c/em\u003e Metabolic modeling of single Th17 cells reveals regulators of autoimmunity. \u003cem\u003eCell\u003c/em\u003e \u003cstrong\u003e184\u003c/strong\u003e, 4168-4185.e21 (2021).\u003c/li\u003e\n\u003cli\u003eJoanito, I. \u003cem\u003eet al.\u003c/em\u003e Single-cell and bulk transcriptome sequencing identifies two epithelial tumor cell states and refines the consensus molecular classification of colorectal cancer. \u003cem\u003eNat Genet\u003c/em\u003e \u003cstrong\u003e54\u003c/strong\u003e, 963\u0026ndash;975 (2022).\u003c/li\u003e\n\u003cli\u003ePatel, A. P. \u003cem\u003eet al.\u003c/em\u003e Single-cell RNA-seq highlights intratumoral heterogeneity in primary glioblastoma. \u003cem\u003eScience (1979)\u003c/em\u003e \u003cstrong\u003e344\u003c/strong\u003e, 1396\u0026ndash;1401 (2014).\u003c/li\u003e\n\u003cli\u003eAlvarez, M. J. \u003cem\u003eet al.\u003c/em\u003e Functional characterization of somatic mutations in cancer using network-based inference of protein activity. \u003cem\u003eNat Genet\u003c/em\u003e \u003cstrong\u003e48\u003c/strong\u003e, 838\u0026ndash;847 (2016).\u003c/li\u003e\n\u003cli\u003eWang, Y. \u003cem\u003eet al.\u003c/em\u003e The 3D Genome Browser: A web-based browser for visualizing 3D genome organization and long-range chromatin interactions. \u003cem\u003eGenome Biol\u003c/em\u003e \u003cstrong\u003e19\u003c/strong\u003e, (2018).\u003c/li\u003e\n\u003cli\u003eFishilevich, S. \u003cem\u003eet al.\u003c/em\u003e GeneHancer: Genome-wide integration of enhancers and target genes in GeneCards. \u003cem\u003eDatabase\u003c/em\u003e \u003cstrong\u003e2017\u003c/strong\u003e, (2017).\u003c/li\u003e\n\u003cli\u003eAbascal, F. \u003cem\u003eet al.\u003c/em\u003e Expanded encyclopaedias of DNA elements in the human and mouse genomes. \u003cem\u003eNature\u003c/em\u003e \u003cstrong\u003e583\u003c/strong\u003e, 699\u0026ndash;710 (2020).\u003c/li\u003e\n\u003cli\u003eBecker, W. R. \u003cem\u003eet al.\u003c/em\u003e Single-cell analyses define a continuum of cell state and composition changes in the malignant transformation of polyps to colorectal cancer. \u003cem\u003eNat Genet\u003c/em\u003e \u003cstrong\u003e54\u003c/strong\u003e, 985\u0026ndash;995 (2022).\u003c/li\u003e\n\u003cli\u003eAbramson, J. \u003cem\u003eet al.\u003c/em\u003e Accurate structure prediction of biomolecular interactions with AlphaFold 3. \u003cem\u003eNature\u003c/em\u003e \u003cstrong\u003e630\u003c/strong\u003e, 493\u0026ndash;500 (2024).\u003c/li\u003e\n\u003cli\u003eYan, Y., Tao, H., He, J. \u0026amp; Huang, S. Y. The HDOCK server for integrated protein\u0026ndash;protein docking. \u003cem\u003eNat Protoc\u003c/em\u003e \u003cstrong\u003e15\u003c/strong\u003e, 1829\u0026ndash;1852 (2020).\u003c/li\u003e\n\u003cli\u003ePersad, S. \u003cem\u003eet al.\u003c/em\u003e SEACells infers transcriptional and epigenomic cellular states from single-cell genomics data. \u003cem\u003eNat Biotechnol\u003c/em\u003e \u003cstrong\u003e41\u003c/strong\u003e, 1746\u0026ndash;1757 (2023).\u003c/li\u003e\n\u003cli\u003eSetty, M. \u003cem\u003eet al.\u003c/em\u003e Characterization of cell fate probabilities in single-cell data with Palantir. \u003cem\u003eNat Biotechnol\u003c/em\u003e \u003cstrong\u003e37\u003c/strong\u003e, 451\u0026ndash;460 (2019).\u003c/li\u003e\n\u003cli\u003eLange, M. \u003cem\u003eet al.\u003c/em\u003e CellRank for directed single-cell fate mapping. \u003cem\u003eNat Methods\u003c/em\u003e \u003cstrong\u003e19\u003c/strong\u003e, 159\u0026ndash;170 (2022).\u003c/li\u003e\n\u003cli\u003eTian, J. \u003cem\u003eet al.\u003c/em\u003e Combined PD-1, BRAF and MEK inhibition in BRAFV600E colorectal cancer: a phase 2 trial. \u003cem\u003eNat Med\u003c/em\u003e \u003cstrong\u003e29\u003c/strong\u003e, 458\u0026ndash;466 (2023).\u003c/li\u003e\n\u003cli\u003eCha, J., Yu, J., Cho, J. W., Hemberg, M. \u0026amp; Lee, I. scHumanNet: a single-cell network analysis platform for the study of cell-type specificity of disease genes. \u003cem\u003eNucleic Acids Res\u003c/em\u003e \u003cstrong\u003e51\u003c/strong\u003e, E8 (2023).\u003c/li\u003e\n\u003cli\u003eDimitrov, D. \u003cem\u003eet al.\u003c/em\u003e LIANA+ provides an all-in-one framework for cell\u0026ndash;cell communication inference. \u003cem\u003eNat Cell Biol\u003c/em\u003e \u003cstrong\u003e26\u003c/strong\u003e, 1613\u0026ndash;1622 (2024).\u003c/li\u003e\n\u003cli\u003eRaredon, M. S. B. \u003cem\u003eet al.\u003c/em\u003e Comprehensive visualization of cell\u0026ndash;cell interactions in single-cell and spatial transcriptomics with NICHES. \u003cem\u003eBioinformatics\u003c/em\u003e \u003cstrong\u003e39\u003c/strong\u003e, (2023).\u003c/li\u003e\n\u003cli\u003eBrowaeys, R., Saelens, W. \u0026amp; Saeys, Y. NicheNet: modeling intercellular communication by linking ligands to target genes. \u003cem\u003eNat Methods\u003c/em\u003e \u003cstrong\u003e17\u003c/strong\u003e, 159\u0026ndash;162 (2020).\u003c/li\u003e\n\u003cli\u003eWang, F. \u003cem\u003eet al.\u003c/em\u003e \u003cem\u003eSingle-Cell and Spatial Transcriptome Analysis Reveals the Cellular Heterogeneity of Liver Metastatic Colorectal Cancer\u003c/em\u003e. (2023).\u003c/li\u003e\n\u003cli\u003eElosua-Bayes, M., Nieto, P., Mereu, E., Gut, I. \u0026amp; Heyn, H. SPOTlight: Seeded NMF regression to deconvolute spatial transcriptomics spots with single-cell transcriptomes. \u003cem\u003eNucleic Acids Res\u003c/em\u003e \u003cstrong\u003e49\u003c/strong\u003e, E50 (2021).\u003c/li\u003e\n\u003cli\u003eButler, A., Hoffman, P., Smibert, P., Papalexi, E. \u0026amp; Satija, R. Integrating single-cell transcriptomic data across different conditions, technologies, and species. \u003cem\u003eNat Biotechnol\u003c/em\u003e \u003cstrong\u003e36\u003c/strong\u003e, 411\u0026ndash;420 (2018).\u003c/li\u003e\n\u003cli\u003eDunsmore, G. \u003cem\u003eet al.\u003c/em\u003e \u003cem\u003eC A N C E R I M M U N O L O G Y Timing and Location Dictate Monocyte Fate and Their Transition to Tumor-Associated Macrophages\u003c/em\u003e. https://www.science.org (2024).\u003c/li\u003e\n\u003cli\u003eBadia-I-Mompel, P. \u003cem\u003eet al.\u003c/em\u003e decoupleR: ensemble of computational methods to infer biological activities from omics data. \u003cem\u003eBioinformatics Advances\u003c/em\u003e \u003cstrong\u003e2\u003c/strong\u003e, (2022).\u003c/li\u003e\n\u003cli\u003eRamavath, H. N. \u003cem\u003eet al.\u003c/em\u003e Zinc induces hephaestin expression via a PI3K-CDX2 dependent mechanism to regulate iron transport in intestinal Caco-2 cells. \u003cem\u003eBiochem Biophys Res Commun\u003c/em\u003e \u003cstrong\u003e626\u003c/strong\u003e, 1\u0026ndash;7 (2022).\u003c/li\u003e\n\u003cli\u003eYang, W. \u003cem\u003eet al.\u003c/em\u003e IL-4-STAT6-induced high Siglec-G/10 expression aggravates the severe immune suppressive tumor microenvironment and impedes the efficacy of immunotherapy in head and neck squamous cell carcinoma. \u003cem\u003eJ Immunother Cancer\u003c/em\u003e \u003cstrong\u003e13\u003c/strong\u003e, (2025).\u003c/li\u003e\n\u003cli\u003eChen, S. H. \u003cem\u003eet al.\u003c/em\u003e Dual checkpoint blockade of CD47 and PD-L1 using an affinity-tuned bispecific antibody maximizes antitumor immunity. \u003cem\u003eJ Immunother Cancer\u003c/em\u003e \u003cstrong\u003e9\u003c/strong\u003e, (2021).\u003c/li\u003e\n\u003cli\u003eSobczak, K. \u003cem\u003eet al.\u003c/em\u003e The unique molecular recognition features of Siglec-10: structural insights into sialoglycan and antibody interactions. Preprint at https://doi.org/10.1101/2025.06.10.658867 (2025).\u003c/li\u003e\n\u003cli\u003eLi, S. \u003cem\u003eet al.\u003c/em\u003e IMM47, a humanized monoclonal antibody that targets CD24, exhibits exceptional anti-Tumor efficacy by blocking the CD24/Siglec-10 interaction and can be used as monotherapy or in combination with anti-PD1 antibodies for cancer immunotherapy. \u003cem\u003eAntib Ther\u003c/em\u003e \u003cstrong\u003e6\u003c/strong\u003e, 240\u0026ndash;252 (2023).\u003c/li\u003e\n\u003cli\u003ePapalazarou, V. \u003cem\u003eet al.\u003c/em\u003e Phenotypic profiling of solute carriers characterizes serine transport in cancer. \u003cem\u003eNat Metab\u003c/em\u003e \u003cstrong\u003e5\u003c/strong\u003e, 2148\u0026ndash;2168 (2023).\u003c/li\u003e\n\u003cli\u003eTseng, D. \u003cem\u003eet al.\u003c/em\u003e Anti-CD47 antibody-mediated phagocytosis of cancer by macrophages primes an effective antitumor T-cell response. \u003cem\u003eProc Natl Acad Sci U S A\u003c/em\u003e \u003cstrong\u003e110\u003c/strong\u003e, 11103\u0026ndash;11108 (2013).\u003c/li\u003e\n\u003cli\u003eHu, T. \u003cem\u003eet al.\u003c/em\u003e Tumor-intrinsic CD47 signal regulates glycolysis and promotes colorectal cancer cell growth and metastasis. \u003cem\u003eTheranostics\u003c/em\u003e \u003cstrong\u003e10\u003c/strong\u003e, 4056\u0026ndash;4072 (2020).\u003c/li\u003e\n\u003cli\u003eBarkal, A. A. \u003cem\u003eet al.\u003c/em\u003e CD24 signalling through macrophage Siglec-10 is a target for cancer immunotherapy. \u003cem\u003eNature\u003c/em\u003e \u003cstrong\u003e572\u003c/strong\u003e, 392\u0026ndash;396 (2019).\u003c/li\u003e\n\u003cli\u003eSathe, A. \u003cem\u003eet al.\u003c/em\u003e Colorectal Cancer Metastases in the Liver Establish Immunosuppressive Spatial Networking between Tumor-Associated SPP1\u0026thorn; Macrophages and Fibroblasts. \u003cem\u003eClinical Cancer Research\u003c/em\u003e \u003cstrong\u003e29\u003c/strong\u003e, 244\u0026ndash;260 (2023).\u003c/li\u003e\n\u003cli\u003eMatusiak, M. \u003cem\u003eet al.\u003c/em\u003e Spatially Segregated Macrophage Populations Predict Distinct Outcomes in Colon Cancer. \u003cem\u003eCancer Discov\u003c/em\u003e \u003cstrong\u003e14\u003c/strong\u003e, 1418\u0026ndash;1439 (2024).\u003c/li\u003e\n\u003cli\u003eMaier, B. \u003cem\u003eet al.\u003c/em\u003e A conserved dendritic-cell regulatory program limits antitumour immunity. \u003cem\u003eNature\u003c/em\u003e \u003cstrong\u003e580\u003c/strong\u003e, 257\u0026ndash;262 (2020).\u003c/li\u003e\n\u003cli\u003eSong, H. S., Ha, S. Y., Kim, J. Y., Kim, M. \u0026amp; Choi, J. H. The effect of genetic variants of SLC22A18 on proliferation, migration, and invasion of colon cancer cells. \u003cem\u003eSci Rep\u003c/em\u003e \u003cstrong\u003e14\u003c/strong\u003e, (2024).\u003c/li\u003e\n\u003cli\u003eScalise, M. \u003cem\u003eet al.\u003c/em\u003e Human OCTN2 (SLC22A5) is down-regulated in virus- and nonvirus-mediated cancer. \u003cem\u003eCell Biochem Funct\u003c/em\u003e \u003cstrong\u003e30\u003c/strong\u003e, 419\u0026ndash;425 (2012).\u003c/li\u003e\n\u003cli\u003eBiswas, S. \u003cem\u003eet al.\u003c/em\u003e Mutational inactivation of TGFBR2 in microsatellite unstable colon cancer arises from the cooperation of genomic instability and the clonal outgrowth of transforming growth factor \u0026beta; resistant cells. \u003cem\u003eGenes Chromosomes Cancer\u003c/em\u003e \u003cstrong\u003e47\u003c/strong\u003e, 95\u0026ndash;106 (2008).\u003c/li\u003e\n\u003cli\u003eGrady, W. M. \u003cem\u003eet al.\u003c/em\u003e Proliferation and Cdk4 expression in microsatellite unstable colon cancers with TGFBR2 mutations. \u003cem\u003eInt J Cancer\u003c/em\u003e \u003cstrong\u003e118\u003c/strong\u003e, 600\u0026ndash;608 (2006).\u003c/li\u003e\n\u003cli\u003eLu, S. \u003cem\u003eet al.\u003c/em\u003e Colorectal cancer with low SLC35A3 is associated with immune infiltrates and poor prognosis. \u003cem\u003eSci Rep\u003c/em\u003e \u003cstrong\u003e14\u003c/strong\u003e, (2024).\u003c/li\u003e\n\u003cli\u003eGuinney, J. \u003cem\u003eet al.\u003c/em\u003e The consensus molecular subtypes of colorectal cancer. \u003cem\u003eNat Med\u003c/em\u003e \u003cstrong\u003e21\u003c/strong\u003e, 1350\u0026ndash;1356 (2015).\u003c/li\u003e\n\u003cli\u003eVasaikar, S. \u003cem\u003eet al.\u003c/em\u003e Proteogenomic Analysis of Human Colon Cancer Reveals New Therapeutic Opportunities. \u003cem\u003eCell\u003c/em\u003e \u003cstrong\u003e177\u003c/strong\u003e, 1035-1049.e19 (2019).\u003c/li\u003e\n\u003cli\u003eClark, D. J. \u003cem\u003eet al.\u003c/em\u003e Integrated Proteogenomic Characterization of Clear Cell Renal Cell Carcinoma. \u003cem\u003eCell\u003c/em\u003e \u003cstrong\u003e179\u003c/strong\u003e, 964-983.e31 (2019).\u003c/li\u003e\n\u003cli\u003eKim, Y. H. \u003cem\u003eet al.\u003c/em\u003e Integrative Multi-omics Analysis Reveals Different Metabolic Phenotypes Based on Molecular Characteristics in Thyroid Cancer. \u003cem\u003eClinical Cancer Research\u003c/em\u003e \u003cstrong\u003e30\u003c/strong\u003e, 883\u0026ndash;894 (2024).\u003c/li\u003e\n\u003cli\u003eLehti\u0026ouml;, J. \u003cem\u003eet al.\u003c/em\u003e Proteogenomics of non-small cell lung cancer reveals molecular subtypes associated with specific therapeutic targets and immune-evasion mechanisms. \u003cem\u003eNat Cancer\u003c/em\u003e \u003cstrong\u003e2\u003c/strong\u003e, 1224\u0026ndash;1242 (2021).\u003c/li\u003e\n\u003cli\u003eSun, C. \u003cem\u003eet al.\u003c/em\u003e Spatially resolved multi-omics highlights cell-specific metabolic remodeling and interactions in gastric cancer. \u003cem\u003eNat Commun\u003c/em\u003e \u003cstrong\u003e14\u003c/strong\u003e, (2023).\u003c/li\u003e\n\u003cli\u003eBustamante Eduardo, M. \u003cem\u003eet al.\u003c/em\u003e \u003cem\u003eA Metabolic Shift to the Serine Pathway Induced by Lipids Fosters Epigenetic Reprogramming in Nontransformed Breast Cells\u003c/em\u003e. \u003cem\u003eSci. Adv\u003c/em\u003e vol. 11 https://maayanlab.cloud/Enrichr/ (2025).\u003c/li\u003e\n\u003cli\u003eRhee, J. K. \u003cem\u003eet al.\u003c/em\u003e Impact of tumor purity on immune gene expression and clustering analyses across multiple cancer types. \u003cem\u003eCancer Immunol Res\u003c/em\u003e \u003cstrong\u003e6\u003c/strong\u003e, 87\u0026ndash;97 (2018).\u003c/li\u003e\n\u003cli\u003eAran, D., Sirota, M. \u0026amp; Butte, A. J. Systematic pan-cancer analysis of tumour purity. \u003cem\u003eNat Commun\u003c/em\u003e \u003cstrong\u003e6\u003c/strong\u003e, (2015).\u003c/li\u003e\n\u003cli\u003eIsella, C. \u003cem\u003eet al.\u003c/em\u003e Stromal contribution to the colorectal cancer transcriptome. \u003cem\u003eNat Genet\u003c/em\u003e \u003cstrong\u003e47\u003c/strong\u003e, 312\u0026ndash;319 (2015). \u003c/li\u003e\n\u003c/ol\u003e"},{"header":"Tables","content":"\u003cp\u003eTables 1 to 9 are available in the Supplementary Files section.\u003c/p\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":"","lastPublishedDoi":"10.21203/rs.3.rs-8267774/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-8267774/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eTumor metabolic reprogramming profoundly influences immune regulation, yet the mechanisms linking solute carrier (SLC) transporter activity to immune suppression remain elusive. Through integrative multi-omics, spatial, and single-cell analyses of colorectal cancer (CRC), we uncover a zinc influx\u0026ndash;driven signaling axis mediated by the SLC39 family that establishes an immune-suppressive tumor ecosystem. Multi-omics clustering of 258 CRC patients identified three SLC-centered archetypes, among which an SLC39-enriched subtype displayed zinc pathway activation and correspondence to the \u0026ldquo;immune-desert\u0026rdquo; CMS2 subtype. Mechanistically, SLC39-mediated zinc influx activated the transcription factor CDX2, promoting enhancer-driven transcription of CD24, an anti-phagocytic \u0026ldquo;don\u0026rsquo;t eat me\u0026rdquo; signal. Zinc-dependent CD24 upregulation occurred independently of CD47 and was restricted to malignant epithelial cells. Single-cell and spatial transcriptomics revealed that CD24-expressing tumor cells interact with SIGLEC10\u003csup\u003e+\u003c/sup\u003e monocyte-derived macrophages, dendritic cells, and resident macrophages, triggering pleiotropic immunoregulatory programs that suppress phagocytosis and remodel adhesion networks. This SLC39\u0026ndash;CD24\u0026ndash;SIGLEC10 axis defined spatially recurrent immune-suppressive niches and was associated with poor survival and resistance to immune checkpoint blockade. Functional assays confirmed that zinc-induced CD24\u0026ndash;SIGLEC10 engagement attenuates macrophage phagocytosis, reversible by CD24 blockade. Furthermore, integrative modeling across five ICI-treated CRC cohorts demonstrated that a seven-gene signature encompassing SLC39 transporters, CDX2, CD24, and SIGLEC10 robustly predicted clinical response to immunotherapy, outperforming established biomarkers. These findings identify SLC39-mediated zinc influx as a regulator of tumor\u0026ndash;immune crosstalk in CMS2-like CRC and highlight the SLC39\u0026ndash;zinc\u0026ndash;CD24\u0026ndash;SIGLEC10 axis as a promising therapeutic target to overcome immune exclusion and immunotherapy resistance.\u003c/p\u003e","manuscriptTitle":"SLC39-driven zinc influx orchestrates pleiotropic tumor–immune crosstalk to establish an immune-suppressive microenvironment in colorectal cancer","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-12-22 09:12:47","doi":"10.21203/rs.3.rs-8267774/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":"8e9936a2-f117-4125-a677-2b53a9c45d1a","owner":[],"postedDate":"December 22nd, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2026-01-22T16:53:20+00:00","versionOfRecord":[],"versionCreatedAt":"2025-12-22 09:12:47","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-8267774","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-8267774","identity":"rs-8267774","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}