Ligand-receptor pair-based signature score Derived from On-treatment Tumor Specimens Predicts Immune Checkpoint Blockade Response in Metastatic Melanoma

Ligand-receptor pair-based signature score Derived from On-treatment Tumor Specimens Predicts Immune Checkpoint Blockade Response in Metastatic Melanoma | 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 Ligand-receptor pair-based signature score Derived from On-treatment Tumor Specimens Predicts Immune Checkpoint Blockade Response in Metastatic Melanoma Huancheng Zeng, Rendong Zhang, Qiongzhi Jiang, Jundong Wu, Zhemin Zhuang, and 1 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-5854916/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted 8 You are reading this latest preprint version Abstract Immune checkpoint blockade (ICB) therapy has transformed the treatment landscape for metastatic melanoma, yet predicting therapeutic response remains a significant challenge. This study hypothesizes that coordinated ligand-receptor (LR) interactions within the tumor microenvironment (TME) critically influence ICB efficacy and proposes that a novel LR pair-based signature score (LRPS) derived from on-treatment samples can predict clinical outcomes. Using transcriptomic data from five independent cohorts, we identified seven LR pairs (FLT3-FLT3LG, LY9-LY9, CD5-CD5, CD40LG-ITGA2B/ITGB3, APP-CD74, TNFRSF17-TNFSF13, FCER2-ITGAV/ITGB3) significantly associated with treatment outcomes. LRPS demonstrated significant predictive power, achieving an area under the curve (AUC) exceeding 0.8 in four cohorts. Based on the LRPS signature, subjects were divided into high- and low-scores groups using the mean score. ICB response rates were higher in the high-scoring cohort subjects than the low-scoring subjects. Patient with high scores tended to have better survival outcomes than did those with low scores. In conclusion, we identified and verified an LRPS signature that provides a theoretical basis for applying such signatures derived from on-treatment tumor samples to predict therapeutic responses to ICB therapies. immune checkpoint blockade metastatic melanoma ligand-receptor pairs biomarker Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Introduction Although immune checkpoint blockade (ICB) has transformed clinical management of advanced malignancies including melanoma, durable therapeutic benefits remain limited to a subset of patients[ 1 – 3 ]. Predictive biomarkers capable of stratifying responders could mitigate adverse effects, prevent acquired resistance, and optimize resource utilization in immunotherapy[ 4 ], underscoring the urgent need for reliable signatures to guide ICB treatment selection. Current predictive frameworks for metastatic melanoma ICB response incorporate diverse biomarkers including but not limited to tumor mutational burden quantification[ 5 ], NLRP3 inflammasome signature [ 6 ], immune infiltrate composition[ 7 , 8 ], inflammatory signature[ 9 ], tumor-associated endothelial gene signature[ 10 ], platelet activation cascades [ 11 ] and immune checkpoint interaction scores (IMPRES)[ 12 ]. Notably, these biomarkers predominantly derive from pre-therapeutic biopsies, potentially overlooking dynamic microenvironmental changes induced by ICB itself. Ligand-receptor (LR) pair interactions in the tumor microenvironment mediate critical cell-to-cell communication processes that drive oncogenesis. These interactions not only regulate tumor cell proliferation and survival but also orchestrate metastatic progression[ 13 ]. Specifically, tumor cells communicate with their microenvironment through direct cell-cell contacts and secretion of soluble mediators such as growth factors, cytokines, and chemokines[ 14 ]. Moreover, stromal components including mesenchymal stem cells, cancer-associated fibroblasts, and tumor-associated macrophages actively modulate malignant phenotypes via bidirectional communication[ 14 ]. For example, CXCL16-CXCR6 axis dysregulation has been shown to promote immunosuppressive macrophage polarization[ 15 ]. Nevertheless, whether dynamic LR pair signatures from on-treatment samples can predict ICB response outcomes in metastatic melanoma remains unexplored. We hypothesize that dynamic LR network rewiring during ICB treatment reflects functional immune-tumor crosstalk and can be translated into a predictive signature (LRPS) with enhanced prognostic value. Through analysis of transcriptomic and clinical data from ICB-treated metastatic melanoma patients, we developed an LRPS using ElasticNet penalized Logistic Regression (ENlR) to predict treatment response in on-treatment specimens. This approach not only provides a clinically actionable biomarker but also identifies targetable communication nodes for rational combination immunotherapy design. Methods Data collection The workflow of this study is illustrated in Fig. 1 . We compiled transcriptomic data from five independent cohorts of on-treatment metastatic melanoma samples. These datasets include the Riaz et al. cohort (GEO: GSE120575) [ 16 ], the Gide et al. cohort (BioProject: PRJEB23709) [ 17 ], the Abril et al. cohort (dbGaP: phs001919.v1.p1) [ 18 ], the Lee et al. cohort (EGA: EGAD00001005738) [ 19 ], and the MGH et al. cohort (GEO: GSE115821 and GSE168204). We compiled transcriptomic data from five independent cohorts of on-treatment metastatic melanoma samples, defined as biopsies collected after initiation of ICB therapy (anti-PD-1/PD-L1 ± anti-CTLA-4). Samples were excluded from analysis based on the following criteria: (1) absence of RNA-seq data; (2) missing clinical response information; or (3) duplicated tumor specimens from the same timepoint in individual patients. Gene expression was quantified using transcripts per million (TPM) for each gene. According to RECIST criteria[ 20 ], responders (R) were defined as individuals who achieved complete response (CR), partial response (PR), or progression-free survival (PFS) of over 180 days. Non-responders (NR) were classified as those with PFS under 180 days accompanied by progressive disease (PD). Ligand-Receptor Pair Selection LR pairs were curated from published literature [ 21 , 22 ]. The selection of these pairs was grounded in the biological relevance of these interactions in the immune response. The comprehensive list of selected LR pairs is presented in Supplementary Table 1. Calculation of signature scores In this study, we utilized the Riaz et al. cohort as the training dataset, while the remaining cohorts served as validation datasets. We computed the score for each LR pair in on-treatment metastatic melanoma samples from the Riaz et al. cohort using the single-sample gene set enrichment analysis (ssGSEA) algorithm. Subsequently, we applied an Elastic-Net penalized logistic regression model to identify the LR pairs most strongly associated with response to ICB therapy. To further refine our analysis, we computed a weighted average for each sample, using the effect size as a weighting factor for the LR pairs most associated with ICB therapy response. This weighted average was defined as the LR pair-based signature score (LRPS). To minimize the risk of overfitting, five-fold cross-validation was performed on the training set[ 23 ]. To correct for class imbalance in our dataset, we incorporated a cost-sensitive learning approach during model training, which was implemented the cv.glmnet, a function from the glmnet package. Specifically, we applied a prior probability-based offset in the binomial logistic regression model. This offset was computed as a log-odds correction term using the empirical class frequencies and a predefined classification threshold (τ = 2/3). The model was trained using cross-validation (nfolds = 5) to optimize the regularization parameter (lambda) with respect to the area under the curve (AUC). By incorporating the correction term, the model was encouraged to assign greater importance to the minority class without modifying the training data or employing resampling techniques. This strategy enabled a more balanced and robust model performance under imbalanced class conditions. The predictive performance of the model was assessed by generating receiver operating characteristic (ROC) curves, with the predictive accuracy of the LRPS quantified by the area under the curve (AUC). Using data from the Riaz et al. cohort, the optimal threshold value was determined using Youden’s index during ROC analysis. Finally, the odds ratios for individual samples were calculated based on their LRPS. Statistical Analysis All statistical analyses and graphical visualizations were conducted using R software (version 4.2.0) and GraphPad Prism (version 8.0). The one-tailed Wilcoxon rank-sum test was applied to assess statistical differences between the R and NR groups. We calculated the sample’s odd ratio based on the signature score. Samples were classified into high and low group use mean value of samples’ odd ratio as cutoff. Kaplan-Meier (KM) survival analysis, combined with the log-rank test, was used to evaluate survival differences between these groups. Cox proportional hazards regression was performed to estimate the hazard ratio (HR) for each cohort. Given the exploratory nature of this biomarker study, we used a significance threshold of 0.05 when reporting p-values, with trends between 0.05 and 0.1 reported to avoid arbitrary dichotomization of significance [ 24 ]. Results 3.1 Patient cohorts In the Riaz et al. dataset, a total of 54 biopsies at the on-treatment timepoint were included in the analysis. Of these, 21 biopsies were from R (Responders), and 33 were from NR (Non-Responders). All patients received anti-PD1 monotherapy. In the Gide et al. dataset, 17 patients with 18 on-treatment biopsies (11 R and 7 NR) received anti-PD1 monotherapy. The median age of the patients was 56 years, with 41% over 60 years old. Additionally, 53% of the patients were male. In the Lee et al. dataset, 23 patients with 35 on-treatment biopsies (6 R and 29 NR) were treated with anti-PD1 monotherapy, including nivolumab or pembrolizumab. The Abril-Rodriguez cohort included 30 on-treatment tumor samples (13 R and 17 NR). All patients received PD-L1 monotherapy (pembrolizumab). The MGH cohort included 31 on-treatment samples (5 R and 26 NR). All patients received anti-PD1/PD-L1 monotherapy. The baseline characteristics of the patients are presented in Supplementary Table 2 . As illustrated in Fig. 1 , the Riaz et al. cohort, comprising 54 on-treatment metastatic melanoma samples, was utilized as the training set, while other cohorts served as validation sets. The ssGSEA algorithm was employed to compute scores for ligand-receptor (LR) pairs within the Riaz et al. training dataset. An elastic net algorithm was then applied to identify LR pairs significantly associated with response to ICB therapy. To mitigate overfitting, we implemented five cross-validation and used a cost-sensitive algorithm (Figs. 2 A and 2 B). Ultimately, as shown in Fig. 2 C, seven LR pairs associated with ICB therapy response were identified: FLT3-FLT3LG, LY9-LY9, CD5-CD5, CD40LG-ITGA2B/ITGB3, APP-CD74, TNFRSF17-TNFSF13, FCER2-ITGAV/ITGB3. Effect size was subsequently used as a weighting factor to calculate the weighted average of the ssGSEA scores for these seven LR pairs, defining the LRPS (Ligand-Receptor Pair Score) for each sample. 3.2. LRPS is an effective predictor of response and prognosis in ICB therapy for on-treatment samples As illustrated in Fig. 3 A, the heatmap presents the distribution of ssGSEA scores for seven LR pairs within the Riaz et al. cohort. The high-LRPS group demonstrated a response rate of 76.2% to ICB therapy, which markedly surpassed the 15.2% observed in the low-LRPS group (Fig. 3 B). Although not statistically significant (p = 0.056), the LRPS in the R group was higher than in the NR group, suggesting a trend toward greater response with elevated LRPS values (Fig. 3 C). ROC analysis further highlighted LRPS’s predictive power, with an AUC of 0.87 in forecasting response to ICB therapy in metastatic melanoma within the training set (Fig. 3 D). Moreover, survival analyses demonstrated that patients with high LRPS experienced extended overall survival (OS) and progression-free survival (PFS) (p < 0.05), underscoring LRPS's prognostic value (Figs. 3 E and 3 F). Additionally, we performed the Global Schoenfeld test to evaluate whether the score acted as a time-dependent covariate. The tests showed p-values greater than 0.05 (Figs. 3 G and 3 H), suggesting that the assumption of proportional hazards was violated. The LRPS score did not vary over time when predicting OS and PFS, indicating it is not a time-dependent covariate in Riaz et al. cohort. 3.3 Validation of LRPS prediction performance We utilized heatmaps to illustrate the distribution of ssGSEA scores for LR pairs within the R and NR groups across validation cohorts (Fig. 4 A). In each cohort, we observed a significantly higher response rate to ICB therapy in the high-LRPS group compared to the low-LRPS group (Fig. 4 B). In all four validation cohorts, LRPS scores were significantly higher in the R group than in the NR group (p < 0.05) (Fig. 4 C). ROC analysis indicated that LRPS was a robust predictor of treatment response, achieving AUC values of 0.8 or higher across the cohorts (Fig. 4 D). Additionally, in the Gide et al. and MGH cohorts, patients with high LRPS experienced longer OS and PFS compared to the low-LRPS group (Figs. 4 E and 4 F), although the OS difference was not statistically significant in the MGH cohort. Additionally, we performed the Global Schoenfeld test to evaluate whether the score acted as a time-dependent covariate. All tests showed p-values greater than 0.05 (Figs. 4 F), suggesting that the assumption of proportional hazards was violated. In summary, the LRPS score did not vary over time when predicting OS and PFS, indicating it is not a time-dependent covariate. We further combined the four validation cohorts, examining LRPS’s ability to predict ICB therapy response across test cohorts. ROC analysis revealed an AUC of 0.78, confirming the predictive strength of LRPS across all test cohorts (Fig. 5 A). Although high LRPS was associated with extended OS in all cohorts, this difference did not reach statistical significance (p = 0.078) (Fig. 5 B). However, high LRPS was significantly associated with improved PFS (p < 0.05) (Fig. 5 C). We performed a chi-square test to assess the association between the LRPS signature and clinical variables, including tumor grade, tumor type (cutaneous vs. non-cutaneous melanoma), mutation type, age, and sex. No statistically significant correlations were found between these variables and the immune-metabolic signature ( Supplementary Table 4 ), indicating that the IMME-ON signature was not significantly influenced by these clinical factors. These findings are limited by the available clinical data, which did not include treatment context, ICB doses, time intervals between diagnosis and sample collection, or disease stage. These factors may potentially affect the signature, but their absence from the dataset prevents further analysis of their impact. 3.4 Impact of Responder Proportion on LRPS Model Performance Although LRPS demonstrated robust predictive performance across validation cohorts, we further examined the impact of responder proportion on model performance metrics to evaluate its clinical applicability. To validate the generalizability of the LRPS model, its performance was evaluated across five independent cohorts, including one training and four test cohorts. As summarized in Supplementary Table 3 , the model achieved robust predictive performance in the training cohort (Riaz et al.) with an AUC of 0.869 and an F1 score of 0.792. In the validation cohorts, the model maintained high AUC values ranging from 0.814 to 0.874. Notably, the best performance was observed in the Gide et al. cohort (AUC = 0.870, F1 = 0.909), which had a relatively balanced responder-to-non-responder ratio (11:7, 61%). In contrast, the Lee and MGH cohorts, both of which exhibited strong class imbalance with responders comprising only 16–17% of samples, showed lower precision (0.375 and 0.385, respectively) and F1 scores (0.545 and 0.556), despite achieving perfect recall (1.000). This suggests that the model successfully identified most responders but at the cost of increased false positives in imbalanced datasets. The Abril-Rodriguez cohort demonstrated consistent performance with an AUC of 0.814, precision of 0.786, and F1 score of 0.815. Collectively, these results indicate that while the LRPS model is generally robust across multiple cohorts, its precision could be affected by responder prevalence, highlighting the importance of class distribution in model evaluation and deployment. 3.5 LRPS is an independent prognostic factor for metastatic melanoma We conducted a multivariate Cox regression analysis in the training dataset from the Riaz et al. cohort, including four variables: LRPS, tumor type, mutation type, and tumor stage, to determine whether LRPS is an independent factor for predicting the prognosis of patients with metastatic melanoma. Our results indicate that LRPS is an independent prognostic marker for OS (HR = 0.002, p = 0.04) and PFS (HR = 0.003, p < 0.01, Figs. 6 A and 6 B). Furthermore, we validated the role of LRPS as an independent prognostic factor for predicting OS and PFS in patients with metastatic melanoma in Gide et al. cohort (Figs. 6 C and 6 D). Discussion Since the approval of ipilimumab, the first checkpoint inhibitor targeting CTLA-4, immunotherapy, particularly with anti-PD-1 and anti-PD-L1 antibodies, has demonstrated remarkable clinical efficacy in treating various cancers and has become a cornerstone of immune checkpoint blockade (ICB) therapy[ 25 ]. H owever, a notable proportion of patients still fail to achieve satisfactory responses to these treatments[ 26 ]. To address this challenge, we developed a scoring system based on seven ligand-receptor (LR) pairs that calculates the Ligand-Receptor Pair Score (LRPS) from on-treatment samples, which may provide insights for predicting responses to ICB therapy in metastatic melanoma. Our findings suggest that patients with high LRPS values exhibit improved treatment response rates across multiple cohorts. ROC analysis revealed that LRPS achieved AUC values exceeding 0.8 in predicting ICB therapy responses in specific cohorts. Furthermore, elevated LRPS was associated with improved prognosis, highlighting its dual role as a predictive and prognostic biomarker. Our study emphasizes the importance of RNA sequencing when on-treatment tumor samples are available. LRPS could help stratify patients into subgroups with differential likelihoods of benefiting from ICB therapy. Specifically, patients with higher LRPS scores may experience superior therapeutic outcomes, whereas those with low LRPS scores may require combination therapies (e.g., chemotherapy or radiotherapy) to remodel the immunosuppressive tumor microenvironment and enhance immune cell activity. Alternative treatment strategies should also be explored for this subgroup. Among the seven LR pairs identified, FLT3-FLT3LG had the highest weight in the elastic net model. FLT3 and its ligand FLT3LG are critical for dendritic cell (DC) development and activation[ 27 ], processes essential for antigen presentation and adaptive immune response initiation. FLT3LG promotes DC differentiation and expansion, potentially enhancing T cell priming and cytotoxic T lymphocyte (CTL) activation against tumors[ 28 ]. This mechanism may improve immune recognition of tumor cells and synergize with ICB therapy. Lymphocyte Antigen 9 (LY9/CD229), highly expressed on NKT cells[ 29 ]. regulates NKT cell differentiation. LY9 deletion promotes NKT cell differentiation and expands innate CD8⁺ T cell populations [ 30 ]. Given the opposing roles of type I (pro-inflammatory) and type II (immunosuppressive) NKT cells in tumor immunity, LY9 upregulation may skew the balance toward type I NKT cells, thereby augmenting anti-tumor responses [ 29 ]. CD5⁺ dendritic cells (DCs) are critical for T cell priming and antitumor immunity. Recent studies demonstrate that CD5⁺ DCs correlate with improved survival in melanoma patients and expand during ICB therapy [ 31 ]. Loss of CD5 impairs T cell-mediated tumor clearance, underscoring its essential role in ICB efficacy. Elevated CD40LG expression in tumors is associated with M2 macrophage infiltration and immunosuppression[ 32 – 34 ]. Concurrently, the integrin heterodimer ITGA2B/ITGB3 (αIIbβ3) may promote tumor cell adhesion to extracellular matrix (ECM) components, creating physical barriers to immune infiltration[ 35 , 36 ]. Together, this axis may foster an immune-evasive microenvironment, reducing ICB efficacy. Amyloid precursor protein (APP) overexpression activates MAPK signaling to drive tumor invasion and metastasis[ 37 ]. Tumor-derived APP binding to CD74 may shift the tumor microenvironment toward immunosuppression[ 38 ]., suggesting that targeting the APP-CD74 axis could restore immune activity and improve ICB outcomes. Elevated TNFRSF17 (BCMA) and TNFSF13 (APRIL) expression in high-LRPS patients aligns with evidence linking plasma cell infiltration to improved immunotherapy outcomes[ 8 ].TNFSF13 binding to BCMA and TACI promotes B cell activation, plasma cell differentiation, and long-term survival[ 39 ], facilitating antibody-dependent cellular phagocytosis (ADCP) of tumor cells[ 39 , 40 ] This mechanism may synergize with ICB to amplify antitumor immunity. FCER2 (CD23) suppresses plasma cell differentiation and antibody production by inhibiting class-switched B cells[ 41 ]. Meanwhile, ITGAV/ITGB3 (αVβ3 integrin) drives TGF-β-mediated stromal remodeling and PD-L1 upregulation[ 35 , 36 ]. Their synergy may establish a feedforward loop enhancing immune evasion. Our study has several limitations that should be acknowledged. First, the retrospective nature of our analysis, relying on aggregated transcriptomic data from public datasets, introduces inherent constraints including potential selection bias and limited patient diversity. Technical variability across sequencing platforms may persist despite normalization efforts, while incomplete annotation of critical clinical covariates (e.g., pretreatment regimens) constrain biological interpretation and global generalizability. Third, while trends were consistent, borderline significance in the training set necessitates validation in larger prospective studies. Fourth, the precision of the LRPS model was reduced in cohorts with severe class imbalance, where perfect recall coincided with elevated false positives, underscoring the need to account for responder prevalence in clinical deployment. Finally, mechanistic insights into specific LR pairs remain incomplete and require experimental validation. In conclusion, the Ligand-Receptor Pair Score (LRPS) derived from on-treatment tumor samples shows promise as a predictive biomarker for ICB therapy response in metastatic melanoma. Patients with high LRPS exhibit consistently improved treatment responses and prognosis across cohorts, supporting its clinical utility. This study provides foundational evidence for integrating LRPS into precision immunotherapy strategies, though further validation is warranted. Declarations Acknowledgements Not applicable. Authors’ contributions Y.T.F and J.D.W participated in the data analysis, organized the article writing, and critically modified the manuscript. H.C.Z and Z.M.Z modified the manuscript, drafted the manuscript and were responsiblefor the acquisition of data; R.D.Z and Q.Z.J contributed to the literature search, and correct language expression. All authors read and approved the manuscript and agree to be accountable for all aspects of the research in ensuring that the accuracy or integrity of any part of the work are appropriately investigated and resolved. Funding This work was supported by the Youth Science Foundation of the Cancer Hospital of Shantou University Medical College (Grant No. 2023A002). Availability of data and materials The datasets analyzed in this study are currently available in the respective online repositories. The specific access links or accession numbers are as follows: Riaz et al. dataset, available in the GEO database (https://www.ncbi.nlm.nih.gov/geo/), accession number GSE120575. Gide et al. dataset, available in the BioProject database (https://www.ncbi.nlm.nih.gov/bioproject), accession number PRJEB23709. Abril et al. dataset, available in the dbGaP database (https://www.ncbi.nlm.nih.gov/gap/), accession number phs001919.v1.p1. Lee et al. dataset, available in the EGA database (https://ega-archive.org/?lang=zh), accession number EGAD00001005738. MGH et al. dataset, available in the GEO database (https://www.ncbi.nlm.nih.gov/geo/), accession numbers GSE115821 and GSE168204. Ethics approval This study does not involve animal or clinical experiments. All data were obtained from public databases, and therefore, it does not require submission for ethical review. Conflicts of interest The authors declare no conflicts of interest related to this study. References Wang, S.J., S.K. Dougan, and M. Dougan, Immune mechanisms of toxicity from checkpoint inhibitors. Trends in Cancer, 2023. 9 (7): p. 543-553. Sinicrope, F.A. and M.J. Turk, Immune checkpoint blockade: timing is everything. Journal for ImmunoTherapy of Cancer, 2024. 12 (8): p. e009722. Lauss, M., et al., Molecular patterns of resistance to immune checkpoint blockade in melanoma. Nature Communications, 2024. 15 (1): p. 3075. Du, K., et al., Pathway signatures derived from on-treatment tumor specimens predict response to anti-PD1 blockade in metastatic melanoma. Nature Communications, 2021. 12 (1): p. 6023. Chan, T., et al., Development of Tumor Mutation Burden as an Immunotherapy Biomarker: Utility for the Oncology Clinic. Annals of oncology : official journal of the European Society for Medical Oncology, 2018. 30 . Ju, M., et al., Pan-cancer analysis of NLRP3 inflammasome with potential implications in prognosis and immunotherapy in human cancer. Briefings in Bioinformatics, 2020. 22 . Xiong, D., Y. Wang, and M. You, A gene expression signature of TREM2hi macrophages and γδ T cells predicts immunotherapy response. Nature Communications, 2020. 11 (1): p. 5084. Chen, S., et al. Functional Gene Expression Signatures from On-Treatment Tumor Specimens Predict Anti-PD1 Blockade Response in Metastatic Melanoma . Biomolecules, 2023. 13 , DOI: 10.3390/biom13010058. Chen, S., et al., Inflammatory response signature score model for predicting immunotherapy response and pan-cancer prognosis. Computational and Structural Biotechnology Journal, 2024. 23 : p. 369-383. Chen, S., et al., A tumor-associated endothelial signature score model in immunotherapy and prognosis across pan-cancers. Frontiers in Pharmacology, 2023. 14 . Chen, S., et al., Comprehensive analysis of glycoprotein VI-mediated platelet activation signaling pathway for predicting pan-cancer survival and response to anti-PD-1 immunotherapy. Computational and Structural Biotechnology Journal, 2023. 21 : p. 2873-2883. Auslander, N., et al., Robust prediction of response to immune checkpoint blockade therapy in metastatic melanoma. Nature Medicine, 2018. 24 (10): p. 1545-1549. Kodet, O., et al. The Abscopal Effect in the Era of Checkpoint Inhibitors . International Journal of Molecular Sciences, 2021. 22 , DOI: 10.3390/ijms22137204. Wilczyński, B., et al., Chemoresistance and the tumor microenvironment: the critical role of cell–cell communication. Cell Communication and Signaling, 2024. 22 (1): p. 486. Lee, S.-C., et al. CXCL16/CXCR6 Axis in Adipocytes Differentiated from Human Adipose Derived Mesenchymal Stem Cells Regulates Macrophage Polarization . Cells, 2021. 10 , DOI: 10.3390/cells10123410. Riaz, N., et al., Tumor and Microenvironment Evolution during Immunotherapy with Nivolumab. Cell, 2017. 171 (4): p. 934-949.e16. Gide, T.N., et al., Distinct Immune Cell Populations Define Response to Anti-PD-1 Monotherapy and Anti-PD-1/Anti-CTLA-4 Combined Therapy. Cancer Cell, 2019. 35 (2): p. 238-255.e6. Abril-Rodriguez, G., et al., PAK4 inhibition improves PD-1 blockade immunotherapy. Nature Cancer, 2020. 1 (1): p. 46-58. Lee, J.H., et al., Transcriptional downregulation of MHC class I and melanoma de- differentiation in resistance to PD-1 inhibition. Nature Communications, 2020. 11 (1): p. 1897. Singla, R., et al., Comparison of RECIST and iRECIST criteria in patients with advanced lung cancer treated with nivolumab. Journal of Cancer Research and Therapeutics, 2023. 19 (5). Shilts, J., et al., A physical wiring diagram for the human immune system. Nature, 2022. 608 (7922): p. 397-404. Hu, C., et al., CellMarker 2.0: an updated database of manually curated cell markers in human/mouse and web tools based on scRNA-seq data. Nucleic Acids Research, 2023. 51 (D1): p. D870-D876. Tibshirani, R., T. Hastie, and J. Friedman, Regularized Paths for Generalized Linear Models Via Coordinate Descent. Journal of Statistical Software, 2010. 33 . Lopez-Gonzalez, L., et al. Exploring Biomarkers in Breast Cancer: Hallmarks of Diagnosis, Treatment, and Follow-Up in Clinical Practice . Medicina, 2024. 60 , DOI: 10.3390/medicina60010168. He, X. and C. Xu, Immune checkpoint signaling and cancer immunotherapy. Cell Research, 2020. 30 (8): p. 660-669. Morad, G., et al., Hallmarks of response, resistance, and toxicity to immune checkpoint blockade. Cell, 2021. 184 (21): p. 5309-5337. Kyrysyuk, O. and K.W. Wucherpfennig, Designing Cancer Immunotherapies That Engage T Cells and NK Cells. Annual Review of Immunology, 2023. 41 (Volume 41, 2023): p. 17-38. Barry, K.C., et al., A natural killer-dendritic cell axis defines checkpoint therapy-responsive tumor microenvironments. Nature Medicine, 2018. 24 (8): p. 1178-1191. Deng, K., et al., Study of LY9 as a potential biomarker for prognosis and prediction of immunotherapy efficacy in lung adenocarcinoma. Peerj, 2024. 12 : p. 1-24. Cuenca, M., et al., Ly9 (SLAMF3) receptor differentially regulates iNKT cell development and activation in mice. European Journal of Immunology, 2018. 48 (1): p. 99-105. He, M.Y., et al., CD5 expression by dendritic cells directs T cell immunity and sustains immunotherapy responses. Science, 2023. 379 (6633): p. 661-+. Wu, Y., et al., Expression of CD40 and growth-inhibitory activity of CD40 ligand in colon cancer ex vivo. Cellular Immunology, 2008. 253 (1): p. 102-109. Kim, S.-M., et al., Glioblastoma-educated mesenchymal stem-like cells promote glioblastoma infiltration via extracellular matrix remodelling in the tumour microenvironment. Clinical and Translational Medicine, 2022. 12 (8): p. e997. Li, W., et al., Pathomics models for CD40LG expression and prognosis prediction in glioblastoma. Scientific Reports, 2024. 14 (1): p. 24350. Bahn, Y.J., et al., TGF-β antagonism synergizes with PPARγ agonism to reduce fibrosis and enhance beige adipogenesis. Molecular Metabolism, 2024. 90 : p. 102054. Sutherland, T.E., D.P. Dyer, and J.E. Allen, The extracellular matrix and the immune system: A mutually dependent relationship. Science. 379 (6633): p. eabp8964. Lee, H.N., M.S. Jeong, and S.B. Jang Molecular Characteristics of Amyloid Precursor Protein (APP) and Its Effects in Cancer . International Journal of Molecular Sciences, 2021. 22 , DOI: 10.3390/ijms22094999. Chen, G., et al., Single-cell transcriptomic analysis reveals that the APP–CD74 axis promotes immunosuppression and progression of testicular tumors. The Journal of Pathology, 2024. 264 (3): p. 250-269. Kroeger, D.R., K. Milne, and B.H. Nelson, Tumor-Infiltrating Plasma Cells Are Associated with Tertiary Lymphoid Structures, Cytolytic T-Cell Responses, and Superior Prognosis in Ovarian Cancer. Clinical Cancer Research, 2016. 22 (12): p. 3005-3015. Ma, X., et al., TNFSF13 insufficiency disrupts human colonic epithelial cell-mediated B cell differentiation. bioRxiv, 2024: p. 2024.09.23.614260. Downs-Canner, S.M., et al., B Cell Function in the Tumor Microenvironment. Annual Review of Immunology, 2022. 40 (Volume 40, 2022): p. 169-193. Additional Declarations No competing interests reported. Supplementary Files SupplementaryTables.xlsx Cite Share Download PDF Status: Under Review Version 1 posted Editorial decision: Revision requested 11 Aug, 2025 Editor assigned by journal 08 May, 2025 Reviews received at journal 21 Apr, 2025 Reviewers agreed at journal 17 Apr, 2025 Reviewers agreed at journal 15 Apr, 2025 Reviewers invited by journal 15 Apr, 2025 Submission checks completed at journal 15 Apr, 2025 First submitted to journal 24 Mar, 2025 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-5854916","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":443337075,"identity":"a2dfd137-7c4e-4653-a991-17a94aa3ce49","order_by":0,"name":"Huancheng Zeng","email":"","orcid":"","institution":"Cancer Hospital of Shantou University Medical College","correspondingAuthor":false,"prefix":"","firstName":"Huancheng","middleName":"","lastName":"Zeng","suffix":""},{"id":443337076,"identity":"1c963bf9-d4ab-45c9-8d94-d703d8386baa","order_by":1,"name":"Rendong Zhang","email":"","orcid":"","institution":"Cancer Hospital of Shantou University Medical College","correspondingAuthor":false,"prefix":"","firstName":"Rendong","middleName":"","lastName":"Zhang","suffix":""},{"id":443337077,"identity":"0fe1278c-6d6d-4740-8d72-0938c17a5dac","order_by":2,"name":"Qiongzhi Jiang","email":"","orcid":"","institution":"Cancer Hospital of Shantou University Medical College","correspondingAuthor":false,"prefix":"","firstName":"Qiongzhi","middleName":"","lastName":"Jiang","suffix":""},{"id":443337078,"identity":"e792f376-323e-4f3b-be16-730a3c88683d","order_by":3,"name":"Jundong Wu","email":"","orcid":"","institution":"Cancer Hospital of Shantou University Medical College","correspondingAuthor":false,"prefix":"","firstName":"Jundong","middleName":"","lastName":"Wu","suffix":""},{"id":443337079,"identity":"6e2dfc3b-1fe3-4757-891a-0e204188be8a","order_by":4,"name":"Zhemin Zhuang","email":"","orcid":"","institution":"Shantou University","correspondingAuthor":false,"prefix":"","firstName":"Zhemin","middleName":"","lastName":"Zhuang","suffix":""},{"id":443337080,"identity":"260d11d7-b2fd-4c2e-99b4-f80854d2fe54","order_by":5,"name":"Yutong Fang","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAwElEQVRIiWNgGAWjYBAC9gYGhgMfKmzk2NibDxCnhecAM+PBGWfSjPl4jiUQrYX5MG/bocR5EjkKRGqRyD8A1HIgvY0hh4HhR8U2YrQkMxycc+5ObhvD2QOMPWduE9ZiD9Ry4E3Zs9w2xr4EZsY2IrSAbDnAw3Y4nY2Zx4B4LQd52g4nsLERrYXnsQEokA3beNgSDhLlFx72xMcfgFEpLz//8cEHPyqI0MIgkIBgHyBCPRDwE6luFIyCUTAKRjAAAKUKQCh/0GF0AAAAAElFTkSuQmCC","orcid":"","institution":"Cancer Hospital of Shantou University Medical College","correspondingAuthor":true,"prefix":"","firstName":"Yutong","middleName":"","lastName":"Fang","suffix":""}],"badges":[],"createdAt":"2025-01-18 11:53:07","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-5854916/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-5854916/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":80790514,"identity":"1e0bdb09-5168-432e-a06b-66e7acfc9b79","added_by":"auto","created_at":"2025-04-17 06:38:59","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":50178,"visible":true,"origin":"","legend":"\u003cp\u003eThe flow chart of LRPS signature score construction.\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-5854916/v1/0d6619ac4f53192c62041241.png"},{"id":80790478,"identity":"ab5dad99-37a7-4bd3-ae14-f6ef3b56916f","added_by":"auto","created_at":"2025-04-17 06:38:56","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":165053,"visible":true,"origin":"","legend":"\u003cp\u003eLRPS signature for on-treatment samples from Riaz et.al. cohort. (a, b). The modelʹs training parameter selection process was used to generate the Riaz et al. on‐treatment samples to generate the LRPS signature. To avoid overfitting, 5‐fold cross‐validation was performed with the parameter setting as ʺtype.measure = auc, family = ‘binomial’.ʺ (c). LRPS signatures consisted of seven selected frames associated with the effect sizes (variable weights) from the elasticnet penalized logistic regression model.\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-5854916/v1/f6cd6fcba3c245122820145d.png"},{"id":80790529,"identity":"8d791e7b-aa6a-4332-b48d-3167ff7e695a","added_by":"auto","created_at":"2025-04-17 06:39:00","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":231691,"visible":true,"origin":"","legend":"\u003cp\u003eLRPS signature for on-treatment samples from Riaz et.al. cohort. (a). Heatmap representing the single sample gene set enrichment analysis value of ontreatment nonresponders (NR) and responders (R) in the Riaz et al. cohort. Nonresponders are presented with the number of samples on the left side, and responders are presented with the number of samples on the right side. (b). Boxplot of LRPS signature scores for on treatment samples from the Riaz et al. datasets. p values were computed via a one-sided Wilcoxon rank-sum test. (c).The response rate to immunotherapy in low and high scores groups stratified by IMME-ON signature core in the Riaz et.al. cohort. (d). Receiver operating correlation curve and area under the curve of LRPS signatures for pretreatment samples from the Riaz et al. cohort. (e,f) Kaplan–Meier curves of PFS and OS for pretreatment samples based on LRPS signature scores for the Riaz et al. cohort. The two-sided log-rank test compared high and low subgroups based on the mean of pre-treatment samples odd ratio as cutoff. Hazard ratio (HR) was calculated and shown with confidence interval (CI). (g-h) Graphical assessment of the proportional hazard assumption of the LRPS signature in on‐treatment samples.\u003c/p\u003e","description":"","filename":"3.png","url":"https://assets-eu.researchsquare.com/files/rs-5854916/v1/7b72634a04dbbf184e3141a6.png"},{"id":80791154,"identity":"3ccb636a-c1e5-4c17-b280-92da9ec876bf","added_by":"auto","created_at":"2025-04-17 06:47:00","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":293474,"visible":true,"origin":"","legend":"\u003cp\u003eLRPS signature for on-treatment samples from validation cohorts. (a). Heatmap representing the single sample gene set enrichment analysis value of on-treatment nonresponders (NR) and responders (R) in the Gide et al., Abril-Rodriguez et al., and Lee et al.cohorts. (b). Boxplot of LRPS signature scores for on-treatment samples from Gide et al., Abril-Rodriguez et al., and Lee et al.cohorts. p values were computed via a one-sided Wilcoxon rank-sum test. (c). The response rate to immunotherapy in low and high scores groups stratified by LRPS signature core in the Gide et al., Abril-Rodriguez et al., and Lee et al. and MGH cohorts. (d). Receiver operating correlation curve and area under the curve of LRPS signatures for on-treatment samples from the Gide et al., Abril-Rodriguez et al., Lee et al. and MGH cohorts. (e). Kaplan–Meier curves of PFS or OS for on-treatment samples based on LRPS signature scores for Gide et al and MGH cohort.The two-sided log-rank test compared high and low subgroups based on the mean of pre-treatment samples odd ratio as cutoff. Hazard ratio (HR) was calculated and shown with confidence interval (CI) (f) Graphical assessment of the proportional hazard assumption of the LRPS signature in on‐treatment samples.\u003c/p\u003e","description":"","filename":"4.png","url":"https://assets-eu.researchsquare.com/files/rs-5854916/v1/ceef6bc0044106a10a73375c.png"},{"id":80790519,"identity":"e4de56cc-de18-4712-8c31-092df759ea95","added_by":"auto","created_at":"2025-04-17 06:38:59","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":87913,"visible":true,"origin":"","legend":"\u003cp\u003eLRPS signature for all on-treatment samples from test cohorts. (a). Receiver operating correlation curve and area under the curve of LRPS signatures for all on-treatment samples from test cohorts. (b). Kaplan–Meier curves of PFS or OS for on-treatment samples based on LRPS signature scores for all on-treatment samples from test cohorts. The two-sided log-rank test compared high and low subgroups based on the mean of pre-treatment samples odd ratio as cutoff. Hazard ratio (HR) was calculated and shown with confidence interval (CI).\u003c/p\u003e","description":"","filename":"5.png","url":"https://assets-eu.researchsquare.com/files/rs-5854916/v1/906376a9add6e5995ad005ef.png"},{"id":80790537,"identity":"4640adbb-c227-4165-b6a7-d56ff83bb394","added_by":"auto","created_at":"2025-04-17 06:39:01","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":141290,"visible":true,"origin":"","legend":"\u003cp\u003eResults of Cox proportional hazards regression for PSS and OS analysis using LRPS signature scores for on-treatment samples. If the p-value is less than 0.05, it is shown in bold font.\u003c/p\u003e","description":"","filename":"6.png","url":"https://assets-eu.researchsquare.com/files/rs-5854916/v1/38898418b089d97b8da8c0e8.png"},{"id":80791166,"identity":"30ad5cd9-7b29-47b1-aa97-c90db4bfe82e","added_by":"auto","created_at":"2025-04-17 06:47:09","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1360865,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-5854916/v1/fc1ae351-427e-40d4-925c-0dbe2f7ccd61.pdf"},{"id":80790531,"identity":"9d405ca1-de43-4a0f-9e04-feb965d20189","added_by":"auto","created_at":"2025-04-17 06:39:00","extension":"xlsx","order_by":0,"title":"","display":"","copyAsset":false,"role":"supplement","size":31956,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryTables.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-5854916/v1/8c68bfac81172e9ae03b1f69.xlsx"}],"financialInterests":"No competing interests reported.","formattedTitle":"Ligand-receptor pair-based signature score Derived from On-treatment Tumor Specimens Predicts Immune Checkpoint Blockade Response in Metastatic Melanoma","fulltext":[{"header":"Introduction","content":"\u003cp\u003eAlthough immune checkpoint blockade (ICB) has transformed clinical management of advanced malignancies including melanoma, durable therapeutic benefits remain limited to a subset of patients[\u003cspan additionalcitationids=\"CR2\" citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e]. Predictive biomarkers capable of stratifying responders could mitigate adverse effects, prevent acquired resistance, and optimize resource utilization in immunotherapy[\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e], underscoring the urgent need for reliable signatures to guide ICB treatment selection.\u003c/p\u003e \u003cp\u003eCurrent predictive frameworks for metastatic melanoma ICB response incorporate diverse biomarkers including but not limited to tumor mutational burden quantification[\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e], NLRP3 inflammasome signature [\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e], immune infiltrate composition[\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e, \u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e], inflammatory signature[\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e], tumor-associated endothelial gene signature[\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e], platelet activation cascades [\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e] and immune checkpoint interaction scores (IMPRES)[\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e]. Notably, these biomarkers predominantly derive from pre-therapeutic biopsies, potentially overlooking dynamic microenvironmental changes induced by ICB itself.\u003c/p\u003e \u003cp\u003eLigand-receptor (LR) pair interactions in the tumor microenvironment mediate critical cell-to-cell communication processes that drive oncogenesis. These interactions not only regulate tumor cell proliferation and survival but also orchestrate metastatic progression[\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e]. Specifically, tumor cells communicate with their microenvironment through direct cell-cell contacts and secretion of soluble mediators such as growth factors, cytokines, and chemokines[\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e]. Moreover, stromal components including mesenchymal stem cells, cancer-associated fibroblasts, and tumor-associated macrophages actively modulate malignant phenotypes via bidirectional communication[\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e]. For example, CXCL16-CXCR6 axis dysregulation has been shown to promote immunosuppressive macrophage polarization[\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e]. Nevertheless, whether dynamic LR pair signatures from on-treatment samples can predict ICB response outcomes in metastatic melanoma remains unexplored.\u003c/p\u003e \u003cp\u003eWe hypothesize that dynamic LR network rewiring during ICB treatment reflects functional immune-tumor crosstalk and can be translated into a predictive signature (LRPS) with enhanced prognostic value. Through analysis of transcriptomic and clinical data from ICB-treated metastatic melanoma patients, we developed an LRPS using ElasticNet penalized Logistic Regression (ENlR) to predict treatment response in on-treatment specimens. This approach not only provides a clinically actionable biomarker but also identifies targetable communication nodes for rational combination immunotherapy design.\u003c/p\u003e"},{"header":"Methods","content":"\u003cp\u003eData collection\u003c/p\u003e \u003cp\u003eThe workflow of this study is illustrated in Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e. We compiled transcriptomic data from five independent cohorts of on-treatment metastatic melanoma samples. These datasets include the Riaz et al. cohort (GEO: GSE120575) [\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e], the Gide et al. cohort (BioProject: PRJEB23709) [\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e], the Abril et al. cohort (dbGaP: phs001919.v1.p1) [\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e], the Lee et al. cohort (EGA: EGAD00001005738) [\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e], and the MGH et al. cohort (GEO: GSE115821 and GSE168204). We compiled transcriptomic data from five independent cohorts of on-treatment metastatic melanoma samples, defined as biopsies collected after initiation of ICB therapy (anti-PD-1/PD-L1\u0026thinsp;\u0026plusmn;\u0026thinsp;anti-CTLA-4). Samples were excluded from analysis based on the following criteria: (1) absence of RNA-seq data; (2) missing clinical response information; or (3) duplicated tumor specimens from the same timepoint in individual patients.\u003c/p\u003e \u003cp\u003eGene expression was quantified using transcripts per million (TPM) for each gene. According to RECIST criteria[\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e], responders (R) were defined as individuals who achieved complete response (CR), partial response (PR), or progression-free survival (PFS) of over 180 days. Non-responders (NR) were classified as those with PFS under 180 days accompanied by progressive disease (PD).\u003c/p\u003e \u003cp\u003eLigand-Receptor Pair Selection\u003c/p\u003e \u003cp\u003eLR pairs were curated from published literature [\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e, \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e]. The selection of these pairs was grounded in the biological relevance of these interactions in the immune response. The comprehensive list of selected LR pairs is presented in \u003cb\u003eSupplementary Table\u0026nbsp;1.\u003c/b\u003e\u003c/p\u003e \u003cp\u003eCalculation of signature scores\u003c/p\u003e \u003cp\u003eIn this study, we utilized the Riaz et al. cohort as the training dataset, while the remaining cohorts served as validation datasets. We computed the score for each LR pair in on-treatment metastatic melanoma samples from the Riaz et al. cohort using the single-sample gene set enrichment analysis (ssGSEA) algorithm. Subsequently, we applied an Elastic-Net penalized logistic regression model to identify the LR pairs most strongly associated with response to ICB therapy. To further refine our analysis, we computed a weighted average for each sample, using the effect size as a weighting factor for the LR pairs most associated with ICB therapy response. This weighted average was defined as the LR pair-based signature score (LRPS).\u003c/p\u003e \u003cp\u003eTo minimize the risk of overfitting, five-fold cross-validation was performed on the training set[\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e]. To correct for class imbalance in our dataset, we incorporated a cost-sensitive learning approach during model training, which was implemented the cv.glmnet, a function from the glmnet package. Specifically, we applied a prior probability-based offset in the binomial logistic regression model. This offset was computed as a log-odds correction term using the empirical class frequencies and a predefined classification threshold (τ\u0026thinsp;=\u0026thinsp;2/3). The model was trained using cross-validation (nfolds\u0026thinsp;=\u0026thinsp;5) to optimize the regularization parameter (lambda) with respect to the area under the curve (AUC). By incorporating the correction term, the model was encouraged to assign greater importance to the minority class without modifying the training data or employing resampling techniques. This strategy enabled a more balanced and robust model performance under imbalanced class conditions.\u003c/p\u003e \u003cp\u003eThe predictive performance of the model was assessed by generating receiver operating characteristic (ROC) curves, with the predictive accuracy of the LRPS quantified by the area under the curve (AUC). Using data from the Riaz et al. cohort, the optimal threshold value was determined using Youden\u0026rsquo;s index during ROC analysis. Finally, the odds ratios for individual samples were calculated based on their LRPS.\u003c/p\u003e \u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003eStatistical Analysis\u003c/h2\u003e \u003cp\u003eAll statistical analyses and graphical visualizations were conducted using R software (version 4.2.0) and GraphPad Prism (version 8.0). The one-tailed Wilcoxon rank-sum test was applied to assess statistical differences between the R and NR groups. We calculated the sample\u0026rsquo;s odd ratio based on the signature score. Samples were classified into high and low group use mean value of samples\u0026rsquo; odd ratio as cutoff. Kaplan-Meier (KM) survival analysis, combined with the log-rank test, was used to evaluate survival differences between these groups. Cox proportional hazards regression was performed to estimate the hazard ratio (HR) for each cohort. Given the exploratory nature of this biomarker study, we used a significance threshold of 0.05 when reporting p-values, with trends between 0.05 and 0.1 reported to avoid arbitrary dichotomization of significance [\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e].\u003c/p\u003e \u003c/div\u003e"},{"header":"Results","content":"\u003cp\u003e3.1 Patient cohorts\u003c/p\u003e \u003cp\u003eIn the Riaz et al. dataset, a total of 54 biopsies at the on-treatment timepoint were included in the analysis. Of these, 21 biopsies were from R (Responders), and 33 were from NR (Non-Responders). All patients received anti-PD1 monotherapy. In the Gide et al. dataset, 17 patients with 18 on-treatment biopsies (11 R and 7 NR) received anti-PD1 monotherapy. The median age of the patients was 56 years, with 41% over 60 years old. Additionally, 53% of the patients were male. In the Lee et al. dataset, 23 patients with 35 on-treatment biopsies (6 R and 29 NR) were treated with anti-PD1 monotherapy, including nivolumab or pembrolizumab. The Abril-Rodriguez cohort included 30 on-treatment tumor samples (13 R and 17 NR). All patients received PD-L1 monotherapy (pembrolizumab). The MGH cohort included 31 on-treatment samples (5 R and 26 NR). All patients received anti-PD1/PD-L1 monotherapy. The baseline characteristics of the patients are presented in \u003cb\u003eSupplementary Table\u0026nbsp;2\u003c/b\u003e.\u003c/p\u003e \u003cp\u003eAs illustrated in Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e, the Riaz et al. cohort, comprising 54 on-treatment metastatic melanoma samples, was utilized as the training set, while other cohorts served as validation sets. The ssGSEA algorithm was employed to compute scores for ligand-receptor (LR) pairs within the Riaz et al. training dataset. An elastic net algorithm was then applied to identify LR pairs significantly associated with response to ICB therapy. To mitigate overfitting, we implemented five cross-validation and used a cost-sensitive algorithm (Figs.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eA and \u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eB). Ultimately, as shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eC, seven LR pairs associated with ICB therapy response were identified: FLT3-FLT3LG, LY9-LY9, CD5-CD5, CD40LG-ITGA2B/ITGB3, APP-CD74, TNFRSF17-TNFSF13, FCER2-ITGAV/ITGB3. Effect size was subsequently used as a weighting factor to calculate the weighted average of the ssGSEA scores for these seven LR pairs, defining the LRPS (Ligand-Receptor Pair Score) for each sample.\u003c/p\u003e \u003cp\u003e3.2. LRPS is an effective predictor of response and prognosis in ICB therapy for on-treatment samples\u003c/p\u003e \u003cp\u003eAs illustrated in Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eA, the heatmap presents the distribution of ssGSEA scores for seven LR pairs within the Riaz et al. cohort. The high-LRPS group demonstrated a response rate of 76.2% to ICB therapy, which markedly surpassed the 15.2% observed in the low-LRPS group (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eB). Although not statistically significant (p\u0026thinsp;=\u0026thinsp;0.056), the LRPS in the R group was higher than in the NR group, suggesting a trend toward greater response with elevated LRPS values (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eC). ROC analysis further highlighted LRPS\u0026rsquo;s predictive power, with an AUC of 0.87 in forecasting response to ICB therapy in metastatic melanoma within the training set (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eD). Moreover, survival analyses demonstrated that patients with high LRPS experienced extended overall survival (OS) and progression-free survival (PFS) (p\u0026thinsp;\u0026lt;\u0026thinsp;0.05), underscoring LRPS's prognostic value (Figs.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eE and \u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eF). Additionally, we performed the Global Schoenfeld test to evaluate whether the score acted as a time-dependent covariate. The tests showed p-values greater than 0.05 (Figs.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eG and \u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eH), suggesting that the assumption of proportional hazards was violated. The LRPS score did not vary over time when predicting OS and PFS, indicating it is not a time-dependent covariate in Riaz et al. cohort.\u003c/p\u003e \u003cp\u003e3.3 Validation of LRPS prediction performance\u003c/p\u003e \u003cp\u003eWe utilized heatmaps to illustrate the distribution of ssGSEA scores for LR pairs within the R and NR groups across validation cohorts (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eA). In each cohort, we observed a significantly higher response rate to ICB therapy in the high-LRPS group compared to the low-LRPS group (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eB). In all four validation cohorts, LRPS scores were significantly higher in the R group than in the NR group (p\u0026thinsp;\u0026lt;\u0026thinsp;0.05) (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eC). ROC analysis indicated that LRPS was a robust predictor of treatment response, achieving AUC values of 0.8 or higher across the cohorts (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eD). Additionally, in the Gide et al. and MGH cohorts, patients with high LRPS experienced longer OS and PFS compared to the low-LRPS group (Figs.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eE and \u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eF), although the OS difference was not statistically significant in the MGH cohort. Additionally, we performed the Global Schoenfeld test to evaluate whether the score acted as a time-dependent covariate. All tests showed p-values greater than 0.05 (Figs.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eF), suggesting that the assumption of proportional hazards was violated. In summary, the LRPS score did not vary over time when predicting OS and PFS, indicating it is not a time-dependent covariate.\u003c/p\u003e \u003cp\u003eWe further combined the four validation cohorts, examining LRPS\u0026rsquo;s ability to predict ICB therapy response across test cohorts. ROC analysis revealed an AUC of 0.78, confirming the predictive strength of LRPS across all test cohorts (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eA). Although high LRPS was associated with extended OS in all cohorts, this difference did not reach statistical significance (p\u0026thinsp;=\u0026thinsp;0.078) (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eB). However, high LRPS was significantly associated with improved PFS (p\u0026thinsp;\u0026lt;\u0026thinsp;0.05) (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eC).\u003c/p\u003e \u003cp\u003eWe performed a chi-square test to assess the association between the LRPS signature and clinical variables, including tumor grade, tumor type (cutaneous vs. non-cutaneous melanoma), mutation type, age, and sex. No statistically significant correlations were found between these variables and the immune-metabolic signature (\u003cb\u003eSupplementary Table\u0026nbsp;4\u003c/b\u003e), indicating that the IMME-ON signature was not significantly influenced by these clinical factors. These findings are limited by the available clinical data, which did not include treatment context, ICB doses, time intervals between diagnosis and sample collection, or disease stage. These factors may potentially affect the signature, but their absence from the dataset prevents further analysis of their impact.\u003c/p\u003e \u003cp\u003e3.4 Impact of Responder Proportion on LRPS Model Performance\u003c/p\u003e \u003cp\u003eAlthough LRPS demonstrated robust predictive performance across validation cohorts, we further examined the impact of responder proportion on model performance metrics to evaluate its clinical applicability. To validate the generalizability of the LRPS model, its performance was evaluated across five independent cohorts, including one training and four test cohorts. As summarized in \u003cb\u003eSupplementary Table\u0026nbsp;3\u003c/b\u003e, the model achieved robust predictive performance in the training cohort (Riaz et al.) with an AUC of 0.869 and an F1 score of 0.792. In the validation cohorts, the model maintained high AUC values ranging from 0.814 to 0.874. Notably, the best performance was observed in the Gide et al. cohort (AUC\u0026thinsp;=\u0026thinsp;0.870, F1\u0026thinsp;=\u0026thinsp;0.909), which had a relatively balanced responder-to-non-responder ratio (11:7, 61%). In contrast, the Lee and MGH cohorts, both of which exhibited strong class imbalance with responders comprising only 16\u0026ndash;17% of samples, showed lower precision (0.375 and 0.385, respectively) and F1 scores (0.545 and 0.556), despite achieving perfect recall (1.000). This suggests that the model successfully identified most responders but at the cost of increased false positives in imbalanced datasets. The Abril-Rodriguez cohort demonstrated consistent performance with an AUC of 0.814, precision of 0.786, and F1 score of 0.815. Collectively, these results indicate that while the LRPS model is generally robust across multiple cohorts, its precision could be affected by responder prevalence, highlighting the importance of class distribution in model evaluation and deployment.\u003c/p\u003e \u003cp\u003e3.5 LRPS is an independent prognostic factor for metastatic melanoma\u003c/p\u003e \u003cp\u003eWe conducted a multivariate Cox regression analysis in the training dataset from the Riaz et al. cohort, including four variables: LRPS, tumor type, mutation type, and tumor stage, to determine whether LRPS is an independent factor for predicting the prognosis of patients with metastatic melanoma. Our results indicate that LRPS is an independent prognostic marker for OS (HR\u0026thinsp;=\u0026thinsp;0.002, p\u0026thinsp;=\u0026thinsp;0.04) and PFS (HR\u0026thinsp;=\u0026thinsp;0.003, p\u0026thinsp;\u0026lt;\u0026thinsp;0.01, Figs.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003eA and \u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003eB). Furthermore, we validated the role of LRPS as an independent prognostic factor for predicting OS and PFS in patients with metastatic melanoma in Gide et al. cohort (Figs.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003eC and \u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003eD).\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eSince the approval of ipilimumab, the first checkpoint inhibitor targeting CTLA-4, immunotherapy, particularly with anti-PD-1 and anti-PD-L1 antibodies, has demonstrated remarkable clinical efficacy in treating various cancers and has become a cornerstone of immune checkpoint blockade (ICB) therapy[\u003cspan class=\"CitationRef\"\u003e25\u003c/span\u003e]. H owever, a notable proportion of patients still fail to achieve satisfactory responses to these treatments[\u003cspan class=\"CitationRef\"\u003e26\u003c/span\u003e]. To address this challenge, we developed a scoring system based on seven ligand-receptor (LR) pairs that calculates the Ligand-Receptor Pair Score (LRPS) from on-treatment samples, which may provide insights for predicting responses to ICB therapy in metastatic melanoma. Our findings suggest that patients with high LRPS values exhibit improved treatment response rates across multiple cohorts. ROC analysis revealed that LRPS achieved AUC values exceeding 0.8 in predicting ICB therapy responses in specific cohorts. Furthermore, elevated LRPS was associated with improved prognosis, highlighting its dual role as a predictive and prognostic biomarker.\u003c/p\u003e\n\u003cp\u003eOur study emphasizes the importance of RNA sequencing when on-treatment tumor samples are available. LRPS could help stratify patients into subgroups with differential likelihoods of benefiting from ICB therapy. Specifically, patients with higher LRPS scores may experience superior therapeutic outcomes, whereas those with low LRPS scores may require combination therapies (e.g., chemotherapy or radiotherapy) to remodel the immunosuppressive tumor microenvironment and enhance immune cell activity. Alternative treatment strategies should also be explored for this subgroup.\u003c/p\u003e\n\u003cp\u003eAmong the seven LR pairs identified, FLT3-FLT3LG had the highest weight in the elastic net model. FLT3 and its ligand FLT3LG are critical for dendritic cell (DC) development and activation[\u003cspan class=\"CitationRef\"\u003e27\u003c/span\u003e], processes essential for antigen presentation and adaptive immune response initiation. FLT3LG promotes DC differentiation and expansion, potentially enhancing T cell priming and cytotoxic T lymphocyte (CTL) activation against tumors[\u003cspan class=\"CitationRef\"\u003e28\u003c/span\u003e]. This mechanism may improve immune recognition of tumor cells and synergize with ICB therapy.\u003c/p\u003e\n\u003cp\u003eLymphocyte Antigen 9 (LY9/CD229), highly expressed on NKT cells[\u003cspan class=\"CitationRef\"\u003e29\u003c/span\u003e]. regulates NKT cell differentiation. LY9 deletion promotes NKT cell differentiation and expands innate CD8⁺ T cell populations [\u003cspan class=\"CitationRef\"\u003e30\u003c/span\u003e]. Given the opposing roles of type I (pro-inflammatory) and type II (immunosuppressive) NKT cells in tumor immunity, LY9 upregulation may skew the balance toward type I NKT cells, thereby augmenting anti-tumor responses [\u003cspan class=\"CitationRef\"\u003e29\u003c/span\u003e].\u003c/p\u003e\n\u003cp\u003eCD5⁺ dendritic cells (DCs) are critical for T cell priming and antitumor immunity. Recent studies demonstrate that CD5⁺ DCs correlate with improved survival in melanoma patients and expand during ICB therapy [\u003cspan class=\"CitationRef\"\u003e31\u003c/span\u003e]. Loss of CD5 impairs T cell-mediated tumor clearance, underscoring its essential role in ICB efficacy.\u003c/p\u003e\n\u003cp\u003eElevated CD40LG expression in tumors is associated with M2 macrophage infiltration and immunosuppression[\u003cspan class=\"CitationRef\"\u003e32\u003c/span\u003e\u0026ndash;\u003cspan class=\"CitationRef\"\u003e34\u003c/span\u003e]. Concurrently, the integrin heterodimer ITGA2B/ITGB3 (\u0026alpha;IIb\u0026beta;3) may promote tumor cell adhesion to extracellular matrix (ECM) components, creating physical barriers to immune infiltration[\u003cspan class=\"CitationRef\"\u003e35\u003c/span\u003e, \u003cspan class=\"CitationRef\"\u003e36\u003c/span\u003e]. Together, this axis may foster an immune-evasive microenvironment, reducing ICB efficacy.\u003c/p\u003e\n\u003cp\u003eAmyloid precursor protein (APP) overexpression activates MAPK signaling to drive tumor invasion and metastasis[\u003cspan class=\"CitationRef\"\u003e37\u003c/span\u003e]. Tumor-derived APP binding to CD74 may shift the tumor microenvironment toward immunosuppression[\u003cspan class=\"CitationRef\"\u003e38\u003c/span\u003e]., suggesting that targeting the APP-CD74 axis could restore immune activity and improve ICB outcomes.\u003c/p\u003e\n\u003cp\u003eElevated TNFRSF17 (BCMA) and TNFSF13 (APRIL) expression in high-LRPS patients aligns with evidence linking plasma cell infiltration to improved immunotherapy outcomes[\u003cspan class=\"CitationRef\"\u003e8\u003c/span\u003e].TNFSF13 binding to BCMA and TACI promotes B cell activation, plasma cell differentiation, and long-term survival[\u003cspan class=\"CitationRef\"\u003e39\u003c/span\u003e], facilitating antibody-dependent cellular phagocytosis (ADCP) of tumor cells[\u003cspan class=\"CitationRef\"\u003e39\u003c/span\u003e, \u003cspan class=\"CitationRef\"\u003e40\u003c/span\u003e] This mechanism may synergize with ICB to amplify antitumor immunity.\u003c/p\u003e\n\u003cp\u003eFCER2 (CD23) suppresses plasma cell differentiation and antibody production by inhibiting class-switched B cells[\u003cspan class=\"CitationRef\"\u003e41\u003c/span\u003e]. Meanwhile, ITGAV/ITGB3 (\u0026alpha;V\u0026beta;3 integrin) drives TGF-\u0026beta;-mediated stromal remodeling and PD-L1 upregulation[\u003cspan class=\"CitationRef\"\u003e35\u003c/span\u003e, \u003cspan class=\"CitationRef\"\u003e36\u003c/span\u003e]. Their synergy may establish a feedforward loop enhancing immune evasion.\u003c/p\u003e\n\u003cp\u003eOur study has several limitations that should be acknowledged. First, the retrospective nature of our analysis, relying on aggregated transcriptomic data from public datasets, introduces inherent constraints including potential selection bias and limited patient diversity. Technical variability across sequencing platforms may persist despite normalization efforts, while incomplete annotation of critical clinical covariates (e.g., pretreatment regimens) constrain biological interpretation and global generalizability. Third, while trends were consistent, borderline significance in the training set necessitates validation in larger prospective studies. Fourth, the precision of the LRPS model was reduced in cohorts with severe class imbalance, where perfect recall coincided with elevated false positives, underscoring the need to account for responder prevalence in clinical deployment. Finally, mechanistic insights into specific LR pairs remain incomplete and require experimental validation.\u003c/p\u003e\n\u003cp\u003eIn conclusion, the Ligand-Receptor Pair Score (LRPS) derived from on-treatment tumor samples shows promise as a predictive biomarker for ICB therapy response in metastatic melanoma. Patients with high LRPS exhibit consistently improved treatment responses and prognosis across cohorts, supporting its clinical utility. This study provides foundational evidence for integrating LRPS into precision immunotherapy strategies, though further validation is warranted.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eAcknowledgements\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNot applicable.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthors’ contributions\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eY.T.F and J.D.W participated in the data analysis, organized the article writing, and critically modified the manuscript. H.C.Z and Z.M.Z modified the manuscript, drafted the manuscript and were responsiblefor the acquisition of data; R.D.Z and Q.Z.J contributed to the literature search, and correct language expression. All authors read and approved the manuscript and agree to be accountable for all aspects of the research in ensuring that the accuracy or integrity of any part of the work are appropriately investigated and resolved.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis work was supported by the Youth Science Foundation of the Cancer Hospital of Shantou University Medical College (Grant No. 2023A002).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAvailability of data and materials\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe datasets analyzed in this study are currently available in the respective online repositories. The specific access links or accession numbers are as follows:\u0026nbsp;Riaz et al. dataset, available in the GEO database (https://www.ncbi.nlm.nih.gov/geo/), accession number GSE120575.\u0026nbsp;Gide et al. dataset, available in the BioProject database (https://www.ncbi.nlm.nih.gov/bioproject), accession number PRJEB23709.\u0026nbsp;Abril et al. dataset, available in the dbGaP database (https://www.ncbi.nlm.nih.gov/gap/), accession number phs001919.v1.p1.\u0026nbsp;Lee et al. dataset, available in the EGA database (https://ega-archive.org/?lang=zh), accession number EGAD00001005738.\u0026nbsp;MGH et al. dataset, available in the GEO database (https://www.ncbi.nlm.nih.gov/geo/), accession numbers GSE115821 and GSE168204.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eEthics approval\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis study does not involve animal or clinical experiments. All data were obtained from public databases, and therefore, it does not require submission for ethical review.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConflicts of interest\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors declare no conflicts of interest related to this study.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n \u003cli\u003eWang, S.J., S.K. Dougan, and M. Dougan, \u003cem\u003eImmune mechanisms of toxicity from checkpoint inhibitors.\u003c/em\u003e Trends in Cancer, 2023. \u003cstrong\u003e9\u003c/strong\u003e(7): p. 543-553.\u003c/li\u003e\n \u003cli\u003eSinicrope, F.A. and M.J. Turk, \u003cem\u003eImmune checkpoint blockade: timing is everything.\u003c/em\u003e Journal for ImmunoTherapy of Cancer, 2024. \u003cstrong\u003e12\u003c/strong\u003e(8): p. e009722.\u003c/li\u003e\n \u003cli\u003eLauss, M., et al., \u003cem\u003eMolecular patterns of resistance to immune checkpoint blockade in melanoma.\u003c/em\u003e Nature Communications, 2024. \u003cstrong\u003e15\u003c/strong\u003e(1): p. 3075.\u003c/li\u003e\n \u003cli\u003eDu, K., et al., \u003cem\u003ePathway signatures derived from on-treatment tumor specimens predict response to anti-PD1 blockade in metastatic melanoma.\u003c/em\u003e Nature Communications, 2021. \u003cstrong\u003e12\u003c/strong\u003e(1): p. 6023.\u003c/li\u003e\n \u003cli\u003eChan, T., et al., \u003cem\u003eDevelopment of Tumor Mutation Burden as an Immunotherapy Biomarker: Utility for the Oncology Clinic.\u003c/em\u003e Annals of oncology : official journal of the European Society for Medical Oncology, 2018. \u003cstrong\u003e30\u003c/strong\u003e.\u003c/li\u003e\n \u003cli\u003eJu, M., et al., \u003cem\u003ePan-cancer analysis of NLRP3 inflammasome with potential implications in prognosis and immunotherapy in human cancer.\u003c/em\u003e Briefings in Bioinformatics, 2020. \u003cstrong\u003e22\u003c/strong\u003e.\u003c/li\u003e\n \u003cli\u003eXiong, D., Y. Wang, and M. You, \u003cem\u003eA gene expression signature of TREM2hi macrophages and \u0026gamma;\u0026delta; T cells predicts immunotherapy response.\u003c/em\u003e Nature Communications, 2020. \u003cstrong\u003e11\u003c/strong\u003e(1): p. 5084.\u003c/li\u003e\n \u003cli\u003eChen, S., et al. \u003cem\u003eFunctional Gene Expression Signatures from On-Treatment Tumor Specimens Predict Anti-PD1 Blockade Response in Metastatic Melanoma\u003c/em\u003e. Biomolecules, 2023. \u003cstrong\u003e13\u003c/strong\u003e, DOI: 10.3390/biom13010058.\u003c/li\u003e\n \u003cli\u003eChen, S., et al., \u003cem\u003eInflammatory response signature score model for predicting immunotherapy response and pan-cancer prognosis.\u003c/em\u003e Computational and Structural Biotechnology Journal, 2024. \u003cstrong\u003e23\u003c/strong\u003e: p. 369-383.\u003c/li\u003e\n \u003cli\u003eChen, S., et al., \u003cem\u003eA tumor-associated endothelial signature score model in immunotherapy and prognosis across pan-cancers.\u003c/em\u003e Frontiers in Pharmacology, 2023. \u003cstrong\u003e14\u003c/strong\u003e.\u003c/li\u003e\n \u003cli\u003eChen, S., et al., \u003cem\u003eComprehensive analysis of glycoprotein VI-mediated platelet activation signaling pathway for predicting pan-cancer survival and response to anti-PD-1 immunotherapy.\u003c/em\u003e Computational and Structural Biotechnology Journal, 2023. \u003cstrong\u003e21\u003c/strong\u003e: p. 2873-2883.\u003c/li\u003e\n \u003cli\u003eAuslander, N., et al., \u003cem\u003eRobust prediction of response to immune checkpoint blockade therapy in metastatic melanoma.\u003c/em\u003e Nature Medicine, 2018. \u003cstrong\u003e24\u003c/strong\u003e(10): p. 1545-1549.\u003c/li\u003e\n \u003cli\u003eKodet, O., et al. \u003cem\u003eThe Abscopal Effect in the Era of Checkpoint Inhibitors\u003c/em\u003e. International Journal of Molecular Sciences, 2021. \u003cstrong\u003e22\u003c/strong\u003e, DOI: 10.3390/ijms22137204.\u003c/li\u003e\n \u003cli\u003eWilczyński, B., et al., \u003cem\u003eChemoresistance and the tumor microenvironment: the critical role of cell\u0026ndash;cell communication.\u003c/em\u003e Cell Communication and Signaling, 2024. \u003cstrong\u003e22\u003c/strong\u003e(1): p. 486.\u003c/li\u003e\n \u003cli\u003eLee, S.-C., et al. \u003cem\u003eCXCL16/CXCR6 Axis in Adipocytes Differentiated from Human Adipose Derived Mesenchymal Stem Cells Regulates Macrophage Polarization\u003c/em\u003e. Cells, 2021. \u003cstrong\u003e10\u003c/strong\u003e, DOI: 10.3390/cells10123410.\u003c/li\u003e\n \u003cli\u003eRiaz, N., et al., \u003cem\u003eTumor and Microenvironment Evolution during Immunotherapy with Nivolumab.\u003c/em\u003e Cell, 2017. \u003cstrong\u003e171\u003c/strong\u003e(4): p. 934-949.e16.\u003c/li\u003e\n \u003cli\u003eGide, T.N., et al., \u003cem\u003eDistinct Immune Cell Populations Define Response to Anti-PD-1 Monotherapy and Anti-PD-1/Anti-CTLA-4 Combined Therapy.\u003c/em\u003e Cancer Cell, 2019. \u003cstrong\u003e35\u003c/strong\u003e(2): p. 238-255.e6.\u003c/li\u003e\n \u003cli\u003eAbril-Rodriguez, G., et al., \u003cem\u003ePAK4 inhibition improves PD-1 blockade immunotherapy.\u003c/em\u003e Nature Cancer, 2020. \u003cstrong\u003e1\u003c/strong\u003e(1): p. 46-58.\u003c/li\u003e\n \u003cli\u003eLee, J.H., et al., \u003cem\u003eTranscriptional downregulation of MHC class I and melanoma de- differentiation in resistance to PD-1 inhibition.\u003c/em\u003e Nature Communications, 2020. \u003cstrong\u003e11\u003c/strong\u003e(1): p. 1897.\u003c/li\u003e\n \u003cli\u003eSingla, R., et al., \u003cem\u003eComparison of RECIST and iRECIST criteria in patients with advanced lung cancer treated with nivolumab.\u003c/em\u003e Journal of Cancer Research and Therapeutics, 2023. \u003cstrong\u003e19\u003c/strong\u003e(5).\u003c/li\u003e\n \u003cli\u003eShilts, J., et al., \u003cem\u003eA physical wiring diagram for the human immune system.\u003c/em\u003e Nature, 2022. \u003cstrong\u003e608\u003c/strong\u003e(7922): p. 397-404.\u003c/li\u003e\n \u003cli\u003eHu, C., et al., \u003cem\u003eCellMarker 2.0: an updated database of manually curated cell markers in human/mouse and web tools based on scRNA-seq data.\u003c/em\u003e Nucleic Acids Research, 2023. \u003cstrong\u003e51\u003c/strong\u003e(D1): p. D870-D876.\u003c/li\u003e\n \u003cli\u003eTibshirani, R., T. Hastie, and J. Friedman, \u003cem\u003eRegularized Paths for Generalized Linear Models Via Coordinate Descent.\u003c/em\u003e Journal of Statistical Software, 2010. \u003cstrong\u003e33\u003c/strong\u003e.\u003c/li\u003e\n \u003cli\u003eLopez-Gonzalez, L., et al. \u003cem\u003eExploring Biomarkers in Breast Cancer: Hallmarks of Diagnosis, Treatment, and Follow-Up in Clinical Practice\u003c/em\u003e. Medicina, 2024. \u003cstrong\u003e60\u003c/strong\u003e, DOI: 10.3390/medicina60010168.\u003c/li\u003e\n \u003cli\u003eHe, X. and C. Xu, \u003cem\u003eImmune checkpoint signaling and cancer immunotherapy.\u003c/em\u003e Cell Research, 2020. \u003cstrong\u003e30\u003c/strong\u003e(8): p. 660-669.\u003c/li\u003e\n \u003cli\u003eMorad, G., et al., \u003cem\u003eHallmarks of response, resistance, and toxicity to immune checkpoint blockade.\u003c/em\u003e Cell, 2021. \u003cstrong\u003e184\u003c/strong\u003e(21): p. 5309-5337.\u003c/li\u003e\n \u003cli\u003eKyrysyuk, O. and K.W. Wucherpfennig, \u003cem\u003eDesigning Cancer Immunotherapies That Engage T Cells and NK Cells.\u003c/em\u003e Annual Review of Immunology, 2023. \u003cstrong\u003e41\u003c/strong\u003e(Volume 41, 2023): p. 17-38.\u003c/li\u003e\n \u003cli\u003eBarry, K.C., et al., \u003cem\u003eA natural killer-dendritic cell axis defines checkpoint therapy-responsive tumor microenvironments.\u003c/em\u003e Nature Medicine, 2018. \u003cstrong\u003e24\u003c/strong\u003e(8): p. 1178-1191.\u003c/li\u003e\n \u003cli\u003eDeng, K., et al., \u003cem\u003eStudy of LY9 as a potential biomarker for prognosis and prediction of immunotherapy efficacy in lung adenocarcinoma.\u003c/em\u003e Peerj, 2024. \u003cstrong\u003e12\u003c/strong\u003e: p. 1-24.\u003c/li\u003e\n \u003cli\u003eCuenca, M., et al., \u003cem\u003eLy9 (SLAMF3) receptor differentially regulates iNKT cell development and activation in mice.\u003c/em\u003e European Journal of Immunology, 2018. \u003cstrong\u003e48\u003c/strong\u003e(1): p. 99-105.\u003c/li\u003e\n \u003cli\u003eHe, M.Y., et al., \u003cem\u003eCD5 expression by dendritic cells directs T cell immunity and sustains immunotherapy responses.\u003c/em\u003e Science, 2023. \u003cstrong\u003e379\u003c/strong\u003e(6633): p. 661-+.\u003c/li\u003e\n \u003cli\u003eWu, Y., et al., \u003cem\u003eExpression of CD40 and growth-inhibitory activity of CD40 ligand in colon cancer ex vivo.\u003c/em\u003e Cellular Immunology, 2008. \u003cstrong\u003e253\u003c/strong\u003e(1): p. 102-109.\u003c/li\u003e\n \u003cli\u003eKim, S.-M., et al., \u003cem\u003eGlioblastoma-educated mesenchymal stem-like cells promote glioblastoma infiltration via extracellular matrix remodelling in the tumour microenvironment.\u003c/em\u003e Clinical and Translational Medicine, 2022. \u003cstrong\u003e12\u003c/strong\u003e(8): p. e997.\u003c/li\u003e\n \u003cli\u003eLi, W., et al., \u003cem\u003ePathomics models for CD40LG expression and prognosis prediction in glioblastoma.\u003c/em\u003e Scientific Reports, 2024. \u003cstrong\u003e14\u003c/strong\u003e(1): p. 24350.\u003c/li\u003e\n \u003cli\u003eBahn, Y.J., et al., \u003cem\u003eTGF-\u0026beta; antagonism synergizes with PPAR\u0026gamma; agonism to reduce fibrosis and enhance beige adipogenesis.\u003c/em\u003e Molecular Metabolism, 2024. \u003cstrong\u003e90\u003c/strong\u003e: p. 102054.\u003c/li\u003e\n \u003cli\u003eSutherland, T.E., D.P. Dyer, and J.E. Allen, \u003cem\u003eThe extracellular matrix and the immune system: A mutually dependent relationship.\u003c/em\u003e Science. \u003cstrong\u003e379\u003c/strong\u003e(6633): p. eabp8964.\u003c/li\u003e\n \u003cli\u003eLee, H.N., M.S. Jeong, and S.B. Jang \u003cem\u003eMolecular Characteristics of Amyloid Precursor Protein (APP) and Its Effects in Cancer\u003c/em\u003e. International Journal of Molecular Sciences, 2021. \u003cstrong\u003e22\u003c/strong\u003e, DOI: 10.3390/ijms22094999.\u003c/li\u003e\n \u003cli\u003eChen, G., et al., \u003cem\u003eSingle-cell transcriptomic analysis reveals that the APP\u0026ndash;CD74 axis promotes immunosuppression and progression of testicular tumors.\u003c/em\u003e The Journal of Pathology, 2024. \u003cstrong\u003e264\u003c/strong\u003e(3): p. 250-269.\u003c/li\u003e\n \u003cli\u003eKroeger, D.R., K. Milne, and B.H. Nelson, \u003cem\u003eTumor-Infiltrating Plasma Cells Are Associated with Tertiary Lymphoid Structures, Cytolytic T-Cell Responses, and Superior Prognosis in Ovarian Cancer.\u003c/em\u003e Clinical Cancer Research, 2016. \u003cstrong\u003e22\u003c/strong\u003e(12): p. 3005-3015.\u003c/li\u003e\n \u003cli\u003eMa, X., et al., \u003cem\u003eTNFSF13 insufficiency disrupts human colonic epithelial cell-mediated B cell differentiation.\u003c/em\u003e bioRxiv, 2024: p. 2024.09.23.614260.\u003c/li\u003e\n \u003cli\u003eDowns-Canner, S.M., et al., \u003cem\u003eB Cell Function in the Tumor Microenvironment.\u003c/em\u003e Annual Review of Immunology, 2022. \u003cstrong\u003e40\u003c/strong\u003e(Volume 40, 2022): p. 169-193.\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":true,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"discover-oncology","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"dion","sideBox":"Learn more about [Discover Oncology](https://www.springer.com/12672)","snPcode":"","submissionUrl":"","title":"Discover Oncology","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Discover Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"immune checkpoint blockade, metastatic melanoma, ligand-receptor pairs, biomarker","lastPublishedDoi":"10.21203/rs.3.rs-5854916/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-5854916/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eImmune checkpoint blockade (ICB) therapy has transformed the treatment landscape for metastatic melanoma, yet predicting therapeutic response remains a significant challenge. This study hypothesizes that coordinated ligand-receptor (LR) interactions within the tumor microenvironment (TME) critically influence ICB efficacy and proposes that a novel LR pair-based signature score (LRPS) derived from on-treatment samples can predict clinical outcomes. Using transcriptomic data from five independent cohorts, we identified seven LR pairs (FLT3-FLT3LG, LY9-LY9, CD5-CD5, CD40LG-ITGA2B/ITGB3, APP-CD74, TNFRSF17-TNFSF13, FCER2-ITGAV/ITGB3) significantly associated with treatment outcomes. LRPS demonstrated significant predictive power, achieving an area under the curve (AUC) exceeding 0.8 in four cohorts. Based on the LRPS signature, subjects were divided into high- and low-scores groups using the mean score. ICB response rates were higher in the high-scoring cohort subjects than the low-scoring subjects. Patient with high scores tended to have better survival outcomes than did those with low scores. In conclusion, we identified and verified an LRPS signature that provides a theoretical basis for applying such signatures derived from on-treatment tumor samples to predict therapeutic responses to ICB therapies.\u003c/p\u003e","manuscriptTitle":"Ligand-receptor pair-based signature score Derived from On-treatment Tumor Specimens Predicts Immune Checkpoint Blockade Response in Metastatic Melanoma","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-04-17 06:38:40","doi":"10.21203/rs.3.rs-5854916/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2025-08-11T09:08:32+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2025-05-08T10:48:24+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-04-21T16:02:38+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"196644665407230333392086931626615039383","date":"2025-04-17T15:50:28+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"115399666506016301880929792845722471359","date":"2025-04-15T14:19:03+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2025-04-15T06:44:32+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2025-04-15T05:00:08+00:00","index":"","fulltext":""},{"type":"submitted","content":"Discover Oncology","date":"2025-03-24T14:02:47+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"discover-oncology","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"dion","sideBox":"Learn more about [Discover Oncology](https://www.springer.com/12672)","snPcode":"","submissionUrl":"","title":"Discover Oncology","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Discover Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"8931a9b4-74c7-4159-9adb-c1f2ece54712","owner":[],"postedDate":"April 17th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[],"tags":[],"updatedAt":"2026-01-05T07:09:38+00:00","versionOfRecord":[],"versionCreatedAt":"2025-04-17 06:38:40","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-5854916","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-5854916","identity":"rs-5854916","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}