A plasma proteomic signature of cancer-related sarcopenia implicates the IGFBP axis in muscle dysfunction

A plasma proteomic signature of cancer-related sarcopenia implicates the IGFBP axis in muscle dysfunction | 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 A plasma proteomic signature of cancer-related sarcopenia implicates the IGFBP axis in muscle dysfunction Filippo Gustavo Dall'Olio, Wael Salem ZRAFI, Xinran SONG, Littisha LAWRANCE, and 20 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8885139/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted 9 You are reading this latest preprint version Abstract Background Cancer-related sarcopenia is associated with poor clinical outcomes but remains difficult to define and quantify in routine oncology practice. Current assessments rely on imaging and functional scales that are time-consuming and provide limited biological insight. We aimed to identify a plasma proteomic signature of cancer-related sarcopenia and to uncover circulating mediators involved in its pathophysiology. Methods Patients were included from two cohorts of the MATCH-R study (NCT02517892): a discovery cohort of advanced cancer patients treated with immunotherapy and an independent validation cohort of metastatic castration-resistant prostate cancer (mCRPC) patients treated with androgen-receptor pathway inhibitors. External validation was performed in the TRACERx cohort of non–small cell lung cancer. Skeletal muscle index at L3 was quantified using imaging, and ECOG performance status served as a functional proxy. Plasma proteomics was performed using the Olink Explore platform. An extreme gradient boosting (XGBoost) model was trained on a high-contrast subset using a neuromuscular-focused protein panel and validated across cohorts. Functional effects of candidate mediators were assessed in differentiating human myoblasts. Results The model generated a continuous sarcopenia probability (SP) score that correlated with muscle mass and functional status and consistently stratified overall survival across cohorts. A reduced four-protein model retained comparable performance, supporting translational applicability. Proteins associated with SP included IGFBP1, IGFBP2, and IL6. IGFBP1 and IGFBP2 impaired myoblast differentiation, while IL6 induced IGFBP1 expression in liver cells. Conclusions Plasma proteomics enables scalable and biologically informed assessment of cancer-related sarcopenia, identifies tumor–host mediators of muscle dysfunction, and supports objective patient stratification for therapeutic intervention. Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Background Sarcopenia was originally described as a progressive decline in skeletal muscle mass, strength, and function associated with aging, but it is now increasingly recognized as a condition that can co-exist across a range of clinical conditions 1 . Sarcopenia represents a major challenge in cancer care. In cancer patients, sarcopenia is associated with impaired quality of life, increased treatment toxicity, reduced tolerance to systemic therapies, and poorer survival, thereby amplifying the overall burden of cancer for both patients and caregivers 2 . Sarcopenia frequently coexists with cancer cachexia and other wasting syndromes, further complicating clinical management 3 . However, reported prevalence estimates range widely 4 , from 5% to nearly 90%, largely reflecting heterogeneity in assessment methods, cutoff definitions, and patient populations. This lack of consensus has limited the comparability and generalizability of existing studies, ultimately slowing progress in the field. Historically, sarcopenia has been primarily studied in geriatric medicine, where its definition and assessment have progressively evolved toward a multidimensional framework incorporating muscle mass, muscle strength, and physical performance 5 . In contrast, oncology research has largely relied on imaging-based quantification of muscle mass, typically derived from routine computed tomography scans. Although this approach has provided important prognostic insights, it captures only a fraction of the biological complexity underlying functional decline and fails to account for systemic and molecular drivers of muscle dysfunction. Cancer-derived sarcopenia is increasingly recognized as a multifactorial condition shaped by nutritional status, systemic inflammation, host metabolism, aging, sex, cancer biology, and anticancer treatments. Yet, the contribution of tumor-driven systemic signals and circulating mediators remains incompletely understood. Emerging evidence suggests that specific tumor biological features and plasma proteomic profiles are associated with the development of cancer derived cachexia, as demonstrated in both resected and advanced non-small cell lung cancer within the TRACERx (TRAcking Cancer Evolution through therapy (Rx)) study 6 . These observations support the concept that circulating factors can act as measurable intermediates linking tumor biology to host functional decline. However, a comprehensive and scalable framework to characterize cancer-derived sarcopenia at the systemic molecular level is still lacking. Skeletal muscle health and maintenance depend both on muscle intrinsic properties but also on coordinated regulation by metabolic, inflammatory, and neural inputs 7 , 8 . Skeletal muscle maintenance and regeneration requires effective myoblast differentiation and repair processes 9 , which are tightly coupled to endocrine signaling and neuromuscular integrity 10 , 11 . Disruption of these integrated systems, rather than isolated muscle atrophy alone, may therefore underlie the development of sarcopenia in cancer. Based on this rationale, we leveraged large scale plasma proteomics to define a molecular signature of physical function and muscle mass across multiple cancer types. We further aimed to identify circulating mediators linking tumor biology and systemic inflammation to impaired myogenesis, thereby providing mechanistic insight into cancer-derived sarcopenia and a foundation for biologically informed patient stratification. We aimed to generate a sarcopenia probability (SP) score that accurately classified sarcopenic status across independent validation cohorts. Methods Patients Patients were selected from two independent cohorts of the MATCH-R (Molecular Analysis for Therapy Choice - Resistance) study 12 (NCT02517892), a prospective, single-center clinical trial conducted at Gustave Roussy between 2015 and 2022. The study was designed to investigate tumor molecular evolution under anticancer therapies and to identify mechanisms of treatment resistance. Patients underwent tumor profiling at baseline and at disease progression, with the aim of characterizing resistance-associated molecular alterations and informing potential treatment strategies. Eligible patients were adults receiving or scheduled to receive systemic anticancer therapy, with at least one tumor lesion accessible for core needle biopsy, who were covered by a social security system and provided written informed consent. The discovery cohort included patients with advanced solid tumors, predominantly lung and bladder cancers, treated with immune checkpoint inhibitors (MATCHR-I). The validation cohort consisted of patients with metastatic castration resistant prostate cancer (mCRPC) treated with androgen-receptor pathway inhibitors 13 . The discovery cohort was used for model training and internal testing, whereas the mCRPC cohort served as an independent validation set (Supplementary Fig. 1). MATCH-R study was approved by the CPP (“Comité de protection des personnes”) and by the ANSM (“Agence nationale de sécurité du médicament et des produits de santé”) and adhered to the principles in the Guideline for Good Clinical Practice and the Declaration of Helsinki. Written informed consent was obtained from all patients included in this study. External validation was performed using the independent TRACERx cohort, whose data have been previously published. Clinical data, proteomic data, and outcome information were retrieved from the publicly available repository described in the original publication 6 . Muscle mass measurement and sarcopenia definition Skeletal muscle mass was quantified using the skeletal muscle index (SMI) derived from computed tomography (CT) or positron emission tomography (PET) scans performed within 60 days of plasma collection. Cross sectional muscle area was measured at the level of the third lumbar vertebra (L3) and normalized to height squared to derive skeletal muscle index (SMI). Low muscle mass was defined using established sex specific cutoffs (52.4 cm²/m² for males and 38.5 cm²/m² for females) 14 . Anthropometer3DNet was also used to extract body composition parameters from CT images. This software, which is available for research purposes on www.oncometer3d.com , can automatically measure several anthropometric parameters including muscle mass (MBM) on multislice CT scans in less than 5 through a deep learning-based segmentation of adipose tissue (visceral and subcutaneous) and muscle voxels 15 . Physical function was approximated using Eastern Cooperative Oncology Group performance status (ECOG PS). To improve phenotype labeling for model training, we defined a high-contrast subgroup. Patients with ECOG PS 0 and high SMI were classified as having a low probability of sarcopenia (LS), whereas patients with low SMI and ECOG PS ≥ 2 were classified as having a high probability of sarcopenia (HS). This subgroup was used for supervised model training, maximizing phenotypic contrast and reducing misclassification. Plasma proteome Plasma proteomic profiling was performed using the Olink Proximity Extension Assay (PEA) technology (Olink Proteomics AB, Uppsala, Sweden). Baseline plasma samples from the discovery cohort were analyzed using the Olink Explore 1536 platform, comprising 1,472 protein assays and 48 controls distributed across four panels (inflammation, oncology, cardiometabolic, and neurology). Sequencing was performed on a NovaSeq 6000 system using two S1 flow cells with 2 × 50 bp read lengths. The validation mCRPC cohort and an external validation cohort from TRACERx were analyzed using the Olink Explore 3072 platform. Raw sequencing counts were converted into normalized protein expression (NPX) values following Olink’s standard quality control and normalization pipeline, including internal extension and inter plate controls. NPX values are reported on a log2 scale, with higher values corresponding to higher relative protein abundance. Assay validation metrics are available from the manufacturer. Cell culture Immortalized human myoblasts (LHCN, AB1190, MB135) were cultured at 37°C and 5% CO₂ in growth medium consisting of DMEM supplemented with 10% fetal bovine serum, Medium 199, basic fibroblast growth factor (0.5 ng/mL), dexamethasone (0.2 µg/mL), and penicillin streptomycin (1%). Absence of mycoplasma contamination was confirmed using MycoAlert assays (Lonza). HepG2 cells were cultured under standard conditions and treated with recombinant human IL-6 (0–30 ng/mL) for 24 or 48 hours. Following treatment, cells were harvested for protein and RNA extraction. Intracellular IGFBP1 protein levels were analyzed by western blotting from whole cell lysates, while secreted IGFBP1 was measured by western blot analysis of culture supernatants collected at the indicated time points. Total RNA was extracted from HepG2 cells and IGFBP1 mRNA expression was quantified by RT-qPCR. Western blot Protein extraction and immunoblotting were performed using standard procedures. Primary antibodies included anti myosin heavy chain (MF20; R&D Systems) and anti β actin (Santa Cruz Biotechnology). Horseradish peroxidase conjugated secondary antibodies and chemiluminescent detection reagents (Millipore) were used for signal visualization. Membranes were imaged using an Amersham Imager 800 system (GE Healthcare). Band intensities were quantified using Fiji software and normalized to β-actin expression. Immunofluorescent staining Differentiated myotubes were washed with phosphate-buffered saline (PBS), fixed with 4% paraformaldehyde for 10 minutes, permeabilized with 0.25% Triton X-100, and blocked with 5% normal goat serum. Cells were incubated with anti myosin heavy chain antibody (R&D Systems; 1:150), followed by Alexa Fluor 488 conjugated secondary antibody (Thermo Fisher Scientific; 1:400). Nuclei were counterstained with Hoechst 33342. Images were acquired using a Cytation 1 Cell Imaging Reader (BioTek) at ×10 magnification. Fusion index was quantified using Fiji software by analyzing six randomly selected fields per condition. Data represent three independent experiments. Transcriptomic analysis RNA sequencing of myoblasts Myoblast experiments were conducted using four biological replicates per condition. Total RNA was extracted using the NucleoSpin RNA isolation kit (Macherey Nagel). RNA integrity was assessed using an Agilent Bioanalyzer 2100. Sequencing libraries were prepared using the NEBNext Ultra RNA Library Prep Kit (New England Biolabs) and sequenced on an Illumina HiSeq 2000 platform. Tumor RNA sequencing RNA from MATCHR-I and mCRPC tumor samples was extracted using the AllPrep DNA/RNA Mini Kit (Qiagen). Libraries were sequenced on Illumina NextSeq 500 or NovaSeq platforms as paired-end reads following standard quality control procedures. Gene expression quantification and enrichment analysis Raw sequencing reads were quality-controlled using Trim Galore. Transcript quantification was performed using Kallisto (v0.44.0) with the GENCODE v27 reference. Transcript-level estimates were aggregated to gene-level counts using TxImport. Genes with low expression (< 2 counts in 10% of samples) were excluded. Differential expression analysis was performed using DESeq2, applying a false discovery rate (FDR) < 0.05 and |log2 fold change| ≥1. Gene set enrichment analysis was conducted using clusterProfiler with Hallmark, Gene Ontology biological process, and KEGG gene sets from the Molecular Signatures Database (MSigDB). Single-sample gene set enrichment analysis To estimate the activity of muscle-related gene pathways at the sample level, we applied single-sample gene set enrichment analysis (ssGSEA). ssGSEA scores were computed from normalized gene expression counts using the GSVA R package (v2.4.1). Predefined gene sets related to skeletal muscle biology were used to derive enrichment scores for each sample, enabling quantitative comparison of pathway activity across samples. Single-cell RNA sequencing Single-nucleus RNA sequencing (snRNA-seq) was performed by Celsius Therapeutics. Nuclei were isolated from frozen tissue samples following standard protocols. Libraries were generated using the 10x Genomics Chromium Single Cell 3′ v3.1 chemistry and sequenced on an Illumina NovaSeq platform. Raw sequencing reads were processed using STARsolo for alignment, barcode processing, and gene counting. Downstream quality control, normalization, and integration were conducted using the Gustave Roussy single-cell analysis pipeline ( https://github.com/gustaveroussy/single-cell/wiki/ ). Quality control metrics were computed for each nucleus, including the number of detected genes, total unique molecular identifier (UMI) counts, and the proportion of reads mapping to mitochondrial genes. Nuclei were retained for downstream analysis if they expressed at least 200 genes, contained a minimum of 1,000 total UMIs, and had ≤ 20% mitochondrial gene expression. Nuclei not meeting these criteria were excluded. Data normalization was performed using SCTransform as implemented in Seurat (v4.0.4), regressing out total UMI counts per nucleus to mitigate technical variability. Principal component analysis (PCA) was applied to the normalized data, and batch effects were corrected using the Harmony algorithm (v0.1.0), with PCA embeddings used as input for integration. The dimensionality and clustering resolution were selected based on visual inspection of Uniform Manifold Approximation and Projection (UMAP) embeddings to identify biologically coherent structures. Cluster stability across parameter combinations was further assessed using the clustree R package (v0.4.3), ensuring robustness of the chosen clustering configuration. Quantitative real-time PCR Total RNA was extracted using the Quick-RNA kit (Zymo Research) and reverse transcribed using the MMLV RT kit (Eurogentec). Quantitative PCR was performed using SYBR Green chemistry on a LightCycler 96 system (Roche). Relative gene expression was calculated using the 2^−ΔΔCt method. All reactions were performed in triplicate. Machine learning and statistical analysis Differential plasma protein expression between patients with high versus low probability of sarcopenia was assessed using the Wilcoxon rank-sum test with false discovery rate correction. Proteins were further selected based on biological relevance and annotation from the Human Protein Atlas 16 , to enhance interpretability and reduce dimensionality in a limited sample size setting. A decision tree based extreme gradient boosting (XGBoost) classification model was developed using the tidymodels framework 17 . The model was trained on the high-contrast subgroup of lung cancer patients using fivefold cross-validation with 20 repetitions, stratified by sarcopenia probability. Hyperparameters were optimized based on classification accuracy. The best-performing model was applied to the full discovery cohort, generating a predicted probability of sarcopenia that was subsequently used in Kaplan Meier survival analyses and Cox proportional hazards models, treating sarcopenia probability as a continuous variable. Spearman correlation was used to assess the relationship between predicted sarcopenia probability and SMI. The model was independently validated in the mCRPC and in the TRACERx cohort. To derive a reduced plasma protein signature, a second XGBoost regression model was trained using predicted sarcopenia probabilities as the outcome on the MATCHR-I and mCRPC using the tidymodels framework and fivefold cross-validation with 20 repetitions Feature selection was performed iteratively based on variable importance, with root mean square error used for model selection. Performance of the extended and reduced signatures was compared in the external validation cohort. Results Patient cohorts and definition of high-contrast sarcopenia groups We included patients from the MATCH-R study at Gustave Roussy. The training cohort comprised 99 patients with advanced cancers receiving immunotherapy, predominantly lung and bladder cancer. Skeletal muscle mass was quantified using the Skeletal Muscle Index (SMI) measured at the L3 vertebral level on CT or PET scans performed within 42 days of plasma collection. Functional status was assessed using ECOG PS. Among patients with available imaging, 51 had low SMI and 34 high SMI. To establish a high-contrast subgroup for model development, we combined SMI with ECOG PS: patients with high SMI and ECOG 0 (functionally unimpaired) were classified as low probability of sarcopenia (LS), whereas patients with low SMI and ECOG ≥ 2 were classified as high probability of sarcopenia (HS). This yielded 36 patients (21 HS, 15 LS) who constituted the training set. Baseline characteristics of the overall and high-contrast groups are summarized in Table 1 .For validation, we used two independent cohorts. The first was an internal validation cohort from MATCH-R, comprising 55 patients with metastatic castration-resistant prostate cancer (mCRPC) treated with enzalutamide, with 88 plasma samples available at baseline and progression. The second was an external validation cohort derived from the TRACERx study, consisting of patients with resected and relapsed non-small cell lung cancer (NSCLC), with proteomic and imaging data available for 151 patients (141 baseline, 114 relapse). These validation sets allowed assessment of the model’s generalizability across different cancer types and disease contexts. Discovery of a proteomic signature of sarcopenia We performed plasma proteomics using the Olink Explore platform in the training cohort. Differential expression analysis between HS and LS patients in the high-contrast subgroup identified 353 proteins enriched in HS, predominantly associated with systemic inflammation, immune dysregulation, matrix remodeling, and metabolic stress, and 64 proteins enriched in LS, many of which were related to muscle or neuronal biology (Supplementary Fig. 1, Supplementary Table 1). To focus on proteins reflecting muscle function rather than tumor biology, we prioritized proteins enriched in LS and underrepresented in HS. A biology-driven selection further narrowed candidates to those implicated in skeletal muscle growth and regeneration (e.g., MSTN, WFIKKN1, CD34, PAMR1, DPP4), neuromuscular connectivity (e.g., CNTN family, adhesion/guidance molecules), extracellular matrix remodeling (e.g., integrins), and autophagy/lysosomal pathways (e.g., LAMP2, IDS). Proteins primarily involved in systemic metabolism, immunity, endocrine regulation, or oncogenic signaling were excluded. This process resulted in a 23-protein panel that served as the basis for the XGBoost classification model (model p23; Supplementary Table S2 ). The model was trained to predict sarcopenia probability (SP) as a continuous variable and classify patients as sarcopenic (SP > 50%) or non-sarcopenic (SP ≤ 50%) within the high-contrast training set. Model performance and association with SMI and ECOG The XGBoost model (p23) was first evaluated within the high-contrast training set. The model achieved an accuracy of 0.893 in classifying patients as sarcopenic or non-sarcopenic, demonstrating robust discrimination between HS and LS cases (Supplementary Fig. 2). When applied to the full training cohort, the predicted probability of sarcopenia (SP) correlated inversely with SMI in both male and female patients (Spearman ρ = -0.39, p = 0.0045 and ρ = -0.42, p = 0.016 respectively; Fig. 1 A), confirming that the model captured meaningful biological variation in muscle mass. SP also reflected functional status: increasing SP was associated with higher ECOG PS scores, indicating concordance between the proteomic signature and clinically assessed physical function (Fig. 1 B). In the internal validation cohort of 55 mCRPC patients, the model maintained a similar correlation with SMI (ρ = -0.41, p = 0.008) and ECOG PS, and in the subset of 18 high-contrast cases, classification accuracy reached 0.889, comparable to the training set (Fig. 3 A). Importantly, in the external TRACERx cohort, which included both baseline and relapse samples, SP also correlated with cross-sectional skeletal muscle area at L3 (for male: baseline ρ = -0.29, p 0.012; relapse ρ = -0.42, p = 0.0018; for female: baseline ρ = -0.24, p 0.073; relapse ρ = -0.35, p = 0.047), confirming the generalizability of the model across independent patient populations (1 C-D). Together, these findings demonstrate that the 23-protein signature provides a robust, scalable measure of sarcopenia probability, consistently reflecting both muscle mass and functional status across multiple cancer types and cohorts. Prognostic impact of the proteomic sarcopenia signature We next investigated the prognostic relevance of the proteomic-based probability of sarcopenia (SP) across cohorts. As expected, in the high-contrast training set, SP strongly stratified overall survival (OS), with sarcopenic patients (SP > 50%) exhibiting a markedly shorter median OS compared with non-sarcopenic patients (3.8 vs 34.8 months; log-rank p < 0.0001; Fig. 2 A). Importantly, this prognostic separation was preserved when extending the analysis to the remaining, lower-contrast patients in the training cohort, in whom sarcopenic patients also experienced significantly worse outcomes (median OS 7.8 vs 22.1 months; p = 0.012; Fig. 2 B). In the internal validation cohort of patients with metastatic castration-resistant prostate cancer (mCRPC), baseline SP similarly identified a subgroup with significantly inferior survival. Sarcopenic patients had a median OS of 8.9 months compared with 21.9 months in non-sarcopenic patients (p < 0.0001; Fig. 2 C), confirming the prognostic robustness of the signature in a cancer type with distinct biology and treatment context. Consistent results were observed in the external TRACERx cohort. Across resected and relapsed NSCLC samples, sarcopenic patients had significantly shorter OS compared with non-sarcopenic patients (median OS 25 vs 46 months p < 0.001 for baseline; median OS 30 vs 44 months, p 0.043 at relapse Fig. 2 D–E). Notably, in this cohort, the proteomic sarcopenia classification only partially overlapped with cachexia status, with most cachectic patients being sarcopenic but a substantial proportion of sarcopenic patients not fulfilling cachexia criteria, reinforcing the biological and clinical distinction between these two syndromes. Multivariable Cox proportional hazards models were used to assess the independent prognostic value of the proteomic sarcopenia probability across cohorts, adjusting for available clinical covariates. In the MATCHR immunotherapy cohort, sarcopenia probability remained independently associated with overall survival after adjustment for skeletal muscle index at L3, ECOG performance status, age, and sex (HR 4.13, 95% CI 1.38–12.3, p = 0.011), while SMI showed only a trend towards significance. In the mCRPC validation cohort, sarcopenia probability and ECOG performance status were both independently associated with survival, whereas age was not. SMI could not be included due to a high rate of missing imaging data (> 50%), reflecting routine clinical follow-up in mCRPC patients, which is primarily based on PSA monitoring rather than systematic radiological assessments.. In the external TRACERx cohort, sarcopenia probability was consistently associated with overall survival both at baseline and at relapse, independently of age, sex, and cross-sectional muscle area, which was not significantly associated with outcome in multivariable analyses. (supplementary table S4) Together, these results demonstrate that the plasma proteomic sarcopenia signature captures clinically meaningful systemic vulnerability that translates into a strong and consistent prognostic impact across tumor types, disease stages, and independent patient cohorts. Longitudinal changes in sarcopenia probability To determine whether the proteomic-based probability of sarcopenia (SP) captures dynamic changes in muscle health over time, we investigated its longitudinal behavior in patients with paired plasma samples collected at baseline and at disease progression or relapse. In the mCRPC validation cohort, paired proteomic samples were available for 33 patients treated with enzalutamide, of whom 20 had contemporaneous clinical documentation of ECOG performance status at both time points. Changes in SP closely mirrored clinical trajectories. Patients whose performance status improved between baseline and progression consistently showed a marked reduction in SP, with both individuals transitioning from sarcopenic to non-sarcopenic classification and exhibiting a mean SP reduction of 58%. In contrast, patients with stable ECOG PS showed minimal variation in SP, whereas those experiencing clinical deterioration displayed a corresponding increase in sarcopenia probability (Fig. 3 B). These findings support the ability of the proteomic signature to track clinically meaningful changes in functional status over time. We next extended this analysis to the external TRACERx cohort, where paired plasma proteomics and quantitative muscle mass assessments were available for 73 patients at baseline and relapse. In this independent dataset, longitudinal changes in SP were significantly correlated with proportional changes in skeletal muscle mass between the two time points (Spearman ρ = −0.32, p = 0.005; Fig. 3 C). Notably, the strength of this association appeared modulated by the disease-free interval separating baseline and relapse, suggesting that longer inter-sample intervals may allow for more pronounced and biologically detectable muscle remodeling. Collectively, these longitudinal analyses demonstrate that the proteomic sarcopenia signature is not merely a static classifier but reflects dynamic changes in muscle health across disease evolution, treatment exposure, and relapse, reinforcing its potential utility for real-time patient monitoring. Development of a reduced model As a final feature selection step, we sought to determine whether a more streamlined proteomic signature, consisting of a reduced number of proteins, could still reproduce the classification performance of our original model. We then trained an Xgboost regression model on MATCH-R cohorts and found that a model comprising a progressively reduced number of proteins maintained a good correlation with initial signature, with a model based on 4 proteins (model P4, based on CNTN3, CBLN4, MSTN, and ITGA11) resulting in an RMSE of 0.053 on the validation fold (supplementary Fig. 2 and supplementary table S5). Model p4 showed similar result in the high contrast groups, and AUC of 1, in the discovery cohort, and an accuracy of 0.83 and AUC of 1 on the prostate cohort. On the overall cohorts, both prediction from the 2 models showed a high correlation of 0.92, 0.98 and up to 0.99. Thus, the results from the classification model were closely approximated by the regression models with lesser plasma proteins Mediators discovery: To identify potential circulating mediators underlying the proteomic signature of sarcopenia, we treated the signature-derived probability of sarcopenia as a continuous variable and investigated its association with individual plasma proteins across cohorts. This strategy allowed us to move beyond classification performance and interrogate biological processes consistently linked to sarcopenia severity. Across the discovery cohort, several plasma proteins showed significant positive correlations with the probability of sarcopenia after false discovery rate correction. When extending this analysis to the internal validation cohort of patients with metastatic castration-resistant prostate cancer and to the external TRACERx cohort, IGFBP1, IGFBP2 and IL6 emerged as the most consistently associated proteins, showing significant positive correlations with sarcopenia probability in all evaluated datasets (Supplementary Table S3 and Fig. 3 D). Among the others, we observed several proteins with known or suspected roles in muscle pathophysiology. These include proteins implicated in muscle disease (e.g., SLC39A14 18 ), markers of muscle damage (CKB, SERPINA3 19 ). Additionally, ITIH3, previously associated with disease activity in myastenia gravis 20 and recently associated with muscle wasting 21 , was also identified. Transcriptomic correlates at tumor level We next investigated the transcriptomic correlates of our sarcopenia signature. Bulk RNA-seq analyses performed in our two internal cohorts revealed no meaningful correlation between IGFBP1/2 plasma levels and their corresponding mRNA expression in tumor tissue. At the pathway level, however, increasing sarcopenia probability was consistently associated with suppression of the Hallmark myogenesis program as well as multiple GO Biological Process pathways related to muscle development and function, supporting a muscle-specific biological signal captured by our signature (Fig. 4 ). To explore the potential cellular origin of putative mediators, we analysed single-cell RNA sequencing data generated from a subset of 16 patients from the MATCHR-I study (clinical characteristics reported in Supplementary Table S6). IGFBP1 expression was predominantly enriched in hepatocytes, together with ITIH3, thus validating in patients what was recently described in murine models 21 . In contrast, IGFBP2 expression was mainly restricted to a subset of tumor cells. IL6 did not map clearly to a specific cellular compartment, a finding that may reflect its multifocal origin, including potential production by skeletal muscle, a compartment not represented in tumor biopsies (Fig. 4 and supplementary Fig. 3). Functional validation of IGFBP-mediated impairment of myogenic differentiation An essential feature of skeletal muscle homeostasis is its capacity for self-repair. Muscle-resident myoblasts spontaneously differentiate to regenerate damaged fibers, a process that is impaired in sarcopenia 22 . We therefore investigated whether proteins identified through our proteomic screen directly interfere with myogenic differentiation in vitro. Immortalized human myoblasts were induced to differentiate and exposed to increasing concentrations of IGFBP1, IGFBP2, or IL-6, as detailed in the Methods. After four days of differentiation, myogenic progression was assessed by quantifying the fusion index (FI) and myosin heavy chain (MHC) expression using immunofluorescence microscopy and western blotting. Addition of IGFBP1 or IGFBP2 to the culture medium resulted in a marked, dose-dependent impairment of myogenic differentiation, with a significant reduction in fusion index (1.5-fold and 2-fold, respectively; Fig. 5 C–D) and decreased MHC protein expression (2-fold and 4-fold, respectively; Fig. 5 E–F). Given that IGFBPs can exert both IGF-1–dependent and IGF-1–independent biological effects²¹, we next evaluated whether increasing IGF-1 availability could rescue this phenotype. Indeed, supplementation with increasing concentrations of IGF-1 fully restored myoblast differentiation in the presence of both IGFBP1 and IGFBP2, supporting a predominantly IGF-1–dependent mechanism whereby IGFBPs sequester IGF-1 and suppress its pro-myogenic signaling (Fig. 5 F–J). Consistent with these phenotypic observations, transcriptomic profiling of differentiating myoblasts exposed to IGFBP1 or IGFBP2 revealed a robust suppression of gene programs related to muscle differentiation and function, including the Myogenesis hallmark and Gene Ontology biological processes associated with muscle contraction and muscle cell development. Importantly, these transcriptional alterations were largely attenuated by concomitant IGF-1 supplementation, further supporting a central role for impaired IGF signaling in mediating the observed differentiation defects. In contrast, exposure to IL-6 did not impair myoblast differentiation or MHC expression, indicating that IL-6 does not act as a direct inhibitor of myogenesis in this experimental context (Supplementary Fig. 4). Based on prior evidence indicating that IL-6 can induce hepatic IGFBP1 expression 23 , we next explored this axis in liver-derived cells. IL-6 exposure led to increased IGFBP1 expression in HepG2 cells, accompanied by elevated levels of secreted IGFBP1 in the culture supernatant, supporting an indirect mechanism whereby systemic inflammation may promote muscle dysfunction through liver-mediated modulation of the IGF axis rather than through direct effects on muscle cells (Supplementary Fig. 4). Discussion In this study, we demonstrate that plasma proteomics can capture a biologically meaningful representation of physical function and muscle mass in oncology patients, offering a potentially scalable and objective alternative to conventional clinical assessments. By focusing on proteins enriched in non-sarcopenic individuals, we identified a proteomic signature reflective of preserved systemic and neuromuscular integrity, rather than advanced cancer-related wasting. This approach enabled the delineation of patient subgroups with distinct muscle masss and functional phenotypes based on molecular features, providing biological resolution that is not achievable using performance status alone. Importantly, our approach estimates a continuous probability of sarcopenia-related muscle dysfunction, rather than assigning a binary diagnosis, reflecting the biological heterogeneity and absence of a single gold-standard definition in oncology. Despite major advances in precision oncology, patient stratification in clinical practice remains largely tumor-centric, with limited incorporation of the host systemic state and its biological determinants. Skeletal muscle mass is increasingly assessed using imaging-based approaches, yet these measures are inconsistently implemented, and provide limited insight into the underlying biology of muscle dysfunction. Functional status is routinely assessed using coarse clinical scales such as the Eastern Cooperative Oncology Group performance status (ECOG PS), which suffer from moderate interobserver agreement 24 , limited resolution, particularly within intermediate categories 25 , and also lack of biological interpretability. Consequently, the molecular processes underlying muscle loss, functional decline and sarcopenia in patients with cancer remain poorly characterized and difficult to quantify at scale. This limitation has important implications for interventional studies in cancer sarcopenia, including trials of exercise, nutritional, or pharmacological interventions, which critically rely on robust, objective, and scalable endpoints. The lack of biologically grounded surrogate markers of muscle function and repair has contributed to heterogeneous trial designs and inconclusive results across the field. A notable aspect of the identified signature is the enrichment of proteins linked to neuronal and neuromuscular biology. In particular, CBLN4 and CNTN3 emerged as major contributors to the model. Both proteins are involved in synaptic organization and plasticity, suggesting that preserved physical function in patients with cancer may depend not only on muscle-intrinsic properties but also on intact neuromuscular and neuronal signaling. This observation aligns with growing evidence that muscle aging and sarcopenia are influenced by alterations at the neuromuscular junction 26 and by central neural regulation 27 , 28 . Moreover, physical exercise, one of the most effective interventions to preserve muscle mass and function, is known to induce robust synaptic and neuronal plasticity 29 , further supporting a biological link between neuromuscular connectivity and functional status. This is also consistent with recent data linking muscle mass and magnetic resonance – assessed brain age 30 . Although our study was not designed to directly assess central nervous system (CNS) involvement, these findings raise the possibility that plasma proteomics may indirectly capture systemic correlates of neuro-muscular health. In this context, plasma proteomic signatures capturing neuromuscular and systemic integrity may represent attractive surrogate endpoints to monitor biological responses to exercise-based or multimodal interventions, particularly when imaging- or strength-based assessments are impractical or insufficiently sensitive. Beyond descriptive stratification, our analyses identified candidate mediators of cancer-related sarcopenia. Among the most consistently correlated proteins across cohorts were IGFBP1 and IGFBP2, both key regulators of the insulin-like growth factor (IGF) axis. While liver-derived IGFBP1 has recently been implicated in muscle wasting 21 , our data extend this paradigm by identifying tumor-derived IGFBP2 as an active contributor to muscle dysfunction. Functional experiments confirmed that IGFBP2 impairs myoblast differentiation, thereby limiting muscle repair and regeneration, an effect that could be reversed by increasing IGF1 availability. These findings support a model in which tumors actively promote sarcopenia through endocrine or paracrine modulation of IGF signaling, reinforcing the concept of a bidirectional tumor host interaction. The broader proteomic landscape associated with the signature further highlights the systemic nature of cancer-related sarcopenia. Several correlated proteins are involved in inflammatory signaling and acute-phase responses, including IL-6, MDK 31 , and C9, consistent with the established role of chronic inflammation in muscle wasting. Others have been linked to insulin resistance at the muscular level, a metabolic alteration increasingly recognized as a contributor to sarcopenia development. Additional proteins associated with extracellular matrix remodeling, tumor proliferation, and invasion likely reflect aggressive tumor biology, which may indirectly exacerbate functional decline through sustained systemic stress rather than direct effects on muscle tissue. Together, these findings suggest that sarcopenia in cancer emerges from the convergence of inflammatory, metabolic, and tumor-derived signals, resembling an accelerated state of inflammaging. Notably, the incomplete overlap between sarcopenia and cachexia observed in the external validation cohort reinforces the concept that they represent overlapping but distinct biological conditions. This distinction highlights the limitation of weight-based definitions and supports the need for muscle-specific, biology-driven stratification approaches. In clinical practice, such a blood-based signature could complement imaging and functional assessments by enabling scalable screening, longitudinal monitoring, and early identification of patients who may benefit from targeted interventions. Beyond its biological insights, our findings illustrate how molecular phenotyping may support the implementation of precision nutrition concepts in oncology. As recently proposed, precision nutrition extends beyond personalization based on static characteristics and relies on the integration of molecular profiling, advanced analytics, and clinically meaningful outcomes to guide targeted interventions 32 . By combining plasma proteomics with machine learning, our approach enables scalable, biologically informed stratification of muscle health and physical function, thereby bridging mechanistic insight with clinical applicability. This framework is consistent with emerging precision nutrition paradigms and may help identify patients most likely to benefit from tailored nutritional, exercise, or anti-inflammatory strategies, rather than relying on one-size-fits-all recommendations. This study has several strengths, including the use of a large, multi-cancer cohort and a broad plasma proteomic panel, supporting the generalizability of the identified biology across tumor types. These results indicate that cancer-related sarcopenia is driven by shared systemic mechanisms rather than tumor-specific processes. Limitations include the absence of direct measurements of muscle strength; however, it is noteworthy that most sarcopenia studies rely primarily on radiological assessments of muscle mass. We mitigated this limitation by incorporating performance status as a functional proxy and by training the model on patients with highly contrasted clinical and radiological phenotypes. Future studies integrating matched tumor, muscle, and plasma samples will be essential to further dissect the causal pathways linking tumor biology to systemic functional decline. In particular, coupling peripheral proteomics with direct assessments of muscle and CNS biology may provide deeper insight into the neuro-muscular and inflammatory circuits that underpin cancer-related sarcopenia. Ultimately, this integrative framework could enable biologically informed patient stratification and support the development of targeted interventions aimed at preserving physical function in oncology patients. More broadly, the identification of plasma-based, mechanism-linked signatures of sarcopenia may help establish objective and scalable surrogate endpoints for future clinical trials in cancer cachexia and sarcopenia, including exercise-based interventions, thereby accelerating therapeutic development in a field long hindered by the lack of robust biomarkers. Declarations Ethics approval and consent to participate The MATCH-R study was approved by the Comité de Protection des Personnes (CPP) and by the Agence nationale de sécurité du médicament et des produits de santé (ANSM), and was conducted in accordance with the principles of the Declaration of Helsinki and the Guideline for Good Clinical Practice. Written informed consent was obtained from all patients prior to inclusion in the study. EudraCT number: 2014-A01147-40; CPP dossier reference: Am7501-4-3183. Consent for publication Non applicable as no individual data are included in the publication. Availability of data and materials The clinical and proteomic data generated in this study contain potentially identifiable personal information and are therefore subject to data protection regulations, including the General Data Protection Regulation (GDPR). For this reason, the datasets are not publicly available. Access to anonymized data may be considered upon reasonable request to the corresponding authors, subject to institutional approval and applicable regulatory constraints. The code used for model development and analysis is available from the corresponding authors upon reasonable request, subject to institutional and data protection regulations. Competing interests Fabrice Barlesi reports institutional relationships (no personal financial interests) with AbbVie, ACEA, Amgen, AstraZeneca, Bayer, Bristol Myers Squibb, Boehringer Ingelheim, Eisai, Eli Lilly Oncology, F. Hoffmann–La Roche Ltd, Genentech, Ipsen, Ignyta, Innate Pharma, Loxo, Novartis, MedImmune, Merck, MSD, Pierre Fabre, Pfizer, Sanofi-Aventis, Summit Therapeutics and Takeda. Caroline Even reports consulting or advisory roles with Innate Pharma, Bristol Myers Squibb, MSD Oncology, Merck Serono, Novartis, F-star Therapeutics, Merus and GlaxoSmithKline (institutional), and travel, accommodation or expenses from MSD Oncology and Merck Serono. Nathalie Lassau reports participation on an advisory board for Jazz Pharmaceuticals. Yohann Loriot reports honoraria from Janssen, Bristol Myers Squibb, Roche, Gilead, MSD and Pfizer; institutional research funding from Amgen, Janssen Oncology, MSD Oncology, Lilly, AstraZeneca, Orion, Exelixis, Incyte, Pfizer, Sanofi, Astellas Pharma, Gilead Sciences, Merck KGaA, Taiho Pharmaceutical, Bristol Myers Squibb, Roche and Tyra Biosciences; and travel, accommodation or expenses from Astellas Pharma, Pfizer, MSD Oncology and AstraZeneca. Antoine Italiano reports research grants from AstraZeneca, Bayer, Bristol Myers Squibb, Merck, MSD and Pharmamar. Carla M. Prado has previously received honoraria and/or paid consultancy from Abbott Nutrition, Nutricia, Nestlé Health Science and Novo Nordisk. Mariam Jamal-Hanjani reports consulting for Astex Pharmaceuticals, Pfizer and Achilles Therapeutics; membership on the Scientific Advisory Board and Steering Committee of Achilles Therapeutics; and speaker honoraria from Pfizer, Astex Pharmaceuticals, Oslo Cancer Cluster, Bristol Myers Squibb and Genentech. Benjamin Besse reports institutional honoraria and speaker’s bureau participation for AbbVie, AstraZeneca, Chugai Pharmaceutical, Daiichi Sankyo, Hedera Dx, Janssen, Merck Sharp & Dohme, Roche, Sanofi Aventis and Springer Healthcare Ltd; consulting or advisory roles (institutional) for AbbVie, BioNTech SE, Bristol Myers Squibb, Chugai Pharmaceutical, CureVac AG, Daiichi Sankyo, F. Hoffmann–La Roche Ltd, Pharmamar, Regeneron, Sanofi Aventis and Turning Point Therapeutics; and institutional research funding from AstraZeneca, BeiGene, Genmab A/S, GlaxoSmithKline, Janssen, Merck Sharp & Dohme, Ose Immunotherapeutics, Pharmamar, Roche-Genentech, Sanofi and Takeda. All other authors declare that they have no competing interests. Funding This work was partially supported by Canceropole Île-de-France (grant number 2024-1-EMERG-06) and by the Fondation Gustave Roussy. Authors’ contributions FGDO, YV, and BB conceived and designed the study. WSZ and MA performed the statistical analyses. XS conducted the in vitro experiments and functional assays. LL, DB, and NL contributed to imaging analyses and radiological data interpretation., CB, YL and KB contributed to proteomic data analysis and biological interpretation. FC, PB, RI, MG, DC, CN, MNC contributed to data collection, clinical annotation, and sample processing. CP and MJH provided critical expertise and contributed to data interpretation. FB, CE, AI, YV, and BB supervised the study. FGDO drafted the manuscript. All authors critically revised the manuscript and approved the final version. Acknowledgements The authors thank Dr. V. Mouly (Institute of Myology, Paris, France) for providing the human immortalized myoblast cell lines LHCN and AB1190, and Dr. S. Tapscott (Fred Hutchinson Cancer Center, Seattle, USA) for the MB135 myoblast cell line. References Kirk B, Cawthon PM, Arai H, et al. The Conceptual Definition of Sarcopenia: Delphi Consensus from the Global Leadership Initiative in Sarcopenia (GLIS). Age Ageing . 2024;53(3):afae052. doi:10.1093/ageing/afae052 Couderc AL, Liuu E, Boudou-Rouquette P, et al. Pre-Therapeutic Sarcopenia among Cancer Patients: An Up-to-Date Meta-Analysis of Prevalence and Predictive Value during Cancer Treatment. Nutrients . 2023;15(5):1193. doi:10.3390/nu15051193 Kiss N, Prado CM, Daly RM, et al. Low muscle mass, malnutrition, sarcopenia, and associations with survival in adults with cancer in the UK Biobank cohort. J Cachexia Sarcopenia Muscle . 2023;14(4):1775-1788. doi:10.1002/jcsm.13256 Williams GR, Dunne RF, Giri S, Shachar SS, Caan BJ. Sarcopenia in the Older Adult With Cancer. J Clin Oncol . 2021;39(19):2068-2078. doi:10.1200/JCO.21.00102 Cruz-Jentoft Aj, Bahat G, Bauer J, et al. Sarcopenia: revised European consensus on on definition and diagnosis. Age and ageing . 2019;48(1). doi:10.1093/ageing/afy169 Al-Sawaf O, Weiss J, Skrzypski M, et al. Body composition and lung cancer-associated cachexia in TRACERx. Nature medicine . 2023;29(4). doi:10.1038/s41591-023-02232-8 Chargé SBP, Rudnicki MA. Cellular and molecular regulation of muscle regeneration. Physiol Rev . 2004;84(1):209-238. doi:10.1152/physrev.00019.2003 Peake JM, Neubauer O, Della Gatta PA, Nosaka K. Muscle damage and inflammation during recovery from exercise. J Appl Physiol (1985) . 2017;122(3):559-570. doi:10.1152/japplphysiol.00971.2016 Mukund K, Subramaniam S. Skeletal muscle: A review of molecular structure and function, in health and disease. Wiley Interdiscip Rev Syst Biol Med . 2020;12(1):e1462. doi:10.1002/wsbm.1462 Zullo A, Mancini FP, Schleip R, Wearing S, Yahia L, Klingler W. The interplay between fascia, skeletal muscle, nerves, adipose tissue, inflammation and mechanical stress in musculo-fascial regeneration. J Gerontology Geriatrics . 2017;65(4):271-283. Lafuste P, Relaix F. Endocrine control of skeletal muscle regeneration and clinical applications. Nat Rev Endocrinol . 2026;22(1):5-6. doi:10.1038/s41574-025-01205-w Vasseur D, Bigot L, Beshiri K, et al. Deciphering resistance mechanisms in cancer: final report of MATCH-R study with a focus on molecular drivers and PDX development. Mol Cancer . 2024;23(1):221. doi:10.1186/s12943-024-02134-4 Menssouri N, Poiraudeau L, Helissey C, et al. Genomic Profiling of Metastatic Castration-Resistant Prostate Cancer Samples Resistant to Androgen Receptor Pathway Inhibitors. Clin Cancer Res . 2023;29(21):4504-4517. doi:10.1158/1078-0432.CCR-22-3736 Prado CMM, Lieffers JR, McCargar LJ, et al. Prevalence and clinical implications of sarcopenic obesity in patients with solid tumours of the respiratory and gastrointestinal tracts: a population-based study. Lancet Oncol . 2008;9(7):629-635. doi:10.1016/S1470-2045(08)70153-0 Decazes P, Ammari S, Belkouchi Y, et al. Synergic prognostic value of 3D CT scan subcutaneous fat and muscle masses for immunotherapy-treated cancer. J Immunother Cancer . 2023;11(9):e007315. doi:10.1136/jitc-2023-007315 Proteomics. Tissue-based map of the human proteome - PubMed. Accessed January 3, 2025. https://pubmed.ncbi.nlm.nih.gov/25613900/ Chen T, Guestrin C. XGBoost: A Scalable Tree Boosting System. Preprint posted online June 10, 2016. doi:10.1145/2939672.2939785 Wang G, Biswas AK, Ma W, et al. Metastatic cancers promote cachexia through ZIP14 upregulation in skeletal muscle. Nat Med . 2018;24(6):770-781. doi:10.1038/s41591-018-0054-2 Gueugneau M, d’Hose D, Barbé C, et al. Increased Serpina3n release into circulation during glucocorticoid-mediated muscle atrophy. J Cachexia Sarcopenia Muscle . 2018;9(5):929-946. doi:10.1002/jcsm.12315 Schroeter CB, Nelke C, Stascheit F, et al. Inter-alpha-trypsin inhibitor heavy chain H3 is a potential biomarker for disease activity in myasthenia gravis. Acta Neuropathol . 2024;147(1):102. doi:10.1007/s00401-024-02754-6 Kaltenecker D, Fisker Schmidt S, Weber P, et al. Functional liver genomics identifies hepatokines promoting wasting in cancer cachexia. Cell . 2025;188(17):4549-4566.e22. doi:10.1016/j.cell.2025.06.039 Blau HM, Cosgrove BD, Ho ATV. The central role of muscle stem cells in regenerative failure with aging. Nat Med . 2015;21(8):854-862. doi:10.1038/nm.3918 Lukowski SW, Fish RJ, Martin-Levilain J, et al. Integrated analysis of mRNA and miRNA expression in response to interleukin-6 in hepatocytes. Genomics . 2015;106(2):107-115. doi:10.1016/j.ygeno.2015.05.001 Blagden SP, Charman SC, Sharples LD, Magee LRA, Gilligan D. Performance status score: do patients and their oncologists agree? Br J Cancer . 2003;89(6):1022-1027. doi:10.1038/sj.bjc.6601231 Sørensen JB, Klee M, Palshof T, Hansen HH. Performance status assessment in cancer patients. An inter-observer variability study. Br J Cancer . 1993;67(4):773-775. doi:10.1038/bjc.1993.140 Kedlian VR, Wang Y, Liu T, et al. Human skeletal muscle aging atlas. Nat Aging . 2024;4(5):727-744. doi:10.1038/s43587-024-00613-3 Zhu XA, Starosta S, Ferrer M, et al. A neuroimmune circuit mediates cancer cachexia-associated apathy. Science . 2025;388(6743):eadm8857. doi:10.1126/science.adm8857 Simoes E, Uchida R, Nucci MP, et al. Cachexia Alters Central Nervous System Morphology and Functionality in Cancer Patients. J Cachexia Sarcopenia Muscle . 2025;16(1):e13742. doi:10.1002/jcsm.13742 de Sousa Fernandes MS, Ordônio TF, Santos GCJ, et al. Effects of Physical Exercise on Neuroplasticity and Brain Function: A Systematic Review in Human and Animal Studies. Neural Plast . 2020;2020:8856621. doi:10.1155/2020/8856621 Raji CA, Meysami S, Lee S, et al. Higher Muscle Volume is Inversely Related to Chronological and Brain Age While Increased Visceral to Muscle Fat Ratio is Positively Related to Chronological and Brain Age. Alzheimer’s & Dementia . 2025;21(S8):e110051. doi:10.1002/alz70862_110051 Ikutomo M, Sakakima H, Matsuda F, Yoshida Y. Midkine-deficient mice delayed degeneration and regeneration after skeletal muscle injury. Acta Histochem . 2014;116(2):319-326. doi:10.1016/j.acthis.2013.08.009 da Silva BR, Brennan L, Horst MA, Wishart DS, Prado CM. Advancing precision nutrition: bridging mechanistic insight and clinical implementation. Nat Rev Endocrinol . 2025;21(9):515-517. doi:10.1038/s41574-025-01141-9 Table Table 1. Clinical characteristics of the MATCH-R cohorts used for model development and validation. Baseline demographic, clinical, and disease characteristics of patients included in the training cohort (MATCH-R immunotherapy) and in the validation cohort (MATCH-R metastatic castration-resistant prostate cancer). Continuous variables are reported as median (interquartile range), and categorical variables as number (percentage). Characteristic Training mCRPC validation Samples = 99 1 Samples = 88 1 Age 64 (56, 71) 69 (65, 77) Sex Female 36 (36%) 0 (0%) Male 63 (64%) 88 (100%) Site Bladder cancer 12 (12%) 0 (0%) Colon cancer MSI 3 (3.0%) 0 (0%) NSCLC adenocarcinoma 66 (67%) 0 (0%) NSCLC undifferentiated/ NOS 4 (4.0%) 0 (0%) NSCLC squamous 11 (11%) 0 (0%) mCRPC 1 (1.0%) 88 (100%) Melanome 1 (1.0%) 0 (0%) Thyroid 1 (1.0%) 0 (0%) ECOG Performance status 0 30 (30%) 26 (46%) 1 42 (42%) 19 (33%) ≥ 2 27 (27%) 12 (21%) Unknown 0 31 Sarcopenia score Probability ≥ 50% 61 (62%) 13 (15%) Probability < 50% 38 (38%) 75 (85%) 1 Median (Q1, Q3); n (%) Additional Declarations Competing interest reported. Fabrice Barlesi reports institutional relationships (no personal financial interests) with AbbVie, ACEA, Amgen, AstraZeneca, Bayer, Bristol Myers Squibb, Boehringer Ingelheim, Eisai, Eli Lilly Oncology, F. Hoffmann–La Roche Ltd, Genentech, Ipsen, Ignyta, Innate Pharma, Loxo, Novartis, MedImmune, Merck, MSD, Pierre Fabre, Pfizer, Sanofi-Aventis, Summit Therapeutics and Takeda. Caroline Even reports consulting or advisory roles with Innate Pharma, Bristol Myers Squibb, MSD Oncology, Merck Serono, Novartis, F-star Therapeutics, Merus and GlaxoSmithKline (institutional), and travel, accommodation or expenses from MSD Oncology and Merck Serono. Nathalie Lassau reports participation on an advisory board for Jazz Pharmaceuticals. Yohann Loriot reports honoraria from Janssen, Bristol Myers Squibb, Roche, Gilead, MSD and Pfizer; institutional research funding from Amgen, Janssen Oncology, MSD Oncology, Lilly, AstraZeneca, Orion, Exelixis, Incyte, Pfizer, Sanofi, Astellas Pharma, Gilead Sciences, Merck KGaA, Taiho Pharmaceutical, Bristol Myers Squibb, Roche and Tyra Biosciences; and travel, accommodation or expenses from Astellas Pharma, Pfizer, MSD Oncology and AstraZeneca. Antoine Italiano reports research grants from AstraZeneca, Bayer, Bristol Myers Squibb, Merck, MSD and Pharmamar. Carla M. Prado has previously received honoraria and/or paid consultancy from Abbott Nutrition, Nutricia, Nestlé Health Science and Novo Nordisk. Mariam Jamal-Hanjani reports consulting for Astex Pharmaceuticals, Pfizer and Achilles Therapeutics; membership on the Scientific Advisory Board and Steering Committee of Achilles Therapeutics; and speaker honoraria from Pfizer, Astex Pharmaceuticals, Oslo Cancer Cluster, Bristol Myers Squibb and Genentech. Benjamin Besse reports institutional honoraria and speaker’s bureau participation for AbbVie, AstraZeneca, Chugai Pharmaceutical, Daiichi Sankyo, Hedera Dx, Janssen, Merck Sharp & Dohme, Roche, Sanofi Aventis and Springer Healthcare Ltd; consulting or advisory roles (institutional) for AbbVie, BioNTech SE, Bristol Myers Squibb, Chugai Pharmaceutical, CureVac AG, Daiichi Sankyo, F. Hoffmann–La Roche Ltd, Pharmamar, Regeneron, Sanofi Aventis and Turning Point Therapeutics; and institutional research funding from AstraZeneca, BeiGene, Genmab A/S, GlaxoSmithKline, Janssen, Merck Sharp & Dohme, Ose Immunotherapeutics, Pharmamar, Roche-Genentech, Sanofi and Takeda. All other authors declare that they have no competing interests. Supplementary Files supplementaryfigures.pdf supplementarytables.xlsx Primaryimages.pdf Graphicalabstract.docx Cite Share Download PDF Status: Under Review Version 1 posted Editorial decision: Revision requested 24 Mar, 2026 Reviews received at journal 24 Mar, 2026 Reviews received at journal 07 Mar, 2026 Reviewers agreed at journal 27 Feb, 2026 Reviewers agreed at journal 23 Feb, 2026 Reviewers invited by journal 21 Feb, 2026 Editor assigned by journal 20 Feb, 2026 Submission checks completed at journal 20 Feb, 2026 First submitted to journal 15 Feb, 2026 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-8885139","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":600155001,"identity":"68b94355-0738-478f-9eaf-de5e4566778c","order_by":0,"name":"Filippo Gustavo Dall'Olio","email":"data:image/png;base64,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","orcid":"","institution":"Institut Gustave Roussy","correspondingAuthor":true,"prefix":"","firstName":"Filippo","middleName":"Gustavo","lastName":"Dall'Olio","suffix":""},{"id":600155002,"identity":"fc94e975-106f-4412-bacc-27706e82ec87","order_by":1,"name":"Wael Salem ZRAFI","email":"","orcid":"","institution":"Institut Gustave Roussy","correspondingAuthor":false,"prefix":"","firstName":"Wael","middleName":"Salem","lastName":"ZRAFI","suffix":""},{"id":600155003,"identity":"d47e5d46-663f-4bd5-bb2a-8292d6acebe1","order_by":2,"name":"Xinran SONG","email":"","orcid":"","institution":"Institut Gustave Roussy","correspondingAuthor":false,"prefix":"","firstName":"Xinran","middleName":"","lastName":"SONG","suffix":""},{"id":600155004,"identity":"fcf364d6-2005-4699-9ee0-b0a44ef36483","order_by":3,"name":"Littisha LAWRANCE","email":"","orcid":"","institution":"Institut Gustave Roussy","correspondingAuthor":false,"prefix":"","firstName":"Littisha","middleName":"","lastName":"LAWRANCE","suffix":""},{"id":600155005,"identity":"3e4623ea-a514-4975-be93-8d1eea87bfd3","order_by":4,"name":"Fei CHEN","email":"","orcid":"","institution":"Institut Gustave Roussy","correspondingAuthor":false,"prefix":"","firstName":"Fei","middleName":"","lastName":"CHEN","suffix":""},{"id":600155006,"identity":"36ab7355-0862-4042-b765-23dde4967cf7","order_by":5,"name":"Rebecca IBRAHIM","email":"","orcid":"","institution":"Institut Gustave Roussy","correspondingAuthor":false,"prefix":"","firstName":"Rebecca","middleName":"","lastName":"IBRAHIM","suffix":""},{"id":600155008,"identity":"7449c42a-0870-4dd5-8a04-ab3691ef8090","order_by":6,"name":"Marie GUINHUT","email":"","orcid":"","institution":"Institut Gustave Roussy","correspondingAuthor":false,"prefix":"","firstName":"Marie","middleName":"","lastName":"GUINHUT","suffix":""},{"id":600155009,"identity":"aa85abab-cf71-46dd-b8ca-2dcf1543ae71","order_by":7,"name":"Pierre BUSSON","email":"","orcid":"","institution":"Institut Gustave Roussy","correspondingAuthor":false,"prefix":"","firstName":"Pierre","middleName":"","lastName":"BUSSON","suffix":""},{"id":600155010,"identity":"4651a402-d7de-4fa6-9531-d1ee2a0c9264","order_by":8,"name":"Catherine BRENNER","email":"","orcid":"","institution":"Institut Gustave Roussy","correspondingAuthor":false,"prefix":"","firstName":"Catherine","middleName":"","lastName":"BRENNER","suffix":""},{"id":600155011,"identity":"29ec7449-51e9-4538-8cc7-04fd3e438627","order_by":9,"name":"Karim BENIHOUD","email":"","orcid":"","institution":"Institut Gustave Roussy","correspondingAuthor":false,"prefix":"","firstName":"Karim","middleName":"","lastName":"BENIHOUD","suffix":""},{"id":600155012,"identity":"0ca8f7d8-8e21-4192-a192-99c377b8a76b","order_by":10,"name":"Fabrice BARLESI","email":"","orcid":"","institution":"Institut Gustave Roussy","correspondingAuthor":false,"prefix":"","firstName":"Fabrice","middleName":"","lastName":"BARLESI","suffix":""},{"id":600155013,"identity":"38f8dd5a-7478-4a0b-8085-9a2fa9b83e09","order_by":11,"name":"Caroline EVEN","email":"","orcid":"","institution":"Institut Gustave Roussy","correspondingAuthor":false,"prefix":"","firstName":"Caroline","middleName":"","lastName":"EVEN","suffix":""},{"id":600155014,"identity":"40c72429-374b-475a-aad4-da099e43b8e9","order_by":12,"name":"Dimitria Brempou","email":"","orcid":"","institution":"University College London","correspondingAuthor":false,"prefix":"","firstName":"Dimitria","middleName":"","lastName":"Brempou","suffix":""},{"id":600155015,"identity":"a4b09f46-88bc-455a-abde-c554f50360ca","order_by":13,"name":"Nathalie LASSAU","email":"","orcid":"","institution":"Institut Gustave Roussy","correspondingAuthor":false,"prefix":"","firstName":"Nathalie","middleName":"","lastName":"LASSAU","suffix":""},{"id":600155016,"identity":"d3d275ba-286d-497a-b6d3-9f3e7b1f2ef3","order_by":14,"name":"Claudio NICOTRA","email":"","orcid":"","institution":"Institut Gustave Roussy","correspondingAuthor":false,"prefix":"","firstName":"Claudio","middleName":"","lastName":"NICOTRA","suffix":""},{"id":600155017,"identity":"b4ad91e4-cf27-4a82-a0a6-d2de0d171679","order_by":15,"name":"Maud NGOCAMUS","email":"","orcid":"","institution":"Institut Gustave Roussy","correspondingAuthor":false,"prefix":"","firstName":"Maud","middleName":"","lastName":"NGOCAMUS","suffix":""},{"id":600155019,"identity":"88390597-a6c8-4697-b450-ff147d21fa07","order_by":16,"name":"Marine AGLAVE","email":"","orcid":"","institution":"Institut Gustave Roussy","correspondingAuthor":false,"prefix":"","firstName":"Marine","middleName":"","lastName":"AGLAVE","suffix":""},{"id":600155020,"identity":"4d48de49-35be-4f6c-abcf-d8346221f549","order_by":17,"name":"Yohann LORIOT","email":"","orcid":"","institution":"Institut Gustave Roussy","correspondingAuthor":false,"prefix":"","firstName":"Yohann","middleName":"","lastName":"LORIOT","suffix":""},{"id":600155021,"identity":"54e59e09-8172-4cf4-8d1a-f1ca42e381f3","order_by":18,"name":"Diana CARDENAS","email":"","orcid":"","institution":"Institut Gustave Roussy","correspondingAuthor":false,"prefix":"","firstName":"Diana","middleName":"","lastName":"CARDENAS","suffix":""},{"id":600155023,"identity":"a7e55ddd-dde3-4716-9e42-fb7e6fadde2b","order_by":19,"name":"Mariam Jamal-Hanjani","email":"","orcid":"","institution":"University College London","correspondingAuthor":false,"prefix":"","firstName":"Mariam","middleName":"","lastName":"Jamal-Hanjani","suffix":""},{"id":600155024,"identity":"60c3c744-d1bf-4822-85ae-f69e9d405a8b","order_by":20,"name":"Carla M Prado","email":"","orcid":"","institution":"University of Alberta","correspondingAuthor":false,"prefix":"","firstName":"Carla","middleName":"M","lastName":"Prado","suffix":""},{"id":600155025,"identity":"50ee3279-a56b-49e5-8398-7faa250b67d7","order_by":21,"name":"Antoine ITALIANO","email":"","orcid":"","institution":"Institut Gustave Roussy","correspondingAuthor":false,"prefix":"","firstName":"Antoine","middleName":"","lastName":"ITALIANO","suffix":""},{"id":600155026,"identity":"ae2d97db-bcef-4cf6-858b-a54d54a3b534","order_by":22,"name":"Yegor Vassetzky","email":"","orcid":"","institution":"Institut Gustave Roussy","correspondingAuthor":false,"prefix":"","firstName":"Yegor","middleName":"","lastName":"Vassetzky","suffix":""},{"id":600155027,"identity":"e6e97844-8e82-4fa4-9f32-eb7a0d2de3f4","order_by":23,"name":"Benjamin BESSE","email":"","orcid":"","institution":"Institut Gustave Roussy","correspondingAuthor":false,"prefix":"","firstName":"Benjamin","middleName":"","lastName":"BESSE","suffix":""}],"badges":[],"createdAt":"2026-02-15 09:53:26","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-8885139/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-8885139/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":104018608,"identity":"8af30046-bb89-4b9a-be31-e53d75dab6b2","added_by":"auto","created_at":"2026-03-05 17:47:24","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":151340,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eAssociation between proteomic-based probability of sarcopenia, muscle mass, and clinical performance\u003cbr\u003e\n(a–d)\u003c/strong\u003e Correlation between the proteomic-based probability of sarcopenia and skeletal muscle index in the training cohort (\u003cstrong\u003ea\u003c/strong\u003e), the mCRPC validation cohort (\u003cstrong\u003eb\u003c/strong\u003e), and the TRACERx cohort at baseline (\u003cstrong\u003ec\u003c/strong\u003e) and at relapse (\u003cstrong\u003ed\u003c/strong\u003e).\u003cbr\u003e\n \u003cstrong\u003e(e,f)\u003c/strong\u003e Box plots showing the distribution of the proteomic-based probability of sarcopenia across ECOG performance status categories in the training cohort (\u003cstrong\u003ee\u003c/strong\u003e) and in the mCRPC validation cohort (\u003cstrong\u003ef\u003c/strong\u003e).\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-8885139/v1/e1943d7f2f4c966aca5a0996.png"},{"id":104018615,"identity":"ba117584-2e36-46b6-8140-01c186edb0e0","added_by":"auto","created_at":"2026-03-05 17:47:24","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":100610,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eAssociation between proteomic-based probability of sarcopenia and overall survival\u003cbr\u003e\n(a,b)\u003c/strong\u003e Overall survival according to the proteomic-based probability of sarcopenia in the high-contrast training cohort (\u003cstrong\u003ea\u003c/strong\u003e) and in the remaining patients from the training cohort with lower contrast (\u003cstrong\u003eb\u003c/strong\u003e).\u003cbr\u003e\n \u003cstrong\u003e(c)\u003c/strong\u003e Overall survival stratified by proteomic-based probability of sarcopenia in the mCRPC validation cohort at baseline.\u003cbr\u003e\n \u003cstrong\u003e(d,e)\u003c/strong\u003e Overall survival according to the proteomic-based probability of sarcopenia in the TRACERx external validation cohort at baseline (\u003cstrong\u003ed\u003c/strong\u003e) and at relapse (\u003cstrong\u003ee\u003c/strong\u003e). Time is expressed in months.\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-8885139/v1/b7a77a770a2a97433a533cb1.png"},{"id":104402344,"identity":"52aef3ec-71eb-4a4e-9478-033eafaa9d85","added_by":"auto","created_at":"2026-03-11 12:15:06","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":137677,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eValidation and longitudinal changes in proteomic-based sarcopenia probability\u003cbr\u003e\n(A)\u003c/strong\u003e Distribution of proteomic-based sarcopenia probability in the high-contrast validation cohort.\u003cbr\u003e\n \u003cstrong\u003e(B)\u003c/strong\u003e Change in proteomic-based sarcopenia probability between paired baseline and progressive disease samples in the MATCH-R prostate cohort, stratified according to changes in ECOG performance status.\u003cbr\u003e\n \u003cstrong\u003e(C)\u003c/strong\u003e Correlation between proportional changes in skeletal muscle index and changes in proteomic-based sarcopenia probability in paired baseline and relapse samples from the TRACERx cohort. Dot color indicates the disease-free interval between the two time points. \u003cstrong\u003e(D)\u003c/strong\u003e Heatmap showing the correlation between candidate circulating mediators and the proteomic-based probability of sarcopenia in the 4 cohorts, namely MATCH-T immuno (MR immuno), MATCH-R metastatic castration resistant prostate cancer (mCRPC), TRACERx recurrence (Rrx_rec) and TRACERx baseline (Rrx_B) .\u003c/p\u003e","description":"","filename":"3.png","url":"https://assets-eu.researchsquare.com/files/rs-8885139/v1/c45eb55abeec204f27dfc37f.png"},{"id":104018613,"identity":"4f0e32d5-d64b-454d-aed6-2120b0a41946","added_by":"auto","created_at":"2026-03-05 17:47:24","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":311822,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eTranscriptomic correlates of the plasma proteomic signature of sarcopenia\u003cbr\u003e\n(a,b)\u003c/strong\u003e Gene set enrichment analysis (GSEA) of Hallmark pathways associated with the plasma proteomic signature of sarcopenia in the MATCH-R immunotherapy cohort (\u003cstrong\u003ea\u003c/strong\u003e) and the MATCH-R metastatic castration-resistant prostate cancer cohort (\u003cstrong\u003eb\u003c/strong\u003e).\u003cbr\u003e\n \u003cstrong\u003e(c)\u003c/strong\u003e Enrichment of the Hallmark Myogenesis gene set in differentiating human myoblasts under control conditions and following treatment with IGFBP1 or IGFBP2, in the presence or absence of IGF1.\u003cbr\u003e\n \u003cstrong\u003e(d,e)\u003c/strong\u003e Gene Ontology Biological Process (GOBP) pathway enrichment analysis showing activation and suppression patterns associated with the proteomic signature of sarcopenia in the MATCH-R immunotherapy (\u003cstrong\u003ed\u003c/strong\u003e) and MATCH-R prostate (\u003cstrong\u003ee\u003c/strong\u003e) cohorts.\u003cbr\u003e\n \u003cstrong\u003e(f)\u003c/strong\u003e Single-cell RNA sequencing analysis showing cell type–specific expression of \u003cstrong\u003eIGFBP1\u003c/strong\u003e in hepatic cells and \u003cstrong\u003eIGFBP2\u003c/strong\u003e in tumor cells.\u003c/p\u003e","description":"","filename":"4.png","url":"https://assets-eu.researchsquare.com/files/rs-8885139/v1/c98a1bece3003aecb800dbfd.png"},{"id":104018610,"identity":"5338ec8e-e741-4687-a2af-27e4a3bc9f01","added_by":"auto","created_at":"2026-03-05 17:47:24","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":570117,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eIGFBP1 and IGFBP2 inhibit myogenic differentiation through an IGF-1–dependent mechanism\u003cbr\u003e\n(A–F)\u003c/strong\u003e LHCN human myoblasts were grown to 100% confluence and induced to differentiate for 96 h in DMEM supplemented with 2% FBS in the presence or absence of increasing concentrations of IGFBP1 or IGFBP2 (0.1, 1, 2.5 and 5 µg/mL). \u003cstrong\u003e(A)\u003c/strong\u003e Representative immunofluorescence images showing the effects of IGFBP1 or IGFBP2 on myogenic differentiation. Nuclei were stained with Hoechst 33342 (blue) and myosin heavy chain (MHC) with an anti-MHC antibody (green). Scale bar, 200 µm. \u003cstrong\u003e(B,C)\u003c/strong\u003e Quantification of the fusion index in LHCN cells treated with IGFBP1 (\u003cstrong\u003eB\u003c/strong\u003e) or IGFBP2 (\u003cstrong\u003eC\u003c/strong\u003e). \u003cstrong\u003e(D)\u003c/strong\u003e Western blot analysis of MHC protein expression in differentiating LHCN cells treated with IGFBP1 or IGFBP2 at the indicated concentrations. \u003cstrong\u003e(E,F)\u003c/strong\u003eDensitometric quantification of MHC protein levels in IGFBP1-treated (\u003cstrong\u003eE\u003c/strong\u003e) and IGFBP2-treated (\u003cstrong\u003eF\u003c/strong\u003e) cells, normalized to β-actin and expressed relative to control conditions (0 µg/mL). **P \u0026lt; 0.01, ***P \u0026lt; 0.001, **\u003cstrong\u003eP \u0026lt; 0.0001 compared with control.\u003c/strong\u003e\u003cbr\u003e\n \u003cstrong\u003e(G–L)\u003c/strong\u003e LHCN cells were grown to confluence and differentiated for 96 h in the presence of IGFBP1 (1 µg/mL) or IGFBP2 (1 µg/mL), with or without increasing concentrations of IGF-1 (0, 50, 150 or 300 ng/mL). Experimental conditions included control (2% FBS only), IGFBP1 or IGFBP2 alone, IGF-1 alone, and combined IGFBP1+IGF-1 or IGFBP2+IGF-1 treatments. \u003cstrong\u003e(G–J)\u003c/strong\u003eRepresentative immunofluorescence images illustrating the modulatory effect of IGF-1 on IGFBP1- or IGFBP2-mediated inhibition of myogenic differentiation. \u003cstrong\u003e(K,L)\u003c/strong\u003eWestern blot analysis and quantification of MHC protein expression demonstrating partial or complete rescue of myogenic differentiation upon IGF-1 supplementation in the presence of IGFBP1 (\u003cstrong\u003eK\u003c/strong\u003e) or IGFBP2 (\u003cstrong\u003eL\u003c/strong\u003e). **#P \u0026lt; 0.05, ##P \u0026lt; 0.01, ####P \u0026lt; 0.0001 compared with control; *P \u0026lt; 0.05, **P \u0026lt; 0.01, ***P \u0026lt; 0.001, **P \u0026lt; 0.0001 compared with IGFBP1-only or IGFBP2-only conditions.\u003c/p\u003e","description":"","filename":"5.png","url":"https://assets-eu.researchsquare.com/files/rs-8885139/v1/2beacd798d525a93b9a3ac6f.png"},{"id":106959921,"identity":"c26ebd2a-6b0a-4677-b8b5-5ef9301d660c","added_by":"auto","created_at":"2026-04-15 09:17:06","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":2500664,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8885139/v1/7b0058c2-4c9c-4347-96b1-a49399ce21d1.pdf"},{"id":104018612,"identity":"1c741345-5bdc-4ca2-801d-fabead95c788","added_by":"auto","created_at":"2026-03-05 17:47:24","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":1237088,"visible":true,"origin":"","legend":"","description":"","filename":"supplementaryfigures.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8885139/v1/8c62fe81b0bf3462e1af364c.pdf"},{"id":104403227,"identity":"aa6c2572-d59b-4c8c-8287-1f8e06a14412","added_by":"auto","created_at":"2026-03-11 12:17:47","extension":"xlsx","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":98893,"visible":true,"origin":"","legend":"","description":"","filename":"supplementarytables.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-8885139/v1/dd33b9786b4ff18543ff2ddb.xlsx"},{"id":104018616,"identity":"35e85588-02ee-40a8-ab98-593390b87e1a","added_by":"auto","created_at":"2026-03-05 17:47:25","extension":"pdf","order_by":3,"title":"","display":"","copyAsset":false,"role":"supplement","size":1991120,"visible":true,"origin":"","legend":"","description":"","filename":"Primaryimages.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8885139/v1/4a5f647df28364120cae1aea.pdf"},{"id":104018614,"identity":"05c06c3c-cb7c-4cae-a38e-ed1da2dca5d0","added_by":"auto","created_at":"2026-03-05 17:47:24","extension":"docx","order_by":4,"title":"","display":"","copyAsset":false,"role":"supplement","size":283355,"visible":true,"origin":"","legend":"","description":"","filename":"Graphicalabstract.docx","url":"https://assets-eu.researchsquare.com/files/rs-8885139/v1/51333dd45d837dfd7dc2b91e.docx"}],"financialInterests":"Competing interest reported. Fabrice Barlesi reports institutional relationships (no personal financial interests) with AbbVie, ACEA, Amgen, AstraZeneca, Bayer, Bristol Myers Squibb, Boehringer Ingelheim, Eisai, Eli Lilly Oncology, F. Hoffmann–La Roche Ltd, Genentech, Ipsen, Ignyta, Innate Pharma, Loxo, Novartis, MedImmune, Merck, MSD, Pierre Fabre, Pfizer, Sanofi-Aventis, Summit Therapeutics and Takeda.\nCaroline Even reports consulting or advisory roles with Innate Pharma, Bristol Myers Squibb, MSD Oncology, Merck Serono, Novartis, F-star Therapeutics, Merus and GlaxoSmithKline (institutional), and travel, accommodation or expenses from MSD Oncology and Merck Serono.\nNathalie Lassau reports participation on an advisory board for Jazz Pharmaceuticals.\nYohann Loriot reports honoraria from Janssen, Bristol Myers Squibb, Roche, Gilead, MSD and Pfizer; institutional research funding from Amgen, Janssen Oncology, MSD Oncology, Lilly, AstraZeneca, Orion, Exelixis, Incyte, Pfizer, Sanofi, Astellas Pharma, Gilead Sciences, Merck KGaA, Taiho Pharmaceutical, Bristol Myers Squibb, Roche and Tyra Biosciences; and travel, accommodation or expenses from Astellas Pharma, Pfizer, MSD Oncology and AstraZeneca.\nAntoine Italiano reports research grants from AstraZeneca, Bayer, Bristol Myers Squibb, Merck, MSD and Pharmamar.\nCarla M. Prado has previously received honoraria and/or paid consultancy from Abbott Nutrition, Nutricia, Nestlé Health Science and Novo Nordisk.\nMariam Jamal-Hanjani reports consulting for Astex Pharmaceuticals, Pfizer and Achilles Therapeutics; membership on the Scientific Advisory Board and Steering Committee of Achilles Therapeutics; and speaker honoraria from Pfizer, Astex Pharmaceuticals, Oslo Cancer Cluster, Bristol Myers Squibb and Genentech.\nBenjamin Besse reports institutional honoraria and speaker’s bureau participation for AbbVie, AstraZeneca, Chugai Pharmaceutical, Daiichi Sankyo, Hedera Dx, Janssen, Merck Sharp \u0026 Dohme, Roche, Sanofi Aventis and Springer Healthcare Ltd; consulting or advisory roles (institutional) for AbbVie, BioNTech SE, Bristol Myers Squibb, Chugai Pharmaceutical, CureVac AG, Daiichi Sankyo, F. Hoffmann–La Roche Ltd, Pharmamar, Regeneron, Sanofi Aventis and Turning Point Therapeutics; and institutional research funding from AstraZeneca, BeiGene, Genmab A/S, GlaxoSmithKline, Janssen, Merck Sharp \u0026 Dohme, Ose Immunotherapeutics, Pharmamar, Roche-Genentech, Sanofi and Takeda.\nAll other authors declare that they have no competing interests.","formattedTitle":"A plasma proteomic signature of cancer-related sarcopenia implicates the IGFBP axis in muscle dysfunction","fulltext":[{"header":"Background","content":"\u003cp\u003eSarcopenia was originally described as a progressive decline in skeletal muscle mass, strength, and function associated with aging, but it is now increasingly recognized as a condition that can co-exist across a range of clinical conditions\u003csup\u003e\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e\u003c/sup\u003e. Sarcopenia represents a major challenge in cancer care. In cancer patients, sarcopenia is associated with impaired quality of life, increased treatment toxicity, reduced tolerance to systemic therapies, and poorer survival, thereby amplifying the overall burden of cancer for both patients and caregivers\u003csup\u003e\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e\u003c/sup\u003e. Sarcopenia frequently coexists with cancer cachexia and other wasting syndromes, further complicating clinical management\u003csup\u003e\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e\u003c/sup\u003e. However, reported prevalence estimates range widely\u003csup\u003e\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e\u003c/sup\u003e, from 5% to nearly 90%, largely reflecting heterogeneity in assessment methods, cutoff definitions, and patient populations. This lack of consensus has limited the comparability and generalizability of existing studies, ultimately slowing progress in the field.\u003c/p\u003e \u003cp\u003eHistorically, sarcopenia has been primarily studied in geriatric medicine, where its definition and assessment have progressively evolved toward a multidimensional framework incorporating muscle mass, muscle strength, and physical performance\u003csup\u003e\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e\u003c/sup\u003e. In contrast, oncology research has largely relied on imaging-based quantification of muscle mass, typically derived from routine computed tomography scans. Although this approach has provided important prognostic insights, it captures only a fraction of the biological complexity underlying functional decline and fails to account for systemic and molecular drivers of muscle dysfunction.\u003c/p\u003e \u003cp\u003eCancer-derived sarcopenia is increasingly recognized as a multifactorial condition shaped by nutritional status, systemic inflammation, host metabolism, aging, sex, cancer biology, and anticancer treatments. Yet, the contribution of tumor-driven systemic signals and circulating mediators remains incompletely understood. Emerging evidence suggests that specific tumor biological features and plasma proteomic profiles are associated with the development of cancer derived cachexia, as demonstrated in both resected and advanced non-small cell lung cancer within the TRACERx (TRAcking Cancer Evolution through therapy (Rx)) study\u003csup\u003e\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e\u003c/sup\u003e. These observations support the concept that circulating factors can act as measurable intermediates linking tumor biology to host functional decline. However, a comprehensive and scalable framework to characterize cancer-derived sarcopenia at the systemic molecular level is still lacking.\u003c/p\u003e \u003cp\u003eSkeletal muscle health and maintenance depend both on muscle intrinsic properties but also on coordinated regulation by metabolic, inflammatory, and neural inputs\u003csup\u003e\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e,\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e\u003c/sup\u003e. Skeletal muscle maintenance and regeneration requires effective myoblast differentiation and repair processes\u003csup\u003e\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e\u003c/sup\u003e, which are tightly coupled to endocrine signaling and neuromuscular integrity\u003csup\u003e\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e,\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e\u003c/sup\u003e. Disruption of these integrated systems, rather than isolated muscle atrophy alone, may therefore underlie the development of sarcopenia in cancer.\u003c/p\u003e \u003cp\u003eBased on this rationale, we leveraged large scale plasma proteomics to define a molecular signature of physical function and muscle mass across multiple cancer types. We further aimed to identify circulating mediators linking tumor biology and systemic inflammation to impaired myogenesis, thereby providing mechanistic insight into cancer-derived sarcopenia and a foundation for biologically informed patient stratification. We aimed to generate a sarcopenia probability (SP) score that accurately classified sarcopenic status across independent validation cohorts.\u003c/p\u003e"},{"header":"Methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e\n \u003ch2\u003ePatients\u003c/h2\u003e\n \u003cp\u003ePatients were selected from two independent cohorts of the MATCH-R (Molecular Analysis for Therapy Choice - Resistance) study\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e12\u003c/span\u003e\u003c/sup\u003e (NCT02517892), a prospective, single-center clinical trial conducted at Gustave Roussy between 2015 and 2022. The study was designed to investigate tumor molecular evolution under anticancer therapies and to identify mechanisms of treatment resistance. Patients underwent tumor profiling at baseline and at disease progression, with the aim of characterizing resistance-associated molecular alterations and informing potential treatment strategies. Eligible patients were adults receiving or scheduled to receive systemic anticancer therapy, with at least one tumor lesion accessible for core needle biopsy, who were covered by a social security system and provided written informed consent.\u003c/p\u003e\n \u003cp\u003eThe discovery cohort included patients with advanced solid tumors, predominantly lung and bladder cancers, treated with immune checkpoint inhibitors (MATCHR-I). The validation cohort consisted of patients with metastatic castration resistant prostate cancer (mCRPC) treated with androgen-receptor pathway inhibitors\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e13\u003c/span\u003e\u003c/sup\u003e. The discovery cohort was used for model training and internal testing, whereas the mCRPC cohort served as an independent validation set (Supplementary Fig.\u0026nbsp;1).\u003c/p\u003e\n \u003cp\u003eMATCH-R study was approved by the CPP (\u0026ldquo;Comit\u0026eacute; de protection des personnes\u0026rdquo;) and by the ANSM (\u0026ldquo;Agence nationale de s\u0026eacute;curit\u0026eacute; du m\u0026eacute;dicament et des produits de sant\u0026eacute;\u0026rdquo;) and adhered to the principles in the Guideline for Good Clinical Practice and the Declaration of Helsinki. Written informed consent was obtained from all patients included in this study.\u003c/p\u003e\n \u003cp\u003eExternal validation was performed using the independent TRACERx cohort, whose data have been previously published. Clinical data, proteomic data, and outcome information were retrieved from the publicly available repository described in the original publication\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e6\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e\n\u003c/div\u003e\n\u003ch3\u003eMuscle mass measurement and sarcopenia definition\u003c/h3\u003e\n\u003cp\u003eSkeletal muscle mass was quantified using the skeletal muscle index (SMI) derived from computed tomography (CT) or positron emission tomography (PET) scans performed within 60 days of plasma collection. Cross sectional muscle area was measured at the level of the third lumbar vertebra (L3) and normalized to height squared to derive skeletal muscle index (SMI). Low muscle mass was defined using established sex specific cutoffs (52.4 cm\u0026sup2;/m\u0026sup2; for males and 38.5 cm\u0026sup2;/m\u0026sup2; for females) \u003csup\u003e\u003cspan class=\"CitationRef\"\u003e14\u003c/span\u003e\u003c/sup\u003e. Anthropometer3DNet was also used to extract body composition parameters from CT images. This software, which is available for research purposes on \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ewww.oncometer3d.com\u003c/span\u003e\u003c/span\u003e, can automatically measure several anthropometric parameters including muscle mass (MBM) on multislice CT scans in less than 5 through a deep learning-based segmentation of adipose tissue (visceral and subcutaneous) and muscle voxels\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e15\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e\n\u003cp\u003ePhysical function was approximated using Eastern Cooperative Oncology Group performance status (ECOG PS). To improve phenotype labeling for model training, we defined a high-contrast subgroup. Patients with ECOG PS 0 and high SMI were classified as having a low probability of sarcopenia (LS), whereas patients with low SMI and ECOG PS\u0026thinsp;\u0026ge;\u0026thinsp;2 were classified as having a high probability of sarcopenia (HS). This subgroup was used for supervised model training, maximizing phenotypic contrast and reducing misclassification.\u003c/p\u003e\n\u003ch3\u003ePlasma proteome\u003c/h3\u003e\n\u003cp\u003ePlasma proteomic profiling was performed using the Olink Proximity Extension Assay (PEA) technology (Olink Proteomics AB, Uppsala, Sweden). Baseline plasma samples from the discovery cohort were analyzed using the Olink Explore 1536 platform, comprising 1,472 protein assays and 48 controls distributed across four panels (inflammation, oncology, cardiometabolic, and neurology). Sequencing was performed on a NovaSeq 6000 system using two S1 flow cells with 2 \u0026times; 50 bp read lengths.\u003c/p\u003e\n\u003cp\u003eThe validation mCRPC cohort and an external validation cohort from TRACERx were analyzed using the Olink Explore 3072 platform. Raw sequencing counts were converted into normalized protein expression (NPX) values following Olink\u0026rsquo;s standard quality control and normalization pipeline, including internal extension and inter plate controls. NPX values are reported on a log2 scale, with higher values corresponding to higher relative protein abundance. Assay validation metrics are available from the manufacturer.\u003c/p\u003e\n\u003ch3\u003eCell culture\u003c/h3\u003e\n\u003cp\u003eImmortalized human myoblasts (LHCN, AB1190, MB135) were cultured at 37\u0026deg;C and 5% CO₂ in growth medium consisting of DMEM supplemented with 10% fetal bovine serum, Medium 199, basic fibroblast growth factor (0.5 ng/mL), dexamethasone (0.2 \u0026micro;g/mL), and penicillin streptomycin (1%). Absence of mycoplasma contamination was confirmed using MycoAlert assays (Lonza). HepG2 cells were cultured under standard conditions and treated with recombinant human IL-6 (0\u0026ndash;30 ng/mL) for 24 or 48 hours. Following treatment, cells were harvested for protein and RNA extraction. Intracellular IGFBP1 protein levels were analyzed by western blotting from whole cell lysates, while secreted IGFBP1 was measured by western blot analysis of culture supernatants collected at the indicated time points. Total RNA was extracted from HepG2 cells and IGFBP1 mRNA expression was quantified by RT-qPCR.\u003c/p\u003e\n\u003ch3\u003eWestern blot\u003c/h3\u003e\n\u003cp\u003eProtein extraction and immunoblotting were performed using standard procedures. Primary antibodies included anti myosin heavy chain (MF20; R\u0026amp;D Systems) and anti \u0026beta; actin (Santa Cruz Biotechnology). Horseradish peroxidase conjugated secondary antibodies and chemiluminescent detection reagents (Millipore) were used for signal visualization. Membranes were imaged using an Amersham Imager 800 system (GE Healthcare). Band intensities were quantified using Fiji software and normalized to \u0026beta;-actin expression.\u003c/p\u003e\n\u003cdiv id=\"Sec8\" class=\"Section2\"\u003e\n \u003ch2\u003eImmunofluorescent staining\u003c/h2\u003e\n \u003cp\u003eDifferentiated myotubes were washed with phosphate-buffered saline (PBS), fixed with 4% paraformaldehyde for 10 minutes, permeabilized with 0.25% Triton X-100, and blocked with 5% normal goat serum. Cells were incubated with anti myosin heavy chain antibody (R\u0026amp;D Systems; 1:150), followed by Alexa Fluor 488 conjugated secondary antibody (Thermo Fisher Scientific; 1:400). Nuclei were counterstained with Hoechst 33342.\u003c/p\u003e\n \u003cp\u003eImages were acquired using a Cytation 1 Cell Imaging Reader (BioTek) at \u0026times;10 magnification. Fusion index was quantified using Fiji software by analyzing six randomly selected fields per condition. Data represent three independent experiments.\u003c/p\u003e\n\u003c/div\u003e\n\u003ch3\u003eTranscriptomic analysis\u003c/h3\u003e\n\u003cp\u003eRNA sequencing of myoblasts\u003c/p\u003e\n\u003cp\u003eMyoblast experiments were conducted using four biological replicates per condition. Total RNA was extracted using the NucleoSpin RNA isolation kit (Macherey Nagel). RNA integrity was assessed using an Agilent Bioanalyzer 2100. Sequencing libraries were prepared using the NEBNext Ultra RNA Library Prep Kit (New England Biolabs) and sequenced on an Illumina HiSeq 2000 platform.\u003c/p\u003e\n\u003cp\u003eTumor RNA sequencing\u003c/p\u003e\n\u003cp\u003eRNA from MATCHR-I and mCRPC tumor samples was extracted using the AllPrep DNA/RNA Mini Kit (Qiagen). Libraries were sequenced on Illumina NextSeq 500 or NovaSeq platforms as paired-end reads following standard quality control procedures.\u003c/p\u003e\n\u003cp\u003eGene expression quantification and enrichment analysis\u003c/p\u003e\n\u003cp\u003eRaw sequencing reads were quality-controlled using Trim Galore. Transcript quantification was performed using Kallisto (v0.44.0) with the GENCODE v27 reference. Transcript-level estimates were aggregated to gene-level counts using TxImport. Genes with low expression (\u0026lt;\u0026thinsp;2 counts in 10% of samples) were excluded. Differential expression analysis was performed using DESeq2, applying a false discovery rate (FDR)\u0026thinsp;\u0026lt;\u0026thinsp;0.05 and |log2 fold change| \u0026ge;1.\u003c/p\u003e\n\u003cp\u003eGene set enrichment analysis was conducted using clusterProfiler with Hallmark, Gene Ontology biological process, and KEGG gene sets from the Molecular Signatures Database (MSigDB).\u003c/p\u003e\n\u003cp\u003eSingle-sample gene set enrichment analysis\u003c/p\u003e\n\u003cp\u003eTo estimate the activity of muscle-related gene pathways at the sample level, we applied single-sample gene set enrichment analysis (ssGSEA). ssGSEA scores were computed from normalized gene expression counts using the GSVA R package (v2.4.1). Predefined gene sets related to skeletal muscle biology were used to derive enrichment scores for each sample, enabling quantitative comparison of pathway activity across samples.\u003c/p\u003e\n\u003ch3\u003eSingle-cell RNA sequencing\u003c/h3\u003e\n\u003cp\u003eSingle-nucleus RNA sequencing (snRNA-seq) was performed by Celsius Therapeutics. Nuclei were isolated from frozen tissue samples following standard protocols. Libraries were generated using the 10x Genomics Chromium Single Cell 3\u0026prime; v3.1 chemistry and sequenced on an Illumina NovaSeq platform. Raw sequencing reads were processed using STARsolo for alignment, barcode processing, and gene counting. Downstream quality control, normalization, and integration were conducted using the Gustave Roussy single-cell analysis pipeline (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://github.com/gustaveroussy/single-cell/wiki/\u003c/span\u003e\u003c/span\u003e\u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e).\u003c/span\u003e\u003c/p\u003e\n\u003cp\u003eQuality control metrics were computed for each nucleus, including the number of detected genes, total unique molecular identifier (UMI) counts, and the proportion of reads mapping to mitochondrial genes. Nuclei were retained for downstream analysis if they expressed at least 200 genes, contained a minimum of 1,000 total UMIs, and had\u0026thinsp;\u0026le;\u0026thinsp;20% mitochondrial gene expression. Nuclei not meeting these criteria were excluded.\u003c/p\u003e\n\u003cp\u003eData normalization was performed using SCTransform as implemented in Seurat (v4.0.4), regressing out total UMI counts per nucleus to mitigate technical variability. Principal component analysis (PCA) was applied to the normalized data, and batch effects were corrected using the Harmony algorithm (v0.1.0), with PCA embeddings used as input for integration.\u003c/p\u003e\n\u003cp\u003eThe dimensionality and clustering resolution were selected based on visual inspection of Uniform Manifold Approximation and Projection (UMAP) embeddings to identify biologically coherent structures. Cluster stability across parameter combinations was further assessed using the clustree R package (v0.4.3), ensuring robustness of the chosen clustering configuration.\u003c/p\u003e\n\u003cdiv id=\"Sec11\" class=\"Section2\"\u003e\n \u003ch2\u003eQuantitative real-time PCR\u003c/h2\u003e\n \u003cp\u003eTotal RNA was extracted using the Quick-RNA kit (Zymo Research) and reverse transcribed using the MMLV RT kit (Eurogentec). Quantitative PCR was performed using SYBR Green chemistry on a LightCycler 96 system (Roche). Relative gene expression was calculated using the 2^\u0026minus;\u0026Delta;\u0026Delta;Ct method. All reactions were performed in triplicate.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec12\" class=\"Section2\"\u003e\n \u003ch2\u003eMachine learning and statistical analysis\u003c/h2\u003e\n \u003cp\u003eDifferential plasma protein expression between patients with high versus low probability of sarcopenia was assessed using the Wilcoxon rank-sum test with false discovery rate correction. Proteins were further selected based on biological relevance and annotation from the Human Protein Atlas\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e16\u003c/span\u003e\u003c/sup\u003e, to enhance interpretability and reduce dimensionality in a limited sample size setting.\u003c/p\u003e\n \u003cp\u003eA decision tree based extreme gradient boosting (XGBoost) classification model was developed using the tidymodels framework\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e17\u003c/span\u003e\u003c/sup\u003e. The model was trained on the high-contrast subgroup of lung cancer patients using fivefold cross-validation with 20 repetitions, stratified by sarcopenia probability. Hyperparameters were optimized based on classification accuracy.\u003c/p\u003e\n \u003cp\u003eThe best-performing model was applied to the full discovery cohort, generating a predicted probability of sarcopenia that was subsequently used in Kaplan Meier survival analyses and Cox proportional hazards models, treating sarcopenia probability as a continuous variable. Spearman correlation was used to assess the relationship between predicted sarcopenia probability and SMI.\u003c/p\u003e\n \u003cp\u003eThe model was independently validated in the mCRPC and in the TRACERx cohort. To derive a reduced plasma protein signature, a second XGBoost regression model was trained using predicted sarcopenia probabilities as the outcome on the MATCHR-I and mCRPC using the tidymodels framework and fivefold cross-validation with 20 repetitions Feature selection was performed iteratively based on variable importance, with root mean square error used for model selection. Performance of the extended and reduced signatures was compared in the external validation cohort.\u003c/p\u003e\n\u003c/div\u003e"},{"header":"Results","content":"\u003cdiv id=\"Sec14\" class=\"Section2\"\u003e \u003ch2\u003ePatient cohorts and definition of high-contrast sarcopenia groups\u003c/h2\u003e \u003cp\u003eWe included patients from the MATCH-R study at Gustave Roussy. The training cohort comprised 99 patients with advanced cancers receiving immunotherapy, predominantly lung and bladder cancer. Skeletal muscle mass was quantified using the Skeletal Muscle Index (SMI) measured at the L3 vertebral level on CT or PET scans performed within 42 days of plasma collection. Functional status was assessed using ECOG PS. Among patients with available imaging, 51 had low SMI and 34 high SMI. To establish a high-contrast subgroup for model development, we combined SMI with ECOG PS: patients with high SMI and ECOG 0 (functionally unimpaired) were classified as low probability of sarcopenia (LS), whereas patients with low SMI and ECOG\u0026thinsp;\u0026ge;\u0026thinsp;2 were classified as high probability of sarcopenia (HS). This yielded 36 patients (21 HS, 15 LS) who constituted the training set. Baseline characteristics of the overall and high-contrast groups are summarized in Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e.For validation, we used two independent cohorts. The first was an internal validation cohort from MATCH-R, comprising 55 patients with metastatic castration-resistant prostate cancer (mCRPC) treated with enzalutamide, with 88 plasma samples available at baseline and progression. The second was an external validation cohort derived from the TRACERx study, consisting of patients with resected and relapsed non-small cell lung cancer (NSCLC), with proteomic and imaging data available for 151 patients (141 baseline, 114 relapse). These validation sets allowed assessment of the model\u0026rsquo;s generalizability across different cancer types and disease contexts.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec15\" class=\"Section2\"\u003e \u003ch2\u003eDiscovery of a proteomic signature of sarcopenia\u003c/h2\u003e \u003cp\u003eWe performed plasma proteomics using the Olink Explore platform in the training cohort. Differential expression analysis between HS and LS patients in the high-contrast subgroup identified 353 proteins enriched in HS, predominantly associated with systemic inflammation, immune dysregulation, matrix remodeling, and metabolic stress, and 64 proteins enriched in LS, many of which were related to muscle or neuronal biology (Supplementary Fig.\u0026nbsp;1, Supplementary Table\u0026nbsp;1).\u003c/p\u003e \u003cp\u003eTo focus on proteins reflecting muscle function rather than tumor biology, we prioritized proteins enriched in LS and underrepresented in HS. A biology-driven selection further narrowed candidates to those implicated in skeletal muscle growth and regeneration (e.g., MSTN, WFIKKN1, CD34, PAMR1, DPP4), neuromuscular connectivity (e.g., CNTN family, adhesion/guidance molecules), extracellular matrix remodeling (e.g., integrins), and autophagy/lysosomal pathways (e.g., LAMP2, IDS). Proteins primarily involved in systemic metabolism, immunity, endocrine regulation, or oncogenic signaling were excluded.\u003c/p\u003e \u003cp\u003eThis process resulted in a 23-protein panel that served as the basis for the XGBoost classification model (model p23; Supplementary Table \u003cspan refid=\"MOESM2\" class=\"InternalRef\"\u003eS2\u003c/span\u003e). The model was trained to predict sarcopenia probability (SP) as a continuous variable and classify patients as sarcopenic (SP\u0026thinsp;\u0026gt;\u0026thinsp;50%) or non-sarcopenic (SP\u0026thinsp;\u0026le;\u0026thinsp;50%) within the high-contrast training set.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec16\" class=\"Section2\"\u003e \u003ch2\u003eModel performance and association with SMI and ECOG\u003c/h2\u003e \u003cp\u003eThe XGBoost model (p23) was first evaluated within the high-contrast training set. The model achieved an accuracy of 0.893 in classifying patients as sarcopenic or non-sarcopenic, demonstrating robust discrimination between HS and LS cases (Supplementary Fig.\u0026nbsp;2). When applied to the full training cohort, the predicted probability of sarcopenia (SP) correlated inversely with SMI in both male and female patients (Spearman ρ = -0.39, p\u0026thinsp;=\u0026thinsp;0.0045 and ρ = -0.42, p\u0026thinsp;=\u0026thinsp;0.016 respectively; Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eA), confirming that the model captured meaningful biological variation in muscle mass.\u003c/p\u003e \u003cp\u003eSP also reflected functional status: increasing SP was associated with higher ECOG PS scores, indicating concordance between the proteomic signature and clinically assessed physical function (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eB).\u003c/p\u003e \u003cp\u003eIn the internal validation cohort of 55 mCRPC patients, the model maintained a similar correlation with SMI (ρ = -0.41, p\u0026thinsp;=\u0026thinsp;0.008) and ECOG PS, and in the subset of 18 high-contrast cases, classification accuracy reached 0.889, comparable to the training set (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eA).\u003c/p\u003e \u003cp\u003eImportantly, in the external TRACERx cohort, which included both baseline and relapse samples, SP also correlated with cross-sectional skeletal muscle area at L3 (for male: baseline ρ = -0.29, p 0.012; relapse ρ = -0.42, p\u0026thinsp;=\u0026thinsp;0.0018; for female: baseline ρ = -0.24, p 0.073; relapse ρ = -0.35, p\u0026thinsp;=\u0026thinsp;0.047), confirming the generalizability of the model across independent patient populations (1 C-D).\u003c/p\u003e \u003cp\u003eTogether, these findings demonstrate that the 23-protein signature provides a robust, scalable measure of sarcopenia probability, consistently reflecting both muscle mass and functional status across multiple cancer types and cohorts.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec17\" class=\"Section2\"\u003e \u003ch2\u003ePrognostic impact of the proteomic sarcopenia signature\u003c/h2\u003e \u003cp\u003eWe next investigated the prognostic relevance of the proteomic-based probability of sarcopenia (SP) across cohorts. As expected, in the high-contrast training set, SP strongly stratified overall survival (OS), with sarcopenic patients (SP\u0026thinsp;\u0026gt;\u0026thinsp;50%) exhibiting a markedly shorter median OS compared with non-sarcopenic patients (3.8 vs 34.8 months; log-rank p\u0026thinsp;\u0026lt;\u0026thinsp;0.0001; Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eA). Importantly, this prognostic separation was preserved when extending the analysis to the remaining, lower-contrast patients in the training cohort, in whom sarcopenic patients also experienced significantly worse outcomes (median OS 7.8 vs 22.1 months; p\u0026thinsp;=\u0026thinsp;0.012; Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eB).\u003c/p\u003e \u003cp\u003eIn the internal validation cohort of patients with metastatic castration-resistant prostate cancer (mCRPC), baseline SP similarly identified a subgroup with significantly inferior survival. Sarcopenic patients had a median OS of 8.9 months compared with 21.9 months in non-sarcopenic patients (p\u0026thinsp;\u0026lt;\u0026thinsp;0.0001; Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eC), confirming the prognostic robustness of the signature in a cancer type with distinct biology and treatment context.\u003c/p\u003e \u003cp\u003eConsistent results were observed in the external TRACERx cohort. Across resected and relapsed NSCLC samples, sarcopenic patients had significantly shorter OS compared with non-sarcopenic patients (median OS 25 vs 46 months p\u0026thinsp;\u0026lt;\u0026thinsp;0.001 for baseline; median OS 30 vs 44 months, p 0.043 at relapse Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eD\u0026ndash;E). Notably, in this cohort, the proteomic sarcopenia classification only partially overlapped with cachexia status, with most cachectic patients being sarcopenic but a substantial proportion of sarcopenic patients not fulfilling cachexia criteria, reinforcing the biological and clinical distinction between these two syndromes.\u003c/p\u003e \u003cp\u003eMultivariable Cox proportional hazards models were used to assess the independent prognostic value of the proteomic sarcopenia probability across cohorts, adjusting for available clinical covariates.\u003c/p\u003e \u003cp\u003eIn the MATCHR immunotherapy cohort, sarcopenia probability remained independently associated with overall survival after adjustment for skeletal muscle index at L3, ECOG performance status, age, and sex (HR 4.13, 95% CI 1.38\u0026ndash;12.3, p\u0026thinsp;=\u0026thinsp;0.011), while SMI showed only a trend towards significance.\u003c/p\u003e \u003cp\u003eIn the mCRPC validation cohort, sarcopenia probability and ECOG performance status were both independently associated with survival, whereas age was not. SMI could not be included due to a high rate of missing imaging data (\u0026gt;\u0026thinsp;50%), reflecting routine clinical follow-up in mCRPC patients, which is primarily based on PSA monitoring rather than systematic radiological assessments..\u003c/p\u003e \u003cp\u003eIn the external TRACERx cohort, sarcopenia probability was consistently associated with overall survival both at baseline and at relapse, independently of age, sex, and cross-sectional muscle area, which was not significantly associated with outcome in multivariable analyses. (supplementary table S4)\u003c/p\u003e \u003cp\u003eTogether, these results demonstrate that the plasma proteomic sarcopenia signature captures clinically meaningful systemic vulnerability that translates into a strong and consistent prognostic impact across tumor types, disease stages, and independent patient cohorts.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec18\" class=\"Section2\"\u003e \u003ch2\u003eLongitudinal changes in sarcopenia probability\u003c/h2\u003e \u003cp\u003eTo determine whether the proteomic-based probability of sarcopenia (SP) captures dynamic changes in muscle health over time, we investigated its longitudinal behavior in patients with paired plasma samples collected at baseline and at disease progression or relapse.\u003c/p\u003e \u003cp\u003eIn the mCRPC validation cohort, paired proteomic samples were available for 33 patients treated with enzalutamide, of whom 20 had contemporaneous clinical documentation of ECOG performance status at both time points. Changes in SP closely mirrored clinical trajectories. Patients whose performance status improved between baseline and progression consistently showed a marked reduction in SP, with both individuals transitioning from sarcopenic to non-sarcopenic classification and exhibiting a mean SP reduction of 58%. In contrast, patients with stable ECOG PS showed minimal variation in SP, whereas those experiencing clinical deterioration displayed a corresponding increase in sarcopenia probability (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eB). These findings support the ability of the proteomic signature to track clinically meaningful changes in functional status over time.\u003c/p\u003e \u003cp\u003eWe next extended this analysis to the external TRACERx cohort, where paired plasma proteomics and quantitative muscle mass assessments were available for 73 patients at baseline and relapse. In this independent dataset, longitudinal changes in SP were significantly correlated with proportional changes in skeletal muscle mass between the two time points (Spearman ρ = \u0026minus;0.32, p\u0026thinsp;=\u0026thinsp;0.005; Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eC). Notably, the strength of this association appeared modulated by the disease-free interval separating baseline and relapse, suggesting that longer inter-sample intervals may allow for more pronounced and biologically detectable muscle remodeling.\u003c/p\u003e \u003cp\u003eCollectively, these longitudinal analyses demonstrate that the proteomic sarcopenia signature is not merely a static classifier but reflects dynamic changes in muscle health across disease evolution, treatment exposure, and relapse, reinforcing its potential utility for real-time patient monitoring.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec19\" class=\"Section2\"\u003e \u003ch2\u003eDevelopment of a reduced model\u003c/h2\u003e \u003cp\u003eAs a final feature selection step, we sought to determine whether a more streamlined proteomic signature, consisting of a reduced number of proteins, could still reproduce the classification performance of our original model. We then trained an Xgboost regression model on MATCH-R cohorts and found that a model comprising a progressively reduced number of proteins maintained a good correlation with initial signature, with a model based on 4 proteins (model P4, based on CNTN3, CBLN4, MSTN, and ITGA11) resulting in an RMSE of 0.053 on the validation fold (supplementary Fig.\u0026nbsp;2 and supplementary table S5). Model p4 showed similar result in the high contrast groups, and AUC of 1, in the discovery cohort, and an accuracy of 0.83 and AUC of 1 on the prostate cohort. On the overall cohorts, both prediction from the 2 models showed a high correlation of 0.92, 0.98 and up to 0.99. Thus, the results from the classification model were closely approximated by the regression models with lesser plasma proteins\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec20\" class=\"Section2\"\u003e \u003ch2\u003eMediators discovery:\u003c/h2\u003e \u003cp\u003eTo identify potential circulating mediators underlying the proteomic signature of sarcopenia, we treated the signature-derived probability of sarcopenia as a continuous variable and investigated its association with individual plasma proteins across cohorts. This strategy allowed us to move beyond classification performance and interrogate biological processes consistently linked to sarcopenia severity.\u003c/p\u003e \u003cp\u003eAcross the discovery cohort, several plasma proteins showed significant positive correlations with the probability of sarcopenia after false discovery rate correction. When extending this analysis to the internal validation cohort of patients with metastatic castration-resistant prostate cancer and to the external TRACERx cohort, IGFBP1, IGFBP2 and IL6 emerged as the most consistently associated proteins, showing significant positive correlations with sarcopenia probability in all evaluated datasets (Supplementary Table \u003cspan refid=\"MOESM3\" class=\"InternalRef\"\u003eS3\u003c/span\u003e and Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eD).\u003c/p\u003e \u003cp\u003eAmong the others, we observed several proteins with known or suspected roles in muscle pathophysiology. These include proteins implicated in muscle disease (e.g., SLC39A14\u003csup\u003e18\u003c/sup\u003e), markers of muscle damage (CKB, SERPINA3\u003csup\u003e19\u003c/sup\u003e). Additionally, ITIH3, previously associated with disease activity in myastenia gravis\u003csup\u003e\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e\u003c/sup\u003e and recently associated with muscle wasting\u003csup\u003e\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e\u003c/sup\u003e, was also identified.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec21\" class=\"Section2\"\u003e \u003ch2\u003eTranscriptomic correlates at tumor level\u003c/h2\u003e \u003cp\u003eWe next investigated the transcriptomic correlates of our sarcopenia signature. Bulk RNA-seq analyses performed in our two internal cohorts revealed no meaningful correlation between IGFBP1/2 plasma levels and their corresponding mRNA expression in tumor tissue. At the pathway level, however, increasing sarcopenia probability was consistently associated with suppression of the Hallmark myogenesis program as well as multiple GO Biological Process pathways related to muscle development and function, supporting a muscle-specific biological signal captured by our signature (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eTo explore the potential cellular origin of putative mediators, we analysed single-cell RNA sequencing data generated from a subset of 16 patients from the MATCHR-I study (clinical characteristics reported in Supplementary Table S6). \u003cem\u003eIGFBP1\u003c/em\u003e expression was predominantly enriched in hepatocytes, together with ITIH3, thus validating in patients what was recently described in murine models\u003csup\u003e\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e\u003c/sup\u003e. In contrast, \u003cem\u003eIGFBP2\u003c/em\u003e expression was mainly restricted to a subset of tumor cells. \u003cem\u003eIL6\u003c/em\u003e did not map clearly to a specific cellular compartment, a finding that may reflect its multifocal origin, including potential production by skeletal muscle, a compartment not represented in tumor biopsies (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e and supplementary Fig.\u0026nbsp;3).\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec22\" class=\"Section2\"\u003e \u003ch2\u003eFunctional validation of IGFBP-mediated impairment of myogenic differentiation\u003c/h2\u003e \u003cp\u003eAn essential feature of skeletal muscle homeostasis is its capacity for self-repair. Muscle-resident myoblasts spontaneously differentiate to regenerate damaged fibers, a process that is impaired in sarcopenia\u003csup\u003e\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e\u003c/sup\u003e. We therefore investigated whether proteins identified through our proteomic screen directly interfere with myogenic differentiation in vitro.\u003c/p\u003e \u003cp\u003eImmortalized human myoblasts were induced to differentiate and exposed to increasing concentrations of IGFBP1, IGFBP2, or IL-6, as detailed in the Methods. After four days of differentiation, myogenic progression was assessed by quantifying the fusion index (FI) and myosin heavy chain (MHC) expression using immunofluorescence microscopy and western blotting.\u003c/p\u003e \u003cp\u003eAddition of IGFBP1 or IGFBP2 to the culture medium resulted in a marked, dose-dependent impairment of myogenic differentiation, with a significant reduction in fusion index (1.5-fold and 2-fold, respectively; Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eC\u0026ndash;D) and decreased MHC protein expression (2-fold and 4-fold, respectively; Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eE\u0026ndash;F). Given that IGFBPs can exert both IGF-1\u0026ndash;dependent and IGF-1\u0026ndash;independent biological effects\u0026sup2;\u0026sup1;, we next evaluated whether increasing IGF-1 availability could rescue this phenotype. Indeed, supplementation with increasing concentrations of IGF-1 fully restored myoblast differentiation in the presence of both IGFBP1 and IGFBP2, supporting a predominantly IGF-1\u0026ndash;dependent mechanism whereby IGFBPs sequester IGF-1 and suppress its pro-myogenic signaling (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eF\u0026ndash;J).\u003c/p\u003e \u003cp\u003eConsistent with these phenotypic observations, transcriptomic profiling of differentiating myoblasts exposed to IGFBP1 or IGFBP2 revealed a robust suppression of gene programs related to muscle differentiation and function, including the \u003cem\u003eMyogenesis\u003c/em\u003e hallmark and Gene Ontology biological processes associated with muscle contraction and muscle cell development. Importantly, these transcriptional alterations were largely attenuated by concomitant IGF-1 supplementation, further supporting a central role for impaired IGF signaling in mediating the observed differentiation defects.\u003c/p\u003e \u003cp\u003eIn contrast, exposure to IL-6 did not impair myoblast differentiation or MHC expression, indicating that IL-6 does not act as a direct inhibitor of myogenesis in this experimental context (Supplementary Fig.\u0026nbsp;4). Based on prior evidence indicating that IL-6 can induce hepatic IGFBP1 expression\u003csup\u003e\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e\u003c/sup\u003e, we next explored this axis in liver-derived cells. IL-6 exposure led to increased IGFBP1 expression in HepG2 cells, accompanied by elevated levels of secreted IGFBP1 in the culture supernatant, supporting an indirect mechanism whereby systemic inflammation may promote muscle dysfunction through liver-mediated modulation of the IGF axis rather than through direct effects on muscle cells (Supplementary Fig.\u0026nbsp;4).\u003c/p\u003e \u003c/div\u003e"},{"header":"Discussion","content":"\u003cp\u003eIn this study, we demonstrate that plasma proteomics can capture a biologically meaningful representation of physical function and muscle mass in oncology patients, offering a potentially scalable and objective alternative to conventional clinical assessments. By focusing on proteins enriched in non-sarcopenic individuals, we identified a proteomic signature reflective of preserved systemic and neuromuscular integrity, rather than advanced cancer-related wasting. This approach enabled the delineation of patient subgroups with distinct muscle masss and functional phenotypes based on molecular features, providing biological resolution that is not achievable using performance status alone.\u0026nbsp;Importantly, our approach estimates a continuous probability of sarcopenia-related muscle dysfunction, rather than assigning a binary diagnosis, reflecting the biological heterogeneity and absence of a single gold-standard definition in oncology.\u003c/p\u003e\n\u003cp\u003eDespite major advances in precision oncology, patient stratification in clinical practice remains largely tumor-centric, with limited incorporation of the host systemic state and its biological determinants.\u0026nbsp;Skeletal muscle mass is increasingly assessed using imaging-based approaches, yet these measures are inconsistently implemented, and provide limited insight into the underlying biology of muscle dysfunction. Functional status is routinely assessed using coarse clinical scales such as the Eastern Cooperative Oncology Group performance status (ECOG PS), which suffer from moderate interobserver agreement\u003csup\u003e24\u003c/sup\u003e, limited resolution, particularly within intermediate categories\u003csup\u003e25\u003c/sup\u003e, and also lack of biological interpretability. Consequently, the molecular processes underlying muscle loss, functional decline and sarcopenia in patients with cancer remain poorly characterized and difficult to quantify at scale.\u0026nbsp;This limitation has important implications for interventional studies in cancer sarcopenia, including trials of exercise, nutritional, or pharmacological interventions, which critically rely on robust, objective, and scalable endpoints. The lack of biologically grounded surrogate markers of muscle function and repair has contributed to heterogeneous trial designs and inconclusive results across the field.\u003c/p\u003e\n\u003cp\u003eA notable aspect of the identified signature is the enrichment of proteins linked to neuronal and neuromuscular biology. In particular, CBLN4 and CNTN3 emerged as major contributors to the model. Both proteins are involved in synaptic organization and plasticity, suggesting that preserved physical function in patients with cancer may depend not only on muscle-intrinsic properties but also on intact neuromuscular and neuronal signaling. This observation aligns with growing evidence that muscle aging and sarcopenia are influenced by alterations at the neuromuscular junction\u003csup\u003e26\u003c/sup\u003e and by central neural regulation\u003csup\u003e27\u003c/sup\u003e,\u003csup\u003e28\u003c/sup\u003e. Moreover, physical exercise, one of the most effective interventions to preserve muscle mass and function, is known to induce robust synaptic and neuronal plasticity\u003csup\u003e29\u003c/sup\u003e, further supporting a biological link between neuromuscular connectivity and functional status. This is also consistent with recent data linking muscle mass and magnetic resonance – assessed brain age\u003csup\u003e30\u003c/sup\u003e. Although our study was not designed to directly assess central nervous system (CNS) involvement, these findings raise the possibility that plasma proteomics may indirectly capture systemic correlates of neuro-muscular health.\u003c/p\u003e\n\u003cp\u003eIn this context, plasma proteomic signatures capturing neuromuscular and systemic integrity may represent attractive surrogate endpoints to monitor biological responses to exercise-based or multimodal interventions, particularly when imaging- or strength-based assessments are impractical or insufficiently sensitive.\u003c/p\u003e\n\u003cp\u003eBeyond descriptive stratification, our analyses identified candidate mediators of cancer-related sarcopenia. Among the most consistently correlated proteins across cohorts were IGFBP1 and IGFBP2, both key regulators of the insulin-like growth factor (IGF) axis. While liver-derived IGFBP1 has recently been implicated in muscle wasting\u003csup\u003e21\u003c/sup\u003e, our data extend this paradigm by identifying tumor-derived IGFBP2 as an active contributor to muscle dysfunction. Functional experiments confirmed that IGFBP2 impairs myoblast differentiation, thereby limiting muscle repair and regeneration, an effect that could be reversed by increasing IGF1 availability. These findings support a model in which tumors actively promote sarcopenia through endocrine or paracrine modulation of IGF signaling, reinforcing the concept of a bidirectional tumor host interaction.\u003c/p\u003e\n\u003cp\u003eThe broader proteomic landscape associated with the signature further highlights the systemic nature of cancer-related sarcopenia. Several correlated proteins are involved in inflammatory signaling and acute-phase responses, including IL-6, MDK\u003csup\u003e31\u003c/sup\u003e, and C9, consistent with the established role of chronic inflammation in muscle wasting. Others have been linked to insulin resistance at the muscular level, a metabolic alteration increasingly recognized as a contributor to sarcopenia development. Additional proteins associated with extracellular matrix remodeling, tumor proliferation, and invasion likely reflect aggressive tumor biology, which may indirectly exacerbate functional decline through sustained systemic stress rather than direct effects on muscle tissue. Together, these findings suggest that sarcopenia in cancer emerges from the convergence of inflammatory, metabolic, and tumor-derived signals, resembling an accelerated state of inflammaging.\u003c/p\u003e\n\u003cp\u003eNotably, the incomplete overlap between sarcopenia and cachexia observed in the external validation cohort reinforces the concept that they represent overlapping but distinct biological conditions. This distinction highlights the limitation of weight-based definitions and supports the need for muscle-specific, biology-driven stratification approaches.\u003c/p\u003e\n\u003cp\u003eIn clinical practice, such a blood-based signature could complement imaging and functional assessments by enabling scalable screening, longitudinal monitoring, and early identification of patients who may benefit from targeted interventions. Beyond its biological insights, our findings illustrate how molecular phenotyping may support the implementation of precision nutrition concepts in oncology. As recently proposed, precision nutrition extends beyond personalization based on static characteristics and relies on the integration of molecular profiling, advanced analytics, and clinically meaningful outcomes to guide targeted interventions\u003csup\u003e32\u003c/sup\u003e. By combining plasma proteomics with machine learning, our approach enables scalable, biologically informed stratification of muscle health and physical function, thereby bridging mechanistic insight with clinical applicability. This framework is consistent with emerging precision nutrition paradigms and may help identify patients most likely to benefit from tailored nutritional, exercise, or anti-inflammatory strategies, rather than relying on one-size-fits-all recommendations.\u003c/p\u003e\n\u003cp\u003eThis study has several strengths, including the use of a large, multi-cancer cohort and a broad plasma proteomic panel, supporting the generalizability of the identified biology across tumor types. These results indicate that cancer-related sarcopenia is driven by shared systemic mechanisms rather than tumor-specific processes. Limitations include the absence of direct measurements of muscle strength; however, it is noteworthy that most sarcopenia studies rely primarily on radiological assessments of muscle mass. We mitigated this limitation by incorporating performance status as a functional proxy and by training the model on patients with highly contrasted clinical and radiological phenotypes.\u003c/p\u003e\n\u003cp\u003eFuture studies integrating matched tumor, muscle, and plasma samples will be essential to further dissect the causal pathways linking tumor biology to systemic functional decline. In particular, coupling peripheral proteomics with direct assessments of muscle and CNS biology may provide deeper insight into the neuro-muscular and inflammatory circuits that underpin cancer-related sarcopenia.\u0026nbsp;Ultimately, this integrative framework could enable biologically informed patient stratification and support the development of targeted interventions aimed at preserving physical function in oncology patients. More broadly, the identification of plasma-based, mechanism-linked signatures of sarcopenia may help establish objective and scalable surrogate endpoints for future clinical trials in cancer cachexia and sarcopenia, including exercise-based interventions, thereby accelerating therapeutic development in a field long hindered by the lack of robust biomarkers.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u003cbr\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003eEthics approval and consent to participate\u003c/p\u003e\n\u003cp\u003eThe MATCH-R study was approved by the Comit\u0026eacute; de Protection des Personnes (CPP) and by the Agence nationale de s\u0026eacute;curit\u0026eacute; du m\u0026eacute;dicament et des produits de sant\u0026eacute; (ANSM), and was conducted in accordance with the principles of the Declaration of Helsinki and the Guideline for Good Clinical Practice. Written informed consent was obtained from all patients prior to inclusion in the study.\u003cbr\u003e\u0026nbsp;EudraCT number: 2014-A01147-40; CPP dossier reference: Am7501-4-3183.\u003c/p\u003e\n\u003cp\u003eConsent for publication\u003c/p\u003e\n\u003cp\u003eNon applicable as no individual data are included in the publication.\u003c/p\u003e\n\u003cp\u003eAvailability of data and materials\u003c/p\u003e\n\u003cp\u003eThe clinical and proteomic data generated in this study contain potentially identifiable personal information and are therefore subject to data protection regulations, including the General Data Protection Regulation (GDPR). For this reason, the datasets are not publicly available. Access to anonymized data may be considered upon reasonable request to the corresponding authors, subject to institutional approval and applicable regulatory constraints.\u003c/p\u003e\n\u003cp\u003eThe code used for model development and analysis is available from the corresponding authors upon reasonable request, subject to institutional and data protection regulations.\u003c/p\u003e\n\u003cp\u003eCompeting interests\u003c/p\u003e\n\u003cp\u003eFabrice Barlesi reports institutional relationships (no personal financial interests) with AbbVie, ACEA, Amgen, AstraZeneca, Bayer, Bristol Myers Squibb, Boehringer Ingelheim, Eisai, Eli Lilly Oncology, F. Hoffmann\u0026ndash;La Roche Ltd, Genentech, Ipsen, Ignyta, Innate Pharma, Loxo, Novartis, MedImmune, Merck, MSD, Pierre Fabre, Pfizer, Sanofi-Aventis, Summit Therapeutics and Takeda.\u003c/p\u003e\n\u003cp\u003eCaroline Even reports consulting or advisory roles with Innate Pharma, Bristol Myers Squibb, MSD Oncology, Merck Serono, Novartis, F-star Therapeutics, Merus and GlaxoSmithKline (institutional), and travel, accommodation or expenses from MSD Oncology and Merck Serono.\u003c/p\u003e\n\u003cp\u003eNathalie Lassau reports participation on an advisory board for Jazz Pharmaceuticals.\u003c/p\u003e\n\u003cp\u003eYohann Loriot reports honoraria from Janssen, Bristol Myers Squibb, Roche, Gilead, MSD and Pfizer; institutional research funding from Amgen, Janssen Oncology, MSD Oncology, Lilly, AstraZeneca, Orion, Exelixis, Incyte, Pfizer, Sanofi, Astellas Pharma, Gilead Sciences, Merck KGaA, Taiho Pharmaceutical, Bristol Myers Squibb, Roche and Tyra Biosciences; and travel, accommodation or expenses from Astellas Pharma, Pfizer, MSD Oncology and AstraZeneca.\u003c/p\u003e\n\u003cp\u003eAntoine Italiano reports research grants from AstraZeneca, Bayer, Bristol Myers Squibb, Merck, MSD and Pharmamar.\u003c/p\u003e\n\u003cp\u003eCarla M. Prado has previously received honoraria and/or paid consultancy from Abbott Nutrition, Nutricia, Nestl\u0026eacute; Health Science and Novo Nordisk.\u003c/p\u003e\n\u003cp\u003eMariam Jamal-Hanjani reports consulting for Astex Pharmaceuticals, Pfizer and Achilles Therapeutics; membership on the Scientific Advisory Board and Steering Committee of Achilles Therapeutics; and speaker honoraria from Pfizer, Astex Pharmaceuticals, Oslo Cancer Cluster, Bristol Myers Squibb and Genentech.\u003c/p\u003e\n\u003cp\u003eBenjamin Besse reports institutional honoraria and speaker\u0026rsquo;s bureau participation for AbbVie, AstraZeneca, Chugai Pharmaceutical, Daiichi Sankyo, Hedera Dx, Janssen, Merck Sharp \u0026amp; Dohme, Roche, Sanofi Aventis and Springer Healthcare Ltd; consulting or advisory roles (institutional) for AbbVie, BioNTech SE, Bristol Myers Squibb, Chugai Pharmaceutical, CureVac AG, Daiichi Sankyo, F. Hoffmann\u0026ndash;La Roche Ltd, Pharmamar, Regeneron, Sanofi Aventis and Turning Point Therapeutics; and institutional research funding from AstraZeneca, BeiGene, Genmab A/S, GlaxoSmithKline, Janssen, Merck Sharp \u0026amp; Dohme, Ose Immunotherapeutics, Pharmamar, Roche-Genentech, Sanofi and Takeda.\u003c/p\u003e\n\u003cp\u003eAll other authors declare that they have no competing interests.\u003c/p\u003e\n\u003cp\u003eFunding\u003c/p\u003e\n\u003cp\u003eThis work was partially supported by Canceropole \u0026Icirc;le-de-France (grant number 2024-1-EMERG-06) and by the Fondation Gustave Roussy.\u003c/p\u003e\n\u003cp\u003eAuthors\u0026rsquo; contributions\u003c/p\u003e\n\u003cp\u003eFGDO, YV, and BB conceived and designed the study.\u003cbr\u003e\u0026nbsp;WSZ and MA performed the statistical analyses.\u003cbr\u003e\u0026nbsp;XS conducted the in vitro experiments and functional assays.\u003cbr\u003e\u0026nbsp;LL, DB, and NL contributed to imaging analyses and radiological data interpretation., CB, YL and KB contributed to proteomic data analysis and biological interpretation.\u003cbr\u003e\u0026nbsp;FC, PB, RI, MG, DC, CN, MNC contributed to data collection, clinical annotation, and sample processing.\u003cbr\u003e\u0026nbsp;CP and MJH provided critical expertise and contributed to data interpretation.\u003cbr\u003e\u0026nbsp;FB, CE, AI, YV, and BB supervised the study.\u003cbr\u003e\u0026nbsp;FGDO drafted the manuscript. All authors critically revised the manuscript and approved the final version.\u003c/p\u003e\n\u003cp\u003eAcknowledgements\u003c/p\u003e\n\u003cp\u003eThe authors thank Dr. V. Mouly (Institute of Myology, Paris, France) for providing the human immortalized myoblast cell lines LHCN and AB1190, and Dr. S. Tapscott (Fred Hutchinson Cancer Center, Seattle, USA) for the MB135 myoblast cell line.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n \u003cli\u003eKirk B, Cawthon PM, Arai H, et al. The Conceptual Definition of Sarcopenia: Delphi Consensus from the Global Leadership Initiative in Sarcopenia (GLIS). \u003cem\u003eAge Ageing\u003c/em\u003e. 2024;53(3):afae052. doi:10.1093/ageing/afae052\u003c/li\u003e\n \u003cli\u003eCouderc AL, Liuu E, Boudou-Rouquette P, et al. Pre-Therapeutic Sarcopenia among Cancer Patients: An Up-to-Date Meta-Analysis of Prevalence and Predictive Value during Cancer Treatment. \u003cem\u003eNutrients\u003c/em\u003e. 2023;15(5):1193. doi:10.3390/nu15051193\u003c/li\u003e\n \u003cli\u003eKiss N, Prado CM, Daly RM, et al. Low muscle mass, malnutrition, sarcopenia, and associations with survival in adults with cancer in the UK Biobank cohort. \u003cem\u003eJ Cachexia Sarcopenia Muscle\u003c/em\u003e. 2023;14(4):1775-1788. doi:10.1002/jcsm.13256\u003c/li\u003e\n \u003cli\u003eWilliams GR, Dunne RF, Giri S, Shachar SS, Caan BJ. Sarcopenia in the Older Adult With Cancer. \u003cem\u003eJ Clin Oncol\u003c/em\u003e. 2021;39(19):2068-2078. doi:10.1200/JCO.21.00102\u003c/li\u003e\n \u003cli\u003eCruz-Jentoft Aj, Bahat G, Bauer J, et al. Sarcopenia: revised European consensus on on definition and diagnosis. \u003cem\u003eAge and ageing\u003c/em\u003e. 2019;48(1). doi:10.1093/ageing/afy169\u003c/li\u003e\n \u003cli\u003eAl-Sawaf O, Weiss J, Skrzypski M, et al. Body composition and lung cancer-associated cachexia in TRACERx. \u003cem\u003eNature medicine\u003c/em\u003e. 2023;29(4). doi:10.1038/s41591-023-02232-8\u003c/li\u003e\n \u003cli\u003eCharg\u0026eacute; SBP, Rudnicki MA. Cellular and molecular regulation of muscle regeneration. \u003cem\u003ePhysiol Rev\u003c/em\u003e. 2004;84(1):209-238. doi:10.1152/physrev.00019.2003\u003c/li\u003e\n \u003cli\u003ePeake JM, Neubauer O, Della Gatta PA, Nosaka K. Muscle damage and inflammation during recovery from exercise. \u003cem\u003eJ Appl Physiol (1985)\u003c/em\u003e. 2017;122(3):559-570. doi:10.1152/japplphysiol.00971.2016\u003c/li\u003e\n \u003cli\u003eMukund K, Subramaniam S. Skeletal muscle: A review of molecular structure and function, in health and disease. \u003cem\u003eWiley Interdiscip Rev Syst Biol Med\u003c/em\u003e. 2020;12(1):e1462. doi:10.1002/wsbm.1462\u003c/li\u003e\n \u003cli\u003eZullo A, Mancini FP, Schleip R, Wearing S, Yahia L, Klingler W. The interplay between fascia, skeletal muscle, nerves, adipose tissue, inflammation and mechanical stress in musculo-fascial regeneration. \u003cem\u003eJ Gerontology Geriatrics\u003c/em\u003e. 2017;65(4):271-283.\u003c/li\u003e\n \u003cli\u003eLafuste P, Relaix F. Endocrine control of skeletal muscle regeneration and clinical applications. \u003cem\u003eNat Rev Endocrinol\u003c/em\u003e. 2026;22(1):5-6. doi:10.1038/s41574-025-01205-w\u003c/li\u003e\n \u003cli\u003eVasseur D, Bigot L, Beshiri K, et al. Deciphering resistance mechanisms in cancer: final report of MATCH-R study with a focus on molecular drivers and PDX development. \u003cem\u003eMol Cancer\u003c/em\u003e. 2024;23(1):221. doi:10.1186/s12943-024-02134-4\u003c/li\u003e\n \u003cli\u003eMenssouri N, Poiraudeau L, Helissey C, et al. Genomic Profiling of Metastatic Castration-Resistant Prostate Cancer Samples Resistant to Androgen Receptor Pathway Inhibitors. \u003cem\u003eClin Cancer Res\u003c/em\u003e. 2023;29(21):4504-4517. doi:10.1158/1078-0432.CCR-22-3736\u003c/li\u003e\n \u003cli\u003ePrado CMM, Lieffers JR, McCargar LJ, et al. Prevalence and clinical implications of sarcopenic obesity in patients with solid tumours of the respiratory and gastrointestinal tracts: a population-based study. \u003cem\u003eLancet Oncol\u003c/em\u003e. 2008;9(7):629-635. doi:10.1016/S1470-2045(08)70153-0\u003c/li\u003e\n \u003cli\u003eDecazes P, Ammari S, Belkouchi Y, et al. Synergic prognostic value of 3D CT scan subcutaneous fat and muscle masses for immunotherapy-treated cancer. \u003cem\u003eJ Immunother Cancer\u003c/em\u003e. 2023;11(9):e007315. doi:10.1136/jitc-2023-007315\u003c/li\u003e\n \u003cli\u003eProteomics. Tissue-based map of the human proteome - PubMed. Accessed January 3, 2025. https://pubmed.ncbi.nlm.nih.gov/25613900/\u003c/li\u003e\n \u003cli\u003eChen T, Guestrin C. XGBoost: A Scalable Tree Boosting System. Preprint posted online June 10, 2016. doi:10.1145/2939672.2939785\u003c/li\u003e\n \u003cli\u003eWang G, Biswas AK, Ma W, et al. Metastatic cancers promote cachexia through ZIP14 upregulation in skeletal muscle. \u003cem\u003eNat Med\u003c/em\u003e. 2018;24(6):770-781. doi:10.1038/s41591-018-0054-2\u003c/li\u003e\n \u003cli\u003eGueugneau M, d\u0026rsquo;Hose D, Barb\u0026eacute; C, et al. Increased Serpina3n release into circulation during glucocorticoid-mediated muscle atrophy. \u003cem\u003eJ Cachexia Sarcopenia Muscle\u003c/em\u003e. 2018;9(5):929-946. doi:10.1002/jcsm.12315\u003c/li\u003e\n \u003cli\u003eSchroeter CB, Nelke C, Stascheit F, et al. Inter-alpha-trypsin inhibitor heavy chain H3 is a potential biomarker for disease activity in myasthenia gravis. \u003cem\u003eActa Neuropathol\u003c/em\u003e. 2024;147(1):102. doi:10.1007/s00401-024-02754-6\u003c/li\u003e\n \u003cli\u003eKaltenecker D, Fisker Schmidt S, Weber P, et al. Functional liver genomics identifies hepatokines promoting wasting in cancer cachexia. \u003cem\u003eCell\u003c/em\u003e. 2025;188(17):4549-4566.e22. doi:10.1016/j.cell.2025.06.039\u003c/li\u003e\n \u003cli\u003eBlau HM, Cosgrove BD, Ho ATV. The central role of muscle stem cells in regenerative failure with aging. \u003cem\u003eNat Med\u003c/em\u003e. 2015;21(8):854-862. doi:10.1038/nm.3918\u003c/li\u003e\n \u003cli\u003eLukowski SW, Fish RJ, Martin-Levilain J, et al. Integrated analysis of mRNA and miRNA expression in response to interleukin-6 in hepatocytes. \u003cem\u003eGenomics\u003c/em\u003e. 2015;106(2):107-115. doi:10.1016/j.ygeno.2015.05.001\u003c/li\u003e\n \u003cli\u003eBlagden SP, Charman SC, Sharples LD, Magee LRA, Gilligan D. Performance status score: do patients and their oncologists agree? \u003cem\u003eBr J Cancer\u003c/em\u003e. 2003;89(6):1022-1027. doi:10.1038/sj.bjc.6601231\u003c/li\u003e\n \u003cli\u003eS\u0026oslash;rensen JB, Klee M, Palshof T, Hansen HH. Performance status assessment in cancer patients. An inter-observer variability study. \u003cem\u003eBr J Cancer\u003c/em\u003e. 1993;67(4):773-775. doi:10.1038/bjc.1993.140\u003c/li\u003e\n \u003cli\u003eKedlian VR, Wang Y, Liu T, et al. Human skeletal muscle aging atlas. \u003cem\u003eNat Aging\u003c/em\u003e. 2024;4(5):727-744. doi:10.1038/s43587-024-00613-3\u003c/li\u003e\n \u003cli\u003eZhu XA, Starosta S, Ferrer M, et al. A neuroimmune circuit mediates cancer cachexia-associated apathy. \u003cem\u003eScience\u003c/em\u003e. 2025;388(6743):eadm8857. doi:10.1126/science.adm8857\u003c/li\u003e\n \u003cli\u003eSimoes E, Uchida R, Nucci MP, et al. Cachexia Alters Central Nervous System Morphology and Functionality in Cancer Patients. \u003cem\u003eJ Cachexia Sarcopenia Muscle\u003c/em\u003e. 2025;16(1):e13742. doi:10.1002/jcsm.13742\u003c/li\u003e\n \u003cli\u003ede Sousa Fernandes MS, Ord\u0026ocirc;nio TF, Santos GCJ, et al. Effects of Physical Exercise on Neuroplasticity and Brain Function: A Systematic Review in Human and Animal Studies. \u003cem\u003eNeural Plast\u003c/em\u003e. 2020;2020:8856621. doi:10.1155/2020/8856621\u003c/li\u003e\n \u003cli\u003eRaji CA, Meysami S, Lee S, et al. Higher Muscle Volume is Inversely Related to Chronological and Brain Age While Increased Visceral to Muscle Fat Ratio is Positively Related to Chronological and Brain Age. \u003cem\u003eAlzheimer\u0026rsquo;s \u0026amp; Dementia\u003c/em\u003e. 2025;21(S8):e110051. doi:10.1002/alz70862_110051\u003c/li\u003e\n \u003cli\u003eIkutomo M, Sakakima H, Matsuda F, Yoshida Y. Midkine-deficient mice delayed degeneration and regeneration after skeletal muscle injury. \u003cem\u003eActa Histochem\u003c/em\u003e. 2014;116(2):319-326. doi:10.1016/j.acthis.2013.08.009\u003c/li\u003e\n \u003cli\u003eda Silva BR, Brennan L, Horst MA, Wishart DS, Prado CM. Advancing precision nutrition: bridging mechanistic insight and clinical implementation. \u003cem\u003eNat Rev Endocrinol\u003c/em\u003e. 2025;21(9):515-517. doi:10.1038/s41574-025-01141-9\u003c/li\u003e\n\u003c/ol\u003e"},{"header":"Table","content":"\u003cp\u003e\u003cstrong\u003eTable 1. Clinical characteristics of the MATCH-R cohorts used for model development and validation.\u0026nbsp;\u003c/strong\u003eBaseline demographic, clinical, and disease characteristics of patients included in the training cohort (MATCH-R immunotherapy) and in the validation cohort (MATCH-R metastatic castration-resistant prostate cancer). Continuous variables are reported as median (interquartile range), and categorical variables as number (percentage).\u003c/p\u003e\n\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\" width=\"520\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd rowspan=\"2\" valign=\"top\" style=\"width: 208px;\"\u003e\n \u003cp\u003eCharacteristic\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 156px;\"\u003e\n \u003cp\u003eTraining\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 156px;\"\u003e\n \u003cp\u003emCRPC validation\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eSamples = 99\u003csup\u003e1\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eSamples = 88\u003csup\u003e1\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eAge\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e64 (56, 71)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e69 (65, 77)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eSex\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp; \u0026nbsp; Female\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e36 (36%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e0 (0%)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp; \u0026nbsp; Male\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e63 (64%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e88 (100%)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eSite\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp; \u0026nbsp; Bladder cancer\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e12 (12%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e0 (0%)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp; \u0026nbsp; Colon cancer MSI\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e3 (3.0%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e0 (0%)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp; \u0026nbsp; NSCLC adenocarcinoma\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e66 (67%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e0 (0%)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 208px;\"\u003e\n \u003cp\u003e\u0026nbsp; \u0026nbsp; NSCLC undifferentiated/ NOS\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 156px;\"\u003e\n \u003cp\u003e4 (4.0%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 156px;\"\u003e\n \u003cp\u003e0 (0%)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp; \u0026nbsp; NSCLC squamous\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e11 (11%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e0 (0%)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp; \u0026nbsp; mCRPC\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e1 (1.0%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e88 (100%)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp; \u0026nbsp; Melanome\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e1 (1.0%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e0 (0%)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp; \u0026nbsp; Thyroid\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e1 (1.0%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e0 (0%)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eECOG Performance status\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp; \u0026nbsp; 0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e30 (30%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e26 (46%)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp; \u0026nbsp; 1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e42 (42%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e19 (33%)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp; \u0026nbsp; \u0026ge; 2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e27 (27%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e12 (21%)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp; \u0026nbsp; Unknown\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e31\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eSarcopenia score\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp; \u0026nbsp; Probability \u0026ge; 50%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e61 (62%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e13 (15%)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp; \u0026nbsp; Probability \u0026lt; 50%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e38 (38%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e75 (85%)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd colspan=\"2\" valign=\"top\" style=\"width: 364px;\"\u003e\n \u003cp\u003e\u003csup\u003e1\u003c/sup\u003e Median (Q1, Q3); n (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 156px;\"\u003e\n \u003cp\u003e\u003csup\u003e\u0026nbsp;\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"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":"molecular-cancer","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"molc","sideBox":"Learn more about [Molecular Cancer](http://gsejournal.biomedcentral.com/)","snPcode":"12943","submissionUrl":"https://submission.nature.com/new-submission/12943/3","title":"Molecular Cancer","twitterHandle":"@SN_Oncology","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"BMC/SO AJ","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"","lastPublishedDoi":"10.21203/rs.3.rs-8885139/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-8885139/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003ch2\u003eBackground\u003c/h2\u003e \u003cp\u003eCancer-related sarcopenia is associated with poor clinical outcomes but remains difficult to define and quantify in routine oncology practice. Current assessments rely on imaging and functional scales that are time-consuming and provide limited biological insight. We aimed to identify a plasma proteomic signature of cancer-related sarcopenia and to uncover circulating mediators involved in its pathophysiology.\u003c/p\u003e\u003ch2\u003eMethods\u003c/h2\u003e \u003cp\u003ePatients were included from two cohorts of the MATCH-R study (NCT02517892): a discovery cohort of advanced cancer patients treated with immunotherapy and an independent validation cohort of metastatic castration-resistant prostate cancer (mCRPC) patients treated with androgen-receptor pathway inhibitors. External validation was performed in the TRACERx cohort of non\u0026ndash;small cell lung cancer. Skeletal muscle index at L3 was quantified using imaging, and ECOG performance status served as a functional proxy. Plasma proteomics was performed using the Olink Explore platform. An extreme gradient boosting (XGBoost) model was trained on a high-contrast subset using a neuromuscular-focused protein panel and validated across cohorts. Functional effects of candidate mediators were assessed in differentiating human myoblasts.\u003c/p\u003e\u003ch2\u003eResults\u003c/h2\u003e \u003cp\u003eThe model generated a continuous sarcopenia probability (SP) score that correlated with muscle mass and functional status and consistently stratified overall survival across cohorts. A reduced four-protein model retained comparable performance, supporting translational applicability. Proteins associated with SP included IGFBP1, IGFBP2, and IL6. IGFBP1 and IGFBP2 impaired myoblast differentiation, while IL6 induced IGFBP1 expression in liver cells.\u003c/p\u003e\u003ch2\u003eConclusions\u003c/h2\u003e \u003cp\u003ePlasma proteomics enables scalable and biologically informed assessment of cancer-related sarcopenia, identifies tumor\u0026ndash;host mediators of muscle dysfunction, and supports objective patient stratification for therapeutic intervention.\u003c/p\u003e","manuscriptTitle":"A plasma proteomic signature of cancer-related sarcopenia implicates the IGFBP axis in muscle dysfunction","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-03-05 17:47:17","doi":"10.21203/rs.3.rs-8885139/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2026-03-25T03:00:10+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-03-24T20:39:02+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-03-07T17:45:29+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"315661523313053902615930650157271275820","date":"2026-02-27T08:39:56+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"204674655411774778195558492279386976666","date":"2026-02-23T07:40:14+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2026-02-21T17:44:27+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2026-02-20T08:20:48+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2026-02-20T08:18:35+00:00","index":"","fulltext":""},{"type":"submitted","content":"Molecular Cancer","date":"2026-02-15T09:44:43+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"molecular-cancer","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"molc","sideBox":"Learn more about [Molecular Cancer](http://gsejournal.biomedcentral.com/)","snPcode":"12943","submissionUrl":"https://submission.nature.com/new-submission/12943/3","title":"Molecular Cancer","twitterHandle":"@SN_Oncology","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"BMC/SO AJ","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"9b0b6ec9-4976-4a46-baf0-d285afc2b294","owner":[],"postedDate":"March 5th, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[],"tags":[],"updatedAt":"2026-04-27T21:23:17+00:00","versionOfRecord":[],"versionCreatedAt":"2026-03-05 17:47:17","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-8885139","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-8885139","identity":"rs-8885139","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}