Enhanced tumor metastasis driven by the crosstalk between melanoma cells and tumor-associated macrophages: insights into the CXCL10/CXCR3-hnRNP C-MIF axis

Abstract

Background

Tumor-associated macrophages (TAMs) are critical drivers in melanoma growth and metastasis orchestrating complex cellular interactions within the tumor microenvironment (TME). While the RNA-binding protein heterogeneous nuclear ribonucleoprotein C1/C2 (hnRNP C) has been linked to melanoma progression and immune evasion, its specific role in macrophage-tumor communication and TME remodeling remains poorly understood.

Methods

Multiple publicly available melanoma datasets (TCGA-SKCM, GSE8401, GSE46517, and single-cell RNA sequencing) were analyzed to map hnRNP C-associated signaling pathways, and macrophage-tumor interactions. Ligand-receptor cell communication analysis was used to characterize networks within the TME. Functional studies were conducted in melanoma cell lines (Buf1379, Buf1402, Buf836) following siRNA-mediated hnRNP C knockdown using wound healing and trans-well assays to assess cell migration and invasion. Macrophage-tumor interactions were comprehensively studied via trans-well co-culture models, flow cytometry, cytokine profiling, and recombinant CXCL10 stimulation assays. Protein-level validation and clinical relevance were assessed using multicohort datasets and immunohistochemistry.

Results

Elevated hnRNP C expression significantly enhanced melanoma cell migration, invasion, and epithelial-mesenchymal transition (EMT)-associated programs. Mechanistically, hnRNP C-high melanoma cells secreted increased levels of the macrophage migration inhibitory factor (MIF), which reprogrammed macrophages toward an interferon-associated TAM (IFN-TAM) phenotype. Moreover, CXCL10 via CXCR3 activation upregulated hnRNP C expression in melanoma cells establishing a bidirectional feedback loop. Single-cell transcriptomic analyses confirmed enriched MIF-mediated signaling between macrophages and melanoma cells. Spatially, this macrophage-derived CXCL10/CXCR3-hnRNP C-MIF signaling axis promoted macrophage remodeling and drove metastatic tumor states. A three-gene prognostic signature (hnRNP C-MIF-CXCR3) showed robust predictive performance across independent melanoma cohorts and strongly correlated with tumor stage, metastatic status, and overall survival. Protein-level validation confirmed elevated hnRNP C and MIF expression in melanoma tissues compared with normal skin.

Conclusions

This study identifies a novel macrophage-tumor signaling circuit in melanoma, wherein CXCL10/CXCR3-driven hnRNP C activation promotes MIF-mediated macrophage reprogramming and subsequent metastatic progression. The identified three-gene signature provides a clinically relevant prognostic biomarker and highlights hnRNP C as a promising therapeutic target for disrupting macrophage-driven tumor progression.

WHAT IS ALREADY KNOWN ON THIS TOPIC

• The tumor microenvironment (TME), particularly tumor-associated macrophages (TAMs), is a central driver of melanoma progression and metastasis, regulating tumor cell migration and shaping immune suppression.

• Tumor cells dynamically reprogram macrophages through secreted cytokines and chemokines, altering macrophage phenotype and function through ligand-receptor signaling within the TME.

• The RNA-binding protein hnRNP C has been linked to tumor progression and poor clinical outcome across several tumor types; however, its specific role in melanoma cell motility and macrophage-tumor communication remains insufficiently defined.

WHAT THIS STUDY ADDS

• This study uncovers a previously uncharacterized macrophage-tumor signaling circuit in melanoma, in which TAM-derived CXCL10 activates CXCR3 signaling in tumor cells, driving hnRNP C upregulation and promoting migration, invasion, and epithelial-mesenchymal transition-like programs.

• hnRNP C-high melanoma cells secrete elevated levels of macrophage migration inhibitory factor (MIF), which promotes macrophage polarization toward an interferon-associated TAM (IFN-TAM) phenotype, reinforcing a pro-tumor, immune-suppressive, and macrophage-rich TME.

• Integrating single-cell transcriptomics, public melanoma cohorts, and functional validation, a CXCL10/CXCR3-hnRNP C-MIF signaling axis was established as a master governor of bidirectional macrophage-tumor crosstalk.

• A novel three-gene signature (hnRNP C-MIF-CXCR3) was developed, demonstrating robust cross-cohort performance and strong clinical associations with advanced tumor stage, metastatic progression, and poor overall survival.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• The CXCL10/CXCR3-hnRNP C-MIF axis provides a novel mechanistic framework for macrophage-driven melanoma progression offering concrete therapeutic targets for disrupting tumor-immune cell communication links.

• Targeting CXCR3 signaling, hnRNP C expression, or MIF-mediated macrophage reprogramming may provide a dual therapeutic strategy to limit tumor dissemination while simultaneously reversing the immune-suppressive TME.

• The three-gene signature represents a clinically relevant prognostic biomarker for melanoma, potentially improving future patient stratification approaches that integrate tumor-intrinsic and immune microenvironment features.

Background

The tumor microenvironment (TME) constitutes a highly complex and dynamic ecosystem composed of tumor cells and diverse stromal and immune cell populations that collectively govern tumor progression, metastatic dissemination, and therapeutic response.^{1 2} Interactions between tumor cells and non-malignant components frequently generate an immunosuppressive milieu characterized by pronounced cellular heterogeneity and functional plasticity, posing major challenges for effective cancer treatment.³ Among these components, tumor-associated macrophages (TAMs) represent one of the most abundant immune cell populations within the TME.⁴ TAMs contribute to multiple aspects of tumor progression, including maintenance of cancer stemness, promotion of angiogenesis, enhancement of tumor cell migration and invasion, and suppression of cytotoxic T cell activity. Although macrophages were historically categorized into pro-inflammatory (M1) and anti-inflammatory (M2) phenotypes,⁵ emerging evidence highlights extensive TAM plasticity within tumors, leading to the identification of distinct functional states, such as interferon-associated (IFN-)TAM, angiogenic (Angio-)TAM, and regulatory (Reg-)TAM phenotypes.⁶ Despite these advances, the specific molecular mechanisms by which TAMs orchestrate tumor cell behavior and metastatic progression remain incompletely understood.

The bidirectional crosstalk between TAMs and tumor cells has emerged as a critical determinant of tumorigenicity. For example, in colorectal cancer, tumor-derived extracellular vesicles induce PD-L1 in TAMs, thereby facilitating immune evasion.⁷ Similarly, CPEB3 modulates macrophage-tumor interaction through the IL-6/STAT3 signaling to suppress epithelial-mesenchymal transition (EMT).⁸ In ovarian cancer, tumor-derived UBR5 promotes TAM recruitment and activation, driving metastatic progression.⁹ These macrophage-mediated interactions are frequently mediated by various cytokines and growth factors, including vascular endothelial growth factor (VEGF),¹⁰ macrophage migration inhibitory factor (MIF),¹¹ and insulin growth factor-binding protein 2 (IGFBP2).¹² However, the complexity and context-dependent nature of these signaling networks remain to be fully elucidated.

Heterogeneous nuclear ribonucleoprotein C1/C2 (hnRNP C) is an RNA-binding protein (RBP) involved in pre-mRNA processing and regulation of gene expression.¹³ Recent studies have implicated hnRNP C in the pathogenesis of multiple malignancies,^{14 15} while our own previous work demonstrated its role in regulating the HLA class I pathway in melanoma.¹⁶ Moreover, hnRNP C has been associated with metastatic progression across multiple cancer types,^{17 18} including melanoma.¹⁹ However, its precise role in mediating tumor-macrophage interactions within the TME has not yet been explored.

In the present study, we investigated the role of hnRNP C in macrophage-tumor crosstalk in melanoma. By integrating public transcriptomic datasets, single-cell RNA sequencing (scRNA-seq) data, and functional in vitro experiments, we identified a novel feedback loop: macrophage-derived CXCL10 activates CXCR3 signaling in melanoma cells, upregulating hnRNP C expression. Elevated hnRNP C subsequently promotes the secretion of MIF, driving macrophage polarization toward an IFN-TAM phenotype. This CXCL10/CXCR3-hnRNP C-MIF signaling axis establishes a bidirectional communication loop that enhances melanoma cell migration, invasion, and metastatic potential. Multi-cohort clinical analyses demonstrated that this axis is enriched in metastatic lesions and correlates with disease progression. Collectively, these findings highlight hnRNP C as a key regulator of macrophage-driven melanoma progression, suggesting that the CXCR3-hnRNP C-MIF axis could serve as a clinically relevant biomarker and potential therapeutic target in melanoma.

Methods and materials

Cell culture

The human metastatic melanoma cell lines Buf1402, Buf1379, and Buf836 were kindly provided by Soldano Ferrone (Department of Surgery, Massachusetts General Hospital, Harvard Medical School, Boston, Massachusetts, USA). Cells were cultured in Roswell Park Memorial Institute 1640 medium (RPMI 1640, Invitrogen, Carlsbad, California, USA) supplemented with 1% penicillin/streptomycin (Sigma-Aldrich, Missouri, USA), 10% fetal bovine serum (FBS) (PAN, Aidenbach, Germany), and 1% L-glutamine (Lonza, Basel, Switzerland) at 37°C in a humidified atmosphere containing 5% CO₂. All cell lines were routinely checked for mycoplasma contamination by qPCR and were only used when negative.

Transfection of siRNA

To explore the function of hnRNP C, melanoma cells (2.4×10⁵ cells/well for 6-well plates, 1×10⁵ cells/well for 24-well or 1.8×10⁵ cells/well for 12-well plates) were transfected with 2 ng/mL siRNA specific for hnRNP C (si hnRNP C, EHU133931, Sigma-Aldrich, St. Louis, Missouri, USA) or a non-targeting negative control (NC, EHUEGFP, Sigma-Aldrich) using 9 µL lipofectamine (RNAiMAX, Invitrogen) according to the manufacturer’s instructions. Transfected cells were collected after 48 hours for subsequent RNA, protein, and functional analyses.

Wound healing and trans-well assays

A scratch wound healing assay was used to determine changes in melanoma cell migration on transfection with si hnRNP C or co-culture with macrophages. As recently described,²⁰ cell monolayers at approximately 90% confluence were scratched using a plastic pipette tip, and images were taken immediately after scratching as well as following culturing at 37°C for 2, 4, 24, 48, or 72 hours, respectively.

For trans-well migration assays, 5×10⁴ transfected cells per well were seeded in 100 µL culture medium containing 1% FBS into the upper chambers of 24-well trans-well inserts (8 µm pore size; Corning, USA), while 500 µL culture medium containing 10% FBS was added to the lower chambers. After 24 hours of incubation, migrated cells on the lower surface of the membrane were quantified.

For trans-well invasion assays, identical numbers of transfected cells were seeded into the upper chamber of trans-well inserts pre-coated with Matrigel/Geltrex LDEV-free basement membrane matrix (A1413201, Thermo Fisher Scientific, USA). After 24 hours incubation, non-invading cells remaining on the upper surface of the membrane were gently removed. For quantitative analysis, the lower side of the insert was incubated with 110 µL CellTiter-Glo 2.0 Reagent (G924A, Promega, Madison, USA) for 10 min, and bioluminescence was detected using a spectrophotometer/microplate reader (TECAN, Männedorf, Switzerland).

For representative microscopic documentation, parallel trans-well inserts from the identical experimental conditions were fixed with 4% paraformaldehyde for 15 min at room temperature and stained with 0.1% crystal violet for 15–20 min. Excess dye was removed by washing with distilled water, and the membranes were air-dried before imaging under a light microscope. Representative images were acquired under comparable imaging settings and are shown to document the migrated or invaded cells on the lower surface of the trans-well membranes.

RNA isolation and real-time quantitative reverse-transcription PCR

Total cellular RNA was isolated using the TRIzol reagent according to the manufacturer’s instructions. RNA was subjected to first strand cDNA synthesis (Thermo Scientific, Rockford, Illinois, USA), followed by real-time quantitative reverse-transcription PCR (RT-qPCR) amplification using SYBR qPCR master mix (Vazyme, Nanjing, PRC) and specific primers (online supplemental table S1). Target gene expression data were normalized to the mRNA levels of the housekeeping genes delta-amino levulinate synthase (ALSA1), glyceraldehyde-3-phosphate dehydrogenase (GAPDH), and β-actin (ACTB) and expressed as mean relative mRNA expression levels from three independent biological experiments.

Protein extraction and western blot analysis

Total protein was extracted from melanoma cells after transfection or co-culture using lysis buffer with protease and phosphatase inhibitors (Thermo Scientific). After quantification of protein concentrations with the Pierce BCA Protein Assay kit (Thermo Scientific), equal protein amounts (25 µg/lane) were resolved on Bolt 4 bis 12% protein gels (Invitrogen) and transferred to membranes via iBlot 2 Transfer Stacks (Invitrogen). Subsequently, membranes were blocked with 5% milk for 1 hour at room temperature and incubated overnight at 4°C with primary antibodies (Ab): anti-β-actin (1:2500, ab6276, Abcam, Cambridge, UK), anti-hnRNP C (1:1000, PA5-24221, Thermo Scientific), and anti-vimentin (1:1000, ab92547, Abcam). Membranes were subsequently incubated with a secondary horseradish peroxidase-conjugated goat anti-mouse/rabbit Ab (CST) for 1 hour at room temperature. Chemiluminescent signals from Elite ECL reagent (12757p, Sigma-Aldrich) were captured on an iBright750 imaging system (Invitrogen). Densitometric analysis was performed using ImageJ, and protein expression levels are expressed as arbitrary units normalized to β-actin.

Isolation of macrophages and co-culture with melanoma cells

Healthy donor buffy coat preparations were obtained from the Blood Bank of the University Hospital of the Martin Luther University Halle-Wittenberg, Halle, Germany. After gradient density purification of peripheral blood mononuclear cells, monocytes were isolated using CD14 microbeads (Miltenyi Biotec, Bergisch Gladbach, Germany) according to the manufacturer’s instructions. Purified monocytes were cultured in medium supplemented with macrophage colony-stimulating factor (M-CSF; 100 ng/mL, 574806, Biolegend, San Diego, California, USA) for 6 days to promote differentiation into M0 macrophages. To generate TAMs, the M0 macrophages were cultured for an additional 48 hours in conditioned medium (supernatant) collected from either transfected or parental melanoma cells, whereas normal medium was used as control.

M0 macrophages and TAMs were either immediately analyzed by flow cytometry for their polarized phenotypes or employed in co-culture experiments to investigate their crosstalk with melanoma cells. For the co-culture setup, Buf1379, Buf1402, and Buf836 cells were seeded in the bottom chamber of a 12-well trans-well system (0.4 µm pore size; Corning, USA). M0 macrophages or TAMs were added in the upper trans-well inserts in duplicate. One set of wells was used to harvest co-culture supernatants and extract mRNA and proteins from the melanoma cells for RT-qPCR and Western blot validations. The parallel set of wells was designated for wound healing assays by introducing a scratch wound in the melanoma layer 24 hours after initiating co-culture, followed by migration monitoring for an additional 24 hours.

For Ab neutralization assay, 2 µg/mL Ab targeting CXCL10 (MAB266), IL8 (MAB208), CXCR1 (MAB330), CXCR2 (MAB331), or CXCR3 (MAB160) (all from R&D Systems, Minneapolis, Minnesota, USA) were added to the co-cultures of melanoma cells and TAMs/M0 macrophages. For ligand stimulation experiments, recombinant CXCL10 (0.8 µg/mL 266-IP, R&D Systems, Minneapolis, Minnesota, USA) or phosphate-buffered saline (PBS) control were added to melanoma cells for 48 hours. Recombinant MIF (5 ng/mL 289-MF/CF, R&D Systems, Minneapolis, Minnesota, USA) was administered to M0 macrophages for 48 hours prior to phenotypic characterization by flow cytometry.

Phagocytosis assay

To evaluate macrophage-mediated phagocytosis of tumor cells, M0 macrophages were first stimulated for 48 hours with conditioned media collected from si hnRNP C, NC, or parental tumor cells. Following stimulation, macrophages were labeled with Carboxyfluorescein succinimidyl ester (CFSE) Dye (1:100, 565082, BD Biosciences, Heidelberg, Germany), while tumor cells were labeled with Cell Proliferation Dye eFluor 670 (1:1000, 65084085, Thermo Fisher Scientific, USA), both according to standard manufacturers’ instructions. Labeled macrophages and tumor cells were then washed and directly co-cultured at a 2:1 ratio (2×10⁴ macrophages and 1×10⁴ tumor cells) for 10 hours. Imaging was performed via fluorescence microscopy using a 10×objective to capture macrophages (green channel) and tumor cells (far-red channel). Phagocytosis was assessed by the presence of red fluorescent tumor-cell signals inside the CFSE-positive macrophages.

Flow cytometry

Macrophage surface marker expression was analyzed by flow cytometry using fluorochrome-conjugated Abs directed against CD86 (562999, BD Biosciences), CD68 (565595, BD Biosciences), CD163 (556018, BD Biosciences), CD206 (564062, BD Biosciences), and HLA-DR (130–123-843, Miltenyi Biotec). After incubation for 15 min at room temperature, the stained cells were acquired on a LSRFortessa flow cytometer (BD Biosciences), and data were analyzed by the FACS Diva (BD Biosciences) software. Viable single-cell macrophage populations were gated and evaluated for expression of surface markers using the corresponding negative controls to set the gating thresholds, which were maintained consistently across matched experimental groups. In addition to the percentages of positively gated cells, the mean fluorescence intensity (MFI) was also used as complementary readouts of marker expression intensity.

Cytokine determination and treatment

Cytokine profiles were compared using the Proteome Profiler Human XL Cytokine Array Kit (ARY022B, R&D Systems) to analyze supernatants (200 µL/sample) from: (1) melanoma cells alone versus co-cultured with TAMs and (2) M0 macrophages alone versus treated with melanoma-conditioned medium. Chemiluminescent blots were developed with Elite ECL (Sigma-Aldrich), captured on an iBright750 system (Invitrogen), and quantified via ImageJ. Target cytokines were validated using ELISA kits from Biolegend—CCL2 (438804), CXCL10 (439904), IL8 (431504), and MIF (438407)—and PeproTech for VEGF (900-TM10).

Transcriptomics and clinical data analysis

HnRNP C expression and the CXCR3-hnRNP C-MIF axis score were analyzed across melanoma datasets. hnRNP C expression was analyzed in primary and metastatic melanoma samples using the GSE8401 from the R2 platform, with immune infiltration assessed via TIMER2.0. The GSE8401 (training) and GSE46517 (validation), TCGA-SKCM data were analyzed in R for CXCR3-hnRNP C-MIF axis score, clinicopathological and overall survival correlations. Group comparisons were performed using Wilcoxon rank-sum or Kruskal-Wallis tests.

Logistic regression estimated odds ratio (ORs) with 95% confidence Interval (CI) for metastatic status. Receiver operating characteristic (ROC) curve analysis (pROC package, DeLong’s method) evaluated diagnostic performance via area under the curve, accuracy, sensitivity, specificity, and F1 score. Protein expression of hnRNP C and MIF was validated using immunohistochemical images and standardized scoring from The Human Protein Atlas.²¹ Staining intensity and the proportion of positive cells were evaluated according to standardized scoring criteria to validate transcriptomic findings at the protein level.

ScRNA-seq data were processed using the Seurat R package. Quality filtering excluded cells with fewer than 200 or more than 6000 detected genes, or with a mitochondrial transcript content >15%. Data were log-normalized, mitochondrial variation was regressed out, and the top 3000 highly variable genes were identified. Following PCA (top 30 components retained), batch effects were corrected using Harmony. Unsupervised clustering via the Louvain algorithm (resolution: 0.2) was visualized using Uniform Manifold Approximation and Projection (UMAP). Cluster-specific marker genes were identified via Seurat’s FindAllMarkers using the Wilcoxon rank-sum test (thresholds: >25% of cells, log2FC >0.25, adjusted p<0.05). Cell types were manually annotated using canonical markers. Cell–cell communication was inferred via CellChat (minimum 5 cells per population) to isolate MIF-associated and CXCL10/CXCR3-associated interactions. A 10× Genomics Xenium human primary melanoma dataset was analyzed using Seurat v5. Cells with <50 or >10 000 detected genes were excluded. Following log-normalization and variable feature selection, PCA embeddings (top 20 components) were used for UMAP visualization and Louvain clustering. Cluster markers were identified via Wilcoxon rank-sum test (>10% of cells, log2FC >0.25, adjusted p<0.05) for manual cell-type and regional annotation. Delineated tumor and macrophage zones were mapped for CXCL10, hnRNP C, MIF, and CD74 spatial distribution.

Statistical quantification

Functional data are presented as mean±SD from at least three independent biological replicates, with values normalized to the negative control. Statistical analyses were conducted in GraphPad Prism and R using Student’s t-test, Wilcoxon test, analysis of variance, or Kruskal-Wallis test. Two-sided p<0.05 indicated significance (∗p<0.05, p<0.01, ∗∗∗p<0.001, ∗∗∗∗p<0.0001).

Results

HnRNP C promotes melanoma migration and EMT-associated invasive phenotypes

To establish the conceptual framework of this study, we first proposed a working model integrating macrophage-tumor communication with clinical and single-cell validation (figure 1A). In this model, IFN-TAM secrete CXCL10, which activates CXCR3 signaling in melanoma cells, subsequently upregulating hnRNP C expression. Elevated hnRNP C, in turn, promotes tumor cell migration and MIF secretion, which contributes to macrophage reprogramming, thus forming a feed-forward tumor-macrophage crosstalk loop. Given that hnRNP C significantly influences the metastatic behavior across various tumor types,^22–26 we first analyzed with R2 genomic analysis tool data from the GSE8401 dataset, comprising 83 cases (31 primary vs 52 metastasis cases). HnRNP C expression levels were significantly higher in metastatic lesions compared with primary melanoma samples (figure 1B). To functionally validate these bioinformatics findings, the melanoma cell lines Buf1379 and Buf1402 were transfected with either a control (NC) or a hnRNP C-specific siRNA (si hnRNP C), generating models with low (hnRNP C^low) and high (hnRNP C^high) expression (online supplemental figure S1A,B). Silencing hnRNP C markedly decreased migration and invasion capabilities of the transfected melanoma cell lines as shown by scratch wound healing assays and trans-well assay (figure 1C–G and online supplemental figure S1C). To qualitatively document the trans-well based functional findings, representative microscopic images of crystal violet-stained migrated and invading cells were performed. Consistent with the quantitative results, hnRNP C knockdown markedly reduced the number of cells capable of passing the trans-well membrane in both Buf1379 and Buf1402 cells compared with the NC groups (figure 1H and online supplemental figure S1D). These representative images further support that hnRNP C contributes to the migratory and invasive phenotype of melanoma cells.

Figure 1. Discovery, validation, and functional characterization of the CXCL10/CXCR3-hnRNP C-MIF axis in melanoma metastasis (A) Schematic overview of the discovery pipeline and multilevel validation strategy for the CXCL10/CXCR3-hnRNP C-MIF axis. Mechanistic studies identified CXCL10 derived from IFN-activated TAMs as an upstream regulator of hnRNP C in melanoma cells via CXCR3 signaling, which promotes cell migration and MIF secretion. The axis was validated across multiple transcriptomic datasets (TCGA-SKCM, GSE46517, and GSE8401), protein expression data from The Human Protein Atlas, scRNA-seq datasets (GSE277165, GSE215120, and GSE269936), as well as experiments with cell lines. The integrated model illustrates a bidirectional tumor-macrophage feedback loop fostering metastatic progression, picture created in BioRender. (B) Comparison of hnRNP C expression in primary (green group, n=31) versus metastatic melanoma (red group, n=52) from the GSE8401 cohort using the R2 platform. (C) Representative wound healing images of Buf1379 melanoma cells left untreated (parental) or transfected with non-targeting control (NC) or hnRNP C-specific siRNA (si hnRNP C), captured at 24, 48, and 72 hours. (D, E) Quantification of relative migration ratios in Buf1379 (D) and Buf1402 (E) cells after hnRNP C knockdown compared with NC controls (mean±SD, n=3 biological replicates). (F, G) Trans-well migration (F) and invasion (G) assays showing reduced migratory capacity and decreased invasive potential following hnRNP C silencing compared with controls. (H) Representative images of crystal violet-stained cells from the trans-well migration assay in Buf1379 and Buf1402 cells after transfection with negative control siRNA (NC) or si hnRNP C. Scale bar: 100 μm. (I, J) EMT-related gene expression analysis in Buf1379 and Buf1402 cells following hnRNP C knockdown. mRNA levels of indicated EMT markers were determined by RT-qPCR and normalized to housekeeping genes (mean±SD, n=3 biological replicates). (K) Representative Western blot analysis of vimentin expression with β-actin as loading control. (L) Quantification of relative vimentin protein levels normalized to β-actin (arbitrary units, mean±SD, n=3 independent experiments). Data are presented as mean±SD. *p<0.05; **p<0.01. EMT, epithelial-mesenchymal transition; MIF, macrophage migration inhibitory factor; RT-qPCR, real-time quantitative reverse-transcription PCR; ROC, receiver operating characteristic; TAM, tumor-associated macrophage.

To explore the underlying mechanisms, the expression of EMT-related markers was evaluated. SiRNA-mediated hnRNP C knockdown resulted in a decrease of vimentin and snail2 mRNA levels (figure 1I,J), which was further confirmed at the protein level for vimentin (figure 1K,L). Taken together, these findings indicate that hnRNP C promotes melanoma metastasis by enhancing invasive and migratory phenotypes at least in part through modulating EMT-associated transcriptional programs. This conclusion was further supported by parallel functional experiments in a third independent melanoma cell line, Buf836, where hnRNP C knockdown reduced migration and invasion, and was accompanied by downregulated mRNA expression of EMT-associated markers like vimentin and snail2, with also reduced protein expression of vimentin (online supplemental figure S1E–K).

TAMs enhance melanoma migration and induce tumor hnRNP C expression

Deconvolution analysis of the TME cell composition within the melanoma dataset GSE8401 revealed a substantial abundance of infiltrating macrophages in both primary and metastatic tumors (online supplemental figure S2), implicating a potential role of macrophages in melanoma progression. To investigate potential crosstalk mechanisms, non-contact trans-well co-culture experiments were performed using human monocyte-derived M0 macrophages or TAMs (figure 2A). Co-culture significantly increased melanoma cell migration in both Buf1379 and Buf1402 cell lines. While unpolarized M0 macrophages moderately enhanced migration, TAMs induced a markedly more potent effect detectable as early as 2 hours post wounding (figure 2B–D and online supplemental figure S3A). As the trans-well system prevents directional contact, these findings suggest that macrophage-secreted soluble factors mediate this pro-migratory response. Notably, co-culture with TAMs significantly increased hnRNP C expression in melanoma cells at both the mRNA and protein levels (figure 2E–G). This was accompanied by increased expression of the EMT marker vimentin, whereas Snail2 and MMP2 mRNA levels remained consistent (figure 2H–K). These data indicate that TAM-derived signals promote melanoma migration at least partly via the induction of tumor-intrinsic hnRNP C. This mechanism was robustly replicated in a third melanoma cell line Buf836, where TAM co-culture, compared with M0 macrophages or parental culture, similarly accelerated scratch migration and upregulated both hnRNP C and vimentin protein expression, accompanied by a positive trend of vimentin and snail2 mRNA levels (online supplemental figure S3B–F).

Figure 2. Tumor-associated macrophages promote melanoma cell migration and EMT via hnRNP C upregulation. (A) Experimental workflow. CD14⁺ monocytes from buffy coats of healthy donors were first differentiated into M0 macrophages on culture with M-CSF and then incubated in melanoma cell-derived conditioned medium to induce TAMs. Melanoma cells were cultured alone (parental) or co-cultured with M0 or TAMs in trans-well systems (0.4 µm pores) for 24 hours prior to analysis. (B) Representative images of the wound healing with Buf1379 melanoma cells cultured alone or co-cultured with M0 or TAMs at the indicated time points. (C, D) Quantification of relative migration ratios of Buf1379 (C) and Buf1402 (D) cells following co-culture with M0 or TAMs (mean±SD, n≥3 biological replicates). (E) RT-qPCR analysis of hnRNP C mRNA expression in melanoma cells after 24-hour co-culture (mean±SD, n≥3 biological replicates). (F) Representative Western blot showing hnRNP C, vimentin, and β-actin expression in melanoma cells following co-culture. (G) Quantification of relative hnRNP C protein levels normalized to β-actin (arbitrary units, mean±SD, n=3 independent experiments). (H–K) EMT marker analysis after co-culture. Shown are the RT-qPCR levels of vimentin (H), Snail2 (J), and MMP2 (K), together with densitometric quantification of vimentin protein levels (I). Data are represented as mean±SD from at least three independent biological replicates. *p<0.05; **p<0.01; ***p<0.001; ****p<0.0001. EMT, epithelial-mesenchymal transition; M-CSF, macrophage colony-stimulating factor; RT-qPCR, real-time quantitative reverse-transcription PCR; TAMs, tumor-associated macrophage.

To determine the reciprocal effects of this crosstalk, M0 macrophages were exposed to melanoma conditioned media and immunophenotypic shifts were tracked by flow cytometry based on the percentage of positively gated macrophage subsets. Conditioned media from NC-transfected melanoma cells induced a macrophage phenotypic shift characterized by an increased frequency of CD86⁺/CD163⁻ and HLA-DR⁺/CD163⁻ populations, consistent with an activated M1-like macrophage state (figure 3A–E and figure 3G–K). Notably, conditioning macrophages with media from hnRNP C-silenced melanoma cells partially reversed this activation state, resulting in significantly reduced frequency of CD86⁺/CD163⁻ and HLA-DR⁺/CD163⁻ macrophages and, in some settings, increased CD86⁻/CD163⁺ cells (figure 3B–E and figure 3H–K). Complementary MFI-based analysis further substantiated the marker intensity dynamics across the groups (figure 3F,L). Consistently, exposure of macrophages to soluble factors derived from parental Buf1379 or Buf1402 cells increased the overall frequency of HLA-DR⁺/CD163⁻ macrophages and reduced the fraction of CD86⁻/CD163⁺ cells relative to baseline M0 macrophages, further confirming a tumor-driven macrophage activation profile (figure 3M–Q). Complementary MFI analysis further supported changes in macrophage marker intensity under these conditions (figure 3R). As an additional functional readout, macrophage phagocytic capacity was assessed following exposure to conditioned media from parental, NC-transfected, or hnRNP C-silenced melanoma cells. Despite differences did not reach statistical significance, macrophages exposed to conditioned media from hnRNP C-silenced melanoma cells showed a modest trend toward increased phagocytic activity compared with the NC group, particularly in the Buf1402 setting (online supplemental figure S3G,H). Together, these findings indicate that melanoma cells secrete soluble mediators in a hnRNP C-dependent manner that actively modulate macrophage phenotypes.

Figure 3. Melanoma-derived hnRNP C-dependent factors modulate macrophage polarization. (A–F) Characterization of Buf1379-conditioned macrophages. (A) Representative flow cytometry plots showing CD86/CD163 and HLA-DR/CD163 expression profiles in M0 macrophages treated with conditioned media from Buf1379 melanoma cells transfected with either control (NC) or hnRNP C-specific siRNA. (B–E) Gated percentage quantification of distinct macrophage subsets: CD86⁺/CD163⁻ (B), CD86⁻/CD163⁺ (C), HLA-DR⁺/CD163⁻ (D), and HLA-DR⁻/CD163⁺ (E) populations. (F) Normalized MFI of surface markers (HLA-DR, CD80, CD206, CD163, CD68, and CD86) across the treatment groups. (G–L) Characterization of Buf1402-conditioned macrophages. (G) Representative flow cytometry plots showing CD86/CD163 and HLA-DR/CD163 expression in M0 macrophages following exposure to conditioned media from Buf1402 melanoma cells transfected with NC or hnRNP C-specific siRNA. (H–K) Percentage quantification of the corresponding CD86⁺/CD163⁻ (H), CD86⁻/CD163⁺ (I), HLA-DR⁺/CD163⁻ (J), and HLA-DR⁻/CD163⁺ (K) populations. Normalized MFI of surface markers (HLA-DR, CD80, CD206, CD163, CD68, and CD86) across the treatment groups (L). (M–R) Phenotypic comparison across M0 cultured or not with supernatants from different melanoma cell lines. (M) Representative flow cytometry plots comparing baseline M0 macrophages to those exposed to melanoma-derived factors from Buf1379 or Buf1402 cells. (N–Q) Relative quantification of the CD86⁺/CD163⁻ (N), CD86⁻/CD163⁺ (O), HLA-DR⁺/CD163⁻ (P), and HLA-DR⁻/CD163⁺ (Q) macrophage populations. (R) Global marker MFI profile under the specified baseline and conditioned states. Data are shown as mean±SD from three independent experiments. *p<0.05; **p<0.01; ***p<0.001. MFI, mean fluorescence intensity.

TAM and melanoma crosstalk via the CXCR3-hnRNP C-MIF axis

To identify the specific hnRNP C-regulated tumor-derived cytokines driving this communication loop, several candidate factors involved in tumor-macrophage communication were examined (figure 4A). Silencing of hnRNP C significantly reduced MIF expression at the mRNA level, whereas VEGF and IGFBP2 remained largely unaffected (figure 4B–D), establishing MIF as a downstream target of hnRNP C. Consistent with this observation, co-culture with TAM didn’t alter IGFBP2 nor VEGF but significantly increased MIF mRNA expression as well as protein secretion in melanoma cells (figure 4E–I), highlighting macrophage-derived signal feedback to stimulate tumor MIF production.

Figure 4. CXCL10-CXCR3-hnRNP C-MIF axis mediates bidirectional melanoma-macrophage crosstalk (A) Cytokine array analysis of supernatants from melanoma cells cultured alone or co-cultured in trans-wells with TAMs. Representative array membranes are shown. (B–D) Impact of hnRNP C knockdown on cytokine expression. RT-qPCR analysis of IGFBP2 (B), VEGFA (C), and MIF (D) mRNA levels in melanoma cells transfected with negative control (NC) or hnRNP C-specific siRNA. (E–G) Co-culture-induced cytokine expression. RT-qPCR analysis of IGFBP2 (E), VEGFA (F), and MIF (G) mRNA expression in melanoma cells recovered after trans-well co-culture with M0 macrophages or TAMs. (H, I) Secreted factor quantification. ELISA quantification of MIF (H) and VEGF (I) levels in co-culture supernatants. (J) Macrophage polarization by recombinant MIF. Flow cytometry analysis showing the normalized MFI of polarization markers in M0 macrophages treated with recombinant MIF relative to untreated controls (mean±SD, n=3). (K) Cytokine array profiling of supernatants derived from M0 and TAMs. (L–N) Secreted chemokines in co-culture. ELISA quantification of CXCL10 (L), CCL2 (M), and IL8 (N) levels in melanoma-macrophage co-culture supernatants. (O, P) Upstream regulation of hnRNP C via chemokine signaling. Western blot analysis (O) and relative densitometric quantification (P) of hnRNP C expression in melanoma cells cultured alone or co-cultured with M0 or TAMs in the presence of neutralizing Abs against CXCL10, CXCR2, or CXCR3. (Q, R) Response to recombinant CXCL10. Western blot analysis (Q) and densitometric quantification (R) of hnRNP C protein levels in melanoma cells following treatment with recombinant CXCL10. Data are presented as mean±SD from three independent biological replicates. *p<0.05; **p<0.01; ***p<0.001. MFI, mean fluorescence intensity; PBS, phosphate-buffered saline; RT-qPCR, real-time quantitative reverse-transcription PCR; TAMs, tumor-associated macrophages; VEGF, vascular endothelial growth factor.

We next evaluated whether MIF could directly recapitulate the macrophages’ phenotypic alterations induced by melanoma conditioned media. Treatment with recombinant MIF did not induce a redistribution of positively gated cells (online supplemental figure S4A–H), but induced a modest increase in CD86 expression intensity (figure 4J). These data suggest that tumor-derived MIF contributes to targeted macrophage activation or an IFN-TAM-like phenotypic remodeling inducing a classical M1/M2 polarization switch.

To further characterize the soluble mediator profile orchestrated by tumor-conditioned macrophages, cytokine profiling was performed. This analysis identified CXCL10, CCL2, and IL8 as major candidate macrophage-derived mediators (figure 4K, online supplemental figure S4I). Quantitative ELISA measurements confirmed a progressive, statistically significant increase in secreted CXCL10 levels under TAM co-culture conditions in both melanoma cell lines (figure 4L). Conversely, CCL2 and IL8 did not show TAM-specific enrichment (figure 4M,N). PD-L1 expression on the surface of melanoma cells was also significantly upregulated following co-culture with TAM (online supplemental figure S5A). Given that CXCL10 selectively signals through the chemokine receptor CXCR3, we next assessed its role in tumor-intrinsic hnRNP C induction. Inhibition of CXCR3, but not of CXCR1, CXCR2, or IL-8 signaling, significantly reduced TAM-induced hnRNP C expression in melanoma cells. In contrast, inhibition of CXCR1, CXCR2, or IL-8 signaling had no detectable inhibitory effect (figure 4O,P and online supplemental figure S5B–E). To confirm a direct regulatory link, treatment with recombinant CXCL10 directly increased hnRNP C expression in both Buf1379 and Buf1402 cells (figure 4Q,R). Additional transcriptomics analysis showed that recombinant CXCL10 increased hnRNP C mRNA expression levels, particularly in Buf1402 cells, demonstrating regulation at the transcriptional level (online supplemental figure S5F). However, CXCL10 did not induce consistent changes in the activation of canonical downstream effectors, such as STAT1, p38/MAPK, or NF-κB either at the mRNA or protein levels under the tested conditions (online supplemental figure S5G–M), suggesting that the downstream signaling cascade linking CXCL10/CXCR3 activation to hnRNP C transcription requires further elucidation. Together, these findings demonstrate that tumor-conditioned macrophages undergo functional remodeling toward a specialized secretory profile and identify macrophage-derived CXCL10 as a key upstream inducer of tumor-intrinsic hnRNP C expression in melanoma cells via CXCR3 signaling.

Clinical and single-cell validation of the CXCR3-hnRNP C-MIF axis

The clinical relevance of this newly identified signaling axis was rigorously evaluated across multiple independent melanoma cohorts. A composite axis score was calculated as the mean of the z-scored expression of hnRNP C, MIF, and CXCR3. In the training dataset GSE8401, both hnRNP C and the composite axis scores accurately distinguished metastatic from primary melanoma samples, a finding further validated by robust diagnostic and predictive performance in the independent GSE46517 cohort (figure 5A,B). The axis score achieved superior accuracy and sensitivity compared with hnRNP C alone (figure 5C), although this was not consistent across all subcohorts (online supplemental figure S6A–D).

Figure 5. Cross-cohort validation, protein expression, and clinical relevance of the hnRNP C-MIF-CXCR3 axis in melanoma (A–C) Diagnostic performance and signature validation. ROC curves (A) showing the predictive performance of hnRNP C and the multigene axis signature in the training (GSE8401) and validation (GSE46517) cohort. Comparison of AUC values (B) with 95% CI. Classification performance metrics (C) including accuracy, F1 score, sensitivity, and specificity for hnRNP C and axis signature in the validation cohort (GSE46517). (D–F) Protein level validation via immunohistochemistry (IHC) from an independent tissue microarray of the Human Protein Atlas. Distribution of IHC staining intensity (D) for hnRNP C and MIF in normal skin and melanoma tissues categorized as low, medium, or high. Quantitative comparison of IHC staining scores (E) between normal (7 cases) and melanoma (11 cases) tissues (Wilcoxon rank-sum test). Representative IHC images (F) showing hnRNP C and MIF expression profiles in normal skin and melanoma tissues. (G–I) Association with clinical stage and metastasis. Correlation between axis signature expression and AJCC stage (G) in TCGA-SKCM. Axis scores are presented as z-normalized expression values. Distribution of axis-high and axis-low tumors (H) across primary and metastatic melanoma samples in TCGA-SKCM (Pearson’s χ² test). Logistic regression analysis (I) evaluating the association between target gene expression and metastatic status. (J) Prognostic value of hnRNP C. Kaplan-Meier overall survival analysis of metastatic melanoma patients stratified by high versus low hnRNP C expression using an optimal cut-off value of 1957.7 (minimum group size of eight patients; log-rank test). AUC, area under the curve; ROC, receiver operating characteristic.

Quantitative immunohistochemistry (IHC) staining in an independent tissue microarray set provided supportive protein-level validation for increased hnRNP C expression in melanoma tissues compared with normal skin controls (figure 5D–F). In the TCGA-SKCM, both the composite axis score and hnRNP C expression significantly correlated with advanced AJCC stage, although a strictly monotonic increase across stages was not observed (figure 5G, online supplemental figure S6E). Furthermore, both signatures were significantly enriched in metastatic lesions compared with primary tumors (figure 5H, online supplemental figure S6F). Multivariate logistic regression confirmed that hnRNP C expression and the axis score were independently associated with metastatic status (figure 5I), and high hnRNP C expression was associated with poorer overall survival in metastatic melanoma patients (figure 5J, online supplemental figure S6G).

Detailed clinical and clinicopathological associations are summarized in online supplemental tables S2–S4 (TCGA-SKCM). To resolve the cellular architecture driving this axis at a single-cell resolution, scRNA-seq datasets were analyzed demonstrating that hnRNP C and MIF expression are predominantly restricted to tumor cells, whereas CXCL10 expression is heavily enriched in M1-inflammatory macrophages (figure 6A–E). Unsupervised ligand-receptor interaction analysis using CellChart verified that MIF signaling is transmitted from tumor cells to infiltrating TME macrophages (figure 6F–H, online supplemental figure S7).

Figure 6. Tumor-derived hnRNP C-MIF axis is linked to inflammatory macrophage activation and CXCL10 signaling (A–C) Single-cell expression profiling of the axis cohorts. (A) Uniform Manifold Approximation and Projection (UMAP) visualization of hnRNP C and MIF expression in tumor cells, alongside CXCL10 in macrophages across three independent melanoma single-cell datasets (GSE277165, GSE269936, and GSE215120). (B) Dot plot showing cell-type-specific expression of CXCL10, CXCR3, hnRNP C, and MIF across tumor and TAM populations in the three datasets. Dot size represents the percentage of expressing cells; color indicates mean expression. (C) Comparative analysis of gene expression showing the preferential enrichment of hnRNP C and MIF in tumor cells versus CXCL10 in TAMs across the three datasets. (D, E) Macrophage subtypes characterization and functional signatures. (D) Violin plot of CXCL10 expression across macrophage subtypes in the GSE115978 dataset. (E) Dot plot showing the expression of MSigDB hallmark gene sets across macrophage subtypes, highlighting inflammatory signatures within M1-like macrophages. Dot size represents the percentage of expressing cells; color indicates mean expression. (F–H) Cellular crosstalk and interactome mapping. (F) Predicted MIF-mediated cell-cell communication networks across melanoma datasets, mapping potential interactions between tumor cells and immune populations. (G, H) Cell–cell communication analysis delineating incoming and outgoing signaling patterns among cell populations highlighting MIF-associated tumor–macrophage interactions in the GSE269936 dataset. TAM, tumor-associated macrophage.

To provide orthogonal spatial support for the predicted melanoma-macrophage communication hub, we further analyzed a publicly available, 10× Genomics Xenium spatial transcriptomics dataset of Formalin-Fixed Paraffin-Embedded (FFPE) human primary dermal melanoma. Unsupervised spatial clustering successfully mapped multiple tissue zones, clearly delineating melanoma-enriched (white line area) and macrophage-enriched (yellow line area) areas based on cell type-specific signature scores (online supplemental figure S8A,B). CXCL10-associated scores were mainly enriched in macrophage-associated spatial regions, whereas the combined hnRNP C/MIF-related score showed a broader diffuse distribution across the malignant tissue zones (online supplemental figure S8C).

Consistent with the predicted MIF-related ligand-receptor network, MIF and its canonical receptor CD74 were detectable within both the melanoma-enriched cluster and neighboring macrophage aggregates (online supplemental figure S8D,E). Spatial visualization further confirmed the distinct in situ distribution of MIF and CD74 transcripts across the melanoma tissue section (online supplemental figure S8F,G). Bivariate co-expression mapping further verified that CD74 was prominent in the macrophage-enriched zones (yellow line area with more blue dots), whereas melanoma-enriched regions displayed co-detection of MIF and CD74 signals (online supplemental figure S8H). Together, these spatial transcriptomics data provide independent tissue-level validation of local MIF-CD74 signaling proximity strongly aligning with our in vitro model of the bidirectional melanoma-macrophage communication network.

Discussion

HnRNP C is a key RBP increasingly recognized for its role in tumor progression and metastasis across various cancer types.^{14 17 27 28} Additionally, the cellular composition of the TME, and particularly TAMs, is known to critically influence metastatic spread and disease outcome. Despite extensive efforts, therapeutic strategies aimed at reprogramming TAMs have thus far yielded disappointing anti-tumor efficacy in clinical trials.^29–31 Prior to this study, a direct functional or regulatory intersection between hnRNP C and TAM dynamics remained completely unexplored.

In this study, significantly higher hnRNP C expression levels were found in melanoma metastases compared with primary lesions, accompanied by increased vimentin expression, a hallmark structural protein indicative of the activation of EMT.^{17 26} Functional co-culture experiments demonstrated that soluble factors derived from both M0 macrophages and TAMs directly upregulate hnRNP C and vimentin expression in melanoma cell lines establishing that macrophage-derived signals enhance pro-metastatic tumor phenotypes. Mechanistically, our data implicate CXCL10/CXCR3 signaling as a major upstream regulator of tumor-intrinsic hnRNP C expression. Ab neutralization and recombinant protein stimulation assays support a model, wherein TAM-derived CXCL10 binds to and activates CXCR3 on melanoma cells thereby enhancing hnRNP C expression. However, the precise intracellular signaling cascades linking CXCL10/CXCR3 signaling to hnRNP C expression remain a focus of ongoing investigations. Beyond these mechanistic findings, our multicohort transcriptomic analyses robustly support the clinical relevance of this signaling axis. Across independent melanoma datasets, both hnRNP C expression and the composite axis score reliably distinguished metastatic from primary tumors. Importantly, elevated axis scores correlated with advanced AJCC stages and enrichment in metastatic lesions within the TCGA-SKCM cohort. Consistently, immunohistochemical analyses confirmed elevated hnRNP C and MIF protein levels in melanoma tissues relative to normal skin. Due to the relatively small sample size of the IHC tissue cohort, these protein-level findings should be interpreted as supportive of the transcriptomic data. Further validation in larger independent melanoma tissue or proteomic cohorts will be needed when such datasets become available. Moreover, patients with high axis scores exhibited significantly reduced overall survival, suggesting that activation of this signaling network underpins aggressive disease biology. Collectively, these findings support the CXCL10/CXCR3-hnRNP C-MIF axis as both a functional pathway involved in melanoma progression and a clinically relevant biomarker framework reflecting macrophage-tumor interactions. Despite increased CXCL10 secretion by TAMs, PD-L1 expression was also upregulated in tumor cells, suggesting that T cell activity remains suppressed within this specific immune context. Phenotype analyses revealed that the infiltrating TAM subset exhibits an IFN-skewed, M1-like profile (CXCL10^high, CD86⁺, MHC-II⁺) rather than a classical, fully activated M2-like state.^{6 32–36}

Silencing hnRNP C in melanoma cells altered the frequencies of marker-defined macrophage subsets, supporting a role of tumor-intrinsic hnRNP C in shaping the macrophage phenotypic states. Coupled with the cytokine profiling and phagocytosis assay, our data suggest that hnRNP C-dependent tumor-derived factors drive macrophage activation and soluble mediator secretion, even if they do not enforce a complete classical macrophage polarization state. Our findings propose a model in which tumor-derived MIF promotes an IFN-TAM-like state, promoting the macrophage feedforward loop to secrete CXCL10. This activates CXCR3 signaling on neighboring melanoma cells, further elevating hnRNP C expression and reinforcing EMT-like metastatic programs. While our controlled co-culture and perturbation experiments support this model, they only represent a simplified model of the complex in vivo TME; thus, validation in more physiologically relevant settings will be crucial. To date, the functional differences between M2-like TAM and IFN-TAM in relation to metastasis have not yet been fully characterized, though the established role of MIF in macrophage reprogramming supports this specialized state.³⁷ This integrated framework may clarify the historically limited efficacy of TAM-targeting strategies, as modulation of macrophage phenotypes without addressing downstream CXCL10/CXCR3-hnRNP C signaling within tumor cells might be therapeutically insufficient. The translational potential of this pathway either as a biomarker for monitoring or predicting metastasis, or as a multipronged therapeutic hub, requires further development. Nevertheless, several avenues warrant investigation, such as attenuating CXCL10/CXCR3 signaling, modulating MIF-driven macrophage states, or directly targeting hnRNP C.

In conclusion, our study identifies a previously unappreciated axis of bidirectional melanoma-macrophage crosstalk mediated by the CXCL10/CXCR3-hnRNP C-MIF pathway, which accelerates metastatic progression while maintaining an active IFN-TAM-like macrophage state with potential pro-metastatic activity. Integrating mechanistic in vivo data, single-cell and spatial transcriptomics, and clinical cohorts highlights the potential of hnRNP C and the composite axis score to act as biomarkers for monitoring disease progression and metastatic risk. While our perturbation experiments and multi-omic data support this axis, its functional relevance in tumor progression and metastasis remains to be confirmed in vivo. Future studies using melanoma metastasis models incorporating targeted hnRNP C modulation, macrophage depletion, and/or MIF blockade are required to validate these findings, elucidate the precise underlying regulatory mechanisms, and assess the clinical utility of this axis for prospective patient stratification.

Data availability statement

All data relevant to the study are included in the article or uploaded as supplementary information.