CXCR4-Mediated Chemokine Signaling Orchestrates the Progression of Conjunctival Melanocytic Lesions

Abstract

Purpose

The biological and microenvironmental differences between benign and malignant conjunctival melanocytic lesions remain poorly understood. This study aims to characterize their microenvironment and reveal the role of chemokine signaling in lesion progression.

Methods

Bulk RNA sequencing and single-cell RNA sequencing were used to delineate the microenvironment of conjunctival melanocytic lesions and identify divergent molecular pathways. CXCR4 expression was validated in PIG1 melanocytes and the conjunctival melanoma (CoM) cell lines (CRMM1, CRMM2, and CM2005.1) using quantitative real-time PCR and Western blotting. Tissue specimens were further analyzed for CXCR4 and CXCL12 expression via immunohistochemistry and immunofluorescence. To assess functional relevance, CoM cell lines were treated with the CXCR4 antagonist AMD3100, followed by analysis of cell viability and downstream PI3K/Akt and MEK/ERK phosphorylation using Western blotting. Additionally, the antitumor efficacy of AMD3100 was validated in vivo.

Results

Transcriptomic profiling revealed marked microenvironmental heterogeneity between benign and malignant lesions. A significant upregulation of chemokine signaling pathway, along with notably elevated CXCR4 expression, was observed in malignant lesions compared to the benign. Exogenous CXCL12 stimulation significantly enhanced the proliferation of CoM cells, whereas AMD3100 treatment induced cytotoxic effects and reduced phosphorylation of Akt and Erk. Additionally, in vivo experiments confirmed the antitumor efficacy of AMD3100.

Conclusions

CXCR4-mediated chemokine signaling pathway is significantly upregulated in CoM. Pharmacological blockade of CXCR4 with AMD3100 exerts cytotoxic effects on CoM cells through inhibition of PI3K/Akt and MEK/ERK pathways, indicating CXCR4 as a promising translational target that warrants further investigation and clinical validation.

Conjunctival melanocytic lesions encompass a spectrum of entities, including benign conditions such as nevi and complexion-associated melanosis, as well as malignant entities, notably conjunctival melanoma (CoM) which is characterized by substantial metastatic potential and poor prognosis.^1–4 Conjunctival melanocytes originate from neural crest cells, migrating and differentiating to ultimately reside in the basal epithelial layer of the conjunctiva.⁵ Despite sharing a common origin, conjunctival melanocytes exhibit remarkably divergent cellular fates, some remain benign, whereas others become aggressively malignant. Furthermore, certain benign melanocytes may eventually undergo malignant transformation. For instance, a subset of CoM has been reported to originate from nevi.¹ Therefore understanding the molecular and cellular mechanisms underlying this divergence in conjunctival melanocyte fate is essential for developing targeted interventions to prevent malignant progression.

Emerging evidence indicates the critical role of the local tissue microenvironment, not limited to malignant lesions, in modulating cellular behavior and disease progression.⁶ Recent single-cell profiling of conjunctiva has resolved distinct cell subsets, offering preliminary insights into their potential interactions and establishing an initial cellular map of the ocular surface niche.⁷ Environmental stressors, such as ultraviolet irradiation, further alter this microenvironment by enhancing local inflammation and extracellular-matrix remodeling, consequently exposing melanocytes to chronic low-grade inflammatory stimuli that promote their propensity for proliferation and phenotypic change.⁸ Collectively, these observations underscore the importance of microenvironmental factors in the initiation and progression of conjunctival melanocytic disease. However, a comprehensive high-resolution characterization of this microenvironment remains to be achieved.

Among microenvironment cues, chemokine signaling axis has been implicated in the pathogenesis and progression of various benign and malignant diseases.^9–11 Prior studies have demonstrated dynamic alterations in chemokine expression during the progression and metastatic spread of conjunctival melanocytic lesions, suggesting a potential role for chemokine signaling in disease development.¹² Although chemokine-targeted inhibitors have entered clinical trials in many diseases,¹⁰ underscoring their promising translational potential, their precise roles and therapeutic efficacy in conjunctival melanocytic lesions remain largely unknown and warrant further investigation.

Here, our study aims to comprehensively delineate the cellular and molecular landscape across the spectrum of conjunctival melanocytic lesions, systematically characterizing the microenvironmental heterogeneity. Moreover, we seek to define the role of chemokine signaling within these diverse microenvironments, clarify their downstream molecular mechanisms, and ultimately identify potential therapeutic targets for precision treatment of conjunctival melanocytic disorders.

Material and Methods

Study Design and Ethical Approval

This study was conducted following the ethical standards outlined in the Declaration of Helsinki. Ethical approval was granted by the Institutional Review Board of Shanghai Jiao Tong University School of Medicine. The requirement for written informed consent was waived due to the retrospective nature of the research.

ScRNA-Seq Data Processing

The single-cell transcriptomic data analyzed in this study were obtained from the publicly available datasets previously generated by our center. For the current analysis, we selected 3 benign samples and 6 malignant in situ samples. All preprocessing and downstream analyses were performed using the Seurat package (version 4.4.0) in R (version 4.1.0). Low-quality cells were filtered out based on the following criteria: (1) a minimum of 200 features, (2) a maximum of 7,500 features, and (3) mitochondrial gene content of less than 25%. Potential doublets were identified by DoubletFinder package (version 2.0.4) and excluded. After filtering, gene expression data were normalized and the top 2000 most variable genes were identified. To correct for potential batch effects driven from patient-specific expression patterns, we applied Harmony package¹³ (version 1.2.0). Dimension reduction was performed using principal-component analysis (PCA) with RunPCA function and the optimal number of principal components was selected using ElbowPlot function. Cell clusters were visualized using the UMAP algorithm.

Major cell types were annotated based on the expression of canonical marker genes as previously described (i.e., EPCAM, KRT5, and KRT8 for epithelial cell; FN1, DCN, COL1A1, and COL1A2 for fibroblast; VWF, PECAM1, and CDH5 for endothelial cell; PMEL, MLANA, and DCT for melanocyte; CD3D, CD3E, NKG7, and KLRC1 for TNK; MZB1, MS4A1, JCHAIN and CD79A for B/Plasma cell; CD68, CD14, LAMP3, TPSAB1, CD1C, and CLEC9A for myeloid cell). Subsequent clustering analysis was conducted within each major lineage to identify finer subpopulations. To distinguish malignant from benign melanocytes, we applied inferCNV (version 1.8.1) to infer large-scale copy number variation (CNV) patterns, using melanocytes from benign samples as the reference. Each melanocyte was assigned a CNV score based on its deviation from the reference baseline. K-means clustering was then used to classify melanocytes according to their CNV profiles. Clusters exhibiting low CNV deviation were annotated as benign, while those with high CNV deviation were considered malignant. Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway activity was estimated in each cell by calculating the average expression of genes within each pathway. Melanocytes were stratified by CNV-based cluster type, and differential pathway activity between these types was assessed using the FindMarkers function. P values were adjusted using the false discovery rate method to identify significantly enriched pathways.

Cell-Cell Communication Analysis

The CellChat (version 1.6.1) was used to analyze the potential intercellular communication.¹⁴ The normalized expression matrix was imported to create a CellChat object using the CellChat() function. The data were then preprocessed with identifyOverExpressedGenes(), identifyOverExpressedInteractions(), and projectData() functions using the default parameters. The computeCommunProb(), filterCommunication(), and computeCommunProbPathway() functions were then conducted to determine any potential ligand–receptor interactions. At last, the cell communication network was aggregated using the aggregateNet function.

RNA Isolation and Library Preparation

Total RNA was isolated utilizing TRIzol reagent (Invitrogen, Carlsbad, CA, USA) following the manufacturer's guidelines. The concentration and purity of the RNA samples were determined using a NanoDrop 2000 spectrophotometer (Thermo Fisher Scientific, Waltham, MA, USA), whereas RNA integrity was verified via the Agilent 2100 Bioanalyzer (Agilent Technologies, Winooski, VT, USA). Library preparation was carried out with the VAHTS Universal V10 RNA-seq Library Prep Kit (Premixed Version), adhering to the provided protocol. Transcriptome sequencing and downstream analysis were performed by OE Biotech Co., Ltd. (Shanghai, China).

RNA Sequencing and Differentially Expressed Genes Analysis

The libraries were sequenced on an Illumina Novaseq 6000 platform and 150 bp paired-end reads were generated. About 48.7M raw reads for each sample were generated. Raw reads of fastq format were first processed using fastp¹⁵ and the low quality reads were removed to obtain the clean reads. Then about 46.6M clean reads for each sample were retained for subsequent analyses. The clean reads were mapped to the reference genome using HISAT2.¹⁶ FPKM of each gene was calculated and the read counts of each gene were obtained by HTSeq-count.^17,18 PCA analyses were performed using R (v 3.2.0) to evaluate the biological duplication of samples.

Differential expression analysis was performed using the DESeq2.¹⁹ Q value < 0.05 and foldchange > 2 or foldchange < 0.5 was set as the threshold for significantly differential expression gene (DEGs). Based on the hypergeometric distribution, GO, KEGG pathway, Reactome and WikiPathways enrichment analysis of DEGs were performed to screen the significant enriched term using R (v 3.2.0), respectively.^20,21 R (v 3.2.0) was used to draw the column diagram, the chord diagram, and bubble diagram of the significant enrichment term.

Cell Culture

The human melanocyte line PIG1 was obtained from the Department of Ophthalmology, Peking University Third Hospital (Beijing, China). Conjunctival melanoma cell lines (CRMM1, CRMM2, and CM2005.1) were generously provided by Prof. Martine J. Jager (Leiden University Medical Center, Leiden, the Netherlands). All cell lines were cultured in Ham's F12K medium (Gibco; Thermo Fisher Scientific) supplemented with 10% fetal bovine serum (Gibco; Thermo Fisher Scientific) and 1% penicillin-streptomycin. Cells were maintained at 37°C in a humidified incubator with 5% CO₂. Cell line authentication was confirmed by short tandem repeat profiling, and all lines tested negative for mycoplasma contamination.

RNA Extraction, Reverse Transcription, and qPCR

Total cellular RNA was isolated using the EZpress RNA Purification Kit (EZBioscience, Roseville, MN, USA) following the manufacturer's guidelines. The cDNA synthesis was carried out with the PrimeScript RT Reagent Kit (Vazyme Biotech Co., Ltd, Nanjing, China). Quantitative real-time PCR (qPCR) was conducted using SYBR Green Master Mix (Thermo Fisher Scientific) under standard cycling conditions. ACTIN served as the reference gene for normalization. Primer sequences are provided in Supplementary Table S1.

Western Blot

Cells were lysed in RIPA buffer (Beyotime Institute of Biotechnology, Jiangsu, China) supplemented with protease inhibitors (Sangon Biotech, Shanghai, China) after washing with PBS. Equal amounts of protein (40–60 µg) were loaded onto 10% SDS-PAGE gels, followed by electrophoretic transfer to PVDF membranes (Millipore, Burlington, MA, USA). Membranes were blocked with 5% nonfat milk in TBST (Sangon Biotech) and subsequently probed with primary antibodies. After washing, membranes were incubated with HRP-conjugated secondary antibodies (1:10,000; Amersham, Piscataway, NJ, USA). The primary antibodies used included ACTIN (1:2000; Proteintech, Rosemont, IL, USA), CXCR4 (1:1000; Cell Signaling Technology, Danvers, MA, USA), Akt (1:1000; Proteintech), phospho-Akt (1:1000; Cell Signaling Technology), Erk (1:1000; Cell Signaling Technology), and phospho-Erk (1:1000; Cell Signaling Technology).

Colony Formation Assay

Cells were seeded at a density of 1000 per well in 12-well plates. AMD3100 or CXCL12 treatments at different concentrations were applied the next day. Culture medium was replaced every three days. After 14 days, colonies were fixed and stained with 1% crystal violet, then allowed to air dry before imaging. Colony formation was quantified using ImageJ software.

Immunohistochemistry (IHC) Staining and H-Score Calculation

Formalin-fixed, paraffin-embedded tissue sections with a thickness of 5 µm were used for immunohistochemistry. Sections were incubated with primary antibodies against CXCR4 and CXCL12 (both 1:2000; Servicebio, Wuhan, China). Visualization was performed using a Panoramic DESK digital slide scanner (3D HISTECH). Immunoreactivity was evaluated using the H-score method, calculated as follows: [0 × percentage of negative cells + 1 × percentage of weakly positive cells + 2 × percentage of moderately positive cells + 3 × percentage of strongly positive cells].

Immunofluoresence

Paraffin-embedded tissue sections were deparaffinized, rehydrated, and fixed, followed by blocking with 5% normal goat serum. The sections were then incubated overnight at 4°C with primary antibodies against CXCR4 (1:400; Abcam, Cambridge, MA, USA), CXCL12 (1:400; Abcam) and α-SMA (1:400; Abcam). After washing, slides were treated with secondary antibodies for one hour. Nuclear staining was performed using DAPI (Sigma-Aldrich Corp., St. Louis, MO, USA) for five minutes. Fluorescent images were captured with a ZEISS Axio Scope A1 upright microscope (Zeiss, Oberkochen, Germany).

Subcutaneous Animal Model

CRMM1 cells 5 × 10⁵ were injected subcutaneously into the left flank of four-week-old female nude mice that were under anesthesia. Two weeks after tumor implantation, the mice were randomly assigned (n = 4 per group) to receive intraperitoneal injections of either sterile saline solution (vehicle) or AMD3100 (5 mg/kg; MedChemExpress, Monmouth Junction, NJ, USA) three times a week. Tumor length (L) and width (W) were measured every three days. On day 10 after the first treatment, mice were euthanized, and the tumors were excised, weighed, and photographed. All procedures were approved by the Animal Experimental Ethics Committee of Shanghai Ninth People's Hospital, Shanghai Jiao Tong University School of Medicine.

Statistical Analysis

All statistical analyses were conducted using R version 4.1.1 and GraphPad Prism 9.0. Each experiment was independently repeated three times to ensure reproducibility. Data are presented as mean ± SEM. Comparisons between two groups were assessed using two-tailed Student's t-tests, while multiple group comparisons were evaluated using one-way ANOVA. A two-sided P value <0.05 was considered indicative of statistical significance.

Results

Single-Cell RNA Sequencing Unveils the Microenvironment of Conjunctival Melanocytic Lesions

To elucidate the cellular architecture of the microenvironment in conjunctival melanocytic lesions, we reanalyzed single-cell RNA sequencing (scRNA-seq) data collected from our previous study.²² Specifically, we selected three benign samples and six cases of CoM. Clinical characteristics of the samples were summarized in Supplementary Table S2. After quality control and doublet exclusion, a total of 29,452 cells were identified and included in the subsequent analysis. Then, unsupervised clustering was performed in the Seurat framework to define major cell populations based on shared transcriptional profiles (Fig. 1A). The scaled expression levels and the proportion of cells expressing cluster-specific markers across each subpopulation are displayed in dot plots (Fig. 1B).

Figure 1.

Single-cell RNA-seq atlas of benign and malignant conjunctival melanocytic lesions. (A) UMAP plot of 29,452 single cells annotated by major cell types. (B) Dot plot of the expression level of cell-type-specific markers. (C) Bar plots comparing cell type fractions across benign and malignant. (D) Hierarchical heatmap showing large-scale CNVs of melanocytes from all samples. Melanocytes from benign sample were included as a control reference. Red: gains; blue: losses. (E) Violin plot illustrating the distribution of CNV scores across melanocyte subtypes identified by clustering based on CNV scores (up). UMAP visualization of melanocyte subtype (down).

To gain deeper insights into the microenvironment landscape, we examined the cellular abundances and relative compositions of major cell lineages across the different sample types (Fig. 1C). Compared to benign samples, which included one melanocytic nevus and two conjunctival tissues, CoM (n = 6) exhibited a marked increase in melanocytic cells, concomitant with a substantial depletion of immune cell populations, including B cells, T cells, NK cells, and mast cells. We further interrogated the scRNA-seq data to infer CNVs within the melanocytic cell populations. Analysis of the inferred CNV profiles revealed both inter-lesional and intra-lesional heterogeneity among CoM samples (Fig. 1D). Melanocytic cells exhibiting CNV signatures were subjected to further clustering based on their CNV scores, enabling the classification into benign and malignant subtypes (Fig. 1E).

Chemokine Pathways Are Profoundly Elevated in Malignant Lesions Compared to Benign

Since conjunctival melanocytic lesions originate from melanocytes, different types of pigmented lesions may result from melanocytic proliferation or benign-to-malignant transformation, we further explored the changes in melanocytes. Based on the functional enrichment analysis of melanocyte subtypes, we identified distinct pathway activities associated with benign and malignant states. Specifically, pathways involved in arachidonic acid metabolism, ribosome function, and vitamin B6 metabolism were significantly upregulated in benign melanocytes. Conversely, malignant melanocytes demonstrated increased activity in canonical oncogenic pathways, including PI3K-Akt signaling and oxidative phosphorylation. Notably, the chemokine signaling pathway was markedly enriched in malignant melanocytes (Fig. 2A).

Figure 2.

Functional enrichment and fibroblast-melanocyte interactions in conjunctival melanocytic lesions. (A) Enrichment functional terms for the two melanocyte's type. (B) Interaction circle plot comparing fibroblast-melanocyte interactions between benign and malignant lesions, with red lines indicating interactions upregulated and blue lines indicating interactions downregulated in malignant condition. (C) Heatmap depicting CXCL signaling pathway interactions between fibroblast and melanocyte under benign and malignant conditions. (D) Bubble plot of significant ligand-receptor pairs in the CXCL signaling pathway mediating interactions between fibroblast subtypes and melanocytes. (E) Violin plot of CXCL12 expression levels across all cell types.

To further elucidate the cellular interactions underpinning these phenotypic differences within the microenvironment, we conducted a comprehensive analysis of cell-cell communication. Our findings indicated that melanocytes predominantly received signaling inputs from fibroblasts in both benign and malignant contexts. Similarly, fibroblasts primarily transmitted signals to melanocytes, emphasizing a critical fibroblast-melanocyte communication axis (Supplementary Figs. S1A–D). Consequently, we focused subsequent analyses on characterizing the interactions specifically between fibroblasts and melanocytes.

Fibroblasts were further classified into distinct subpopulations, namely Fib-PI16, Fib-CXCL12, Fib-FAP, and Fib-Smooth muscle-ACTA2 (Supplementary Fig. S2D, S2E). Interaction analyses revealed that all fibroblast subtypes exhibited significantly increased interactions with melanocytes in malignant lesions compared to benign counterparts (Fig. 2B). Given our prior observation of elevated chemokine signaling in malignant melanocytes, we specifically interrogated fibroblast-to-melanocyte interactions within the chemokine signaling axis. Strikingly, fibroblast-melanocyte chemokine interactions were minimal in benign lesions but were markedly amplified in malignant lesions (Fig. 2C).

Subsequent ligand-receptor pair analysis identified CXCL12/CXCR4 as the dominant chemokine axis mediating fibroblast-melanocyte interactions in malignant contexts (Fig. 2D). Single-cell transcriptomic validation further confirmed significant upregulation of CXCR4 expression specifically in malignant melanocytes (Supplementary Figs. S2A–C). Given the potential for other cell types to express CXCL12, we next sought to determine its cellular origin. Expression profiling across all cell populations revealed that fibroblasts are the predominant source of CXCL12 secretion (Fig. 2E), thereby reinforcing the importance of fibroblast-derived CXCL12/CXCR4 signaling in promoting the malignant progression of melanocytic lesions.

Bulk RNA Sequencing Reveals Pronounced Activation of Chemokine Pathways in Conjunctival Melanoma

To systematically define the transcriptomic distinctions between benign and malignant melanocytic lesions, we next performed bulk RNA sequencing (RNA-seq) to validate and extend our single-cell observations, enabling a global evaluation of gene expression programs. Bulk RNA-seq was conducted on melanocytic nevi (n = 3) and melanomas (n = 3). Detailed clinical information for these patients was provided in Supplementary Table S2. Differential expression analysis, as depicted in the volcano plot, identified 591 genes significantly upregulated and 498 genes downregulated in CoM relative to nevus (Fig. 3A). Prominently, multiple members of the chemokine family were among the most upregulated transcripts in CoM.

Figure 3.

Transcriptomic profiling of melanocytic nevi and conjunctival melanoma by bulk RNA sequencing. (A) Volcano plots of differentially expressed genes in CoM and Nevus. (B) GO term enrichment analysis of the downregulated genes. The vertical coordinates are the enriched GO terms, and the horizontal coordinates are the numbers of the downregulated genes in these GO terms. The green columns represent the biological process GO terms; the purple columns represent the cellular component GO terms; the orange columns represent the molecular function GO terms. (C) Bubble plot for KEGG pathway enrichment analysis. The y-axis shows pathway terms, whereas the x-axis denotes gene ratio. The size of each circle represents the number of genes. The hue of the circles symbolizes various q values. (D) Bubble plot for Reactome pathway enrichment analysis. The y-axis shows pathway terms, whereas the x-axis denotes gene ratio. The size of each circle represents the number of genes. The hue of the circles symbolizes various q values. (E) Representative enrichment plot obtained from CoM and nevus.

Gene ontology (GO) enrichment analysis revealed significant activation of the chemokine signaling pathway and CXCR chemokine receptor activity in CoM (Fig. 3B). Consistent with these findings, bubble plot visualization of KEGG enrichment indicated a pronounced enhancement of chemokine–receptor interactions in malignant lesions (Fig. 3C). Reactome enrichment analysis further corroborated these results, demonstrating upregulation of the “Chemokine receptors bind chemokines” pathway in CoM (Fig. 3D).

Gene Set Enrichment Analysis (GSEA) provided additional confirmation, demonstrating strong enrichment of the chemokine signaling pathway (normalized enrichment score [NES] = 2.61, P < 0.001) and “Chemokine receptors bind chemokines” pathway (NES = 2.62, P < 0.001) in CoM (Fig. 3E). Collectively, these integrative analyses highlight a central role for chemokine-mediated signaling in the molecular reprogramming associated with the progression of conjunctival melanocytic lesions.

CXCR4 Exhibits High Expression in Conjunctival Melanoma

We next evaluated CXCR4 expression at both the mRNA and protein levels in CoM cell lines compared with the control melanocytic cell line, PIG1. RT-qPCR and Western blotting analyses revealed that CXCR4 expression was significantly upregulated in CoM cell lines (Figs. 4A, 4B), indicating transcriptional and translational activation of this chemokine receptor in malignant melanocytes.

Figure 4.

CXCR4 and CXCL12 are significantly upregulated in conjunctival melanoma. (A) RT-qPCR shows the relative mRNA expression of chemokine and chemokine receptors. (B) Western blotting of protein expression of CXCR4 in PIG1, CRMM1, CRMM2, and CM2005.1 cell lines. ACTIN is control. (C) Representative H&E staining and IHC images of CXCR4 and CXCL12 expression in Nevus and CoM, accompanied by statistical analysis. ****, P < 0.0001. Scale bar: 20 µm. (D) Representative immunofluorescence (IF) images of CXCR4 and CXCL12 in Nevus and CoM. Scale bar: 10 µm.

To further validate these findings in clinical specimens, we performed IHC staining on patient-derived tissue samples. Conjunctival nevus was used as a representative model of benign melanocytic cells. Consistent with the in vitro findings, both CXCR4 and its cognate ligand CXCL12 were markedly overexpressed in CoM tissues compared to melanocytic nevi (Fig. 4C), suggesting that upregulation of the CXCL12–CXCR4 axis is a feature of malignant transformation.

Additionally, immunofluorescence assays were conducted to investigate the subcellular distribution of CXCR4 and CXCL12. The results demonstrated robust cytoplasmic localization of both proteins in CoM samples, with significantly higher fluorescence intensity relative to benign controls (Fig. 4D). To further validate cellular localization, we performed immunofluorescence staining of tumor sections using α-SMA as a fibroblast marker, which confirmed that CXCL12 and CXCR4 are spatially co-localized with fibroblasts (Supplementary Fig. S3). These findings collectively suggest that activation of the CXCL12–CXCR4 signaling axis may play a crucial role in promoting the malignant phenotype of melanocytic lesions.

AMD3100 Effectively Suppresses Conjunctival Melanoma Proliferation

To investigate the therapeutic potential of CXCR4 inhibition in CoM, we treated CoM cell lines and PIG1 with the CXCR4 antagonist AMD3100. Colony formation assay demonstrated that AMD3100 had no significant cytotoxic effects on PIG1 cells across a range of concentrations, indicating its relative specificity for malignant melanocytes. In contrast, CoM cell lines exhibited a marked, dose-dependent increase in cytotoxicity following AMD3100 treatment (Figs. 5A, 5B). Building on these observations, we next investigated whether stimulation with CXCL12 could promote CoM proliferation. The results showed that exogenous CXCL12 markedly enhanced colony formation, and this proliferative effect was weakened by co-treatment with AMD3100, underscoring the functional importance of the CXCL12/CXCR4 axis (Figs. 5C, 5D). Given that previous studies have implicated AMD3100 in the suppression of the PI3K/Akt and MEK/ERK signaling pathways in various tumor models, we next assessed the activity of these pathways in CoM cells after CXCR4 inhibition. Western blotting analyses revealed a significant reduction in the phosphorylation levels of Akt and ERK proteins on AMD3100 treatment (Fig. 5E), confirming effective inhibition of both the PI3K/Akt and MEK/ERK cascades. To substantiate these findings in vivo, we established subcutaneous CoM xenograft tumor models. AMD3100 significantly impeded tumor progression, yielding smaller volumes and lower final tumor weights compared with the vehicle controls (Figs. 5F–I).

Figure 5.

AMD3100 acts on conjunctival melanoma by inhibiting the PI3K-Akt and MEK/ERK pathways. (A) Colony formation assay of PIG1, CRMM1, CRMM2, and CM2005.1 cell lines under the treatment of AMD3100 with the concentration of 0, 1, 5, and 10 nmol. (B) Statistical analysis of colony formation assay. **P < 0.01. (C) Colony formation assay of CRMM1 and CM2005.1 cell lines under the treatment of CXCL12 (40 ng/mL) and AMD3100 (10 nmol). (D) Statistical analysis of colony formation assay. *P < 0.05. (E) Western blotting of protein expression of Akt, p-Akt, Erk, and p-Erk of CRMM1 and CM2005.1 cell lines under the treatment of AMD3100 with the concentration of 0, 1, and 10 µM. (F) Schematic presentation of xenograft tumor model. (G) The image of tumor burden of the two treatment groups. (H, I) Quantification data of tumor volume and tumor weight between the two groups, *P < 0.05, **P < 0.01.

In summary, relative to nevi, CoM is characterized by enhanced CXCL12 secretion from cancer-associated fibroblasts (CAFs) and elevated CXCR4 expression in melanoma cells, collectively indicating a strengthened CXCL12–CXCR4 signaling axis. Targeting this axis with AMD3100 exhibited promising antitumor activity in CoM, highlighting CXCR4 as a potential translational target in the malignant progression of melanocytic lesions (Fig. 6).

Figure 6.

The schematic illustrates the role of the chemokine CXCR4 in conjunctival melanocytic lesions.

Discussion

This study provides the integrative transcriptomic characterization that comprehensively delineates the microenvironmental landscape of conjunctival melanocytic lesions. Specifically, we revealed substantial microenvironmental heterogeneity between benign and malignant conjunctival melanocytic lesions. A pronounced elevation of chemokine signaling pathways was observed in malignant lesions compared to their benign counterparts, particularly implicating the CXCL12/CXCR4 chemokine axis. Pharmacological inhibition of CXCR4 was shown to exert anti-tumor effects on CoM cells, primarily through modulation of the PI3K-Akt and MEK/ERK signaling cascades.

The ocular surface is a highly specialized and complex anatomical system, with the conjunctiva covering approximately two-thirds of its total area, extending from the corneal limbus to the eyelid margins.²³ In addition to its principal components, including immune cells, conjunctival epithelium, meibomian glands, lacrimal glands, and the neural network, other elements such as stromal cells, endocrine factors, and the ocular microbiota play essential roles in maintaining tissue homeostasis.^24–26 By profiling the conjunctival microenvironment at single-cell resolution, we showed that disease progression followed a clear ecological transition: benign lesions reside in a lymphocyte-rich, stromally balanced niche, whereas malignant melanoma occupies an immunosuppressive niche enriched with tumor-associated macrophages. This pattern parallels previous observations in cutaneous melanocytic lesions, where intact tissue architecture and active immune surveillance typify benign states,^27,28 whereas malignant transformation involves tumor-driven remodeling, recruitment, and polarization of tumor-associated macrophages to reinforce immune escape.^29–31 Furthermore, cell–cell communication analysis revealed that melanocyte–fibroblast crosstalk is the dominant interaction axis in both benign and malignant settings. In homeostatic or nevus tissues, fibroblasts secrete trophic factors, such as FGF2, HGF and SCF to regulate melanocyte survival and pigment production.³² During malignant progression, this dependency persists but undergoes some changes: melanoma cells release TGF-β, PDGF-BB, and related cues, converting neighboring fibroblasts into CAFs. These CAFs, in turn, secrete CXCL12, IL-6, and matrix remodeling enzymes, establishing a pro-invasive, immunosuppressive stroma that reciprocally supports tumor growth. Consequently, our ligand-receptor analysis records high communication scores in both states. What changes in the two states is the message content, shifting from homeostatic support in benign lesions to invasion and immune evasion in melanoma.

Chemokine signaling, particularly involving CXCL12 and its receptor CXCR4, has been widely implicated in cancer biology by promoting tumor cell survival, migration, and metastatic dissemination.^33–36 CXCR4 was originally identified as an orphan G-protein–coupled receptor capable of inducing intracellular Ca²⁺ flux upon engagement with its sole ligand, CXCL12.³⁴ In this study, both single-cell and bulk RNA-seq datasets revealed a consistent up-regulation of chemokine-related genes in malignant melanocytes, with ligand-receptor analysis pinpointed CXCL12/CXCR4 as the dominant axis. CXCR4 overexpression was confirmed at the transcript and protein levels in patient specimens, while tracing the upstream ligand showed that fibroblasts, particularly the CAFs, were confirmed as the primary source of CXCL12, in accordance with previous reports from multiple cancer types.^37,38 In tissue specimens, we further confirmed the spatial co-localization of fibroblasts with components of the CXCL12/CXCR4 axis. Future experiments using co-culture models could provide additional insights into the direct mechanisms underlying fibroblast–melanocyte interactions. Although CAFs dominate CXCL12 production, we also detected low-level CXCL12 expression in endothelial cells at the single cell level, which may modestly shape local CXCL12/CXCR4 gradients and signaling. We further elucidated that CXCL12/CXCR4 engagement activates PI3K-Akt and MEK/ERK signaling cascades in melanoma cells. These pathways converge to up-regulate matrix metalloproteinases and promote invasiveness, a synergy previously documented in other cancer models.³⁴ Collectively, our data expand the role of CXCL12/CXCR4 beyond immune-cell recruitment by establishing a tumor-intrinsic, fibroblast-driven signaling loop that sustains conjunctival melanoma progression, thereby highlighting the axis and the fibroblast–melanocyte interactions as a promising translational target.

Clinically, the CXCR4 antagonist AMD3100 (plerixafor) is already approved by the Food and Drug Administration for stem-cell mobilization and has shown antitumor activity in preclinical and clinical settings across various solid tumors.³⁹ Consistent with prior studies, our results confirmed AMD3100's efficacy in suppressing conjunctival melanoma proliferation and reducing AKT and ERK phosphorylation, validating effective blockade of CXCL12/CXCR4 signaling axis. These findings indicate that targeted inhibition of CXCR4 may offer therapeutic benefits for patients with conjunctival melanoma, warranting further exploration in translational studies.

Although our exploratory study provides important insights, it also has limitations. The relatively modest sample size suggests that further validation in larger, independent cohorts encompassing a broader clinical spectrum of conjunctival melanocytic lesions is warranted. Additionally, integrating spatial transcriptomic analyses in future investigations could provide deeper insight into the spatial organization and heterogeneity of the microenvironment, thereby strengthening the generalizability and clinical applicability of our findings.

In conclusion, our study provides the integrative transcriptomic characterization using bulk and single-cell RNA sequencing to elucidate the microenvironmental distinctions between benign and malignant conjunctival melanocytic lesions. Our results highlight the significant upregulation of the chemokine signaling pathway, especially the CXCL12/CXCR4 axis, in CoM compared to both melanocytic nevi and normal conjunctiva. By demonstrating the prominent expression and functional relevance of CXCR4 in tumor progression, our results deepen the understanding of the conjunctival melanocytic lesions’ microenvironment and highlight the translational potential of targeting CXCR4 to modulate PI3K–Akt and MEK/ERK signaling. Further preclinical and clinical investigations are warranted to evaluate the therapeutic feasibility of this approach in CoM.