Identification of cancer‑associated fibroblast biomarkers in cutaneous squamous cell carcinoma

Introduction

Cutaneous squamous cell carcinoma (cSCC), originating from epidermal keratinocytes, is the second most common tumor in humans (1). The estimated lifetime incidence of cSCC is between 7 and 11%, although its incidence has been rising in recent years. In addition to cosmetic concerns, advanced cSCC poses a notable risk of mortality (2,3). A previous study has shown that the metastasis risk for patients with cSCC is 3.7%, and the disease-specific mortality risk is 2.1% (3). Tumor diameter of ≥2 cm, invasion beyond fat, poor differentiation, perineural invasion and location on the ear, temple or anogenital area are risk factors associated with poor prognosis (3). Molecular targeted therapy has emerged as a promising treatment for cSCC (4,5), offering new hope to patients. However, cSCC often exhibits variable responses to targeted therapies (4,6), contributing to worse prognosis. Several molecular targets have been identified in cSCC, including the EGFR pathway, the Hedgehog pathway and the MAPK pathway (7). Clinical trials targeting these pathways have shown promising results (8–11). Upregulation of EGFR, observed in 35 to 56% of cSCC cases, leads to increased cell proliferation and survival (12–14). EGFR activation triggers signaling cascades, including the MAPK and PI3K/Akt pathways, promoting cell proliferation and inhibiting apoptosis (15). Uncontrolled EGFR signaling in cSCC drives uncontrolled cell proliferation and resistance to apoptosis, contributing to tumor growth and spread. Cetuximab and erlotinib effectively treat advanced cSCC by inhibiting EGFR, which is frequently upregulated in cSCC tumors (16). Most current treatment strategies target isolated pathways within cSCC, with limited consideration of the complex interactions between tumor cells and their microenvironment. Although studies have focused on immune cells and vascular endothelial cells within the cSCC microenvironment (1,17), there is limited research on whether cancer-associated fibroblasts (CAFs) serve a key role in the tumor stroma. While some studies have identified CAF marker-expressing cells in cSCC (18–21), they have not comprehensively described the impact of CAFs on cSCC and its microenvironment. Importantly, previous research has often relied on cell surface markers; however, the surface markers for subpopulations have not yet been standardized, which has introduced complexity to subsequent research.

To address this gap, an independent prioritization was performed using bulk datasets and intersection with GeneCards. In addition, tissue-level validation of selected CAF-associated markers and observational multiplex immunofluorescence (mIF) assessment were conducted across normal skin (NL), actinic keratosis (AK) and cSCC samples. The genes identified in the present study facilitate the recognition of tumor-associated fibroblasts and enhance understanding of their roles and molecular mechanisms in cSCC carcinogenesis and progression.

Materials and methods

Microarray data

GEO (http://www.ncbi.nlm.nih.gov/geo) (22) is a publicly accessible repository for functional genomics data, including high-throughput gene expression profiles from chips and microarrays. The GSE191334 (GPL11154) (23,24) and GSE139505 (GPL18573) (24) datasets, were retrieved from the GEO. Probe identifiers were converted to official gene symbols using platform-specific annotation files. The GSE191334 dataset comprised 7 cSCC tissues and 7 non-cancerous controls, while the GSE139505 dataset included 9 cSCC samples and 7 non-cancerous controls. Additionally, cSCC-related genes were identified using the GeneCards database (https://www.genecards.or) (25). During the gene screening process, the advanced search function was not enabled and only a basic keyword search on the GeneCards homepage was performed. By entering the keyword ‘cutaneous squamous cell carcinoma’, a gene list containing 6,422 genes was obtained (retrieved on October 23, 2024).

Identification of DEGs

DEGs between cSCC and non-cancerous samples were identified using GEO2R (http://www.ncbi.nlm.nih.gov/geo/geo2r), an interactive web-based tool for comparing gene expression across experimental conditions. To balance the identification of statistically significant genes and control false positives, P-values adjusted by the Benjamini-Hochberg false discovery rate (FDR) were used. Probes lacking gene symbols were excluded, and those mapping to the same gene were averaged. Given the use of bulk sequencing data, genes with |log fold change|≥1 and an adjusted P<0.05 were considered statistically significant to capture biologically relevant changes.

To obtain robust and reproducible cSCC-related candidate genes, the intersection of DEGs screened from the two datasets as target hub genes were extracted. Cross-dataset batch correction and DEG meta-analysis were not implemented in this workflow: Raw expression data from two cohorts were never integrated, eliminating inter-dataset batch interference; the intersection filtering strategy was prioritized over meta-analysis to screen genes with consistent differential expression trends in two independent sample batches.

Kyoto encyclopedia of genes and genomes (KEGG) and gene ontology (GO) enrichment analyses

Functional annotation of DEGs was performed using the Database for Annotation, Visualization and Integrated Discovery (DAVID) bioinformatics resource (http://david.ncifcrf.gov; version 2021) (26), which offers integrated biological data and tools for gene functional annotation. KEGG (27) and GO (28) analyses were conducted to uncover high-level functional and pathway-related insights. To avoid false positive outcomes induced by multiple simultaneous tests of numerous gene sets, Benjamini-Hochberg FDR multiple testing correction was applied for all enrichment results. Terms with FDR <0.05 were considered statistically significant.

PPI network construction and module analysis

PPI networks were generated using STRING (http://string-db.org; version 12.0) (29). Interactions with a combined score >0.4 were deemed significant. The resulting networks were visualized and analyzed in Cytoscape (version 3.10.2) (30). Dense network modules were identified using the MCODE plugin (version 2.0.3) (31) with the following parameters: MCODE score >5, degree cut-off=2, node score cut-off=0.2, max depth=100 and k-score=2. Functional enrichment of genes within the key module was further assessed via KEGG and GO using DAVID.

Hub-protein identification

Network topology was evaluated using CytoHubba (version 0.1) (32) to identify hub proteins. Both local [degree, maximum clique centrality (MCC), maximum neighborhood component and density of maximum neighborhood component] and global (edge percolated component, bottleneck, eccentricity, closeness, radiality, betweenness and stress) ranking methods were applied. Hub proteins were determined based on the MCC scores, focusing on the top 10 nodes.

IHC staining

After ethical approval, between September 2020 and March 2024, samples were collected from 53 patients (18 men and 35 women; median age, 62 years; range 47–72 years) in the outpatient operating room of the Department of Dermatology at the First Affiliated Hospital of Kunming Medical University (Kunming, China). These samples included 21 normal skin specimens and 32 skin specimens diagnosed as cSCC.

The inclusion criteria were as follows: i) Voluntary participation in the present study, consent for the collected samples to be used for scientific research and publication in public journals, and signed informed consent forms; and ii) normal skin or cSCC lesions located on sun-exposed areas of the face and neck. The exclusion criteria were as follows: i) Severe systemic diseases or other non-squamous cell neoplasms; ii) patients who had undergone liver, kidney or bone marrow transplantation; iii) systemic use of glucocorticoids, immunosuppressants, radiotherapy or chemotherapy within 1 month, or topical treatment of local skin lesions within 1 week; iv) patients with long-term exposure to carcinogenic substances such as radiation, asphalt, coal tar or xylene, or those with squamous cell carcinoma secondary to trauma, scarring, viral warts or chronic ulcers; and v) patients with psychiatric disorders.

Formalin-fixed, paraffin-embedded tissue sections (4 µm) were baked at 65°C for 2 h, deparaffinized and rehydrated through a graded alcohol series. Specifically, xylene I and xylene II were applied for 10 min each at room temperature, and 100, 95, 80 and 70% ethanol were applied for 5 min each at room temperature. Antigen retrieval was performed by heating sections in 10 mM citrate buffer (pH 6.0) using a microwave. Endogenous peroxidase activity was blocked with peroxidase blocking solution (3% H2O2, at room temperature) for 10 min, followed by incubation with 10% normal goat serum (cat. no. ab138478; Abcam) for 1 h to minimize non-specific binding. Sections were then incubated overnight at 4°C with primary antibodies against MCAM (1:250; cat. no. ab75769; Abcam), ACTA2 (1:100; cat. no. ab7817; Abcam), TAGLN (1:200; cat. no. ab213273; Abcam), RGS5 (1:200; cat. no. ab314671; Abcam) and PDGFRB (1:100; cat. no. ab313777; Abcam). After washing, slides were treated with an IHC enhancer for 20 min and incubated with an HRP-conjugated secondary antibodies (1:200; cat. nos. A0208 and A0192; Beyotime Biotechnology) at 37°C for 1 h. Detection was performed using DAB, and sections were counterstained with hematoxylin, dehydrated, cleared and mounted. Whole-slide digital images of histological sections were acquired using CaseViewer software (V2.4.0.119028; 3DHISTECH, Ltd.). Images at ×40 objective magnification were exported, and the corresponding scale bar length for this magnification was 20 µm. All quantitative morphological analyses were performed with reference to the 20 µm scale standard. IHC results were evaluated using the H-score semiquantitative scoring method (33,34). H-score was calculated as the product of the staining intensity score (0–3) and positive cell percentage score (0–4). Intensity: 0: colorless; 1: light yellow; 2: brownish yellow; and 3: tan. Percentage score: 0, 0–5%; 1, 6–25%; 2, 26–50%; 3, 51–75%; and 4, 76–100%. For each sample, five random non-overlapping high-power fields (×40) were evaluated. The final H-score per sample was the mean of the five fields, ranging from 0 to 12. All IHC slides were independently evaluated by two experienced dermatopathologists in a double-blinded manner to reduce subjective scoring bias, and group data are reported as mean ± SD.

mIF

The following primary antibodies were used: Rabbit monoclonal anti-ACTA2 (1:250; cat. no. ab124964; Abcam) and rabbit polyclonal anti-TAGLN (1:250; cat. no. ab14106; Abcam). As both primary antibodies were raised in the same host species (rabbit), tyramide signal amplification (TSA) was employed for sequential detection without cross-reactivity. The secondary antibody was HRP-conjugated goat anti-rabbit IgG (1:1,000; cat. no. ab6721; Abcam). TSA substrates were Cy3-tyramide (for ACTA2, red channel) and AF488-tyramide (for TAGLN, green channel). Nuclei were counterstained with DAPI (blue channel).

Tissue samples

Between January 2024 and May 2024, samples were collected from sun-exposed areas of the face and neck of 18 patients (7 men and 11 women; median age 65 years; range 49–80 years) in the outpatient operating room of the Department of Dermatology at the First Affiliated Hospital of Kunming Medical University. These samples included 6 normal skin specimens, 6 specimens diagnosed as AK and 6 specimens diagnosed as cSCC.

Multiplex staining procedure

Tissue sections were deparaffinized, rehydrated and subjected to heat-induced antigen retrieval (same method as aforementioned). Permeabilization was performed with 0.5% Triton X-100 in PBST, followed by blocking with 10% goat serum (at room temperature, 1 h).

i) Round 1 (ACTA2). Sections were incubated overnight at 4°C with the anti-ACTA2 antibody. After washing, they were incubated with the HRP-conjugated secondary antibody for 30 min, followed by Cy3-tyramide (1:500) for 10 min in the dark.

For antibody stripping, to remove non-covalently bound antibody complexes, slides were subjected to a second round of antigen retrieval (identical conditions as the first retrieval). After naturally cooling to room temperature, the sections were washed three times with PBS for 5 min each, the tissue was ready for the next round.

ii) Round 2 (TAGLN). Slides were re-blocked and incubated with the anti-TAGLN antibody overnight at 4°C. After washing, slides were incubated with the same HRP-conjugated secondary antibody (1:1,000; 30 min; at room temperature), followed by AF488-tyramide (1:200 in amplification buffer) for 10 min in the dark.

iii) Nuclear counterstaining and mounting. DAPI solution (1 µg/ml in PBS) was applied for 10 min at room temperature. Slides were briefly rinsed in distilled water, air-dried and mounted with an anti-fade mounting medium, then stored at 4°C until imaging.

Image acquisition and quantification

mIF images were acquired using a Pannoramic MIDI II whole-slide scanner (3DHISTECH Ltd.). The following fluorescence channels were used: DAPI (excitation, 358 nm; emission: 461 nm, blue channel), AF488 (excitation, 495 nm; emission, 519 nm; green channel for TAGLN) and Cy3 (excitation: 550 nm; emission: 570 nm; red channel for ACTA2). Image visualization was performed using CaseViewer (version 2.4.0) software. Fluorescence images from the red channel (ACTA2) and green channel (TAGLN) were merged using CaseViewer software. Co-localization of ACTA2 and TAGLN appeared as yellow in the merged images. For each section, five random fields were captured at ×40 magnification. The absolute number of ACTA2+TAGLN+ double-positive cells per high-power field (×40) was quantified.

Statistical analysis

Statistical analyses were performed using GraphPad Prism (version 9.5.1; Dotmatics). The IHC H-score data from each group were first subjected to the Kolmogorov-Smirnov normality test. If the test result was P>0.05, the data were considered to follow a normal distribution, and comparisons between two groups were performed using an independent samples t-test. If the test result was P≤0.05, the data were considered not to follow a normal distribution, and comparisons between two groups were performed using the Mann-Whitney U test. All data in the present study were normally distributed, and a P<0.05 was considered to indicate a statistically significant difference.

Results

Identification of DEGs in cSCC

Standardized analysis of the two microarray datasets identified DEGs (11,788 in GSE139505 and 7,031 in GSE191334), intersecting with the 6,422 cSCC-related genes in GeneCards, yielding 1,391 genes (Fig. 1A).

Figure 1. Overlapping genes among three sets and their functional enrichment. (A) Venn diagram of two Gene Expression Omnibus datasets and one GeneCards dataset. (B) Kyoto Encyclopedia of Genes and Genomes enrichment analysis of differentially expressed genes. ECM, extracellular matrix; AGE-RAGE, advanced glycation end-products-receptor for advanced glycation end-products.

KEGG and GO enrichment analysis of DEGs

To investigate biological processes in cSCC, pathway enrichment analysis was conducted using DAVID. GO analysis of DEGs revealed significant enrichment in biological processes related to cell proliferation, angiogenesis and ECM (Fig. S1). Cellular component enrichment highlighted cytoplasmic, nuclear and ECM changes. Molecular function analysis indicated enrichment in protein binding, ATP binding, DNA binding, cytoskeletal components and ECM. KEGG pathway analysis indicated enrichment in the ‘PI3K-Akt signaling pathway’, ‘cell cycle’, ‘focal adhesion’, ‘ECM-receptor interaction’ and the ‘p53 signaling pathway’ (Fig. 1B).

PPI network construction and module analysis

The PPI network of DEGs (Table SI) was constructed, and critical modules were identified using Cytoscape (Fig. 2A). Functional analysis of module genes indicated enrichment in MAPK, PI3K-Akt and ECM pathways (Fig. 2B). Further analysis with the Cytoscape Cytohubba plugin identified the top 20 genes by MCC score, emphasizing genes involved in DNA replication (Fig. 2C), suggesting bulk sequencing may reflect tumor cell biology, with microenvironmental changes less prominent in bulk data.

Figure 2. PPI network analysis results. (A) Display of MCODE analysis module genes. (B) Kyoto Encyclopedia of Genes and Genomes enrichment analysis of MCODE module genes and CytoHubba MCC TOP 20 genes. (C) Display of CytoHubba Maximum Clique Centrality Top 20 genes.

Analysis and identification of candidate CAF biomarkers

Reviewing single-cell studies in PubMed and Google Scholar, 50 reported CAF marker genes in cSCC were identified (Table I; Fig. 3A) (18–21). Cross-referencing these with the 1,391 DEGs yielded 20 overlapping genes (Fig. 3A), including CXCL13, COL1A1, FN1, WNT5A, MYH11, MCAM, TGFB1, STMN1, MMP11, COL3A1, LUM, TNC, OGN, FAP, TGFB3, TOP2A, COL4A1, MKI67, COL6A3 and DCN. Further IHC validation revealed higher expression of MCAM in cSCC compared with NL (Figs. 3B and S2A). Genes frequently identified in single-cell studies but not in bulk sequencing, such as ACTA2, TAGLN, RGS5 and PDGFRB, also displayed elevated expression in cSCC compared with NL (Figs. 3C and S2B), highlighting the utility of single-cell technology in uncovering complex biological variations within tumors. Meanwhile, mIF also demonstrated the presence of a CAF subset co-expressing the genes ACTA2 and TAGLN throughout the pathological evolution of human skin tissue from the normal state to cSCC (Fig. 3D). In normal samples, double-positive cells had been relatively scarce.

Figure 3. Candidate gene identification. (A) Intersection of DEGs and CAFs from the cSCC single-cell study. (B) Immunohistochemical analysis statistics of MCAM in cSCC and normal samples. (C) Immunohistochemical analysis statistics of ACTA2, TAGLN, RGS5 and PDGFRB in cSCC and NL. *P<0.05 and **P<0.01. (D) Results of multiplex immunofluorescence for NL, AK and cSCC. Magnification, ×40. CAFs, cancer-associated fibroblast markers; cSCC, cutaneous squamous cell carcinoma; NL, normal skin; AK, actinic keratosis; MCAM, melanoma cell adhesion molecule; of ACTA2, actin α2 smooth muscle; TAGLN, transgelin; RGS5, regulator of G protein signaling 5 and PDGFRB, platelet derived growth factor receptor β; DEGs, differentially expressed genes.

Table I. Fibroblast marker genes summarized in single-cell cutaneous squamous cell carcinoma studies.

Table I.

Fibroblast marker genes summarized in single-cell cutaneous squamous cell carcinoma studies.

Cluster	Markers	Annotation	(Refs.)
Fib1	RGS5, KCNJ8, ACTA2 and MCA	Angiogenesis	(18)
Fib2	COL1A1/2, LUM and DC	Classic fibroblasts	(18)
Fib3	ACTA2, MYH11 and AOC3	Myofibroblasts	(18)
Fib4	MGST1, MFAP5, WISP2 and PTGDS	Reticular fibroblasts	(18)
Fib5	ASPN, OGN and TNN	Mesenchymal fibroblasts	(18)
Fib6	STMN1, MKI67 and TOP2A	Proliferating fibroblasts	(18)
CAFs	LRP1, SDC1, TFPI, FN1, THBS1, TGFB1 and TGFB3	-	(19)
mCAFs	RGS5, DCN, COL6A2, COL1A1, COL1A2	-	(21)
iCAFs	DCN, COL6A2, COL1A1, COL1A2	-	(21)
IL7R+ CAFs	IL7R, IL1B, IL6, CXCL1, CXCL3, CXCL5, CXCL6, CXCL8, CXCL13, CXCL14, DCN, COL6A2, COL1A1, COL1A2, COL4A1, COL6A1, COL6A2, COL6A3, FN1, TNC and THY1	Tumor recurrence	(21)
CAFs	ACTA2, FAP	-	(20)
iCAF	COL6A2, COL6A3, FN1, TNC and THY1	-	(20)
myCAF	COL3A1, C3, IGF1, ADM, CXCL12, COL3A1, MMP11, WNT5A and ANGPTL2	-	(20)

[i] Fib, fibroblast; CAFs, cancer-associated fibroblasts; myCAF, myofibroblastic cancer-associated fibroblast; iCAF, inflammatory cancer-associated fibroblast; mCAF, matrix cancer-associated fibroblast; RGS5, regulator of G protein signaling 5; KCNJ8, potassium inwardly rectifying channel subfamily J member 8; ACTA2, actin α2, smooth muscle; MCAM, melanoma cell adhesion molecule; TAGLN, transgelin; PDGFRB, platelet derived growth factor receptor β; LUM, lumican; DCN, decorin; MYH11, myosin heavy chain 11; AOC3, amine oxidase copper containing 3; MGST1, microsomal glutathione S-transferase 1; MFAP5, microfibril associated protein 5; WISP2, WNT1 inducible signaling pathway protein 2; PTGDS, prostaglandin D2 synthase; ASPN, asporin; OGN, osteoglycin; TNN, tenascin N; STMN1, stathmin 1; MKI67, marker of proliferation Ki-67; TOP2A, DNA topoisomerase II α; LRP1, LDL receptor related protein 1; SDC1, syndecan 1; TFPI, tissue factor pathway inhibitor; FN1, fibronectin; THBS1, thrombospondin 1; CXCL, C-X-C motif chemokine ligand; FAP, fibroblast activation protein; IGF1, insulin growth factor 1; C3, complement C3; ADM, adrenomedullin; THY1, Thy-1 cell surface antigen; ANGPTL2, angiopoietin like-2; TNC, tenascin C.

Discussion

cSCC is a common epithelial malignancy with an estimated lifetime incidence of 6 to 11% (35–37) and AK has been considered its precancerous lesion. Up to 5% of cSCC cases progress to metastatic cSCC, and this rate is higher in types of cancer with multiple high-risk factors. It has been reported that male sex, advanced age and immunocompromised status are associated with a higher risk of metastasis (38). Advanced cSCC is often fatal. cSCC typically exhibits a high tumor mutational burden (39–41), driven by an accumulation of risk genes, primarily induced by UV radiation (42), as well as interactions with the stroma and local immune modulation, which collectively drive tumor progression (43–45). AK is a precancerous lesion of cSCC (46,47), with ~10% of AK cases progressing to cSCC within 2 years (48). Although multiple risk factors have been identified, cumulative UV exposure with advancing age remains the greatest risk factor, particularly on sun-exposed areas of the face and neck. As a prototypic keratinocyte carcinoma, cSCC arises through a stepwise progression from healthy epidermal keratinocytes to precancerous lesions and subsequent cSCC, involving cumulative mutations in pathways that regulate cell division, communication, apoptosis and differentiation, along with abnormal interactions with the tumor microenvironment (TME) and immune surveillance (49–52). These findings deepen the understanding of the complex mechanisms underlying cSCC pathogenesis. In the present study, two cSCC datasets were integrated from the GEO database and cSCC-related genes from GeneCards, along with previously reported CAF marker genes in cSCC from PubMed and Google Scholar, as well as IHC and mIF data under different pathological conditions. The candidate genes identified in the present study facilitate recognition of tumor-associated fibroblasts and provide insights into their potential role and molecular mechanisms in cSCC carcinogenesis and progression. Moreover, the present study emphasizes the necessity of further exploration into cellular heterogeneity and intercellular interactions within cSCC, aspects that will be investigated in subsequent studies.

During tumor formation, progression and dissemination, cancer cells interact with diverse cell types within the surrounding stroma, forming the TME, which is key to disease progression (53). The TME is marked by high concentrations of CAFs, which support tumor advancement by remodeling the surrounding ECM and secreting various factors that promote tumorigenesis (18,54,55). The advent and standardization of single-cell transcriptomics has now enabled comprehensive study of various cell types, including CAFs, as well as CAF heterogeneity across different tumor types (18,54,55). Notably, the subpopulations identified in various studies, even within the same disease type, do not consistently align, including the marker genes (18–21). The present study revealed that MCAM not only demonstrated high gene expression in bulk sequencing data, but was also validated by IHC. MCAM serves a role in cell adhesion and cohesion within endothelial monolayers at vascular and tissue intercellular junctions. It is highly expressed across various types of cancer (56–58). In colorectal cancer, MCAM+ fibroblasts interact with IL-1 receptor 1, enhancing NF-κB/IL-34/C-C motif chemokine ligand 8 signaling to recruit macrophages and form a pro-TME (59). The role and mechanism of MCAM within the cSCC TME warrants further investigation.

Certain genes, identified as CAF markers in multiple single-cell cSCC studies (Table I) (18–21), were not among the DEGs identified in bulk sequencing data, but demonstrated their high expression in cSCC through IHC. This highlights the advantage of single-cell techniques in exploring complex biological changes within tumors. For example, the gene ACTA2 is a known marker of myofibroblasts in cSCC and oral squamous cell carcinoma and is associated with worse prognosis (18,60). In glioblastoma, it marks mature CAFs (61), while in colorectal cancer, ACTA2+ fibroblasts can induce immunosuppression (59). The gene TAGLN has been reported as upregulated across various tumors. For instance, this gene mediates ECM stiffness in ovarian cancer progression through the Ras homolog family member A/Rho-associated protein kinase pathway (62), enhances stem cell characteristics in head and neck squamous cell carcinoma via the TGFBI-TAGLN axis (63) and promotes the hypoxic response and maintenance of glioblastoma stem cells through p53 acetylation (64). In the present study, the mIF results revealed the presence of ACTA2+TAGLN+ cell subsets in all AK and cSCC samples. However, their contribution to cSCC progression requires further functional and mechanistic studies.

The gene RGS5 is implicated in angiogenesis across various tumors (65). In liver cancer, it mediates Wnt/β-catenin signaling for tumor progression (66), and in triple-negative breast cancer, it is a marker of extracellular matrix CAFs, potentially serving a role in metabolic pathways (67,68). In breast cancer, a CAF subset expressing integrin α11 and PDGFRβ promotes tumor invasion via PDGFRβ/JNK signaling and tenascin C upregulation, and targeting this axis suppresses CAF-induced invasiveness (67). In the present study, IHC further validated that these genes show a trend toward higher expression in cSCC.

Leveraging large-scale bulk transcriptomic datasets, GeneCards-based cross-analysis and multi-tissue validation, the present study identified a panel of candidate genes of CAF in cSCC, namely MCAM, ACTA2, TAGLN, RGS5 and PDGFRB, with a particular focus on the ACTA2+TAGLN+ CAF subpopulation. These observations establish a foundation for unraveling the contributions of CAFs to cSCC biology. Nonetheless, the present study has certain limitations that merit consideration. The mIF assessment, although informative, was constrained by a modest sample size (six paired samples), which may limit extrapolation to broader patient populations. Furthermore, the approach does not permit causal inference regarding the functional impact of the identified genes on CAF behavior or tumor progression. Additionally, the study was conducted at a single time point, precluding analysis of the temporal dynamics of CAF activation and plasticity during cSCC evolution. To overcome these limitations, future investigations will incorporate single-cell transcriptomic profiling to resolve CAF heterogeneity at the single-cell level, followed by functional characterization, including gain- and loss-of-function assays, lineage tracing and co-culture models, to delineate the precise mechanisms by which ACTA2+TAGLN+ CAFs influence tumor invasion, immune evasion and even therapeutic resistance. Ultimately, such endeavors may inform the rational design of CAF-targeted interventions for cSCC.

Supplementary Material

Supporting Data

Supporting Data

Acknowledgements

Not applicable.

Funding

The present work was supported by the National Natural Science Foundation of China (grant no. 82260517).

Availability of data and materials

The data generated in the present study may be requested from the corresponding author.

Authors' contributions

YD contributed to conceptualization, visualization, and the writing and revision of the original draft. JQ, YC and QH made substantial contributions to the acquisition, analysis and interpretation of data, including the collection of human tissue samples, as well as the preparation, staining and reading of pathology slides. YT contributed to conceptualization. LH contributed to conceptualization, resources, funding acquisition, supervision, and the review and editing of the manuscript. All authors read and approved the final version of the manuscript. YD and LH confirm the authenticity of all the raw data.

Ethics approval and consent to participate

The present study protocol was approved by the Ethics Committee of the First Affiliated Hospital of Kunming Medical University [approval no. (2020)-L-29], and written informed consent was obtained from all patients. All procedures performed in the present study involving human participants were in accordance with the Declaration of Helsinki (as revised in 2013).

Patient consent for publication

Not applicable.

Competing interests

The authors declare that they have no competing interests.

References