Chondroitin sulfate proteoglycan 4 increases invasion of recessive dystrophic epidermolysis bullosa-associated cutaneous squamous cell carcinoma by modifying transforming growth factor-β signalling

Abstract

Background

Recessive dystrophic epidermolysis bullosa (RDEB) is a rare genetic skin-blistering disorder that often progresses to metastatic cutaneous squamous cell carcinoma (cSCC) at chronic wound sites. Chondroitin sulfate proteoglycan 4 (CSPG4) is a cell-surface proteoglycan that is an oncoantigen in multiple malignancies, where it modulates oncogenic signalling, drives epithelial-to-mesenchymal transition (EMT) and enables cell motility.

Objectives

To evaluate CSPG4 expression and function in RDEB cSCC.

Methods

RDEB cSCC cell lines were used to assess CSPG4-dependent changes in invasive potential, transforming growth factor (TGF)-β1-stimulated signal activation and clinically relevant cytopathology metrics in an in vitro full-thickness tumour model. CSPG4 expression in RDEB cSCC and non-RDEB cSCC tumours was analysed via immunohistochemistry and single-cell RNA sequencing (scRNA-Seq), respectively.

Results

Inhibiting CSPG4 expression reduced invasive potential in multiple RDEB cSCC cell lines and altered membrane-proximal TGF-β signal activation via changes in SMAD3 phosphorylation. CSPG4 expression was uniformly localized to basal layer keratinocytes in fibrotic RDEB skin and tumour cells at the tumour–stroma interface at the invasive front in RDEB cSCC tumours in vivo. Analysis of published scRNA-Seq data revealed that CSPG4 expression was correlated with an enhanced EMT transcriptomic signature in cells at the tumour–stroma interface of non-RDEB cSCC tumours. Cytopathological metrics, for example nucleus : cell area ratio, were influenced by CSPG4 expression in in vitro tumour models.

Conclusions

We determined that CSPG4 expression in RDEB cSCC cell lines enhanced the invasive potential of tumours. Mechanistically, CSPG4 was found to enhance membrane-proximal TGF-β-stimulated signalling via SMAD3, which is a key mediator of EMT in RDEB cSCC. The implication of these studies is that CSPG4 may represent a therapeutic target that can be leveraged for the clinical management of patients with RDEB cSCC.

Chondroitin sulfate proteoglycan 4 (CSPG4)-dependent modulation of the canonical TGF-β signalling pathway was identified through SMAD3. CSPG4 expression was correlated to enhanced epithelial–mesenchymal transition transcriptional signatures in non-RDEB cutaneous squamous cell carcinoma (cSCC) cells located at the tumour–stroma boundary. CSPG4 expression is strongly expressed at the invasive front of RDEB cSCC tumours. RDEB cSCC may benefit from therapeutic strategies that target CSPG4 and in development for other malignancies.

What is already known about this topic?

• Cutaneous squamous cell carcinoma (cSCC) in recessive dystrophic epidermolysis bullosa (RDEB) is a highly metastatic disease with few effective treatments.

• Chondroitin sulfate proteoglycan 4 (CSPG4) has known roles in promoting the malignant progression of other cancers.

What does this study add?

• We identified CSPG4-dependent modulation of the canonical transforming growth factor (TGF)-β signalling pathway via SMAD3.

• We correlated CSPG4 expression to enhanced epithelial-to-mesenchymal transition (EMT) transcriptional signatures in non-RDEB cSCC cells located at the tumour–stroma boundary.

• We demonstrated that CSPG4 is strongly expressed in basal keratinocytes in fibrotic RDEB skin and at the invasive front of RDEB cSCC tumours.

What is the translational message?

• CSPG4 expression in RDEB skin is associated with underlying dermal fibrosis and precedes RDEB cSCC development, where it promotes a proinvasive tumour cell phenotype.

• CSPG4 expression may serve as a predictive biomarker for cSCC development in RDEB skin, and as a potentially targetable biomarker of invasion in RDEB cSCC.

Skin is composed of an epidermis containing stratified layers of keratinocytes overlaying a dermis comprised of fibroblasts, immune cells and vasculature. The epidermis and dermis interface through the protein-rich basement membrane zone (BMZ), which ensures the structural integrity of the skin. Epidermolysis bullosa (EB) is a class of rare genetic skin disorders characterized by extremely fragile skin and mechanically induced blisters at the BMZ. The most severe form of EB is generalized severe recessive dystrophic EB (RDEB), which occurs in 3 in every 1 million live births.¹ Type VII collagen – a secreted protein responsible for linking cell-surface hemidesmosome units to the BMZ – is fully absent or dysfunctional in RDEB and leads to a wide array of complications, including chronic wounds, dermal fibrosis, infection, joint contractures, and dental and gastrointestinal malformations.¹

A major complication of RDEB is the development of aggressive and often metastatic cutaneous squamous cell carcinoma (cSCC), with a cumulative risk of developing cSCC reaching 76.1% by 35 years of age.² RDEB-associated cSCC (RDEB cSCC) is the major cause of death in patients with RDEB; almost 80% of patients succumb to RDEB cSCC within 5 years of diagnosis.^2–4 RDEB cSCC typically arises at chronic wound sites, where disease progression is driven by chronic inflammation and associated fibrosis mediated, in part, by excessive secretion of transforming growth factor (TGF)-β.⁵ TGF-β signalling contributes to many oncogenic processes, including fibrosis, angiogenesis and epithelial-to-mesenchymal transition (EMT).^6–10 Importantly, TGF-β stimulates tumorigenic capacity in some RDEB cSCC cell lines.^11,12 To date, curative treatments in non-RDEB cSCC – conventional chemotherapy, monoclonal antibodies and/or surgical resection – have been largely ineffective for RDEB cSCC.^4,13,14 Thus, there exists an urgent need for new insights into mechanisms associated with RDEB cSCC development and progression to bolster the repertoire of therapeutic approaches.

Chondroitin sulfate proteoglycan 4 (CSPG4), a cell-surface glycoprotein, has emerged as a therapeutic target in multiple treatment-resistant malignancies.^15–23 In healthy tissues, the expression of CSPG4 is restricted to immature progenitor cells.²⁴ In healthy skin, CSPG4 expression is limited to keratinocyte stem cell populations, where it promotes cell–cell adhesion and stem cell clustering.²⁵ However, in malignancies CSPG4 functions as an oncoantigen whose upregulated expression is repeatedly associated with metastasis and/or decreased progression-free survival.^22,24,26

In tumours and certain tumour-derived cell lines, CSPG4 expression is heterogeneously distributed and positively regulated by inflammation.^24,27 Cancer cells expressing CSPG4 are highly invasive and motile, and more readily undergo EMT.^22,28–31 In cancer cell lines, CSPG4 has also been shown to drive drug resistance.^22,32 As a transmembrane scaffolding protein, CSPG4 does not have intrinsic signalling capacity, but instead functions to enhance membrane-proximal oncogenic signalling pathways, including focal adhesion kinase (FAK), extracellular signal-regulated kinase (ERK) 1/2 and certain oncogenic receptor tyrosine kinases [e.g. fibroblast growth factor (FGF) and platelet-derived growth factor-α].^33,34

Given the association of CSPG4 expression with malignant progression, we posit that elevated CSPG4 expression may contribute to the invasiveness of RDEB cSCC. Our goals were to evaluate whether CSPG4 expression is altered in the complex pathobiology of RDEB disease progression and to determine whether CSPG4 has a role in driving a malignant phenotype in RDEB cSCC. We show that CSPG4 expression in basal keratinocytes within uninjured RDEB skin is indistinguishable from patterns observed in normal skin. However, CSPG4 expression is greatly enhanced in RDEB basal keratinocytes that interface with a chronically inflamed/fibrotic dermis. Finally, we report that this high level of CSPG4 expression is maintained in RDEB cSCC cells at the tumour–stroma boundary, where it functions to promote a mesenchymal/invasive phenotype, in part by enhancing TGF-β-mediated signalling.

Materials and methods

Cell lines and tissues

RDEB cSCC lines were derived from whole tumours isolated from patients with RDEB (Table S1; see Supporting Information).^35,36 These lines were a generous gift from Dr Andrew South (Thomas Jefferson University). Tissue samples (Table S2; see Supporting Information) and primary keratinocytes/fibroblasts were derived from punch biopsies from healthy donors or patients with RDEB. There were five RDEB donors (labelled 1–5) and four healthy donors (labelled A–D). Cell lines were screened for mycoplasma using an EZ-PCR Mycoplasma Detection Kit (Biological Industries, Cromwell, CT, USA) and positive lines were treated with Plasmocin (InvivoGen, San Diego, CA, USA).

RDEB cSCC and primary keratinocytes were cultured in EpiLife medium supplemented with EpiLife Defined Growth Supplement (Gibco, ThermoFisher Scientific, Waltham, MA, USA) at 37 °C in 5% CO₂. Primary keratinocytes were grown on type I collagen (5 µg cm^–2; Advanced BioMatrix, Carlsbad, CA, USA). Primary fibroblasts were cultured in ‘MEM α, GlutaMAX^TM Supplement, no nucleosides’ supplemented with 1× nonessential amino acids (Gibco, ThermoFisher Scientific); 10% fetal bovine serum; 5 ng mL^–1 FGF (PeproTech, ThermoFisher Scientific); and 10 ng mL^–1 epidermal growth factor (Sigma-Aldrich, St. Louis, MO, USA) at 37 °C in 5% CO₂.

Small interfering RNA transfection

Small interfering RNA (siRNA) targeting CSPG4 (CUUCUCCUCCUCUCAUGACUU)²⁸ and AllStars negative control siRNA were obtained from QIAGEN (Hilden, Germany). Cells were transfected with siRNA using lipofectamine RNAiMAX transfection reagent (Invitrogen, Carlsbad, CA, USA), according to the manufacturer’s recommendations. Cells were harvested after 48 h for cell characterization assays or to confirm knockdown with Western blot.

Invasion assays

RDEB cSCC cells (5.0 × 10⁴ cells per 0.5 mL) in EpiLife growth medium were seeded into Matrigel^®-coated invasion chambers (8 µm polyethylene terephthalate; Corning, Glendale, AZ, USA) after siRNA transfections. Lower wells were filled with EpiLife growth medium with 10 ng mL^–1 TGF-β1 (PeproTech). After 48 h, invaded cells were stained using a Differential Quik III Stain Kit (Electron Microscopy Sciences, Hatfield, PA, USA). Membranes were mounted on glass slides, then counted at × 20 magnification from five random fields/membrane. Each experiment was repeated three times.

Signalling experiments

SCC2 and SCC4 cell lines were treated with CSPG4-targeted or negative control siRNAs as described. After transfection, cell lines were suspended (2.5 × 10⁵ cells mL^–1 in basal EpiLife media) in six-well plates coated with 0.8 mg cm^–2 poly(2-hydroxyethyl methacrylate) (Sigma-Aldrich) and cultured for 24 h at 37 °C in 5% CO₂. After 24 h, 10 ng mL^–1 TGF-β1 or basal media was spiked into wells and cells were harvested for protein extraction after 0.5, 2 and 6 h.

Western blots

Western blot lysate was prepared with supplemented RIPA buffer (Santa Cruz Biotechnology, Santa Cruz, CA, USA). Antibodies are described in Table S3 (see Supporting Information). All secondaries used were fluorescent IRDye products (1 : 5000), and nitrocellulose membranes were imaged using the Odyssey XF system and quantified with Image Studio (version 3.2; LI-COR Biosciences, Lincoln, NE, USA).

Histological evaluation

RDEB (donors 1 and 2) and normal (donor A) fresh frozen paraffin embedded (FFPE) tissue and full-thickness model (FTM) sections (4 µm) were subjected to antigen retrieval (Reveal Decloaker; Biocare Medical, Pacheco, CA, USA). After quenching and blocking (Background Sniper; Biocare Medical), sections were incubated overnight at 4 °C with primary CSPG4 antibody (LSBio, Seattle, WA, USA; Table S3). The slides were incubated with detection reagent (Novocastra Novolink Polymer Kit; Leica Biosystems, Deer Park, IL, USA) and diaminobenzidine. Sections were counterstained with Harris haematoxylin. Histological sections were evaluated in a semiquantitative manner using a scale ranging from 0 to +3 (0 = negative; + 1 = weak; + 2 = moderate; + 3 = strong).

Single-cell RNA sequencing analysis

Published single-cell RNA sequencing (scRNA-Seq) datasets from healthy skin and patient-matched non-RDEB cSCC tumours (n = 10 donors)³⁷ were accessed in a gene-barcode matrix via Gene Expression Omnibus (SuperSeries accession number: GSE144240; SubSeries accession number: GSE144236). scRNA-Seq analyses were performed using R (version 4.1.0; R Foundation for Statistical Computing, Vienna, Austria). Data were processed with the ‘Seurat’ package (version 4.0.5)³⁸ and cells were annotated by comparing their highly expressed genes to cell type-specific expression reported in the Human Protein Atlas.^39,40 Adapting a pipeline outlined in Ji et al.,³⁷ tumour epithelial cells were scored for EMT gene expression signatures using a hallmark EMT gene set (n = 200 genes), which did not include CSPG4.⁴¹ The ‘AddModuleScore’ command in Seurat was used to score the transcriptomes of each tumour epithelial cell for expression of EMT-related genes (EMT score) relative to all other tumour epithelial cells.

Statistical analysis

Statistical analysis was performed with Prism 9 (version 9.5.0; GraphPad, La Jolla, CA, USA) or R (version 4.1.0). Statistical techniques are described in the figure legends. P-values < 0.05 were considered statistically significant.

Material and product information is provided in Table S4 (see Supporting Information). Additional methods are described in Appendix S1 (see Supporting Information).

Results

CSPG4 is expressed on the surface of RDEB cSCC cells in vitro and in vivo

We evaluated the expression level of CSPG4 in six RDEB cSCC cell lines and cultured primary keratinocytes from RDEB patients and normal volunteers (Table S1). All RDEB cSCC cell lines expressed CSPG4 at various levels (Figure 1a), which has been described in other solid tumours.^17,22,42 RDEB keratinocytes tended to express higher levels of CSPG4 vs. normal keratinocytes (Figure 1a, b). In RDEB cSCC cell lines with high levels of CSPG4 expression (SCC2 and SCC4), a mean (SD) of 62.1% (5.9) and 57.8% (6.4) of the total cells in each population expressed cell-surface CSPG4, respectively (Figure 1c).

Figure 1.

Primary keratinocytes, recessive dystrophic epidermolysis (RDEB) cutaneous squamous cell carcinoma (cSCC) cell lines and RDEB tissues express chondroitin sulfate proteoglycan 4 (CSPG4) at variable levels. (a) Western blot showing cultured primary keratinocytes isolated from punch biopsies from healthy donors (N), from RDEB donors (R) or from RDEB cSCC cell lines derived from whole tumours (Table S1). The CSPG4 banding pattern shows a non-chondroitin sulfate-modified core protein at 250 kDa (the dominant isoform present in all samples) and a chondroitin sulfate-modified isoform that migrates as a diffuse band near 400 kDa. (b) Quantification of both 250- and 400-kDa CSPG4 bands in Western blot shown in (a) normalized to glyceraldehyde 3 phosphate dehydrogenase (GAPDH). (c) Flow cytometry showing the percentage of total population expressing CSPG4 on the cell surface in RDEB cSCC cell lines 2, 4 and 53 (n = 3 biologic replicates). Bars indicate mean, with error bars representing the SD. (d) Fresh frozen paraffin-embedded sections of healthy skin, RDEB skin with a histopathological finding of extensive dermal fibrosis and RDEB cSCC with a histopathological finding of extensive dermal fibrosis stained with haematoxylin and eosin (H&E; top panels). Immunohistochemistry (IHC) against CSPG4 (bottom panels) was evaluated in a semiquantitative manner: healthy skin = +1 (weak and focal); fibrotic RDEB skin = +2 (membranous); RDEB cSCC tumour = +2 (membranous and cytoplasmic). Epidermis (E) and dermis (D) are indicated. Pockets of CSPG4-expressing cells in healthy skin are highlighted with black arrows (bottom left panel). CSPG4 expression on the invasive front of RDEB cSCC are indicated by black arrows (bottom right panel). Inset IHC images show membranous staining of CSPG4 (bottom panels). Scale bars range from 200 to 250 µm. (e) Immunofluorescence (IF) and haematoxylin and eosin staining of fresh frozen samples from healthy skin and clear RDEB skin without a clinical history of wounding. Scale bars = 50 µm in IF images. Haematoxylin and eosin images are at ×40 magnification.

Elevated CSPG4 expression has been described in dysplasia preceding melanoma and head-and-neck SCC.^17,43 Therefore, we examined CSPG4 expression patterns in healthy and RDEB skin and RDEB cSCC tumours to determine whether a progressive increase in CSPG4 expression was evident. RDEB skin often acquires extensive dermal fibrosis and scar formation as a result of repeated injury and chronic inflammation.^5–7 To account for this unique feature of RDEB skin pathology, our RDEB skin sample pool contained clear skin without a history of injury or scar formation, skin with histopathological confirmation of extensive fibrosis and RDEB cSCC with underlying fibrosis (Table S2).

We first analysed CSPG4 expression in healthy skin, fibrotic RDEB skin and RDEB-cSCC using immunohistochemistry (Figure 1d). Consistent with previous reports,^24,25 healthy skin had interspersed clusters of basal keratinocytes with faint membranous CSPG4 expression (Figure 1d, bottom left panel, black arrows). In contrast, fibrotic RDEB skin had strong uniform membranous CSPG4 expression in basal keratinocytes (Figure 1d, bottom centre panel). To determine whether robust enrichment of CSPG4 expression in basal keratinocytes was a consistent feature of RDEB skin, we used immunofluorescence to visualize CSPG4 expression in clear RDEB skin without fibrosis. We found that CSPG4 was expressed by basal keratinocytes oriented directly above BMZ marker laminin-332 in both healthy and clear RDEB skin (Figure 1e). Notably, CSPG4 expression patterns were similar between healthy skin (Figure 1d, left column; Figure 1e, top row) and clear RDEB skin (Figure 1e, bottom row), with interspersed clusters of CSPG4-expressing basal keratinocytes.

We detected CSPG4 expression at the tumour–stroma interface of RDEB cSCC and in stromal tumour islands, where strong membranous CSPG4 expression at the invasive front was apparent (Figure 1d, bottom right panel, black arrows). Staining patterns in healthy skin, clear RDEB skin, fibrotic RDEB skin and RDEB cSCC tumours were consistent across additional samples analysed with multiple CSPG4-targeted antibodies (Figure S1, Table S3; see Supporting Information). Collectively, these results suggest that increased CSPG4 expression in subpopulations of RDEB keratinocytes or RDEB cSCC cells is associated with a chronically inflamed fibrotic dermis and may affect localized invasion of primary tumours.

Full-thickness models of RDEB cSCC reveal that CSPG4 expression maintains malignant phenotypic properties independent of underlying fibrosis

We developed a RDEB cSCC FTM to evaluate specifically the impact of dermal fibroblasts on modifying the malignant phenotype of RDEB cSCC cells expressing or lacking CSPG4 (Figure S2; see Supporting Information). In contrast to models that utilize fibroblasts seeded into fibrin or collagen gels, fibroblasts in our FTMs synthesized their own three-dimensional ECM.

Using CRISPR/Cas9, we created a stable CSPG4-knockout cell line derived from parental SCC4. Sequence validation confirmed polyallelic disruption, including total deletion of start codons (Figure S3a, b; see Supporting Information) and loss of CSPG4 protein expression in the knockout cell line (–/–; Figure 2a). We generated and analysed FTMs as described in Appendix S1 (Figure 2b; Figure S2a–c).

Figure 2.

In vitro full-thickness models of recessive dystrophic epidermolysis (RDEB) cutaneous squamous cell carcinoma (cSCC) reveal chondroitin sulfate proteoglycan 4 (CSPG4)-dependent changes in tumour cytopathology. (a) Western blot showing total loss of CSPG4 protein expression in SCC4-derived knockout cells (–/–). (b) Representative haematoxylin and eosin-stained images of a full-thickness model assembled from parental SCC4 (+/+) or knockout (–/–) SCC4 cell lines and RDEB or healthy primary fibroblasts (scale bars = 20 µm). Insets show dyskeratotic cells (scale bars = 5 µm; n≥ 3 models per cSCC/fibroblast combination). (c) Average number of dyskeratotic cells per image (n≥ 9 images per SCC/fibroblast combination). (d) Average total cSCC cell count, including dyskeratotic cells, per image (n≥ 9 images per cSCC/fibroblast combination). (e) Average percentage of dyskeratotic cells per total cSCC cells per image (n≥ 9 images per cSCC/fibroblast combination). (f) Average nucleus area (µm²) and (g) average cell area (µm²) of randomly selected cSCC cells, excluding dyskeratotic cells, for each cSCC/fibroblast combination (n ≥ 180 cells). (h) Ratio of nucleus : cell areas for each randomly selected cSCC cell. Bars represent mean (SD) values. P-values were calculated using one-way Anova with Šídák correction for multiple comparisons. GAPDH, glyceraldehyde 3 phosphate dehydrogenase.

We approached FTM analysis from a cytopathological perspective and evaluated two parameters indicative of malignancy: dyskeratotic/apoptotic cells (increased in malignancy) and nuclear size.^44–46 Dyskeratotic/apoptotic cells are indicators of rapid cell turnover (Figure 2b, insets). Increased chromatin volume leading to increased nuclear size and nucleus : cell area ratio are well-established measures of malignant progression and are useful in tumour grading.⁴⁴ We hypothesized that CSPG4-expressing tumours would display increased dyskeratosis and have larger nucleus : cell area ratios than CSPG4-null tumours.

We first performed pairwise comparisons of FTMs by CSPG4 expression status (+/+ vs. –/–) to determine whether cytopathology parameters differed depending on dermis origin (RDEB vs. healthy) of fibroblasts incorporated into the model. We found that ‘dermis’ origin did not significantly influence the average number of dyskeratotic cells per section, the proportion of dyskeratotic cells per total cSCC cells, or nuclear and cell areas of either cSCC cell type (Figure S3c–h).

Next, we compared cSCC cytopathology for each FTM by ‘dermis’ origin (Figure 2c–h). The average number of dyskeratotic CSPG4 (+/+) cSCC cells per section was consistently higher than that of FTMs containing CSPG4 (–/–) cells (Figure 2c). When dyskeratotic cell count was normalized to total number of cSCC cells (Figure 2d), CSPG4 (+/+) FTMs had significantly increased percentages on both ‘dermis’ origins relative to (–/–) FTMs (Figure 2e). The nuclear area of CSPG4 (+/+) cells was consistently larger than (–/–) FTMs on both ‘dermis’ origins (Figure 2f), despite no significant differences in cell area (Figure 2g), which led to significantly larger nucleus : cell area ratios in (+/+) FTMs compared with (–/–) FTMs (Figure 2h).

Overall, there were clear differences in clinically relevant histopathological metrics among CSPG4-expressing and CSPG4-null FTMs, indicating that CSPG4 expression mediates sustained histopathological features of RDEB cSCC cells that are independent of underlying stromal fibroblasts in these models. However, this result does not account for changes in CSPG4 expression that might result from complex changes in the microenvironment associated with chronic inflammation, a hallmark of RDEB.

CSPG4 expression in non-RDEB cSCC is associated with an EMT transcriptomic signature

We used published scRNA-Seq data to assess the relationship between CSPG4 expression in non-RDEB cSCC cells and transcriptome signatures that accompany invasion.³⁷ Keratinocytes from healthy skin (Figure 3a, top panel) and cSCC tumours (Figure 3a, centre panel) were grouped into distinct populations based on gene expression signatures. Basal, cycling and differentiating keratinocytes were identified among both normal and tumour cells (Figure 3a, top and centre panels), but tumour cells contained a unique population of stroma-interfacing, tumour-specific keratinocytes (TSK),³⁷ with gene markers inconsistent with other keratinocytes populations (Figure 3a, b).

Figure 3.

CSPG4 expression in cutaneous squamous cell carcinoma (cSCC) tumour cells is enriched in the tumour-specific keratinocyte (TSK) population and is correlated with an epithelial-to-mesenchymal transition (EMT) transcriptomic phenotype. Analyses of healthy skin (n = 10 participants; n = 9779 total cells) and cSCC tumour (n = 7 patients; n = 4210 total cells) datasets originally published in Ji et al. were replicated as per their methods.³⁷ (a) Uniform Manifold Approximation and Projection (UMAP) plots showing keratinocyte populations in healthy (top panel) and tumour (centre panel) tissue. Bottom panel shows distribution of CSPG4-expressing cells (n = 229 cells) in the tumour dataset. (b) Heatmap showing the average expression level of hallmark gene markers and CSPG4 (left text) for each keratinocyte cluster (top text). (c) Cell counts in each tumour keratinocyte population (left panel) split by negative (grey) or positive (red) CSPG4 expression status. Proportion of CSPG4⁺ cells in each population (right panel). (d) CSPG4 expression level [log(normalized counts)] for every cell (left panel) or only CSPG4-expressing cells (right panel) in each tumour keratinocyte population. (e) EMT score for bulk tumour keratinocyte populations (left panel) or split by CSPG4 expression status (right panel). (f) Correlation plot between EMT score and CSPG4 expression level [log(normalized counts)] for all CSPG4⁺ cells. P-values determined by Student’s t-test with Bonferroni correction.

A minority of cSCC cells expressed CSPG4 (5.4% of total cells; Figure 3a, bottom panel). Although TSKs comprised 7.5% of total tumour cells (Figure 3c, left panel), they constituted 46.7% of CSPG4-expressing cells (P = 1.69 × 10^–64, Fisher’s exact test). Of the TSKs, one-third expressed CSPG4 vs. < 5% of cells in other keratinocyte populations (Figure 3c, right panel). The TSK population had significantly higher average CSPG4 expression compared with basal, cycling and differentiating keratinocytes (Figure 3d, left panel). Among CSPG4-expressing cells, TSKs did not have significantly different average CSPG4 expression from basal and differentiating keratinocytes (P = 0.60 and P > 0.99, respectively), but had significantly higher expression levels than cycling keratinocytes (Figure 3d, right panel).

In accordance with Ji et al.,³⁷ we found that TSKs had significantly greater enrichment of EMT-related genes (EMT score) compared with basal, cycling and differentiating tumour keratinocytes (Figure 3e, left panel). CSPG4-expressing keratinocytes had higher EMT scores than their population-matched CSPG4-negative counterparts (Figure 3e, right panel). This difference was statistically significant in all populations (P ≤ 0.002), with the exception of differentiating keratinocytes (P = 0.40), where a small number of CSPG4-positive cells (n = 5/1165 total cells) reduced the statistical power. Notably, CSPG4 expression level (log-transformed normalized counts) had a significantly positive correlation with EMT score (Figure 3f). These results indicate that CSPG4 expression is enriched in a unique population of stroma-interfacing tumour cells, where it is associated with an EMT transcriptomic phenotype and supports our identification of strong CSPG4 expression at the tumour–stroma boundary in RDEB cSCC.

CSPG4 enhances the invasive potential of RDEB cSCC cells in vitro

We used Matrigel invasion assays to determine whether CSPG4 expression promoted RDEB cSCC cell invasion. RDEB cSCC cell lines expressing high levels of CSPG4 were transfected with negative control or CSPG4-targeted siRNAs and evaluated for knockdown of CSPG4 (Figure 4a). After CSPG4 knockdown, both SCC2 and SCC4 cell lines exhibited a significant decrease in invasion through Matrigel-coated membranes in response to TGF-β1 stimulation [mean (SD) 88.1% (2%) and 41.3% (8%) reduction, respectively] compared with CSPG4-expressing cells (Figure 4b). A similarly significant reduction in invasive response to TGF-β1 was also observed in genetically engineered CSPG4-null (–/–) RDEB cSCC cells [mean (SD) 42% (7.7); Figure S3c], demonstrating the proinvasive effect of CSPG4 is independent of gene-inhibition technologies.

Figure 4.

Chondroitin sulfate proteoglycan 4 (CSPG4) enhances invasive potential in recessive dystrophic epidermolysis bullosa (RDEB) cutaneous squamous cell carcinoma (cSCC) cell lines. (a) Western blot showing knockdown of CSPG4 following transfection with CSPG4-targeted or negative control small interfering RNA (siRNA). (b) Invasion assay using SCC2 and SCC4 cell lines transfected with negative control (+ CSPG4) or CSPG4-targeted (– CSPG4) siRNAs. Plot shows the invasive capacity in response to transforming growth factor (TGF)-β1 (10 ng mL^–1). Bars represent the mean (SD) number of invaded cells from five random fields/well from ≥ 6 replicates. P-values were determined by a Student’s t-test with Holm–Šídák correction. (c) Proliferation data from SCC2 (left panel) and SCC4 (right panel) after transfection with negative control (+ CSPG4) or CSPG4-target (– CSPG4) siRNAs, serum-starved to stall proliferation, and then stimulated with or without TGF-β1 in complete growth medium for 48 h. Absorbance at 492 nm is directly proportional to cell count. Bars represent mean (SD) absorbance values (n = 6 replicates). P-values determined with a Student’s t-test. Ctrl, control; GAPDH, glyceraldehyde 3 phosphate dehydrogenase; Neg, negative.

We next assessed whether CSPG4 expression alters TGF-β1-driven proliferation in SCC2 and SCC4. We found that TGF-β1 promoted a slight (SCC2) to significant (SCC4) increase in cell numbers in CSPG4-expressing cells compared with CSPG4-knockdown cells (Figure 4c). However, the effect of CSPG4 expression on TGF-β1-stimulated growth did not completely account for differences observed in the invasion assays, especially in the SCC2 cell line, where loss of CSPG4 resulted in an 88.1% inhibition of invasion (Figure 4b). Overall, these results demonstrate that CSPG4 enhances the invasive potential of RDEB cSCC cell lines when stimulated by TGF-β1, a major contributor to RDEB cSCC invasion.

CSPG4 expression in RDEB cSCC enhances TGF-β1-stimulated SMAD3 activation

We assessed the impact of CSPG4 expression on membrane-proximal components of TGF-β signalling pathways. Although multiple SMAD proteins affect cSCC progression in context-dependent ways,⁴⁷ we chose to focus on SMAD3 as its activation primarily mediates TGF-β-induced inflammation.^48,49 Furthermore, SMAD3 activation has been associated with the secretion of ECM-degrading proteases and TGF-β1-stimulated migration of RDEB cSCC cell lines.⁵⁰

RDEB cSCC cell lines were subjected to signalling analysis as described in the ‘Materials and methods’. We compared levels of activated, phosphorylated SMAD3 (pSMAD3) with total SMAD3 levels via Western blot in SCC2 (Figure 5a) and SCC4 cells (Figure 5b). We quantified the ratio of pSMAD3 : SMAD3 for all conditions in both cell lines (Figure 5c, d). A linear regression model was fitted to log-transformed pSMAD3 : SMAD3 band intensity ratios, which allowed for the calculation of estimated marginal means (EMMs) for each siRNA and stimulation condition. Using the EMMs, the magnitude of SMAD3 activation in TGF-β1-stimulated conditions relative to baseline was ascertained as a function of CSPG4 expression.

Figure 5.

SMAD3 phosphorylation is enhanced in chondroitin sulfate proteoglycan 4 (CSPG4)-expressing SCC2 and SCC4 cell lines. Representative Western blots against phosphorylated SMAD3 (pSMAD3), total SMAD3 and glyceraldehyde 3 phosphate dehydrogenase (GAPDH) loading control in (a) SCC2 cells and (b) SCC4 cells 0.5, 2 and 6 h after transforming growth factor (TGF)-β1 stimulation. Cells were transfected with either negative control (Neg Ctrl) or CSPG4-targeted (CSPG4) small interfering RNAs (siRNAs); three replicates for each siRNA and TGF-β1 stimulation (+/–) condition. Quantification of pSMAD3 : SMAD3 band intensity ratios in (c) SCC2 cells and (d) SCC4 cells. pSMAD3 and SMAD3 band intensities were first normalized to GAPDH loading control, and then a ratio of pSMAD3 : SMAD3 was calculated for each sample. Samples transfected with Neg Ctrl siRNA (+ CSPG4) are indicated by red bars and CSPG4-targeted siRNA (– CSPG4) are indicated by blue bars, with each siRNA set containing samples without (‘SFM’) or with (‘TGF-β1’) TGF-β1 stimulation. Bars represent mean (SEM) band intensity ratio from three replicates. The magnitude of SMAD3 activation in TGF-β1-stimulated samples (‘TGF-β1’) relative to baseline (‘SFM’) was calculated at each timepoint for Neg Ctrl siRNA (+ CSPG4) and CSPG4-targeted siRNA (– CSPG4) conditions in (e) SCC2 cells and (f) SCC4 cells. These calculations were performed by fitting a linear regression model to log-transformed pSMAD3 : SMAD3 band intensity ratios. A magnitude value of 1 indicates no change in SMAD3 activation after TGF-β1 stimulation relative to baseline. Using the linear regression model, estimated marginal means were calculated for each siRNA and stimulation condition and post hoc statistical comparisons were performed to obtain P-values. Bars represent back-transformed means (SEM).

CSPG4-expressing SCC2 (Figure 5e) and SCC4 cells (Figure 5f) exhibited significantly increased levels of TGFβ1-induced SMAD3 activation compared with CSPG4-negative cells. These findings support a role of CSPG4-mediated activation of TGF-β signal transduction via SMAD3, which is linked to TGF-β1-stimulated RDEB cSCC invasion.⁵⁰

Discussion

CSPG4 is of prognostic and therapeutic value in many malignancies,^15–23,29 and impacts tumour invasion and progression in several ways. It facilitates ECM degradation;^51,52 acts as a co-receptor to affect major signalling pathways that regulate EMT;^29,53 and is associated with decreased sensitivity to chemotherapeutic agents.^22,32 Our study is the first to demonstrate that CSPG4 can function to promote TGF-β1-mediated invasion and signalling in RDEB cSCC. Furthermore, we identified several mechanisms unique to RDEB cSCC that may help to explain how CSPG4 can promote localized invasion and metastasis, adding to existing evidence supporting CSPG4-mediated enhancement of mesenchymal transition in invasive carcinomas.

As previously shown, we have also demonstrated that CSPG4 expression is restricted to subpopulations of basal keratinocytes in healthy skin.²⁵ Furthermore, this interspersed pattern of CSPG4 expression in basal keratinocytes was also seen in noninjured RDEB skin. This indicates that causative COL7A1 mutations in RDEB keratinocytes have no detectable impact on CSPG4 expression in the absence of injury.

Notably, increases in basal keratinocyte CSPG4 expression (in terms of average intensity and frequency) are observed in chronically inflamed/damaged RDEB skin, where RDEB cSCC typically develops.⁵ These areas are characterized by chronic inflammation and impaired wound healing prior to tumour formation, resulting in fibrosis marked by increased secretion of protumorigenic cytokines (e.g. tumour necrosis factor-α, interleukin-6 and TGF-β1) and hypoxia.^6,7,54 These and other features associated with an inflamed and fibrotic dermis may induce the uniform localization of CSPG4-expressing cells at the stromal interface in fibrotic RDEB skin and in RDEB cSCC tumours. This pattern – compared with heterogeneous, interspersed expression of CSPG4 in basal keratinocytes within nonfibrotic RDEB skin and healthy skin – emphasizes the unique and important influence of a chronically inflamed and fibrotic dermis on keratinocyte CSPG4 expression and function in RDEB disease progression (Figure 6).

Figure 6.

Chronic inflammation and fibrosis increase chondroitin sulfate proteoglycan 4 (CSPG4) expression in recessive dystrophic epidermolysis bullosa (RDEB) skin and promotes RDEB cutaneous squamous cell carcinoma (cSCC) progression. RDEB skin without a history of chronic injury has similar CSPG4 expression patterns to healthy skin, with expression limited to clusters of basal-layer epidermal stem cells (left panel). However, upon cycles of chronic injury at the dermal–epidermal junction (orange bolts) and inflammation because of innate defects in adhesion and wound-healing programmes, RDEB skin becomes increasingly fibrotic (centre panel). This dermal phenotype is marked by a stiff extracellular matrix (collagens and α-smooth muscle actin) assembled primarily by myofibroblasts and excessive secretion of inflammatory cytokines [e.g. transforming growth factor (TGF)-β1, interleukin (IL)-6, tumour necrosis factor (TNF)-α] by infiltrating immune cells. At the onset of fibrosis, CSPG4 expression in basal-layer keratinocytes transitions from heterogeneous to homogeneous. This change may be because CSPG4 expression is upregulated in response to factors and conditions present in fibrotic tissue,^24,27 or due to expansion/proliferation of CSPG4-expressing cells in response to inflammatory factors (see Figure 4c). When RDEB skin with underlying fibrosis develops RDEB cSCC, a strong uniform expression of CSPG4 by stroma-interfacing cells is retained (right panel) – potentially due to epigenetic modifications,¹⁷ or stimulation by inflammatory factors. The increased bioavailability of TGF-β1 in RDEB stroma before and after RDEB cSCC development functions to promote tumour cell invasion and/or metastasis through, in part, CSPG4-dependent mechanisms described in this work.

A chronically inflamed dermis is a well-recognized contributing factor to the development of RDEB cSCC, and its cellular and secretory composition changes as it undergoes cycles of damage and unresolved inflammation. These cycles include the synthesis of numerous inflammatory mediators that affect multiple cell types present in the dermis, such as infiltrating immune cells, vascular endothelial cells and fibroblasts. The resulting fibrotic environment consists of ECM-depositing myofibroblasts and a stiff, cross-linked collagenous matrix, into which RDEB cSCC cells readily invade.^3,5 Therefore, we developed an in vitro FTM as a tool to evaluate specifically the ability of dermal fibroblasts (in the absence of an inflammatory response) to modify the malignant phenotype of RDEB cSCC cells expressing or lacking CSPG4. In contrast to other models that utilize fibroblasts seeded into fibrin or collagen gels,^55–57 fibroblasts within these FTMs synthesize an ECM reminiscent of their tissue of origin and unique to RDEB pathology. We determined that CSPG4-expressing RDEB cSCC cells exhibited cytopathological criteria (e.g. increased levels of dyskeratosis and increased nucleus : cell area ratios)^44–46,58 consistent with a sustained malignant phenotype vs. CSPG4-null cells. The results demonstrate that CSPG4 expression and not the fibroblast source (RDEB or healthy dermis) is responsible for sustaining the malignant phenotype of RDEB cSCC cells over the extended culture period of the model.

CSPG4 is a ‘part-time’ proteoglycan, with various levels of the CSPG4 core protein expressed with or without chondroitin sulfate modification. This is a phenomenon we also saw in the RDEB cSCC cells used in this study (Figure 1a). The CSPG4 core protein does not signal intrinsically, but functions as an adaptor to indirectly enhance the activation of multiple oncogenic signalling pathways.²⁹ For example, CSPG4 enhances α4β1 integrin-mediated FAK activation in melanoma indirectly by the binding of the CSPG4 core protein cytoplasmic domain to syntenin.³⁰ Syntenin is an intracellular adaptor protein that promotes the progression and invasion of multiple tumour types.^59–62 It has two tandemly associated PDZ domains that can interact with a variety of membrane-localized proteins to regulate the architecture of the cell membrane, signal transduction and endocytosis.⁶³

Previously, syntenin has been shown to affect TGF-β1-induced SMAD2/3 activation and mesenchymal transition in cancer cells.⁶⁴ There, the loss of syntenin inhibited TGF-β1-stimulated cell migration and SMAD2/3-mediated transcriptional responses. The authors demonstrated that the carboxyl terminus of TGF-βRI binds syntenin and that knockdown of syntenin promoted caveolin-mediated internalization of TGF-βRI but not TGF-βRII.⁶⁴ Thus, syntenin could enhance the activation of SMADs by limiting TGF-βRI internalization. Because TGF-βRI is directly responsible for phosphorylating and activating SMAD3, we propose that a CSPG4–syntenin–TGF-β receptor complex may indirectly enhance SMAD3 activation by reducing the ligand-induced endocytosis of the TGF-β receptor complex.^65,66

In summary, we have demonstrated that keratinocyte CSPG4 expression in RDEB disease pathobiology is tightly related to chronic injury and inflammation, resulting in dermal fibrosis that both precedes and drives RDEB cSCC progression. Further, we have shown that CSPG4 in RDEB cSCC can mediate an invasive/prometastatic tumour cell phenotype potentially by mediating changes in membrane-proximal TGF-β1-mediated signalling. This work collectively supports the concept of RDEB as a stroma-driven disease and provides new insights into mechanisms by which CSPG4 can mediate RDEB cSCC development and progression.

Funding sources

This research received clinical specimen assistance from the University of Minnesota’s Biorepository and Laboratory Services programme. This research was conducted with funding support from National Institutes of Health grant NHLBI R01 AR063070 (J.T.) and the NIH National Center for Advancing Translational Sciences grants UL1TR002494 (Biorepository and Laboratory Services, C.L.E.) and KL2TR002492 (C.L.E.). J.B.M. is funded by the Atwater Fund, Elsa U. Pardee Foundation and Chairman’s Fund Professorship in Cancer Research. The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH or the NIH National Center for Advancing Translational Sciences.

Data availability

Additional data and code are available upon reasonable request. Single-cell RNA sequencing data are publicly available in the National Center for Biotechnology Information Gene Expression Omnibus (https://www.ncbi.nlm.nih.gov/geo/) with Gene Expression Omnibus accession number GSE144236.

Ethics statement

This study was approved by the University of Minnesota Institutional Review Board (protocol MT2013-01R).

Patient consent

Primary fibroblasts and keratinocytes were derived from punch biopsies from healthy donors or patients with recessive dystrophic epidermolysis bullosa following informed consent.

Supporting Information

Additional Supporting Information may be found in the online version of this article at the publisher’s website.