Suppression of TGFβ and Angiogenesis by Type VII Collagen in Cutaneous SCC

Abstract

Background:

Individuals with severe generalized recessive dystrophic epidermolysis bullosa (RDEB), an inherited blistering disorder caused by mutations in the COL7A1 gene, develop unexplained aggressive squamous cell carcinomas (SCC). Here we report that loss of type VII collagen (Col7) in SCC results in increased TGFβ signaling and angiogenesis in vitro and in vivo.

Methods:

Stable knockdown (KD) of Col7 was established using shRNA, and cells were used in a mouse xenograft model. Angiogenesis was assessed by immunohistochemistry, endothelial tube-forming assays, and proteome arrays. Mouse and zebrafish models were used to examine the effect of recombinant Col7 on angiogenesis. Findings were confirmed in anonymized, archival human tissue: RDEB SCC tumors, non-EB SCC tumors, RDEB skin, normal skin; and two human RDEB SCC cell lines. The TGFβ pathway was examined using immunoblotting, immunohistochemistry, biochemical inhibition, and siRNA. All statistical tests were two-sided.

Results:

Increased numbers of cross-cut blood vessels were observed in Col7 KD compared with control xenografts (n = 4 to 7 per group) and in RDEB tumors (n = 21) compared with sporadic SCC (n = 24, P < .001). Recombinant human Col7 reversed the increased SCC angiogenesis in Col7 KD xenografts in vivo (n = 7 per group, P = .04). Blocking the interaction between α2β1 integrin and Col7 increased TGFB1 mRNA expression 1.8-fold and p-Smad2 levels two-fold. Increased TGFβ signaling and VEGF expression were observed in Col7 KD xenografts (n = 4) compared with control (n = 4) and RDEB tumors (TGFβ markers, n = 6; VEGF, n = 17) compared with sporadic SCC (TGFβ markers, n = 6; VEGF, n = 21). Inhibition of TGFβ receptor signaling using siRNA resulted in decreased endothelial cell tube formation (n = 9 per group, mean tubes per well siC = 63.6, SD = 17.1; mean tubes per well siTβRII = 29.7, SD = 6.1, P = .02).

Conclusions:

Type VII collagen suppresses TGFβ signaling and angiogenesis in cutaneous SCC. Patients with RDEB SCC may benefit from anti-angiogenic therapy.

Type VII collagen (Col7) is the major component of anchoring fibrils, structures anchoring the skin epidermis to the dermis, important in wound healing and as a barrier to epithelial cancer invasion (1,2). In RDEB, a genetic skin disease characterized by blistering, scarring, and aggressive SCC, expression of Col7 and anchoring fibrils is greatly decreased (3,4). Data from a Ras/IκBα-driven tumorigenesis model suggested that the noncollagenous (NC1) domain of Col7 was necessary for tumor formation by RDEB keratinocytes (5). However, some RDEB tumors do not have Col7 expression (6). We have previously shown, by transient KD of Col7 in a 3D model, that loss of Col7 promotes SCC invasion and disorganized differentiation in vitro (7). Two recent studies have established that Col7 directly regulates the dermal extracellular matrix (ECM) composition in RDEB (8,9).

Expression of Col7 is regulated by TGFβ at the transcriptional level (10,11), and in a hypomorphic mouse model of RDEB (10% Col7) expression of dermal active TGFβ1 is enhanced (12). TGFβ is known to promote epithelial-mesenchymal transition (EMT) and induce angiogenesis (13). Binding of TGFβ activates TGFβ receptor II (TβRII), which then recruits and phosphorylates TGFβ receptor I (TβRI) (14). Latent TGFβ1 is activated by αvβ6 integrin for presentation to TGFβ receptors (15). Conversely, TGFβ1 induces expression of αvβ6 integrin in keratinocytes (16). The αvβ6 integrin plays a critical role in promoting TGFβ-dependent EMT (17) and invasion (18,19).

In the current study, we developed stable Col7 knockdown SCC cells and performed xenografts of 3D cultures in SCID nude mice to establish the effect of loss of Col7 in SCC in vivo. We found that loss of Col7 enhances angiogenesis in SCC in mouse xenografts, RDEB skin, and SCC tumors. Increased tumor angiogenesis was reversed by human recombinant Col7 protein therapy in both mouse and zebrafish xenograft models. Blocking the interaction between α2β1 integrin and Col7 enhanced TGFβ signaling. Increased TGFβ signaling was observed in RDEB tumors compared with sporadic SCC. Inhibition of TGFβ receptor signaling resulted in decreased endothelial cell tube formation and decreased VEGF secretion. These data suggest that Col7 acts as a tumor suppressor by reducing TGFβ signaling to inhibit angiogenesis.

Methods

Patients

The use of anonymized, archival human tumor specimens was conducted according to the Declaration of Helsinki principles and approved by the appropriate local ethics committees.

Ethical Regulations

All mouse experiments were performed in accordance with UK Home Office regulations and the European Legal Framework for the protection of animals used for scientific purposes (European Directive 86/609/EEC).

Antibodies and Recombinant Proteins

The rabbit polyclonal anti-Col7 antibody was purified as described before (20). Antibodies used in this work are described in Supplementary Table 1 (available online). Human recombinant full-length type VII collagen protein (hrCol7) was purified as described previously (21). Mouse type IV collagen protein (Col4) was purchased from BD Biosciences (Oxford, UK).

Generation of shRNA-Transduced Cell Lines

Stable knockdown of type IV and type VII collagens was established in a cutaneous SCC cell line as described in the Supplementary Methods (available online). Sequences of different shRNA clones and quantitative polymerase chain reaction (qPCR) primers are in Supplementary Tables 2 and 3, respectively (available online).

In Vitro Epidermal Models

Organotypic cultures on collagen:matrigel gels were performed as previously described with some modifications (Supplementary Methods, available online) (22).

Xenografting of Collagen:Matrigel Gels Onto Hairless SCID Mice

For xenografting of collagen:matrigel gels in mice, in vitro organotypic cultures on collagen:matrigel gels with shCOL7, shCOL4 and shC cells were performed as described above. After a seven-day incubation period in vitro, collagen:matrigel gels were xenografted subcutaneously onto hairless SCID Crl:SHO-Prkdc^scidHr^hrmice (four replicates per condition). Mice were kill after six weeks, and gels (including surrounding fibrotic tissue and tumors formed) were excised, cut in half, and either snap frozen or fixed in formal saline for histological analysis. Treatment with hrCol7 was performed by injecting 30 μg of hrCol7 (or PBS as a control) using a 30-gauge needle around the edge of the implanted gel at days 7 and 21 postimplantation. Seven replicates per condition were performed.

Fluorescence and DAB Immunostaining

Archival skin sections were obtained of formalin-fixed and paraffin-embedded human RDEB and non-EB SCCs (Supplementary Table 4, available online). Tissue microarray blocks were generated previously from RDEB and non-EB SCC (23). Immunostaining is described in the Supplementary Methods (available online).

In Vitro HUVEC Tube Formation Assay

An endothelial cell tube formation assay in Matrigel was performed as described in the Supplementary Methods (available online).

Human Angiogenesis Proteome Array

A Proteome Profiler Human Angiogenesis Antibody Array (R&D Systems, UK) was used to detect expression of angiogenesis-related proteins in shC and shCOL7 SCC-IC1 cells as described in the Supplementary Methods (available online).

siRNA Transfection

For TβRI and TβRII knockdown, the EB3K cell line was transfected with synthetic siRNAs (Dharmacon, UK), targeting TβRI or TβRII (Supplementary Table 5 and Supplementary Methods, available online).

FACS Labeling of α2 Integrin in SCC and α2 Integrin Inhibition

Labeling of the α2 integrin receptor by FACS and use of a neutralizing antibody are described in the Supplementary Methods (available online).

TGF-β–Dependent Reporter Assay

Cells were cotransfected with CAGA₁₂-Luciferase (containing Smad binding elements [SBE]) and pRenilla (pRr-TK) (24) at a 10:1 ratio using Fugene 6 (Promega, Southampton, UK). After 24 hours, transfected cells were stimulated for 18 hours with 2ng/mL TGFβ1. Luciferase and Renilla luminescence were read using a dual luciferase reporter assay kit (Promega) following the manufacturer’s instructions. Luciferase was normalized to the Renilla control to give relative luminescence for each sample expressed in arbitrary units (A/U).

VEGF ELISA

Secreted VEGF was detected using the Human VEGF Quantikine ELISA Kit (R&D Systems, Abingdon, UK). Conditioned media was collected from cells treated for 24 hours with the TGF-beta type I receptor inhibitor, SD-208 (2 µM) or DMSO vehicle control. VEGF levels were compared with a standard curve of recombinant VEGF.

Zebrafish Embryo Xenograft Studies

EB3K cells stained with CMTMR (Invitrogen, Paisley, UK) were injected into the yolk sac of 24-hour-old transgenic fli1:EGFP (enhanced green fluorescent protein) embryos as described in the Supplementary Methods (available online).

Statistical Methods

The Student’s t test was used for two variables analyses. Fisher’s Exact test was used for comparisons between groups. All statistical tests were two-sided, and a P value of less than .05 was considered statistically significant. Statistical analysis was performed using Microsoft Excel and IBM SPSS Statistics for Windows, Version 22.0. Armonk, NY: IBM Corp.

Results

The Effect of Loss of Col7 on Angiogenesis

A cutaneous SCC cell line, SCC-IC1, was transduced with shRNA lentiviral particles targeting Col7 mRNA (shCOL7), Col4 mRNA (shCOL4), or a nontargeting control (shC). The shRNA shCOL7_3 statistically significantly reduced expression of Col7 in SCC-IC1 (Supplementary Figure 1, A and B, available online). Stable knockdown of Col4 was also established (Supplementary Figure 1, C and D, available online). Cells with KD of Col7 became more elongated with loss of cell-cell adhesion (Supplementary Figure 1E, available online) suggesting EMT. To establish the effect of absent epithelial Col7 in SCC invasion, organotypic cultures with shCOL7 and shC SCC cells were xenografted onto SCID mice (Supplementary Figure 2A, available online). Surprisingly, we observed that shCOL7 xenografts were pinker, with an increased number of tortuous blood vessels (BV) entering the gel six weeks postimplantation, compared with shC or shCOL4 xenografts (Figure 1A). Hematoxylin and eosin staining demonstrated that shCOL7 cells displayed increased invasion. Involucrin and Keratin 10 staining were reduced in shCOL7, but not in shCOL4 xenografts, suggesting that loss of Col7 reduces differentiation (Supplementary Figure 2B, available online).

Figure 1.

The effect of loss of type VII collagen on blood vessel formation in squamous cell carcinoma (SCC) xenografts. A) Photomicrographs of nontargeting control (shC), type VII collagen knockdown (shCOL7), and type IV collagen knockdown (shCOL4) gels excised from mice showing blood vessel formation. The scale bars represent 5mm. B) Immunofluorescence staining of shC, shCOL7, and shCOL4 xenografts with an antibody to the blood vessel marker Meca-32 and Cell Profiler output. DAPI (4’,6-diamidino-2-phenylindole) was used as a nuclear stain. C) Quantification of vessel number and size using Cell Profiler. Data are mean ± SD, n = 45 sections from four animals per group and are representative of two independent experiments. Statistical analysis was performed using two-tailed Student’s t tests. **P < .01, ***P < .001. The scale bar represents 500 µm. n.s. = not statistically significant.

Immunostaining using Meca-32, a murine endothelial cell marker, demonstrated that shCOL7 xenografts were more vascular than shC and shCOL4 (Figure 1, B and C). To see if this was clinically relevant, immunostaining of the endothelial marker CD31 in 21 RDEB SCC tumors that lacked Col7 (Supplementary Table 4, available online) was performed and demonstrated that RDEB SCC tumors had statistically significantly increased numbers of cross-cut BVs (mean BV number, RDEB SCC: 198, SD = 14; non-EB SCC: 107, SD = 12; P < .001) that were also larger in size compared with 24 non-EB SCC tumors (mean BV diameter, RDEB SCC: 53, SD = 7; non-EB SCC: 34, SD = 4; P = .025) (Figure 2, A and B). Interestingly, RDEB patient scarred skin was also more vascular than normal skin (mean BV number, RDEB skin: 121, SD = 21; non-EB skin: 40, SD = 7; P < .001) (n = 8 of each) (Figure 2, C and D).

Figure 2.

Blood vessel size and number in recessive dystrophic epidermolysis bullosa (RDEB) squamous cell carcinoma (SCC) and RDEB scarred skin. A) Representative sections of RDEB and non-epidermolysis bullosa (Non-EB) squamous cell carcinoma tumors stained with an antibody to CD31. B) Quantification of vessel number and size using Cell Profiler. Data are mean ± SD; n = 21 (RDEB SCC tumors); n = 24 (Non-EB SCC tumors). C) Representative sections of RDEB patient and normal skin stained with an antibody to CD31. D) Quantification of vessel number and size using Cell Profiler data are mean ± SD; n = 8 (RDEB skin); n = 8 (Non-EB skin). Statistical analysis was performed using two-tailed Student’s t tests. ***P < .001. The scale bars represent 100 μm. Non-EB = non-epidermolysis bullosa; RDEB = recessive dystrophic epidermolysis bullosa; SCC = squamous cell carcinoma.

Loss of Col7 and Epithelial VEGF Expression

To assess whether BV formation was promoted directly by SCC shCOL7 cells, we performed an endothelial tube formation assay in vitro using HUVEC cells. Media from SCC-IC1 shCOL7 cells increased tube formation (mean tubes per well, plain media: 19.9, SD = 2.6; shC conditioned media: 22.2, SD = 4.5; shCOL7 conditioned media: 31.8, SD = 4.3; P = .047), suggesting a possible role for secreted growth factors (Figure 3, A and B). An angiogenesis proteome array identified possible angiogenic factors in shCOL7 cells, including VEGF, confirmed by quantitative real-time PCR (qRT-PCR) (Figure 3C; Supplementary Figure 3A, available online). Immunostaining showed that VEGF expression was markedly increased in RDEB SCC epithelium, as well as at the invasive front in shCOL7 xenografts, which may explain increased angiogenesis in the context of loss of Col7 (Figure 3D; Supplementary Figure 3, B-D, available online). Interestingly, RDEB patients have elevated levels of urinary basic fibroblast growth factor (bFGF) (25), a known regulator of VEGF-induced angiogenesis (26).

Figure 3.

Loss of Col7 and epithelial vascular endothelial growth factor (VEGF) expression. A) Endothelial tube formation assay in vitro showing brightfield images and the Cell Profiler output of human umbilical vein endothelial cells (HUVECs) after incubation with conditioned media (CM) from type VII collagen knockdown (shCOL7) or nontargeting control (shC) cells compared with unconditioned media (plain media). B) The number of tubes per well was quantified using Cell Profiler. Data are mean ± SD; n = 9. C) A Proteome Profiler Human Angiogenesis Antibody Array comparing expression of angiogenesis-related proteins between shCOL7 and shC cells in duplicate. Quantification shown in the bottom panel and proteins with observed increases are highlighted (VEGF, SerpinE1 [SERPE1], Thrombospondin-1 [TSP-1], Urokinase plasminogen activator [uPA], Tissue inhibitor of metalloproteinase-1 [TIMP-1], Amphiregulin [AREG]). D) Representative recessive dystrophic epidermolysis bullosa squamous cell carcinoma (RDEB SCC), n = 17 and non-epidermolysis bullosa (Non-EB) SCC (n = 21) tumors stained with an antibody to VEGF. DAPI (4’,6-diamidino-2-phenylindole) was used as a nuclear stain. The scale bars represent 100 μm. Statistical analysis was performed using two-tailed Student’s t tests. **P < .01. Non-EB = non-epidermolysis bullosa; RDEB = recessive dystrophic epidermolysis bullosa; SCC = squamous cell carcinoma; VEGF = vascular endothelial growth factor.

Recombinant Col7 Therapy in Mouse and Zebrafish Models

Human recombinant Col7 (hrCol7) is incorporated into the basement membrane (BM) upon intradermal injection and is a proposed treatment for RDEB (27). We injected hrCol7 diluted in PBS into shCOL7 SCC xenografts at weeks 1 and 3 after grafting and compared with shCOL7 SCC xenografts injected with PBS alone. We observed that hrCol7 injection resulted in a statistically significant reduction in BV size in shCOL7 xenografts (mean BV size, shC: 96.3, SD = 17.0; shCOL7: 202.3, SD = 42.8; P = .04; shCOL7 + hrCol7: 98.5, SD = 28.5; P = .04) (Figure 4, A and B). Additionally, in an in vivo zebrafish xenograft rescue model (28,29), Col7-null EB3K cells were microinjected with or without hrCol7 into the yolk sac of one-day-old embryos (n = >13 per group). At Day 3 after injection, EB3K cells stimulated angiogenesis as demonstrated by intratumoral vascularization and hrCol7 reversed this phenotype (Figure 4C; Supplementary Figure 3, E and F, available online).

Figure 4.

Human recombinant Col7 (hrCol7) therapy in mouse and zebrafish models. A) Immunofluorescence staining of sections of nontargeting control (shC), type VII collagen knockdown (shCOL7), and shCOL7 + hrCol7 xenografts with an antibody to the blood vessel marker, Meca-32. Levels of type VII collagen (Col7) expression in the different samples is demonstrated in the lower panel. DAPI (4’,6-diamidino-2-phenylindole) was used as a nuclear stain. The scale bars represent 100 μm. B) Quantification of vessel size using Cell Profiler. Statistical analysis was performed using two-tailed Student’s t tests. Data are mean ± SD; n = 10 sections per animal from seven animals per group. C) Three-dimensional micrographs of confocal images of four-day-old zebrafish embryos injected with recessive dystrophic epidermolysis bullosa squamous cell carcinoma (RDEB squamous cell carcinoma [SCC]) EB3K cells (red) into the yolk sac at Day 1. Blood vessels are shown in green. Embryos were injected either with hrCol7 protein (bottom panel, n = 13) or phosphate-buffered saline (PBS) (top panel, n = 15). Statistical analysis was performed using two-tailed Student’s t tests. *P < .05. The scale bars represent 70 μm.

Interruption of the Col7-α2β1 Integrin Interaction and TGFβ Signaling

Type VII collagen binds to the α2β1 integrin on human dermal fibroblasts (30). The α2 integrin receptor is expressed on the surface of cutaneous SCC cells (Figure 5A), and an adhesion assay demonstrated that an α2-neutralizing antibody reduced RDEB SCC adhesion to hrCol7 by 50%, suggesting a direct role in cell attachment to Col7 (relative cell adhesion normalized to BSA-α2 Integrin, hrCol7-α2 Integrin: 54%, SD = 7.4%, P = .006; Col4-α2 Integrin: 66%, SD = 6.6%, P < .001) (Figure 5B). Adhesion to hrCol7, but not BSA or Col4, reduced expression of p-Smad2 and αvβ6 integrin and kindlin 2, known components of the TGFβ signaling pathway in RDEB SCC (Figure 5C; Supplementary Figure 4D, available online) (31). No change was observed in TβRII expression. Blocking the α2 integrin increased p-Smad2 and αvβ6 integrin expression on hrCol7 but not on BSA or Col4. Similar changes were demonstrated in a SBE luciferase reporter assay (Figure 5D). Increased SBE luciferase activity was also observed in shCol7 cells compared with shC with or without TGFβ1, the latter because of a response to autocrine TGFβ1 (luciferase luminescence normalized to renilla control, shC: 2.6, SD = 0.05; shCOL7: 4.7, SD = 0.046, P < .001, shC+TGFβ: 17.2, SD = 1.54; shCOL7+TGFβ: 45.9, SD = 4.17, P < .001) (Figure 5E). TGFB1 mRNA expression decreased 1.7-fold in RDEB SCC cells plated on hrCol7 compared with BSA. Blocking α2 integrin increased TGFB1 expression 1.8-fold on hrCol7 but not BSA (Figure 5F). These data suggest that binding of Col7 to the α2 integrin in SCC keratinocytes downregulates TGFβ1 signaling, a known regulator of cell invasion, EMT, and angiogenesis.

Figure 5.

Interruption of the Col7-α2β1 integrin interaction and TGFβ signaling. A) Fluorescence-activated cell sorting plots showing relative phycoerythrin fluorescence in squamous cell carcinoma (SCC)–IC1 cells stained with immunoglobulin G (IgG) isotype control and α2 integrin antibody. B) Adhesion of recessive dystrophic epidermolysis bullosa squamous cell carcinoma (RDEB SCC) cells to human recombinant collagen VII (hrCol7), type IV collagen (Col4), or bovine serum albumin (BSA) substrates in the presence of an α2 integrin–blocking antibody (Col7 – α2 integrin; Col4 – α2 integrin; and BSA – α2 integrin); n = 3 from two independent experiments. All samples are normalized to IgG control. C) Western blotting analysis of p-SMAD2, SMAD2, and transforming growth factor receptor II (TβRII) in EB3K cells plated on BSA, hrCol7, or Col4 in the presence of an α2 integrin–blocking antibody or control IgG. GAPDH expression was used as an internal control of protein loading. For densitometric analysis, results were normalized to GAPDH and are expressed as fold induction over BSA-IgG. This is a representative blot of three independent experiments. D) Assay showing changes in CAGA₁₂ (Smad-binding element)-luciferase SMAD reporter activity in EB3K cells plated on BSA, hrCol7, or Col4 in the presence of an inhibitory antibody to the α2 subunit of α2β1 integrin or control IgG. Luciferase luminescence was normalized to the Renilla transfection control. E) Assay showing changes in CAGA₁₂-luciferase SMAD reporter activity in shC and shCOL7 cells in response to 2ng/mL of TGFβ1. F) Real-time quantitative reverse transcription analysis of TGFB1 in RDEB SCC EB3K cells incubated with a neutralizing antibody to α2 integrin and then plated on Col7 or BSA for 2.5 hours. Statistical analysis was performed using two-tailed Student’s t tests. *P < .05, **P < .01, and ***P < .001.

TGFβ Signaling in shCOL7 SCC Cells and RDEB Tumors

Analysis of TGFβ signaling and EMT in vivo demonstrated increased expression of active TGFβ1, its receptor, TβRI, as well as p-Smad2/3 in invasive cells in the stroma, in shCOL7 xenografts (Supplementary Figure 4A, available online). Vimentin, a marker of EMT, was notably increased in stromal invasive epithelial cells, colocalizing with GFP, in shCOL7 xenografts (Supplementary Figure 4, A and B, available online). Expression of αvβ6 integrin and fibronectin, a ligand of αvβ6 (19,32), was also increased in shCOL7 but not in shCOL4 xenografts (Figure 6A), suggesting that the TGFβ-driven increased expression of fibronectin and αvβ6 integrin in RDEB SCC is specific to loss of Col7. The increase in αvβ6 integrin expression was also confirmed using a second shRNA clone targeting Col7 (Supplementary Figure 4C, available online).

Figure 6.

The effect of loss of type VII collagen on TGFβ signaling in xenograft and organotypic models. A) Immunofluorescence staining of sections of nontargeting control (shC), type VII collagen knockdown (shCOL7), and type IV collagen knockdown (shCOL4) squamous cell carcinoma (SCC)–IC1 xenografts with antibodies to the TGFβ targets fibronectin and αvβ6 integrin. DAPI (4’,6-diamidino-2-phenylindole) was used as a nuclear stain. Upper panel shows hematoxylin and eosin staining of representative sections. Immunofluorescence staining of sections of shC and shCOL7 SCC-IC1, and EB2K and EB3K cell line–derived organotypic cultures with antibodies to αvβ6 integrin (B) and fibronectin (C), incubated with the transforming growth factor receptor I (TβRI) inhibitor SD-208 (2 µM) or with the vehicle control, dimethyl sulfoxide (DMSO). DAPI was used as a nuclear stain. D) Invasion index for organotypic cultures on collagen:matrigel gels with shC and shCOL7 SCC-IC1 and EB2K and EB3K cells. Cultures were incubated with either the TβRI inhibitor SD-208 (2 µM) or the vehicle control DMSO. Data are mean ± SD; n = 3. Statistical analysis was performed using two-tailed Student’s t tests. *P < .05 and **P < .01. The scale bars represent 100 µm in (A), (B), and (C). H&E = hematoxylin and eosin; n.s. = not statistically significant.

To understand whether TGFβ mediates SCC invasion in RDEB, we performed in vitro 3D organotypic cultures with SCC-IC1 shC and shCOL7 cells, and RDEB SCC cell lines EB2K (reduced Col7 expression) and EB3K (no Col7 expression), in the presence of SD-208, a specific TβRI inhibitor known to inhibit primary tumor growth and metastasis in vivo (33). Expression of αvβ6 integrin and its ligand fibronectin was increased in shCOL7 organotypics compared with shC, as well as in two RDEB SCC cell lines, and reduced by SD-208 (Figure 6, B and C). The shCOL7 cells had a higher invasion index compared with shC cells, which was statistically significantly reduced by SD-208. Additionally, SD-208 statistically significantly reduced invasion of the two RDEB cell lines (Figure 6D).

To explore the clinical relevance of increased TGFβ signaling in human tumors, immunostaining of TβRI, Twist, fibronectin, and αvβ6 integrin was performed in SCC tumors from six RDEB patients lacking Col7 expression and six sporadic SCC patients. All markers were expressed at high levels in RDEB SCC tumors and were absent, or present at low levels, in non-EB SCC tumors (Figure 7A). Conditioned media from the EB3K cell line with transient knockdown of TβRII resulted in a statistically significant decrease in tube formation (mean tubes per well, plain media: 40.9, SD = 7.6; siC conditioned media: 63.6, SD =17.1; siTβRI conditioned media: 55.8, SD = 5.5; siTβRII conditioned media: 29.7, SD = 6.1; P = .02; siTβRI+ siTβRII conditioned media: 21.9, SD = 6.4; P = .02) (Figure 7B; Supplementary Figure 4, E and F, available online), supporting the notion that TGFβ signaling from tumor cells is regulating angiogenesis in Col7-null SCC. Increased secreted VEGF was observed in SCC-IC1 shCOL7 cell media, which was inhibited by SD-208, suggesting that increased VEGF-driven angiogenesis is mediated by TGF-β (Figure 7C). Finally, TGF-β–driven nuclear Smad4 signaling is a known mediator of angiogenesis (34,35). Immunostaining of ShC xenografts showed intense cytoplasmic staining and some nuclear staining of Smad4. In contrast, in shCOL7 xenografts there was less cytoplasmic Smad4 and nuclear Smad4 was present at the invading edge of the tumor. Injection of hrCol7 partially restored cytoplasmic Smad4 staining throughout the tumor. (Figure 7D). Immunostaining of RDEB and non-EB SCC tumors showed increased nuclear Smad4 in RDEB tumor sections (Figure 7E) consistent with our hypothesis that loss of Col7 is driving TGF-β–induced angiogenesis in RDEB.

Figure 7.

TGFβ signaling in recessive dystrophic epidermolysis bullosa (RDEB) tumors and effect of modulation of TGFβ on endothelial tube formation and VEGF expression. A) Immunofluorescence staining of squamous cell carcinoma (SCC) skin sections from three representative recessive dystrophic epidermolysis bullosa (RDEB) SCC tumors (total n = 6) with antibodies to transforming growth factor receptor I (TβRI), twist, fibronectin, and αvβ6 integrin. Non-epidermolysis bullosa (Non-EB) SCC tumors were used as a control (n = 6). DAPI (4’,6-diamidino-2-phenylindole) was used as a nuclear stain. Staining was quantified and relative integrated intensity is shown for each antibody. A/U = arbitrary units. B) Endothelial tube formation assay in vitro with conditioned media (CM) from RDEB SCC EB3K cells with knockdown of transforming growth factor receptors, TβRI (siTβRI), TβRII (siTβRII), or double knockdown (siTβRI+II). Data are mean ± SD; n = 9 per group, from three independent experiments. C) Vascular endothelial growth factor (VEGF) ELISA of conditioned media from type VII collagen knockdown (shCOL7) and nontargeting control (shC) cells incubated with the TβRI inhibitor, SD-208 (2 µM) or with the vehicle control, dimethyl sulfoxide (DMSO). Data are mean ± SD; n = 3. D) Staining of sections of shC, shCOL7, and shCOL7 + hrCol7 (human recombinant collagen 7) xenografts with an antibody to SMAD4. E) Representative sections of RDEB SCC (n = 17) and Non-EB SCC (n = 10) tumors stained with an antibody to SMAD4. The scale bars in (A), (D), and (E) represent 100 μm. Statistical analysis was performed using two-tailed Student’s t tests. *P < .05, **P < .01, and ***P < .001. n.s. = not statistically significant. Non-EB = non-epidermolysis bullosa; RDEB = recessive dystrophic epidermolysis bullosa; SCC = squamous cell carcinoma; VEGF = vascular endothelial growth factor.

Discussion

Type VII collagen is an important component of epithelial basement membranes. Here, we describe that Col7 is a major suppressor of angiogenesis in skin SCC. The increased expression of TβRI, Twist, αvβ6 integrin, and its ligand fibronectin, an EMT inducer (36), in RDEB SCC tumors and shCol7 xenografts indicates a major role for TGFβ signaling in RDEB tumorigenesis, which is regulated by Col7. This is further confirmed by the suppression of p-Smad2, αvβ6, and kindlin2 expression, mediators of TGFβ signaling, in Col7-negative RDEB SCC plated on hrCol7 matrix.

Limitations of this study include the use of an shRNA model in UV-induced SCC cell lines for knockdown of Col7 in the mouse xenografts. However, the validation of our results in RDEB SCC cells and tumors and a zebrafish model would suggest that these data are robust.

In RDEB SCC, markedly truncated or absent Col7 protein will interrupt Col7-α2 integrin binding, resulting in increased αvβ6 integrin signaling. Small changes in TGFβ1 will upregulate αvβ6 integrin, resulting in further autocrine TGFβ1 expression to further perpetuate EMT and angiogenesis. Both Twist and Smad4 are known to upregulate VEGF and angiogenesis (37,38). The neurotrophic factor Artemin, also upregulated in the proteome array, can promote de novo angiogenesis via Twist and VEGF-A (39). Thrombospondin, a potent negative regulator of angiogenesis (40), was also upregulated, possibly as a negative feedback response to increased VEGF. Increased expression of amphiregulin, an EGF receptor ligand known to promote tumor angiogenesis, was also observed in shCol7 cells (41).

Chronic blistering in RDEB induces TGFβ expression, which in turn promotes scarring. A recent study of monozygotic twins with phenotypic variability in RDEB found that both canonical and noncanonical TGFβ pathways were basally more activated in the fibroblasts of the more severely affected twin (42). Previous transcriptomic data from our group and others showed that TGFβ signaling is important in RDEB (7,43). Moreover, SCC arising in areas of chronic inflammation such as burn scars are more likely to metastasise (44). Because epithelial cancer invasion results in loss of the BM, including Col7, and our data show that loss of Col7 is a potent pro-invasive, pro-angiogenic activator, Col7 must have a tumor suppressor role in skin. Consistent with this, dominant dystrophic EB patients who have one normal COL7A1 allele rarely develop SCC (4). Wood et al. (45) showed that the COL7A1 gene is mutated and a possible tumor suppressor gene in breast cancer. Furthermore, hypermethylation of COL7A1 in breast cancer resulted in loss of Col7 expression in tumors correlating with a poor prognosis (46). Our data may be relevant to tumors from other organs in which Col7 is expressed, including breast, gastrointestinal, and head and neck cancers.

Intradermal hrCol7 protein injection has already been used successfully as a therapy for RDEB in mice, resulting in correction of subepidermal blistering with incorporation of Col7 at the basement membrane (27). Normal dermal fibroblast therapy has also been successfully used to restore basement membrane Col7 expression in RDEB patients (47). For patients with RDEB for whom Col7 protein therapy may be considered in the future, a potential anti-angiogenic effect of Col7 replacement should be considered. For patients with RDEB SCC, these data also suggest that clinical trials of treatment with anti-angiogenic therapy may be appropriate.

Funding

Support for this work was provided by research grants from DEBRA Ireland, DEBRA International-DEBRA UK, and Barts Charity (EOT). AT and VMK are supported by the Academy of Finland (project 137687), the Finnish Cancer Research Foundation, Sigrid Jusélius Foundation, and Turku University Hospital EVO grant (project 13336).

The study sponsors had no role in the design of the study; the collection, analysis, or interpretation of the data; the writing of the manuscript; nor the decision to submit the manuscript for publication. We acknowledge the BALM facility and Nuzhat Baksh for isolating the RDEB squamous cell carcinoma keratinocytes. The authors have no conflict of interest to declare.