Type VII Collagen Is Required for Ras-Driven Human Epidermal Tumorigenesis

See allHide authors and affiliations

Science  18 Mar 2005:
Vol. 307, Issue 5716, pp. 1773-1776
DOI: 10.1126/science.1106209


Type VII collagen defects cause recessive dystrophic epidermolysis bullosa (RDEB), a blistering skin disorder often accompanied by epidermal cancers. To study the role of collagen VII in these cancers, we examined Ras-driven tumorigenesis in RDEB keratinocytes. Cells devoid of collagen VII did not form tumors in mice, whereas those retaining a specific collagen VII fragment (the amino-terminal noncollagenous domain NC1) were tumorigenic. Forced NC1 expression restored tumorigenicity to collagen VII–null epidermis in a non–cell-autonomous fashion. Fibronectin-like sequences within NC1 (FNC1) promoted tumor cell invasion in a laminin 5–dependent manner and were required for tumorigenesis. Tumor-stroma interactions mediated by collagen VII thus promote neoplasia, and retention of NC1 sequences in a subset of RDEB patients may contribute to their increased susceptibility to squamous cell carcinoma.

Epidermal squamous cell carcinoma (SCC) is the second most common cancer in the United States, with approximately 250,000 new cases yearly (1). Early-onset, invasive SCCs characterize RDEB, a blistering skin disorder caused by mutations in the COL7A1 gene that encodes the keratinocyte-secreted type VII collagen epidermal basement membrane zone (BMZ) protein (24). More than 55% of RDEB patients die from SCC by age 40 (5). A subset of RDEB patients, however, does not develop SCCs (5). Whether this phenotypic difference correlates with specific functional defects in collagen VII is unknown. Furthermore, there is no compelling evidence that other inherited chronic blistering diseases are associated with a substantially increased incidence of SCC (5, 6), which indicates that skin fragility alone is insufficient to induce cancer. Collagen VII is expressed in sporadic SCCs arising in a variety of epithelial tissues (7), which indicates that collagen VII loss is not required for tumorigenesis; however, it is unclear whether collagen VII expression in non-RDEB SCCs limits tumor invasion or facilitates tumor growth.

To investigate the role of collagen VII in tumorigenesis, we studied SCCs obtained from 10 consecutive, unrelated RDEB patients. All of these tumors displayed collagen VII expression, which indicates that collagen VII loss is not characteristic of SCC in RDEB. In contrast, 4 out of 10 skin tissue biopsies from a different group of RDEB patients without SCC lacked detectable collagen VII (Fig. 1A). To explore collagen VII function in this setting, we examined the oncogenic potential of primary epidermal cells from 12 additional unrelated RDEB patients. Tumorigenesis was examined by coexpressing oncogenic Ras and the NF-κB inhibitor IκBα in primary keratinocytes by retroviral transduction to produce cells that generate human epidermal tumors indistinguishable from SCC upon engraftment to immunodeficient mice (8). Cells from four RDEB patients formed no tumors after subcutaneous injection, whereas cells from the remaining eight patients generated tumors that were all collagen VII–positive (Fig. 1, B and C, and fig. S1), which indicates the existence of tumorigenic and nontumorigenic RDEB subsets.

Fig. 1.

Relationship of epidermal tumorigenesis in RDEB to type VII collagen expression. (A) Immunostaining with antibodies to the collagen VII NC1 domain in dermally invasive sporadic SCCs from both non-RDEB and RDEB patients. Data are representative of strong NC1 detection observed in SCCs from 10 unrelated RDEB patients and 10 otherwise normal non-RDEB patients (NL). Note the complete absence of NC1 at the dermal-epidermal junction in noncancerous RDEB skin (RDEBNull) as a negative staining control (scale bar, 100 μm). (B) Tumor growth kinetics in nude mice after subcutaneous injection of primary RDEB keratinocytes coexpressing oncogenic Ras and IκBα. NL, keratinocytes from normal patients without SCC; RDEB1 and RDEB4 correspond to patients studied below; n = 5 mice per patient; error bars, +/– SD). (C) Collagen VII immunoblots and tumorigenicity of keratinocytes from 12 unrelated consecutive RDEB patients. Patients 4, 6, 8, and 10 lack collagen VII protein fragments, as detected by pAbs, and are denoted as RDEBNull. The bar graph depicts the percentage of mice forming Ras-IκBα–driven tumors for each of the 12 RDEB patients (n = 5 mice per group).

Collagen VII, as detected by polyclonal antibodies, was entirely absent in the nontumorigenic subset (termed RDEBNull) (Fig. 1C). In contrast, cells from all eight tumorigenic RDEB patients retained expression of the 145-kD amino-terminal noncollagenous domain (NC1, denoted RDEBNC1) (Fig. 1C). This was unexpected, because mutations in these eight patients generate a variety of premature termination codons that are C terminal to NC1 (table S1). Forced expression of three collagen VII mutants truncated at amino acids 1260 (NC1), 1726 (M1), and 2199 (M2) in RDEBNull cells, however, also produced a 145-kD band in each case (fig. S1), which is consistent with the observed stability of NC1 protein fragments (9). Keratinocytes from RDEB patients thus fall into two distinct phenotypes, a tumorigenic subset retaining expression of a collagen VII fragment and a nontumorigenic subset devoid of collagen VII.

To determine whether RDEBNull cells regained tumorigenicity upon collagen VII restoration, we first reintroduced full-length collagen VII directly into keratinocytes using an integrating plasmid approach (10). Collagen VII re-expression restored Ras-driven tumorigenicity in mice to primary RDEBNull keratinocytes (Fig. 2A). We next determined whether collagen VII could act in trans to rescue Ras-driven tumorigenesis by RDEBNull keratinocytes. Coinjection of normal fibroblasts, which also secrete small amounts of collagen VII, restored tumor formation, but RDEBNull fibroblasts did not (Fig. 2B). Collagen VII therefore restores tumor formation to RDEBNull keratinocytes in a non–cell-autonomous manner.

Fig. 2.

Type VII collagen effects on the tumorigenic potential of RDEBNull keratinocytes in mice. (A) Tumor growth kinetics of RDEBNull keratinocytes expressing Ras and IκBα in which full-length collagen VII has been restored (RDEBNull:COL7A1) (n = 5 mice per group +/– SD). (B) Fibroblasts either from normal individuals confirmed to secrete full-length collagen VII (NL Fb) or from RDEBNull patients (RDEBNull Fb) were coinjected with Ras-IκBα–expressing primary RDEBNull keratinocytes (n = 5 mice per group +/– SD).

We next expressed a series of collagen VII fragments in RDEBNull keratinocytes to identify the collagen VII sequences responsible for tumor formation in mice (Fig. 3, A and B). The 1260 amino acid NC1 domain restored tumorigenicity to RDEBNull cells, whereas more C-terminal sequences did not (Fig. 3C), which is consistent with NC1 retention in the tumorigenic RDEB subset. A region of NC1 that spans amino acids 761 to 1050 and that encompasses fibronectin III–like repeats (FNC1) was chosen for further study because it contains sequences that bind laminin 5 (9), another extracellular protein expressed in SCC. Deletion of FNC1 sequences from NC1 (ΔNC1) abolished tumorigenesis; conversely, introduction of FNC1 sequences alone, which led to secretion of an FNC1-containing protein fragment, restored tumorigenicity to RDEBNull cells (Fig. 3C). Thus, FNC1 sequences are both necessary and sufficient for Ras-driven epidermal tumorigenesis.

Fig. 3.

Type VII collagen sequences in epidermal tumorigenesis. (A) Collagen VII domain constructs. Numbers denote amino acid positions. NC1, noncollagenous domain 1; TH, triple helical domain; NC2, noncollagenous domain 2. (B) Verification of collagen VII domain expression in RDEBNull keratinocytes by immunoblotting using pAbs to collagen VII. (C) Tumor formation in nude mice 5 weeks after subcutaneous injection of Ras-IκBα–expressing RDEBNull keratinocytes engineered to express either full-length collagen VII or the sequences noted (n = 5 mice per group). (D) mAbs and pAbs to collagen VII NC1 sequences. Recombinant proteins blotted are noted at left, and antibodies are noted above each lane. (E) Volumes of tumors formed by Ras-IκBα–expressing normal primary keratinocytes treated with the antibodies noted +/– SD (n = 5 mice per group).

These findings raised the possibility that FNC1 sequences may be an important determinant of the tumorigenic potential of non-RDEB epidermal cells. Accordingly, we tested whether antibody-mediated inhibition of FNC1 would inhibit subcutaneous tumor formation by normal human keratinocytes. Polyclonal antibodies (pAbs) to FNC1 were raised, and their specificity was tested with other available NC1 antibodies (11, 12) by using recombinant NC1, ΔNC1, and FNC1 protein (Fig. 3D). A monoclonal antibody (mAb) to NC1, mAb 32, bound FNC1 sequences, whereas mAb 161 and control pAb ΔNC1 bound NC1 outside FNC1 (Fig. 3D). Intraperitoneal injection of either pAb or mAb targeting FNC1 at the time of subcutaneous injection of tumorigenic normal cells prevented tumor formation, whereas ΔNC1 antibodies did not (Fig. 3E). These results indicate that FNC1 sequences may be important in common epidermal tumorigenesis and may represent a potential therapeutic target.

In principle, FNC1 could affect a variety of neoplastic processes, such as tumor cell proliferation, survival, and/or invasion. Defects in any of these processes are likely to abolish subcutaneous tumor formation. To distinguish among these potential mechanisms, we used a surface skin regeneration approach on severe combined immunodeficient (SCID) mice (13), which requires that neoplastic cells penetrate through human BMZ and dermal matrix to form tumors. As observed previously (8), epidermal cells in skin regenerated from normal primary human keratinocytes coexpressing Ras and IκBα penetrated the dermal-epidermal junction and aggressively invaded the underlying mesenchyma in vivo (Fig. 4A). Tumor cell invasion was unaffected by intraperitoneal injection of laminin 10–blocking antibodies (14) but was entirely blocked by FNC1 antibodies (Fig. 4A). FNC1 antibodies did not inhibit tumor cell proliferation or trigger apoptosis, as assessed by mitotic index, histology, and terminal deoxynucleotidyl transferase–mediated deoxyuridine triphosphate nick end labeling (TUNEL) staining. In vitro invasion assays confirmed that antibodies to FNC1, but not ΔNC1, blocked enhanced invasion of normal cells coexpressing Ras and IκBα and that invasiveness of RDEBNull cells was restored by FNC1 (figs. S2 and S3). These data indicate that FNC1 sequences act primarily by facilitating invasion rather than by altering other processes important in neoplasia. Consistent with this, injection of FNC1 antibodies arrested expansion of established subcutaneous tumors (Fig. 4B).

Fig. 4.

FNC1 sequences enhance tumor cell invasion. (A) FNC1 antibodies block dermal invasion in vivo. Ras-IκBα–expressing and LacZ control–expressing human skin was regenerated on SCID mice using devitalized human dermis, with mice receiving FNC1 mAb 32 or mAb to laminin 10 (Lam10) (14) by weekly intraperitoneal injection starting 1 week after grafting. Representative micrographs are shown at 5 weeks (n = 5 mice per group). Note widespread dermal invasion of keratinocytes in controls in contrast to FNC1 antibody–treated mice, where epidermal tissue is hyperplastic but where cells do not penetrate the dermal-epidermal junction (arrows) (scale bar, 100 μm). (B) FNC1 antibodies block expansion of established tumors. Mice growing tumors established by subcutaneous injection of Ras-IκBα–expressing normal keratinocytes were injected intraperitoneally at weekly intervals (arrows) with antibodies to either FNC1 or laminin-10 control. (C) Laminin 5–NC1 interactions are disrupted by tumor-blocking antibodies to FNC1. Solid-phase binding assay of purified laminin-5 protein with recombinant NC1 in the presence of the antibodies noted. (D) Antibodies to laminin 5 block FNC1-enhanced invasion. Normal primary human keratinocytes expressing Ras and IκBα in the presence of FNC1 protein were treated with the antibodies shown, and cellular invasion was assessed. Data represents triplicate independent experiments +/– SD and are quantitated as a percentage of basal invasion by RDEBNull cells alone. (E) Laminin 5–null, collagen VII–null, and TGase1-null primary human keratinocytes from patients with junctional epidermolysis bullosa, RDEB, and lamellar ichthyosis, respectively, were transduced with Ras and IκBα retrovectors, and invasion was assessed in the presence of FNC1. Error bars in all panels represent SD.

FNC1 sequences encompass a region of collagen VII that was previously demonstrated to bind to laminin 5 (9, 15), another extracellular adhesion protein expressed in human epidermal neoplasia (7, 16, 17). To study whether laminin 5 plays a role in FNC1-mediated invasion, we examined interactions between FNC1 and purified laminin 5 using a solid-phase ligand-binding assay. Consistent with previous studies (18), recombinant FNC1 bound laminin 5, but not laminin 1, in a dose-dependent manner (fig. S4). Moreover, tumor-blocking FNC1 antibodies disrupted physical interactions between the entire NC1 domain and laminin 5. In contrast, ΔNC1 antibodies, which had no effect on tumor formation, also had no effect on this protein-protein interaction (Fig. 4C). Moreover, antibodies to laminin 5, but not to laminin 10, inhibited the stimulatory effect of FNC1 on tumor cell invasion (Fig. 4D). In addition, FNC1 protein enhanced invasiveness in lamellar ichthyosis keratinocytes lacking transglutaminase 1 (TGase1), but it had no effect on laminin 5–null keratinocytes from patients with junctional epidermolysis bullosa, which indicates that FNC1-mediated invasion requires laminin 5 (Fig. 4E).

In summary, we have demonstrated that collagen VII is required for Ras-driven human epidermal tumorigenesis, specifically by enhancing tumor cell invasion in a laminin 5–dependent manner. We propose a model in which physical interactions between laminin 5 and collagen VII that occur outside the cell facilitate cellular invasion. Given this extra-cellular mode of action, FNC1 would theoretically be accessible to therapeutics such as humanized mAbs. Our preliminary data also suggest that RDEB patients retaining NC1 might be at higher risk of SCC. Of interest, the single RDEBNull patient in our series who was over age 30 had not developed SCC by age 45, whereas all three of the RDEBNC1 patients who were over age 30 had experienced early-onset SCCs by their thirties (table S1). However, whether NC1 retention is a reliable prognostic marker for epidermal cancer in RDEB is a question that can only be addressed through large prospective studies. The present work may have possible safety-monitoring implications for collagen VII gene- and protein-therapy studies planned in RDEB.

Supporting Online Material

Materials and Methods

Figs. S1 to S4

Table S1


References and Notes

View Abstract

Stay Connected to Science

Navigate This Article