Non-hematopoietic PAR-2 is essential for matriptase-driven pre-malignant progression and potentiation of ras-mediated squamous cell carcinogenesis.

The membrane-anchored serine protease, matriptase, is consistently dysregulated in a range of human carcinomas, and high matriptase activity correlates with poor prognosis. Furthermore, matriptase is unique among tumor-associated proteases in that epithelial stem cell expression of the protease suffices to induce malignant transformation. Here, we use genetic epistasis analysis to identify proteinase-activated receptor (PAR)-2-dependent inflammatory signaling as an essential component of matriptase-mediated oncogenesis. In cell-based assays, matriptase was a potent activator of PAR-2, and PAR-2 activation by matriptase caused robust induction of nuclear factor (NF)κB through Gαi. Importantly, genetic elimination of PAR-2 from mice completely prevented matriptase-induced pre-malignant progression, including inflammatory cytokine production, inflammatory cell recruitment, epidermal hyperplasia and dermal fibrosis. Selective ablation of PAR-2 from bone marrow-derived cells did not prevent matriptase-driven pre-malignant progression, indicating that matriptase activates keratinocyte stem cell PAR-2 to elicit its pro-inflammatory and pro-tumorigenic effects. When combined with previous studies, our data suggest that dual induction of PAR-2-NFκB inflammatory signaling and PI3K-Akt-mTor survival/proliferative signaling underlies the transforming potential of matriptase and may contribute to pro-tumorigenic signaling in human epithelial carcinogenesis.

Introduction

Dysregulated pericellular proteolytic signaling is a major epigenetic driver of the sequential transformation of a normal human epithelial stem cell to a tumor cell (reviewed in (1)). In this regard, a wealth of evidence has coalesced within the past decade to implicate the aberrant expression and activity of the epithelial membrane-anchored serine protease, matriptase, to the genesis and progression of human epithelial cancers. Matriptase is near ubiquitously expressed in the epithelial compartment of both pre-malignant and malignant human tissues, where its activity is derailed in a variety of manners that include overexpression, misexpression, and loss of posttranscriptional regulation by cognate protease inhibitors (2–15), reviewed in (16, 17). For example, during multi-stage squamous cell carcinogenesis, matriptase expression undergoes a remarkable translocation from post-mitotic suprabasal cells to proliferative cells of the basal compartment, including epithelial stem cells with tumorigenic potential (13, 18). Importantly, we have previously shown that mimicking this translocation by mis-expressing matriptase at low levels in the basal keratinocyte compartment suffices to induce malignant transformation and dramatically potentiates carcinogen-induced tumor formation in transgenic mice (19).

Proteinase-activated receptor (PAR)-2 is a G-protein coupled receptor with pleiotropic functions in vertebrate development and in the maintenance of postnatal homeostasis. Dysregulated PAR-2 activation is believed to contribute to a vast number of common human diseases, owed in particular to the ability of the receptor to induce key inflammatory signaling pathways (20–23). PAR-2 was originally identified as a protease activated receptor that was distinct from PAR-1 in that it responded poorly to thrombin, but was efficiently activated by the digestive enzyme trypsin (24). Subsequently, a large number of trypsin-like serine proteases were shown to be able to activate PAR-2 in various cell-based assays. These include coagulation factors FVIIa and FXa, facilitated by FVIIa binding to tissue factor (25–27), mast cell alpha and beta tryptases, kallikreins 4, 5, 6, and 14 (28–30), as well as the membrane-anchored serine proteases, human airway trypsin-like serine protease (31), TMPRSS2 (32), prostasin (21), hepsin (21), and matriptase (33). Despite the profusion of proteases displaying the capacity to activate PAR-2 ex vivo, the specific protease or repertoire of proteases that activates the receptor in many physiological and pathological contexts in vivo remains to be established.

PAR-2 is expressed by both primary keratinocytes and by established keratinocyte cell lines (34, 35) where its activation can elicit a range of cellular responses. These include inositol phospholipid hydrolysis and calcium mobilization (34), activation of c-Jun N-terminal kinase, p38 mitogen-activated protein kinase and Rho (36, 37), induction of NFκB activity (37), secretion of inflammatory cytokines, including granulocyte-macrophage colony-stimulating factor (GM-CSF), interleukin-6 (38), interleukin-8/CXCL1/Gro-1 (35), and thymic stromal lymphopoietin (39), as well as release of prostaglandin (PG)E2 and PGF2a (40), and expression of intercellular cell adhesion molecule-1 (41).

In this study, we determined the specific contribution of PAR-2 to the oncogenic activity of matriptase, because matriptase-induced squamous cell carcinogenesis is preceded by a signature chronic inflammation, and because the protease and the G-protein coupled receptor are co-expressed in normal squamous epithelium and in squamous cell carcinoma (15, 19, 42). We took advantage of the fact that PAR-2 deficiency in mice causes partial embryonic lethality, but that born offspring is healthy and display normal long-term survival, thus enabling the use of a definitive genetic epistasis analysis to address this issue (21, 43). Importantly, we show that both matriptase-driven pre-malignant progression and the potentiation of carcinogen-induced squamous cell carcinogenesis by matriptase are entirely dependent on non-hematopoietic PAR-2. Furthermore, we show that matriptase activation of PAR-2 leads to pro-tumorigenic inflammatory cytokine release via activation of NFκB through Gαi. The study identifies matriptase as a candidate in vivo activator of PAR-2 in the context of human tumorigenesis and demonstrates the importance of posttranscriptional deregulation of PAR-2 signaling to epithelial malignant transformation.

Results

Loss of PAR-2 prevents matriptase-induced pre-malignant progression of mouse squamous epithelium

Matriptase is exclusively expressed in the suprabasal compartment of homeostatic mouse epidermis, but the protease becomes expressed in the stem cell compartment upon exposure to tumor-promoting agents (13, 18). Mimicking this translocation of matriptase expression by low-level expression of a mouse matriptase cDNA under the control of a keratin-5 promoter in transgenic mice (hereafter termed K5-Mat mice) leads to spontaneous multi-stage squamous cell carcinogenesis with tumor formation beginning at one year of age (19). PAR-2 is expressed in the keratinocyte compartment and has been hypothesized to be an important mediator of skin inflammation following its activation by trypsin-like serine proteases (see Introduction). Therefore, to determine the possible contribution of PAR-2 to matriptase-mediated squamous cell carcinogenesis, we interbred K5-Mat mice and PAR-2-deficient (F2rl1−/−) mice to generate cohorts of K5-Mat;F2rl1−/− and their associated littermate K5-Mat;F2rl1+/+, F2rl1−/−, and wildtype mice. The outward appearance and health of the mice were followed for 52 weeks, after which the mice were sacrificed for histological examination. As reported previously, K5-Mat mice at this age presented with alopecia secondary to follicular metaplasia and ichthyosis (Figure 1A, left panel and data not shown), epidermal hyperplasia (Figure 1B) and hyperproliferation (Figure 1C), multifocal dysplasia (Figure 1D, example with arrowhead), and expression of the stress-associated marker, keratin-6 in the interfollicular epidermis (Figure 1H). As also reported previously, dermal fibrosis with increased dermal cellularity was prominent (Figure 1L). Unexpectedly, however, even at this advanced age, K5-Mat;F2rl1−/− mice did not display alopecia (Figure 1A, second panel from left), epidermal hyperplasia (Figure 1B), epidermal hyperproliferation (Figure 1C), dysplasia (Figure 1E), interfollicular keratin-6 expression (Figure 1I), or dermal fibrosis (Figure 1M) and, in fact, were outwardly and microscopically indistinguishable in all respects from their F2rl1−/−, and wildtype littermates (Figure 1A, compare second panel from left with third and fourth panel, Figure 1B and C, compare Figure 1E with F and G, I with J and K, and M with N and O). Taken together, these data reveal that matriptase-mediated pre-malignant progression is entirely PAR-2-dependent.

Figure 1

Matriptase-mediated pre-malignant progression of mouse squamous epithelium is PAR-2-dependent

(A). Representative examples of the outward appearance of littermate K5-Mat;F2rl1+/+ (left panel), K5-Mat;F2rl1−/− (second panel from left), F2rl1−/− (second panel from right), and wildtype mice (right panel) at 52 weeks of age. Matriptase-induced alopecia and ichthyosis are prevented by genetic elimination of PAR-2. (B). Enumeration of epidermal thickness in littermate K5-Mat;F2rl1+/+ (N=11, purple circles), K5-Mat;F2rl1−/− (N=7, green circles), wildtype (N=5, red circles), and F2rl1−/− mice (N=6, grey circles) at 52 weeks of age. *** P < 0.0006 (Mann-Whitney U test, two-tailed). N.S. = not significant (C). Epidermal keratinocyte proliferation rates in littermate K5-Mat;F2rl1+/+ (N=6, purple circles), K5-Mat;F2rl1−/− (N=7, green circles), wildtype mice (N=3, red circles), and F2rl1−/− (N=3, grey circles) at 52 weeks of age. Horizontal bars in C and B indicate median values. * P < 0.02, ** P < 0.004, *** P < 0.0006, N.S. = not significant (Mann-Whitney U test, two-tailed). (D–O). Representative histological appearance of dorsal skin of littermate K5-Mat;F2rl1+/+ (D, H, and L), K5-Mat;F2rl1−/− (E, I, and M), wildtype mice (F, J, and N) and F2rl1−/− (G, K, and O) mice at 52 weeks of age stained with H&E (D–G), keratin-6 antibodies (H–K), and Masson’s trichrome (L–O). Multifocal dysplasia (examples with arrowhead in D), interfollicular epidermal keratin-6 expression (examples with arrowhead in H), and dermal fibrosis (example with star in L) are completely prevented by genetic elimination of PAR-2. Size bars = 50 µm.

Matriptase fails to potentiate chemical carcinogenesis in the absence of PAR-2

In addition to initiating spontaneous malignant progression, dysregulated matriptase potentiates ras-dependent squamous cell carcinogenesis induced by topical application of the carcinogen, 7,12-dimethylbenzanthracene (DMBA) (19). To establish the contribution of PAR-2 to matriptase-mediated potentiation of chemically-induced squamous cell carcinogenesis, we exposed K5-Mat;F2rl1−/− mice and their associated K5-Mat;F2rl1+/−, F2rl1−/−, and F2rl1+/− littermates to repeated DMBA applications, and then observed the formation of tumors by outward inspection of mice for up to 63 weeks (Figure 2A, examples of individual mice in C–F). Because F2rl+/− mice are phenotypically normal, and because phenotypic differences observed between F2rl+/− and F2rl−/− mice can be assumed to be identical to or greater than differences observed between F2rl and F2rl−/− mice, K5-Mat;F2rl+/− and F2rl+/− mice were used as controls to simplify the complex breeding scheme needed to littermate control the study. As expected, tumor formation was strongly accelerated in K5-Mat mice when compared to littermate controls (median tumor-free survival of 25 weeks for K5-Mat;F2rl1+/− versus 48 weeks for F2rl1+/−; P <0.0001, log-rank test, two-tailed). K5-Mat mice also had to be euthanized faster than littermate control mice due to the tumor burden and the tumor morbidity reaching study endpoints (Figure 2B) (median time to euthanization, 41 weeks for K5-Mat;F2rl1+/− versus 62 weeks for F2rl1+/−, P <0.004, Wilcoxon rank sum test, two-tailed). Dysregulated matriptase, however, failed to accelerate DMBA-induced squamous cell carcinogenesis when PAR-2 was genetically eliminated. Thus, tumor latency (median tumor-free survival of 45 weeks for K5-Mat;F2rl1−/− versus 48 weeks for F2rl1−/− and 48 weeks for F2rl1+/−; P = N.S., Kaplan-Meier log rank test), and time to euthanization (median time to euthanization of 61 weeks for K5-Mat;F2rl1−/− versus 56 weeks for F2rl1−/−, and 62 weeks for F2rl1+/−; P = N.S., Wilcoxon rank sum test, two-tailed) was similar to control and to PAR-2-deficient mice (Figure 2A and B). Total tumor burden at euthanization (Figure 2B) and histological appearance of tumors (Figure 2G–N), did not differ between mice of different genotypes, with all mice presenting with squamous cell carcinoma of variable degrees of differentiation.

Figure 2

Dysregulated matriptase does not promote chemical carcinogenesis in the absence of PAR-2

Kaplan-Meier analysis of tumor-free survival (A), and time to sacrifice and total tumor burden at sacrifice (B) of littermate K5-Mat;F2rl1+/− (N=11, purple squares), K5-Mat;F2rl1−/− (N=11, green diamonds), F2rl1+/− (N=10, red circles), and F2rl1−/− (N=7, grey triangles) mice. The mice were treated with DMBA every 3 weeks and were followed for up to 63 weeks. *** P < 0.0001, K5-Mat;F2rl1+/+ versus other genotypes in A (log-rank test, two-tailed). Vertical and horizontal bars in B indicate standard error of the mean. C–F. Representative outward appearance of littermate K5-Mat;F2rl1+/− (left panel), K5-Mat;F2rl1−/− (second panel from left), F2rl1+/− (second panel from right), and F2rl1−/− (right panel) mice at 26 weeks of age. Example of carcinoma (white star) and papilloma (white triangle) in K5-Mat;F2rl1+/− mouse is shown. G–N. Representative of H&E-stained sections of squamous cell carcinoma at low (G–J) and high (K–N) magnification in a 25-week-old K5-Mat;F2rl1+/− (left panels) mouse, and in 45 to 55-week-old K5-Mat;F2rl1−/− (second panels from left), F2rl1+/− (second panels from right), and F2rl1−/− (right panels) mice showing similar histological appearance of tumors irrespective of genotype. Stars in G–J show examples of invasion, while arrowheads and arrows in K–N shows examples of cells invading vessels and atypical mitosis, respectively. Size bars. G–J = 100 µm. K–N = 30 µm.

Non-hematopoietic cell-derived PAR-2 mediates matriptase-driven pre-malignant progression

PAR-2 is expressed not only by keratinocytes, but also by a variety of leukocyte populations that infiltrate the skin during pre-neoplastic progression. We therefore next examined if the failure of PAR-2-deficient mice to promote the stereotypic pro-tumorigenic proliferative responses that are associated with dysregulated epidermal matriptase were caused by loss of keratinocyte matriptase-PAR-2 signaling or, rather, were an indirect consequence of intrinsic defects of PAR-2-deficient leukocytes. For this purpose, we performed bone marrow chimeric studies in which K5-Mat and wildtype littermate mice were lethally irradiated and thereafter reconstituted with histocompatible bone marrow from either F2rl1+/+ or F2rl1−/− mice. Successful bone marrow transplantation was ensured by FACS analysis for expression of donor-specific hematopoietic markers (Supplementary Figure 1). After recovery from grafting, the bone marrow chimeric mice were followed for 11 weeks and then sacrificed and examined (Figure 3). As predicted, K5-Mat mice grafted with wildtype bone marrow displayed epidermal hyperplasia, epidermal hyperproliferation, and dermal fibrosis when compared to wildtype mice grafted with wildtype bone marrow (Figure 3A and B, compare example in C with E, and example in G with I). No diminution of hyperplasia, hyperproliferation, and fibrosis was apparent when K5-Mat mice were grafted with PAR-2-deficient bone marrow (Figure 3A and B, compare example in D with F, and example in H with J). Thus, epithelial hyperplasia, hyperproliferation and dermal collagen density all were similar in K5-Mat mice grafted with either PAR-2-deficient or PAR-2-sufficient bone marrow. These data demonstrate that dysregulated matriptase activates PAR-2 on cells of non-hematopoietic origin to elicit its tumor-promoting effects.

Figure 3

Hematopoietic PAR-2 is dispensable for matriptase-mediated pre-malignant progression

(A) Enumeration of epidermal thickness in 17-week-old lethally-irradiated littermate K5-Mat;F2rl1+/+ mice (orange and blue circles) reconstituted with either F2rl1+/+ (N=5, orange circles) or F2rl1−/− bone marrow (N=7, blue circles), and wildtype mice (brown and grey circles) reconstituted with F2rl1+/+ (N=5, brown circles) or F2rl1−/− bone marrow (N=5, grey circles). Horizontal bars indicate median values. * P < 0.01 and **P < 0.007, respectively, N.S. = not significant (Mann-Whitney U test, two-tailed). (B) Epidermal keratinocyte proliferation rates in 17-week-old lethally irradiated littermate K5-Mat;F2rl1+/+ mice (orange and blue circles) reconstituted with either F2rl1+/+ (N=5, orange circles) or F2rl1−/− bone marrow (N=4, blue circles), and wildtype mice (brown and grey circles) reconstituted with either F2rl1+/+ (N=6, brown circles) or F2rl1−/− bone marrow (N=5, grey circles). Horizontal bars indicate median values. ** P < 0.003, N.S. = not significant (P < 0.06) (Student’s t-test, two-tailed). (C–J) Representative examples of H&E (C–F) and Masson’s trichrome (G–J) -stained epidermal sections from 13-weeks-old lethally irradiated littermate K5-Mat;F2rl1+/+ mice (C, D, G, and H) reconstituted with F2rl1+/+ (C and G) or F2rl1−/− bone marrow (D and H), and wildtype mice (E, F, I, and J) reconstituted with F2rl1+/+ (E and I) or F2rl1−/− (F and J) bone marrow. Stars in G and H indicate dermal fibrosis. Size bars. C–F = 50 µm. G–J = 100 µm.

Dysregulated matriptase fails to elicit dermal inflammatory cell recruitment in the absence of PAR-2

Matriptase-induced pre-malignant progression is associated with a signature accumulation of inflammatory cells in the dermis (13, 19). To explore if dysregulated matriptase induces inflammatory cell recruitment through PAR-2, we enumerated dermal mast cells in young (8–10 week old) littermate and older (52-week-old) littermate K5-Mat;F2rl1+/+, K5-Mat;F2rl1−/−, F2rl1+/+, and F2rl1−/− mice (Figure 4). In both younger (Figure 4A–D, and I) and older (Figure 4E–H, and J) mice, a significant increase in mast cell accumulation was observed in K5-Mat mice compared to wildtype littermates. Interestingly, however, elimination of PAR-2 completely negated the matriptase-induced increase in mast cell accumulation, with the number of dermal mast cells in K5-Mat;F2rl1−/− mice being similar to wildtype and F2rl1−/− littermates (compare example in Figure 4B with C and D, I, and Figure 4F with G and H, J). This data suggest that dysregulated matriptase mediates inflammatory cell recruitment through PAR-2.

Figure 4

Matriptase-induced dermal inflammatory cell accumulation is PAR-2-dependent

A–H. Representative examples of toluidine blue-stained dorsal skin sections from 8 to 10-week-old-littermate (A–D) and 52-week-old littermate (E–H) K5-Mat;F2rl1+/+ (left panels), K5-Mat;F2rl1−/− (second panels from left), F2rl1+/+ (second panels from right), and F2rl1−/− (right panels) mice. Examples of dermal mast cells (blue) are indicated with arrows. Size bars = 50 µm. (I and J) Enumeration of mast cell accumulation in 8 to 10-week-old (left panel) and 52-week-old (right panel) K5-Mat;F2rl1+/+ (N=9 to 12, purple circles), K5-Mat;F2rl1−/− (N=7 to 9, green circles), F2rl1+/+ (N=6 to 7, red circles), and F2rl1−/− mice (N=6 to 7, grey circles). * P < 0.01. ** P < 0.001, N.S. = not significant (Mann-Whitney U test, two-tailed).

Dysregulated matriptase induces inflammatory cytokine release through PAR-2

We next explored if the dramatic impact of global PAR-2 ablation on matriptase-driven pre-malignant progression could be related to the elimination of a keratinocyte-intrinsic matriptase-PAR-2-mediated inflammatory signaling pathway. For this purpose, we performed qPCR analysis of skin cells to determine the mRNA levels of a number of keratinocyte-produced inflammatory cytokines and inflammatory mediators in young (8 to10-week-old) K5-Mat;F2rl1−/− mice and their associated K5-Mat;F2rl1+/+, F2rl1−/−, and wildtype littermates. These included mRNAs encoding the murine interleukin 8 equivalent, chemokine (C-X-C motif) ligand 1 (Cxcl1), interleukin-1β (Il1b), thymic stromal lymphopoietin (Tslp), colony stimulating factor 2 (Csf2), tumor necrosis factor (Tnf), interferon-γ (Ifnγ), and intercellular adhesion molecule 1 (Icam1) (Figure 5A–G). mRNA levels of Cxcl1, Il1b, Tslp, Csf2, Tnf and Icam1 all were significantly increased (2–171 fold) in young K5-Mat mice when compared to their wildtype littermates (Figure 5A–E, G), and Ifng mRNA displayed a 3.6 fold increase without reaching statistical significance (Figure 5F). This result demonstrates that dysregulated matriptase induces expression of skin inflammatory mediators. Importantly, loss of PAR-2 from K5-Mat mice reduced the level of expression of Cxcl1, Il1b, Tslp, Csf2, and Tnf mRNA species to levels similar to those observed in wildtype and F2rl1−/− mice (Figure 5A–E) suggesting that keratinocyte PAR-2 is required for matriptase-induced expression of these cytokines. To substantiate this hypothesis, we next measured mRNA levels of Cxcl1, Il1b, Tslp, Csf2, Tnf, Ifng, and Icam1 in 17-week-old K5-Mat mice reconstituted with, respectively, PAR-2-deficient and -sufficient bone marrow (Figure 5H–N). As expected, mRNA levels of Cxcl1, Il1b, Tslp, Tnf, Ifng, and Icam1 (Figure 5H–J, L–N) all were significantly increased in K5-Mat mice reconstituted with wildtype bone marrow when compared to control mice (wildtype mice reconstituted with either wildtype or PAR-2-deficient bone marrow), while Csf2 mRNA was 1.7-fold increased without reaching statistical significance (Figure 5K). Il1b, Tslp, Tnf, Ifng, and Icam mRNA levels remained significantly increased in K5-Mat mice reconstituted with PAR-2-deficient bone marrow (Figure 5I, J, L, M, and N), while Cxcl1 and Csf2 mRNA were, respectively, 1.8 and 1.7-fold increased (Figure 5H and K) without reaching statistical significance, showing that hematopoietic PAR-2 is dispensable for the induction of expression of skin inflammatory mediators by dysregulated matriptase through PAR-2. Analysis of skin from 52-week-old mice showed a significant increase in Cxcl1, Il1b, Tslp, and Csf2 mRNA levels of K5-Mat mice when compared to wildtype littermates (Figure 5O–R). The levels of these four cytokine mRNA species were reduced, 24-, 8-, 35-, and 47–fold, respectively, by PAR-2 elimination, reaching significance for Cxcl1, while not reaching significance for Il1b, Tslp, and Csf2 (p < 0.13, 0.065, and 0.070, respectively). Taken together, the data indicate that dysregulated matriptase induces early expression of key pro-tumorigenic inflammatory mediators through keratinocyte PAR-2.

Figure 5

Dysregulated matriptase induces persistent skin inflammatory cytokine production through non-hematopoietic PAR-2

(A–G). Skin mRNA levels of Cxcl1 (A), Il1b (B), Tslp (C), Csf2 (D), Tnf (E), Ifng (F), and Icam1 (G), in 8 to 10-week-old littermate K5-Mat;F2rl1+/+ (N=5 to 7, purple circles), K5-Mat;F2rl1−/− (N=4 to 6), green circles), wildtype (N=3 to 4, red circles), and F2rl1−/− mice (N=3 to 4, grey circles). * P < 0.01, **, P < 0.001, *** P < 0.01, N.S. = not significant (Mann-Whitney U test, two-tailed). (H–N). Skin mRNA levels of Cxcl1 (H), Il1b (I), Tslp (J), Csf2 (K), Tnf (L), Ifng (M), and Icam1 (N) in 17-week-old lethally-irradiated littermate K5-Mat;F2rl1+/+ mice (orange and blue circles) reconstituted with F2rl1+/+ (N=5, orange circles) or F2rl1−/− bone marrow (N=5 to 7, blue circles), and control (F2rl1+/+ and F2rl1−/−) mice (N=12 to 13, brown circles) reconstituted with either F2rl1+/+ or F2rl1−/− bone marrow. Horizontal bars indicate median values. * P < 0.01, **, P < 0.001, N.S. = not significant (Mann-Whitney U test, two-tailed). (O–U). Skin mRNA levels of Cxcl1 (O), Il1b (P), Tslp (Q), Csf2 (R), Tnf (S), Ifng (T), and Icam1 (U) in 52-week-old littermate K5-Mat;F2rl1+/+ (N=4 to 12, purple circles), K5-Mat;F2rl1−/− (N=4 to 5, green circles), wildtype (N=5, red circles), and F2rl1−/− mice (N=3 to 5, grey circles). * P < 0.01, ** P < 0.001, N.S. = not significant (Mann-Whitney U test, two-tailed).

Matriptase activates NFκB through PAR-2 and Gαi

Because NFκB is a key epidermal regulator of inflammatory cytokine production in both non-neoplastic and neoplastic contexts (44, 45) and because many of the cytokines whose mRNAs were increased in K5-Mat mice are regulated by NFκB, we investigated if matriptase activation of PAR-2 could induce NFκB activity in epithelial cells. For this purpose, we used a recently established reconstituted cell-based assay (46, 47) in which HEK293 cells were transfected with a PAR-2 expression vector and either a serum response element-luciferase reporter plasmid or NFκB-luciferase reporter plasmid. The transfected cells then were exposed to either a PAR-2 agonist, a dual PAR-1 and PAR-2 agonist or to soluble recombinant matriptase. As described recently (47), a dose-dependent increase in serum response element activity was observed when PAR-2-transfected cells were exposed to a PAR-2 agonist peptide (Figure 6A, bars 1–5) or the dual PAR-1 and -2 agonist peptide (Figure 6A, bar 6). A similar dose-dependent increase in serum response element activity was observed when PAR-2 transfected cells (Figure 6B, bars 1–3) but not mock transfected cells (Figure 6B, bar 4) were exposed to matriptase. Importantly, exposure of PAR-2 transfected cells to the PAR-2 agonist peptide (Figure 6C, bars 1–5), the dual PAR-1 and -2 agonist peptide (Figure 6C, bar 6) or to matriptase (Figure 6D, bar 1–3) also elicited dose-dependent NFκB activity. When PAR-2 was substituted with PAR-1, the dual PAR-1 and -2 agonist (Figure 6E, bar 1–5), but not the PAR-2 agonist (Figure 6E, bar 6) or soluble recombinant matriptase (Figure 6F, bar 1–3, compare with bar 4) induced a dose-dependent increase in serum response element activity, showing that soluble recombinant matriptase selectively activates PAR-2. Although PAR-1 induced serum response element activity in response to the dual PAR-1 and -2 agonist (Figure 6E, bars 1–5, compare to bar 6), the activated receptor failed to induce NFκB activity (Figure 6G, bars 1–5, compare to bar 6) or, as expected, when exposed to soluble recombinant matriptase (Figure 6H, bar 1–3, compare with bar 4). To further explore this matriptase-PAR-2-NFκB signaling pathway, we next attempted to identify the G protein that couples to matriptase-activated PAR-2 to elicit NFκB activation. Co-introduction of dominant-negative RGS-βARK (Figure S3A, B, E and G) and PDZrhoGEF (Figure S3C, D, F and H) mutants that target, respectively, Gαq/11 and Gα12/13, to the PAR-2 expression vector- and NFκB reporter plasmid-transfected cells did not affect the induction of NFκB activity by either the PAR-2 agonist or by recombinant soluble matriptase. To specifically determine the involvement of Gαi, HEK293 cells were mock transfected (Figure 6I, bar 1 and 2) or transfected with the PAR-2 expression vector and the NFκB reporter plasmid were either untreated (Figure 6I, bar 3 and 4) or exposed to PAR-2 agonist (Figure 6I, bars 5 and 6) or soluble recombinant matriptase (Figure 6I, bars 7 and 8) and treated with either vehicle (Figure 6I, bars 1, 3, 5, and 7) or pertussis toxin (PTX, Figure 6I, bars 2, 4, 6, and 8). As observed in Figure 6C and D), both the PAR-2 agonist and matriptase induced NFκB activity (Figure 6I, compare bar 3 with bars 5 and 7). Pertussis toxin strongly reduced both PAR-2 agonist (Figure 6I, compare bar 5 with bar 6) and matriptase-induced (Figure 6I, compare bar 7 with bar 8) NFκB activity. Interestingly, using the identical experimental setup, pertussis toxin had only a minor effect on the ability of activated PAR-2 to induce serum response element activity (Figure 6J). Taken together, these data demonstrate that matriptase-activated PAR-2 selectively signals through Gαi to activate NFκB.

Figure 6

Matriptase can activate NFκB through PAR-2 and Gαi

(A–H) HEK293 cells were co-transfected with PAR-2 expression vector (PAR-2, A–D) or PAR-1 expression vector (PAR-1, E–H) or with empty expression vector and either serum response element-luciferase (SRE) (A, B, E, and F) or NFκB -luciferase reporter plasmids (NFκB) (C, D, G, and H), as indicated. The transfected cells in A and C were either treated for 6 h with the PAR-2 agonist, 2-furoyl-LIGRLO-amide (PAR-2 ag., 0.25, 1, 2.5 and 10 µM in bars 2–5 respectively) or with 10 µM dual PAR-1 and -2 agonist, TRAP-6 (lane 6). The transfected cells in E and G were either treated with TRAP-6 (0.25, 1, 2.5 and 10 µM in bars 2–5, respectively) or with 10 µM 2-furoyl-LIGRLO-amide (lane 6) for 6 h. Cells in B, F, D, and H were treated with recombinant human matriptase serine protease domain for 6 h (1 or 15 nM in bars 2–4) or with vehicle (bar 4). I and J. HEK293 cells were co-transfected with empty vector (bars 1 and 2) or PAR-2 expression vector (bars 3–8) in combination with either serum response element-luciferase (I) or NFκB -luciferase reporter (J) plasmids. The transfected cells were treated overnight with vehicle (lanes 1, 3, 5, and 7) or pertussis toxin (PTX) (lanes 2, 4, 6, and 8), and then were treated with vehicle (lanes 3 and 4) or were treated with either 20 µM PAR-2 agonist (bars 1 and 2, 5 and 6) or 15 nM recombinant human matriptase (lanes 7 and 8). All cells were transfected with a pRL-Renilla luciferase reporter plasmid as an internal control for transfection efficiency. Data are shown as the mean ± standard deviation of the mean of triplicate transfections of firefly luciferase light units/Renilla luciferase light units. The data are representative of three similar experiments.

We recently identified pro-HGF activation as a critical mechanism by which dysregulated matriptase promotes squamous cell carcinogenesis, based on the failure of matriptase to accelerate tumor progression in mice with specific ablation of the HGF receptor, c-Met, from the keratinocyte stem cell compartment (13). Interestingly, however, unlike PAR-2 ablation, loss of c-Met failed to prevent mast cell accumulation in response to dysregulated matriptase, with the number of mast cells being similar in c-Met-sufficient K5-Mat and c-Met-deficient K5-Mat mice (Supplementary Figure 3A–D). Furthermore, activation of PAR-2, by targeting cells with soluble recombinant matriptase for up to 3 h failed to induce transactivation of c-Met, as evidenced by the absence of autophosphorylation of c-Met and phosphorylation of Gab-1 (Supplementary Figure 3E and F). These data indicate that activation of PAR-2 and c-Met by matriptase induces separate pro-tumorigenic signaling pathways and that each is required for the promotion of squamous cell carcinogenesis by the membrane-anchored serine protease.

Discussion

Multiple studies have linked dysregulated matriptase to human epithelial carcinogenesis, but the molecular mechanisms by which the membrane protease promotes carcinogenesis are only beginning to be elucidated. Here we employed genetic epistasis analysis to specifically query the contribution of one candidate substrate, PAR-2, to matriptase-driven squamous cell carcinogenesis. Unexpectedly, we found that genetically eliminating PAR-2 sufficed to negate all pro-tumorigenic effects of dysregulated epidermal matriptase. Thus, matriptase-driven pre-malignant progression did not occur in the absence of PAR-2, and dysregulated matriptase failed to potentiate DMBA-induced carcinogenesis in the absence of the G-protein coupled receptor. Specific ablation of PAR-2 from hematopoietic cells did not affect matriptase-mediated pre-malignant progression, suggesting that dysregulated matriptase activates epidermal (most likely keratinocyte) PAR-2 to promote malignant progression. Because matriptase and PAR-2 has been shown to be consistently co-expressed in the epithelial compartment of human squamous cell carcinoma (15), the current study provides strong evidence that a matriptase-PAR-2 signaling axis contributes to human squamous cell carcinogenesis.

We have previously shown that matriptase-driven squamous cell carcinogenesis is preceded by a conspicuous recruitment of large numbers of mast cells to the underlying dermis (19). The current study provides evidence that this effect of matriptase dysregulation is related to PAR-2-dependent NFκB activation via Gαi and the associated release of NFκB-dependent inflammatory cytokines. Firstly, the matriptase-mediated increase in the mRNA levels of several NFκB-induced inflammatory cytokines was entirely dependent on non-hematopoietic PAR-2. Secondly, in a reconstituted cell-based system, matriptase activation of PAR-2 induced a robust activation of an NFκB reporter gene via Gαi. Because the causal association between chronic inflammation and epithelial carcinogenesis is well established, it is possible that the elicitation of pro-inflammatory cytokines via NFκB activation could account for all the pro-tumorigenic effects associated with untimely activation of PAR-2 by matriptase. The current study, however, does not exclude other PAR-2 activation-induced effects, such as the direct stimulation of proliferation, cell death protection or motility enhancement. Our findings are aligned with a previous study showing that soluble matriptase can induce cytokine release in cultured human endothelial cells through PAR-2 and NFκB (48).

By using a genetic epistasis approach similar to the current study, we have previously shown that dysregulated matriptase promotes squamous cell carcinogenesis by activating a c-Met-Akt-signaling pathway by proteolytic conversion of pro-HGF to active HGF on the keratinocyte cell surface (13). Interestingly, in this study, imposition of either PAR-2 or c-Met deficiency on K5-Mat transgenic mice negated the ability of dysregulated matriptase to potentiate DMBA-induced tumorigenesis and blocked matriptase-driven pre-malignant progression, including epidermal hyperproliferation, dysplasia, follicular metaplasia, expression of stress-associated epidermal keratins, and induction of dermal fibrosis. Analysis of archived tissues from c-Met-sufficient K5-Mat and c-Met-deficient K5-Mat mice did, however, reveal that loss of c-Met signaling, unlike loss of PAR-2 signaling, failed to prevent matriptase-induced inflammatory cell recruitment. Furthermore, although PAR-2 has been proposed to transactivate c-Met in the context of hepatocellular carcinoma invasion (49), we found no evidence that PAR-2 activation was associated with increased c-Met activity using a reconstituted cell-based assay using HEK cells or using immortalized keratinocytes. An intriguing unanswered question that arises from these studies, therefore, is why matriptase-induced inflammatory PAR-2 signaling does not promote malignant progression in the absence of activated c-Met and, conversely, why matriptase-induced c-Met-Akt signaling does not promote malignant progression in the absence of activated PAR-2.

In summary, by performing genetic epistasis analysis in an animal model, in which the signature dysregulation of matriptase in human carcinomas is mimicked by low level mis-expression of matriptase in the epidermis (19), we have provided definitive evidence that matriptase proteolytically cleaves both PAR-2 and pro-HGF in vivo to activate, respectively, tumor promoting G-protein coupled and receptor tyrosine kinase signaling pathways. Furthermore, we have shown that both pathways must be activated simultaneously by matriptase for the protease to exert its tumor promoting effects on squamous epithelium. The study underscores that pericellular serine proteases may be excellent therapeutic targets for the prevention and treatment of cancer, due to their convenient extracellular location and their capacity to regulate fundamental pro-tumorigenic cellular signaling pathways.

Materials and Methods

Animals

All experiments were performed in an Association for Assessment and Accreditation of Laboratory Animal Care International-accredited vivarium following Institutional Guidelines and standard operating procedures. The generation of mice overexpressing matriptase under the keratin-5 promoter (K5-Mat) has been reported previously (19). Mice carrying a F2rl1 null allele (50) were purchased from the Jackson Laboratories (Bar Harbor, ME). K5-Mat; F2rl1−/− and K5-Mat;F2rl1+/+, K5-Mat; F2rl1+/−, F2rl1+/+, F2rl1+/− and F2rl1−/− mice in a mixed (FVB/NJ/C57BL/6J/Black Swiss) background were generated by interbreeding. The study was strictly littermate controlled to avoid genetic background differences from confounding data interpretation.

Chemical carcinogenesis

DMBA was diluted to a final concentration of 250 µg/ml in acetone and 100 µl was applied to the dorsal skin of the mice every three weeks.

Bone marrow chimeras

Bone marrow was isolated from either C57BL/6J-F2rl1+/+ mice or from C57BL/6J-F2rl1−/− mice and was injected intravenously into lethally irradiated (900 rads) six-week-old K5-Mat and wild-type littermate control F1 offspring from FVB/NJ-K5-Mat mice crossed to C57BL/6J mice. DMBA treatment was performed 8 weeks after transplantation as described above.

Histological and immunohistochemical analysis

Skin and tumor tissues were fixed for 24 h in 4% paraformaldehyde (PFA) in phosphate buffered saline (PBS), processed into paraffin, sectioned into sagittal 3-µm sections and stained with hematoxylin and eosin, Masson's trichrome or toluidine blue prior to histopathological assessment. Immunohistochemical analysis was performed as described previously (51). Antibodies recognizing keratin-6 (Covance, Chantilly, VA), and F4/80 (AbD Serotec, Raleigh, NC) were used for staining. Epithelial cell proliferation was visualized by intraperitoneal injection of 10 µg/g of 5-bromo-2′-deoxyuridine (BrdU), 2 h prior to euthanasia. BrdU incorporation was detected with a mouse anti-BrdU antibody (Accurate, Westbury, NY). Bound antibodies were visualized with the Vectastain ABC peroxidase kit as recommended by the manufacturer (Vector Laboratories, Burlingame, CA) and developed with the addition of the 3,3’-diaminobenzidine substrate (Sigma, St Louis, MO). Slides were digitalized using a ScanScope (Aperio, Vista, CA) and the height of the epidermis (epithelial thickness), mast cell count per 105 µm2 and the number of BrdU positive basal cells (basal proliferation rates) were quantified.

RNA preparation and real-time PCR

Whole skin was snap-frozen in liquid nitrogen and ground to a fine powder with mortar and pestle. Total RNA was prepared by extraction in Trizol reagent (Life Technologies, Frederick, MD) as recommended by the manufacturer. Reverse transcription and PCR amplification were performed using the RETROscript™ Kit (Ambion, Inc., Austin, TX), as recommended by the manufacturer. First strand cDNA synthesis was performed using an Oligo dT primer. The following primer pairs were used for qPCR: Tnf; 5’-CAGCCTCTTCTCATTCCTGC-3’ and 5’- AGGGTCTGGGCCATAGAACT-3’. Ifng; 5’-AGCGGCTGACTGAACTCAGATTGTAG-3’ and 5’- GTCACAGTTTTCAGCTGTATAGGG-3’. Il1b; 5’-GGGCCTCAAAGGAAAGAATC-3’ and 5’-TACCAGTTGGGGAACTCTGC-3’. Gapdh; 5'-GTGAAGCAGGCATCTGAGG-3' and 5'-CATCGAAGGTGGAAGAGTGG-3'. Annealing temperature for these primer sets was 55°C. Expression levels were normalized to Gapdh in each sample. Tslp, Cxcl1, Icam1 and Csf2 qPCRs were done using TaqMan probes following the manufacturer’s instructions (Life Technologies).

PAR-2 activation assay

The assay was performed as described before (46). Briefly, HEK293 cells were plated on poly-L-lysine-coated 24-well plates and grown in Dulbecco's Modified Eagle Medium (DMEM) (Life Technologies, Frederick, MD) supplemented with 10% fetal bovine serum (FBS) for 24 h. Cells were co-transfected either with pSRE-firefly luciferase (50 ng) (Stratagene, Cedar Creek, TX) or NFκB-firefly luciferase (50 ng) reporter (Stratagene, La Jolla, CA), combined with pRL-Renilla luciferase (20 ng), pcDNA 3.1-PAR-2 (50 ng) or pcDNA 3.1-PAR-1 (50 ng) (Missouri S&T cDNA Resource Center, Rolla, MO) using Lipofectamine according to the manufacturer’s instructions (Life Technologies). The medium was changed after 12 h, and 36 h after transfection, the cells were serum-starved overnight and then stimulated with either matriptase (R&D, Minneapolis, MN), PAR-2 agonist peptide ([3H]2-furoyl-LIGRL-NH2) (Tocris, Minneapolis, MN), PAR-1 and-2 agonist peptide (tcY-NH2) (Tocris), or vehicle for 6 h. Cells were washed twice in PBS and luciferase activity was determined using the dual luciferase assay kit (Promega, Madison, WI) according to the manufacturer's instructions. Luminescence was measured using a Microtiter Plate Luminometer (Dynex Technologies, Chantilly, VA) and the SRE activation was determined as the ratio of firefly to Renilla luciferase counts.

G-protein assay

HEK293T cells were plated in 24 well plates and transfected with either SRE-luciferase or NFkB-luciferase reporter together with pRL-Renilla and PAR-2. 36 h post-transfection the cells were serum starved ON with pertussis toxin (50 ng/ml) or with vehicle added to the media in a final concentration. For all three above mentioned G-protein experiments, the cells were stimulated with 10 uM PAR-2 agonist or 15 nM recombinant human matriptase for 6h at 37°C before the luciferase activity was determined.