Ras proteins play a causal role in human cancer by activating multiple pathways that promote cancer growth and invasion. However, little is known about how Ras induces the first diagnostic features of invasion in solid tumors, including loss of epithelial integrity and breaching of the basement membrane (BM). In this study, we found that oncogenic Ras strongly promotes the activation of hepsin, a member of the hepsin/TMPRSS type II transmembrane serine protease family. Mechanistically, the Ras-dependent hepsin activation was mediated via Raf–MEK–ERK signaling, which controlled hepsin protein stability through the heat shock transcription factor-1 stress pathway. In Ras-transformed three-dimensional mammary epithelial culture, ablation of hepsin restored desmosomal cell–cell junctions, hemidesmosomes, and BM integrity and epithelial cohesion. In tumor xenografts harboring mutant KRas, silencing of hepsin increased local invasion concomitantly with accumulation of collagen IV. These findings suggest that hepsin is a critical protease for Ras-dependent tumorigenesis, executing cell–cell and cell–matrix pathologies important for early tumor dissemination.
These findings identify the cell-surface serine protease hepsin as a potential therapeutic target for its role in oncogenic Ras-mediated deregulation of epithelial cell–cell and cell–matrix interactions and cohesion of epithelial structure.
During malignant tumor progression, the onset and maintenance of invasive phenotype not only require cytoskeletal alterations to help cell motility, but also disassembly of the cell–cell and cell–matrix junctions to facilitate detachment of the cells or cell clusters from the tumor bulk. These cell–cell and cell–matrix pathologies also include dissolution of the basement membrane (BM), which normally separates epithelial tissues from the stroma (1). While loss of the epithelial tissue structure and dissolution of the BM are diagnostic hallmarks of all advanced epithelial cancers, the dynamics and mechanisms causing these histopathologic changes are still poorly understood. At the cellular level, the intricate cell–cell and cell–matrix remodeling processes during invasion appear to involve concomitant or consequential cycles of loss and reestablishment of cell–cell and cell–matrix interactions to support better motility, survival, and proliferation (2, 3). It is generally held that these abrupt protumorigenic alterations emerge from oncogene-driven erratic intracellular cytokine and growth factor signaling mechanisms, which act in conjunction with dysregulated pericellular protease systems (4).
Oncogenic-activating mutations in Ras proteins are found in about one-third of human cancers; these mutations lock Ras into its active GTP-bound state, amplifying signaling through the Raf–MEK–ERK cascade (also known as MAPK pathway; refs. 5, 6; cBioPortal). In addition to point mutations, Ras gene amplifications, mutation-independent gain of function, and other mechanisms also commonly feed the downstream pathway in an abnormally persistent manner in epithelial cancers (6–12). The oncogenicity of Ras proteins was historically recognized in the context of cell transformation, typified by Ras-induced morphologic changes in cultured fibroblasts because of altered cytoskeletal dynamics and cell adhesions (13, 14). In two-dimensional (2D) monolayer cultures of epithelial cells, persistent Ras–MAPK signaling also diminishes cell adhesion by damaging epithelial-specific cell–cell junctions. For example, Ras destabilizes adherence junctions by downregulating E-cadherin and β-catenin (15). In addition, Ras-induced loss of occludin, claudin-1, and ZO-1 interferes with tight junction formation (16) and Ras-dependent dissociation of desmoplakin from the cell periphery disrupts desmosomes (DM; ref. 17).
In three-dimensional (3D) organotypic epithelial cultures, a hyperactive Ras pathway leads to loss of apicobasal cell polarity and disruption of epithelial architecture, which associate not only with deficiencies in adherence junctions and tight junctions, but also with loss of the α6-integrin and laminin components of the BM. The Ras-transformed disorganized epithelial 3D structures have been shown to feature groups of motile cells that form multicellular protrusions, resembling invasive processes (18–21).
While oncogenic activation of the Ras pathway alone is sufficient to break epithelial cohesion and induce dissolution of BM, little is known how Ras establishes these cell–cell and cell–matrix pathologies during epithelial tumor progression. Oncogenic Ras is known to activate a number of proteases, including matrix metalloproteinases (MMP), plasminogen activators, and cathepsins (22, 23). In particular, downstream of Ras, the MAPK pathway has been coupled to activation of MMP-9 and consequent cleavage of the BM protein, laminin-111 (21). However, besides possible involvement of MMPs, little is known about the mechanisms that couple active Ras to the pericellular cancer degradome.
Here, we found that oncogenic Ras proteins induce, via the MAPK-mediated heat shock transcription factor 1 (HSF1) stress pathway, a strong activation of the type II transmembrane serine protease, hepsin, which is an epithelial protease widely overexpressed in prostate, breast, ovarian, renal cell, and endometrial cancer (24–29). Strikingly, ablation of hepsin in Ras-transformed 3D epithelial structures leads to restoration of epithelial cohesion, desmosomal junctions, and an intact BM. In addition, we show that silencing of hepsin reduces invasive behavior of the mutant KRas–harboring tumor xenografts. These findings expose serine protease hepsin as a critical upstream regulator of the Ras-dependent cancer degradome and point out hepsin as a potential therapeutic target for Ras pathway intervention strategies.
Materials and Methods
See the Supplementary Materials and Methods in the Supplementary Data for the ethical permits (Supplementary Table S1), list of reagents (Supplementary Table S2), antibodies (Supplementary Table S3), cell lines and culture media composition (Supplementary Table S4), primer sequences (Supplementary Table S5), short hairpin RNA (shRNA) sequences (Supplementary Table S6), and vector constructs (Supplementary Table S7).
Human breast cancer samples and tissue from human reduction mammoplasties
Breast cancer specimens were derived from material sent to the pathology laboratory of the University of Helsinki Central Hospital (Helsinki, Finland) in conjunction with elective surgery, and the set was published previously (30). The Helsinki University Hospital Ethics Committee (Helsinki, Finland) approved this study. License for the use of human breast cancer samples was granted by the National Supervisory Authority for Welfare and Health, and for the use of noncancerous samples by the Ministry of Social Affairs and Health of Finland. Paraffin-embedded 5-μm-thick sections of noncancerous and breast cancer tumor tissue blocks were cut and plated onto objective glasses. Hepsin or phospho-ERK (pERK) IHC on sections was performed as described previously (27, 30). See Supplementary Table S1 for complete information about ethical permits related to the use of human material in this study.
Genetically engineered mouse models
Experimental animal work was approved under a license (see Supplementary Table S1) granted by The National Animal Ethics Committee of Finland. KrasLSL−G12D (B6.129S4-Krastm4Tyj/J) (here Kras; C57Bl/6J background) and Trp53fl/fl mice (B6.129P2-Trp53tm1Brn/J) (here p53; C57Bl/6J background) were crossed to produce Kras;p53 mouse line, a model for non–small cell lung cancer with advanced adenocarcinoma. To initiate tumor formation, mice were intranasally infected with adenovirus-CMV-Cre constructs (Gene Transfer Vector Core, University of Iowa, Iowa City, IA). After the mice were sacrificed, their lungs were fixed and processed for IHC. See also Supplementary Materials and Methods. For primary mouse mammary epithelial cell (MMEC) isolations and culturing see Supplementary Materials and methods.
MDA-MB-231 breast cancer cells were transduced with recombinant lentiviral particles carrying shRNA hepsin, expanded, and validated for hepsin knockdown. Thereafter, 8 × 105 cells were transplanted subcutaneously in two to three sites on the back skin of NOD.CB17-Prkdcscid/J (NOD SCID) mice (a kind gift from Dr. Kari Alitalo, University of Helsinki, Helsinki, Finland). The endpoint for tumor take rate and growth measurements was 33 days after the transplantation.
3D culture models
Briefly, human cell lines were 3D cultured in egg white, whereas mouse primary mammary epithelial cells were embedded in Matrigel (Becton Dickinson). See Supplementary Materials and Methods for details.
Cell culture and the cycloheximide chase assay
Authenticated cell lines were obtained from the ATCC (breast epithelial cell lines) or Invitrogen (293FT cells). MPE600 cell line was a gift from Dr. Outi Monni, University of Helsinki (Helsinki, Finland). Immediately upon arrival, cells were expanded and aliquoted into cell banks. The cell lines were cultured at +37°C and 5% CO2, except for MDA-MB cell lines that were kept at +37°C and atmospheric air conditions. The cells were in general passaged twice per week with two media changes per week. The exact culture conditions are described in the Supplementary Data. Our general laboratory practice allowed maximum of 30 passages per each vial thawed from the cell banks. All cell lines were regularly tested for Mycoplasma contamination once every 2 months with MycoAlert PLUS Detection Kit (Lonza, catalog no., LT07-710) according to the manufacturer's instructions. For cycloheximide chase experiment, the cells were grown in petri dishes until 80% confluent. Subsequently, the cells were treated with 50 μg/mL cycloheximide (ready-made solution in DMSO, Merck) or DMSO control for indicated times in normal cell culture conditions. Cell lysates were collected as described below in the “protein analysis” section.
For 3D immunofluorescence (IF), the 3D cultures were fixed with 2% paraformaldehyde, permeabilized with 100% methanol (−20°C), and blocked with 10% normal goat serum in IF buffer (0.1% BSA, 0.2% Triton-X, and 0.05% Tween-20 in PBS). The structures were stained with primary antibodies, and diluted in blocking buffer overnight at 4°C. The structures were then washed in IF buffer and incubated for 45 minutes at room temperature with the appropriate Alexa Fluor secondary antibody diluted into the blocking buffer. For additional details and 2D methods see Supplementary Materials and Methods.
Serine protease activity assays
To quantitate serine protease activity, the cells were treated with PBS (intact cells) or 1% Triton-X 100-PBS buffer (permeabilizing conditions) and, thereafter, fluorogenic tertbutoxycarbonyl-Gln-Arg-Arg-AMC (BOC-QRR-AMC) substrate was administered into the cells. Peptide cleavage was measured at 37°C using an ELISA plate reader at wavelengths 350em/450ex nm. See Supplementary Materials and Methods for other details.
Retroviruses were produced in the Phoenix Ampho packaging cells (gift from Garry Nolan, Stanford University, Stanford, CA) by using a total of 2 μg of recombinant vector plasmid per 6-well plate (5 × 105 cells were seeded 24 hours previously), and viral particles were collected after 72 hours. Target cell transductions and recombinant lentivirus production were performed as described previously (31).
Immunoblot and immunoprecipitation analyses
For protein extraction, the cells were lysed on ice using 1% Triton-X 100 with or without phosphatase inhibitors. Immunoblotting was performed as described in (30). Quantitation of loading normalized band intensity of at least three biological replicates was performed with ImageJ software (version: 2.0.0-rc-69/1.52n, RRID:SCR_003070). For those experiments not quantitated, a representative Western blot analysis result exemplary of at least three biological replicates is shown. For immunoprecipitation assays, the cell lysates were prepared as above. Immunoprecipitation was performed with Dynabeads Protein G (Invitrogen/Thermo Fisher Scientific) system according to the manufacturer's instructions.
qRT-PCR, cloning, transfections, image acquisition, image analysis, and quantitation
The methods have been described in Supplementary Materials and Methods.
Oncogenic Ras induces expression and proteolytic activity of hepsin
Hepsin protease activity has been commonly interpreted through proxies, which include elevated expression, processing of hepsin to its active 31-kDa fragment, or a redistributed hepsin expression pattern. We found in an earlier study that ectopic expression of hepsin associates with rapid downmodulation of hepsin's cognate inhibitor, hepatocyte growth factor activator inhibitor-1 (HAI-1; ref. 27), which prompted us to determine whether the normalized hepsin-HAI-1 (h-H) ratio offers a more sensitive proxy for hepsin activity than processed hepsin levels alone. We used MCF10A-hepsinIND20 cells expressing doxycycline-inducible hepsin (27) to compare noninduced (endogenous hepsin) and doxycycline-induced states, and found that the h-H ratio yields an improved sensitivity (11-fold) in comparison with the analysis of the 31 kDa hepsin fragment alone (1.6-fold; Fig. 1A).
Next, we analyzed the h-H ratio in MCF10A cells engineered to express common oncogenic proteins (Fig. 1B; Supplementary Fig. S1A and S1B) and found that oncogenic HRasV12, and to a lesser extent E2F-1, prominently increased the h-H ratio (Fig. 1B; Supplementary Fig. S1C). In addition to HRasV12, KRasV12 and ectopic wild-type KRas also increased the h-H ratio (Fig. 1C; Supplementary Fig. S1D and S1E). In addition to MCF10As, HRasV12 also augmented the h-H ratio when stably expressed in two other nonmalignant breast epithelial cell lines (Fig. 1D; Supplementary Fig. S1F and S1G). HRasV12 also enhanced the total proteolytic activity of hepsin, supported by evidence of enhanced cleavage of the preferred hepsin peptide substrate, BOC-QRR-AMC, in both intact and permeabilized MCF10A cells (Fig. 1E and F; see also Fig. 1A).
Ras and PI3K pathway mutations are mutually exclusive in breast cancer (32), which led us to explore the h-H ratio in a panel of seven genetically profiled breast cancer cell lines (33), three of which harbored an activating PIK3CA mutation, but wild-type HRAS and KRAS alleles (PIK3CAmut cells) and three others harboring activating mutations in HRAS or KRAS, but wild-type PIK3CA and PTEN (Ras-mutant cells). One cell line with an inactivating RB1 mutation (RB1-mutant cells) was examined, because deficiency in RB1 unleashes E2F proteins (33).
We measured a high (>10) h-H ratio in MCF10A-HRasV12 cells (positive control), in two of the three Ras-mutant breast cancer cell lines and in the RB1-mutant cells (Fig. 1G; Supplementary Fig. S1H). On the contrary, the h-H ratio was low (<1) in all PI3KCAmut cell lines. Furthermore, introduction of an ectopic wild-type KRas into the T47D PIK3CAmut cell line with low baseline h-H ratio increased the level of 31 kDa hepsin and decreased HAI-1, thus substantially increasing the h-H ratio (Fig. 1H; Supplementary Fig. S1I; see also Fig. 1G). Taken together, oncogenic Ras proteins strongly upregulate hepsin protein expression and enhance its proteolytic activity in nontransformed and cancerous cells.
Oncogenic Ras extends the half-life of hepsin protein
The effect of HRasV12 on h-H ratio was observed in all studied nontransformed mammary epithelial cell lines and in the majority of Ras-mutant breast cancer cell lines (Fig. 1B–D, G, and H; Supplementary Fig. S1C–S1I). However, the oncogenic Ras proteins did not alter hepsin mRNA levels (Fig. 2A), suggesting altered translation or protein stability. In support of enhanced stability, HRasV12 extended the half-life of hepsin in conditions where new protein translation was blocked with cycloheximide (Fig. 2B; Supplementary Fig. S1J).
Ubiquitin tagging is a common denominator in the targeting of proteins to all three major protein degradation pathways in mammalian cells: the proteasome, the lysosome, and the autophagosome (34). Consistent with the notion that oncogenic Ras reduces hepsin turnover, we found that KRasV12 expression reduces steady-state hepsin ubiquitination (Fig 2C, quantitation; Supplementary Fig. S1K). To define the major cellular degradation pathway for hepsin, we inhibited proteasome-mediated degradation pathways with three different compounds in three mammary epithelial cell lines with low or intermediate hepsin protein expression (MCF10A, T47D, and MCF7). However, proteasome inhibition did not affect hepsin turnover (Supplementary Fig. S1L). In contrast, inhibition of the lysosomes with bafilomycin B1 or concanamycin A led to a clear accumulation of hepsin protein in all cell lines tested (Fig. 2D). We also found that hepsin colocalizes with the lysosomal protein, lamp2 (Fig. 2E). The colocalization was observed in untreated cells and mostly visible in the cells treated with bafilomycin B1. These results suggest that the hepsin turnover is predominantly regulated by the lysosomal pathway.
To define the impact of HRasV12 on the lysosomal localization of hepsin, we quantitated the colocalization pattern of hepsin and lamp 2 in MCF10A-HRasV12 cells (Fig. 2E). While the total colocalization (colocalized pixel value/cell) was increased (Fig. 2E), the lysosome-size normalized colocalization of hepsin and lamp2 was decreased in MCF10A-HRasV12 cells as compared with the control MCF10A cells (Fig. 2F). We also analyzed the total lysosomal activity in HRasV12-expressing cells, and found that the total lysosomal activity was increased by Ras (Supplementary Fig. S1M). These observations are consistent with the earlier studies on the lysosomal effects of Ras (35, 36). Thus, we conclude that HRasV12-expressing cells have active lysosomes able to import hepsin, but we also note that HRasV12 may slightly reduce the lysosomal import of hepsin. This finding is congruent with the notion that HRasV12 reduces the steady-state ubiquitylation of hepsin.
The Ras–MAPK–HSF1 pathway upregulates hepsin
We generated MCF10A cells expressing a doxycycline-inducible HRasV12 to explore the acute effects of HRasV12 on hepsin and HAI-1 (Supplementary Fig. S2A). In MCF10A-Ind20-HRasV12 cells, doxycycline induction of HRasV12 elevated the levels of active 31 kDa hepsin fragment and enhanced the proteolytic QRRase activity within 48 hours without affecting hepsin mRNA levels (Supplementary Fig. S2B). Furthermore, long-term (24 days) induction of HRasV12 persistently upregulated the 31 kDa active hepsin (Fig. 3A). The level of doxycycline-induced hepsin (hepsinInd20 + doxycycline; see Fig. 1A) was about 1.6-fold more than the control and the level of hepsin expression induced by HRasV12 (Ind20-HRasV12 + doxycycline) was on average 2.8-fold more than the control (Fig. 3A). Surprisingly, a long-term HRasV12 induction was required for downregulation of HAI-1 levels, suggesting that HRasV12 regulates hepsin and HAI-1 via separate signaling mechanisms, rather than hepsin being directly causal to HAI-1 downmodulation (Fig. 3A and B). Hepsin can directly downregulate HAI-1 via a shedding mechanism (27), but it appears that HRasV12 regulates the balance of hepsin and HAI-1 in a more complex manner. We did not follow up further on the mechanisms of HAI-1 downmodulation; instead, we sought more insight into the mechanisms whereby HRasV12 regulates hepsin protein stability.
GTP-bound Ras recruits more than seven distinct Ras effector pathways for signaling, including the Raf–MEK–ERK MAPK cascade, PI3K–AKT–mTOR, and the RalGEF-Ral small GTPase signaling networks (37). We examined which HRasV12 pathways are responsible for hepsin upregulation by using seven small-molecule inhibitors targeting either the MAPK or PI3K pathways (Fig. 3C). The experiments revealed that only inhibition of MEK or ERK suppresses the HRasV12-induced upregulation of hepsin and its proteolytic activity (Fig. 3D–H; Supplementary Fig. S2C).
The Ras–MAPK pathway is classically pictured as a kinase signaling cascade, which results in translocation of a pool of ERK molecules to the nucleus (38). Nuclear ERK targets multiple transcription factors, including ELK1 and c-MYC, followed by transcriptional induction of mitogen-induced gene sets (39). While none of the classical MAPK-associated functions provided tangible rationale to explain hepsin stabilization, we were intrigued by recent findings linking Ras–MAPK signaling to MEK-induced phosphorylation of HSF1 (40). HSF1 is a stress-sensing transcription factor, which induces expression of HSPs that act as molecular chaperones to counteract protein misfolding and proteotoxicity (41). Notably, MEK-HSF1 targets include many plasma membrane receptors and proteins (40). We observed that both constitutive and inducible HRasV12 upregulate the HSF1 protein expression and induce HSF1 phosphorylation at its activating residue, Ser326 (Fig. 3I; Supplementary Fig. S2D). This HSF1 activation was inhibited by MEK or ERK inhibitor, indicating MAPK pathway as a mediator (Supplementary Fig. S2E). In addition, ectopic coexpression of hepsin and HSF1 increased hepsin levels, which shows that HSF1 can regulate hepsin turnover (Fig. 3J; Supplementary Fig. S2F). In vertebrates, heat shock induces HSF1 activation, and HSF1 together with other heat shock transcription factors coordinate the upregulation of heat shock genes (42). We used heat shock as a physiologic stimulus to activate endogenous HSF1, and found that heat shock leads to accumulation of hepsin concomitantly with S326-phosphorylated HSF1 (Fig. 3K). This heat shock induction of hepsin was blocked by KRIBB11, a small-molecule inhibitor that interferes with HSF1′s transcription factor function (Fig. 3K; ref. 43). We also silenced HSF1 by shRNA, finding that loss of HSF1 downmodulates hepsin protein levels in all tested cells expressing either ectopically or endogenously mutant Ras (Fig. 3L and M; Supplementary Fig. S2G–S2I). Finally, we silenced HSF1 by shRNA in cells that were engineered to transiently overexpress hepsin and KRasV12. We observed that the mutant Ras–induced diminished hepsin ubiquitination was rescued by loss of HSF1 (Fig. 3N; Supplementary Fig. S2J; see also Fig. 2C). These findings demonstrate that the mechanisms whereby Ras-MAPK stabilizes hepsin include HSF1 stress signaling pathway, which can regulate protein turnover via enhanced chaperoning activity through the action of HSPs (Supplementary Fig. S2K).
Oncogenic Ras disrupts hepsin-enriched epithelial junctions and induces DM-to-cytosol redistribution of hepsin
In the normal human breast tissue samples obtained via reduction mammoplasty, hepsin expression was strictly restricted to the basal and basolateral sides of the alveolar epithelial cells (Fig. 4A; ref. 27), which contrasts to breast tumor samples where hepsin is typically redistributed to the cytoplasm (27, 28, 44). In agreement, examination of a panel of 49 breast cancer samples showed strong hepsin immunoreactivity in as many as 41 samples (84%; top histogram) and the staining pattern was diffuse in >70% of the samples (bottom histogram, Fig. 4A).
Earlier studies have exposed hepsin colocalization with both desmosomal and hemidesmosomal proteins in nontransformed cells and some cancer cell types (26, 27). Desmosomes (DM) and hemidesmosomes (HDM) are both anchoring junctions that stabilize intercellular and cell–BM adhesions, respectively. Interestingly, similar to tumor samples, the Ras-mutant and RB1-mutant triple-negative breast cancer (TNBC) cell lines all showed diffuse cytoplasmic hepsin staining in monolayer 2D cultures. These cell lines lacked pericellular desmoplakin, indicating loss of DMs (Fig. 4B; Supplementary Fig. S3A). In contrast to mutant Ras–expressing cells, hepsin was expressed in the pericellular or pericellular/cytoplasmic regions in PIK3CAmut cells with strong overall colocalization with desmoplakin (Fig. 4B; Supplementary Fig. S3A). The proteolytic activity of hepsin corresponding to the cell lines shown in Fig. 4B is shown on the right side of the images (Fig. 4B). These results show a correlation between the oncogenic Ras activity and DM-to-cytosol redistribution of hepsin in breast cancer cells. Furthermore, we found that introduction of ectopic oncogenic HRasV12 into nontransformed mammary epithelial cells (MCF10A, MCF12A, and HMEC-hT) induced DM-to-cytosol redistribution of hepsin, resembling the expression pattern observed in the mutant Ras–expressing breast cancer cell lines (Fig. 4C; Supplementary Fig. S3B). Ectopic HRasV12, KRasV12, and wild-type KRas all resulted in similar impact on hepsin in isogenic MCF10A background (Fig. 4C and D). Moreover, all Ras-transformed cells consistently showed loss of pericellular desmoplakin (Fig. 4C and D).
Oncogenic Ras dissolves the hepsin-enriched BM
In 3D MCF10A cultures, epithelial cells form normal breast alveolus-like structures. In these cultures, hepsin localizes predominantly to BM as it does in the alveoli of normal human breast (Fig. 4A). Previous (27) and current IF studies indicate colocalization of hepsin with the HDM markers, laminin-332 and α6-integrin, in the nontransformed mammary epithelial 3D cultures (Supplementary Fig. S3C). The analysis of the breast cancer cell lines, which were able to form 3D structures in the egg white matrix (see below for rationale), showed that the PIK3CAmut cell lines, MCF7 and T47D, formed alveolus-like structures with basolateral hepsin expression (Fig. 4E). The Ras-mutant cell lines, Hs578T, MDA-MB-231, and MDA-MB-436, formed much less cohesive structures with cytosolic staining of hepsin. When MCF10A cultures with or without oncogenic Ras were compared, the controls showed basolateral expression of hepsin, whereas the HRasV12- or KRasV12-transformed structures showed cytosolic hepsin in poorly cohesive 3D structures (Fig. 4E). We also compared the effects of two classical cooperating oncogenes, MYC and HRasV12, and found that only HRasV12 induced cytoplasmic redistribution of hepsin and loss of cohesion in 3D structures (Fig. 4F, note the loss of HDM markers, laminin-332 and α6-integrin; Supplementary Fig. S3D). Immunoblot analysis of MCF10A-HRasV12 lysates showed cleavage of the laminin-332 β3 chain (Fig. 4G), which is a demonstrated proteolytic target of hepsin (27, 45). Notably, we used egg white matrix in all 3D experiments because this matrix does not contain matrix-derived BM components (like laminin-rich Matrigel) and is, therefore, expectedly more sensitive scaffold for analysis of invasive potential than artificially reconstituted BM gels (46). All basal BMs come from the cells that form epithelial structure in the egg white–based 3D cultures (27). Specifically, in the egg white–cultured HRasV12-transformed structures, the dissolution of BM correlated with diminished overall cohesion and occasional detachment of the cells from colonies, phenotypes that are reminiscent of the early-invasive processes observed in vivo (Fig. 4F).
Impact of endogenous KrasG12D on hepsin
To determine whether the endogenous levels of oncogenic Ras upregulate hepsin and induce its redistribution, we investigated mammary epithelial cells freshly isolated from mice harboring a conditional Lox-Stop-Lox (LSL)-KrasG12D allele (47). The expression of KrasG12D was induced ex vivo by infecting the isolated cells with Cre-expressing adenoviruses (AdCre), and subsequently, the cells were cultured in 3D (Supplementary Fig. S4A). Similar to the effects observed in MCF10A-HRasV12 culture experiments in Fig. 4F, endogenous active KrasG12D also promoted development of abnormally large and misshapen epithelial 3D structures, increased 31 kDa hepsin expression level (in five of seven mammary epithelial cell preparations isolated from different mice), and induced redistribution of hepsin from basolateral membranes to the cytosol (Fig. 4H and I; Supplementary Fig. S4A–S4C). Closer inspection revealed individual cells bulging out from the cultured 3D structures (Supplementary Fig. S4D). Next, we examined whether the physiologic expression level of KrasG12D induces redistribution of hepsin in vivo. KrasG12D was induced in a p53-null background by nasal inhalation of AdCre, which leads to formation of focal lung adenocarcinomas (48, 49). About 65% of the tumors manifested strong cytosolic hepsin immunostaining, contrasting to weak or absent hepsin expression in healthy lungs (Fig. 4J; Supplementary Fig. S4E). Thus, both the ectopic and endogenous expression of oncogenic Ras induces cytosolic redistribution and upregulation of 31 kDa active hepsin concomitantly with compromised epithelial integrity, dissolution of BM, and cellular bulging phenotype never observed in the control epithelial structures.
Ras effects on subcellular expression pattern of hepsin are mediated via the MAPK pathway
Because the Ras-induced biochemical effects, such as stabilization and proteolytic activity of hepsin, were mediated via MAPK pathway (Fig. 3D, E, and H; Supplementary Fig. S2C), we asked whether MAPK pathway is also involved in mediating the subcellular redistribution of hepsin. Indeed, MEK or ERK inhibition fully prevented Ras from inducing hepsin mislocalization and these inhibitors restored the desmosomal expression pattern of hepsin (Fig. 5A). In the 3D cultures of MCF10A-HRasV12 cells, MEK and ERK inhibitors restored the symmetric morphology, basal expression pattern of hepsin, as well as colocalization of laminin-332 and α6-integrin at the BM (Fig. 5B). Thus, intervention of the oncogenic Ras pathway via the inhibitors of MAPK pathway fully restores the normal expression pattern and activity of hepsin.
Ras mutations are rare in breast cancer, but a number of other mechanisms eliciting overactive Ras–MAPK signaling have been reported in breast cancer, for example, Ras gene amplification, overexpression of EGFR, HER2, and other receptor tyrosine kinases, and loss of pathway inhibition via mutated RAS GAPs (Supplementary Fig. S5A; refs. 7, 9–11, 50). To determine whether overactive MAPK pathway associates with high levels of hepsin in breast cancer, we determined the hepsin and pERK expression in serial sections cut from paraffin-embedded tumor samples corresponding to 57 patients with breast cancer (Supplementary Fig. S5B; ref. 30). We found that about 80% of the breast cancer samples showed intense pERK staining and two-thirds (66%) of the samples were double positive for pERK and hepsin (hepsinhigh/pERKhigh; Fig. 5C).
Hepsin is required for the Ras-induced loss of epithelial integrity
To explore the role of hepsin in Ras-mediated disruption of epithelial cohesion, we silenced hepsin by an shRNA approach in MCF10A-HRasV12 cells. Knockdown of hepsin diminished hepsin activity and restored HAI-1 expression (Fig. 6A and B; Supplementary Fig. S5C). In 2D petri dish cultures of MCF10A-HRasV12 cells, the hepsin knockdown partially rescued DM, and the residual low levels of hepsin, which escaped knockdown, predominantly colocalized with desmoplakin in the pericellular regions (Fig. 6C). In 3D cultures of MCF10A-HRasV12 cells, hepsin knockdown had a dramatic effect both on morphology and HDMs; hepsin shRNAs robustly rescued the symmetrical glandular morphology of the HRasV12-transformed structures and restored the basal expression of α6-integrin and the β3 and γ2 chains of laminin-332 (Fig. 6D). Moreover, silencing of hepsin inhibited the proteolytic processing of laminin-332 β3, which occurred in control MCF10A-HRasV12 cells (Fig. 6E). In 3D culture, HRasV12 expression perturbed apical polarity, as indicated by disorientation of the apical Golgi marker, GM130, but this apical polarity defect could not be rescued by knockdown of hepsin (Fig. 6F).
To confirm the results suggesting a critical role for hepsin as a mediator of the loss of epithelial cohesion caused by oncogenic Ras, we employed a hepsin-neutralizing antibody (Ab25) for the experiments (51, 52). The MCF10A-Ind20-HRasV12 cells were cultured for 4 days in an exogenous BM-free egg white hydrogel to form acinar structures, followed by 3-day doxycycline treatment to acutely induce HRasV12. The doxycycline-induced HRasV12 expression, like chronic HRasV12 expression, resulted in a disruption of epithelial integrity and loss of HDMs (Fig. 6G). These effects were fully prevented by the hepsin function–blocking antibody, Ab25, that inhibited peptide cleavage activity in 3D as well (Fig. 6G; Supplementary Fig. S5D). Together, these findings indicate a key role for hepsin as a mediator of Ras-dependent disruption of the epithelial integrity, involving loss of adhesive DM and HDM structures.
Silencing of hepsin reduces tumor take rate and invasive growth of KRasG13D MDA-MB-231 xenografts
We examined the role of hepsin in tumor growth by using an MDA-MB-231 xenograft model of TNBC. Hepsin was silenced by shRNA in the MDA-MB-231 breast cancer cells, which carry a mutant KRas; KRasG13D (cBioPortal). After validation of hepsin knockdown, the cells were subcutaneously transplanted into the host mice. We observed a lower take rate for the cells transduced with shRNA for hepsin (Fig. 7A). In addition, tumor growth was somewhat slower in the groups with silenced hepsin, although the results did not reach statistical significance (Fig. 7B). We investigated the local tumor invasion pattern from formalin-fixed sections subjected to hematoxylin and collagen IV staining, and found that the tumor growth pattern was clearly different between the tumors originating from control or hepsin shRNA–transduced cells. The control samples typically showed a tight tumor bulk (T) predominantly composed of tumor cells and a surrounding region showing a mixture of tumor cells and muscle cells (T/M) or tumor cells and fat cells (T/F; Fig. 7C). Those areas showing a mixture of cells represented the invasive tumor front (ITF). The ITF was significantly smaller in tumors originating from shHepsin #1 or #5 transduced MDA-MB-231 cells in comparison with the controls (Fig. 7C). We also analyzed the expression pattern of collagen IV, which is a major BM component, and found that loss of hepsin led to a significant increase in the collagen IV levels within the tumor tissue (Fig. 7D). These results suggest an important role for hepsin in the establishment of invasive front by the TNBC cells of tumor grafts.
Together, our findings suggest a previously unanticipated functional role for the type II transmembrane serine protease, hepsin, as a component of the oncogenic Ras pathway (Fig. 7E). The Ras–MAPK–HSF1 pathway enhances hepsin activity to disrupt junctional integrity, BM, and epithelial cohesion, which are cellular events that commonly associate with the earliest stages of tumor invasion.
In this study, we show that oncogenic Ras proteins upregulate the expression and activity of the transmembrane serine protease, hepsin, via a set of molecular events that act downstream of the Ras–MEK–ERK pathway. In the context of Ras-transformed epithelial tissue, dysregulated hepsin has a crucial role in mediating disruption of epithelial integrity that involves loss of desmosomal structures important for cell–cell adhesion and loss of hemidesmosomal structures important for cell–BM adhesion.
Earlier studies have proposed that the functional changes observed in cancer cell proteomes, which ultimately lead to enhanced enzyme activity, are commonly because of altered posttranscriptional and posttranslational mechanisms rather than transcriptional changes (53, 54). Here, we show evidence that Ras stabilizes hepsin and upregulates its proteolytic activity through the MEK–HSF1 stress signaling pathway. The MEK pathway has been previously shown to upregulate HSF1 via MEK-dependent phosphorylation of HSF1 at S326 and upregulated HSF1 affects the stability of number of proteins via HSPs (40). For example, in response to acute stress, HSF1 transcriptionally induces the expression of HSP27 that augments the basal chaperoning activity provided by noninducible HSC70 and HSP90 proteins (55). The HSF1 response is normally transient due to negative feedback loops, but in cancer is often constitutively active (55). We found that HSF1 is required for hepsin upregulation in the presence of active Ras signaling and that in the absence of oncogenic Ras, both the ectopic HSF1 expression and heat shock treatment via HSF1 are sufficient to upregulate the protein levels of hepsin.
We also demonstrated that oncogenic Ras decreases the steady-state ubiquitination of hepsin and that this effect was mediated via HSF1. Ubiquitin tagging can target proteins to all three major protein degradation pathways in mammalian cells: the proteasome, the lysosome, and the autophagosome (34), and our experiments suggest that the turnover of hepsin is primarily regulated by the lysosomal degradation pathway. These results, altogether, would be consistent with a model that Ras–MEK–HSF1 pathway reduces the steady-state ubiquitination of hepsin, which then inhibits transit of hepsin from the membranes via endosomes to the lysosomes. Indeed, for many transmembrane receptors, ubiquitination is a signal that triggers endocytosis and transport to the lysosome for degradation (56). However, while HRasV12-transformed cells showed proportionally less hepsin in their lysosomes than control cells, the difference was small, and the lysosomes were highly active in the HRasV12-transformed cells. Therefore, the “suppressed lysosomal transport” model, at this stage, is still only speculative and requires further investigations. The alternative possibility, also supported by the current data, is that hepsin is accumulated outside the lysosomal compartment because of enhanced chaperoning activity instated by oncogenic Ras–HSF1 pathway (Supplementary Fig. S2K). In this model, stabilization of hepsin is part of the adaptive cellular response to Ras-induced chronic oncogenic stress. While the natural stress response primarily occurs to alleviate proteotoxic stress in circumstances of acute stress signaling, chronic oncogenic stress may lead to a number of hazardous protumorigenic events, such as dysregulated expression and activity of hepsin. While the exact mechanisms coupling HSF1 to upregulation of hepsin remain to be clarified, this study clearly adds hepsin to a list of protumorigenic proteins upregulated by the MAPK–HSF1 stress branch of Ras signaling.
In both normal breast epithelium and mammary epithelial 3D culture, hepsin most visibly localizes to the BM and basolateral membranes. More specifically, data from this and earlier studies have shown that in the basolateral membranes, hepsin colocalizes with lateral DM proteins, and with HDM in the BM (26, 27, 30). Strikingly, in Ras-transformed 3D cultures, which show loss of epithelial integrity and severe loss of HDM, the inhibition of hepsin by shRNA or with a function-blocking antibody nearly completely rescued basal α6-integrin and laminin-332 expression, as well as the BM integrity. The apical polarity marker, GM130, was not rescued, whereas DMs, which overlap with the site of hepsin expression, were also rescued in HRas-transformed cultures. It remains somewhat an open question whether Ras upregulates the hepsin activity first, and then this event is causal to loss of DMs, HDMs, and epithelial integrity. Alternatively, Ras could activate hepsin and induce its redistribution more indirectly, by disrupting junctional protein complexes that normally determine the spatial localization of hepsin. However, in this latter case, hepsin inhibition might not have such dramatic epithelial integrity normalizing actions as we showed in this study. In support of the “hepsin first” model, earlier findings have shown that ectopic doxycycline-induced overexpression and enhanced activity of hepsin alone induce similar deteriorating effects on DMs and HDMs as we have shown here with Ras (27). It should be noted, however, that even if hepsin is crucially important for the Ras-induced loss of epithelial cohesion, it is unlikely that hepsin performs all epithelial cohesion eroding activities alone. For example, earlier studies have implicated a role for MMP-9 as a downstream mediator of Ras in those events that disrupt polarity, the laminin-111 component of the BM, and epithelial integrity in 3D culture (21). Moreover, hepsin has been linked to activation of pro-MMPs in osteoarthritis and prostate cancer models (57, 58).
Collectively, our findings suggest that overactive hepsin is an integral part of the Ras-dependent degradome, promoting the loss of desmosomal and hemidesmosomal junction integrity, and contributing to the loss of BM structure and overall epithelial integrity. Such proteolytic events contribute to processes that provide tumor cells invasive potential. In support of a role for hepsin in local tumor invasion, we found that ablation of hepsin in KRas-mutant MDA-MB-231 cells generated tumor with diminished invasive potential. Hepsin-deficient tumors displayed increased deposition of collagen IV, which is a major component of BM. These results are consistent with the proposed role of hepsin in regulation of the BM homeostasis. We believe that hepsin, as an extracellular and hence drug- or antibody-accessible protease, could offer an attractive target for future anticancer drug development, aiming to target not only dysfunctional cell proliferation machinery, but also early stages of tumor dissemination.
T.A. Tervonen reports grants from Academy of Finland, Business Finland, EU H2020 RESCUER, Finnish Cancer Organizations, Sigrid Juselius Foundation, and iCAN Digital Precision Cancer Medicine Flagship and nonfinancial support from Genentech during the conduct of the study, as well as other compensation from UPM Biomedicals (payment for research service) outside the submitted work. S.M. Pant reports grants from Instrumentarium Science Foundation, Orion Research Foundation, K. Albin Johanssons Stiftlse, Finnish Cancer Foundation, Biomedicum Foundation, Ida Montinin Säätiö, and Paulon Säätiö during the conduct of the study. J. Klefström reports grants from Finnish Cancer Organizations, The Academy of Finland, Business Finland, HiLIFE, EU H2020 RESCUER, and iCAN Digital Precision Cancer Medicine Flagship and grants and personal fees from Sigrid Juselius Foundation during the conduct of the study, as well as grants and personal fees from AbbVie, Orion Pharma, Roche/Genentech, and UPM Biochemicals and personal fees from AstraZeneca, MSD, and Pfizer outside the submitted work. No disclosures were reported by the other authors.
T.A. Tervonen: Conceptualization, resources, data curation, formal analysis, supervision, funding acquisition, validation, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing. S.M. Pant: Resources, data curation, validation, investigation, methodology, writing–review and editing. D. Belitškin: Resources, data curation, formal analysis, investigation, methodology. J.I. Englund: Resources, data curation, formal analysis, validation, investigation, visualization, methodology, writing–original draft. K. Närhi: Resources, formal analysis, investigation, methodology, writing–original draft. C. Haglund: Resources, methodology. P.E. Kovanen: Resources, data curation, formal analysis, investigation, visualization, methodology. E.W. Verschuren: Resources, investigation, methodology. J. Klefström: Conceptualization, resources, data curation, formal analysis, supervision, funding acquisition, validation, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing.
The authors thank Dr. Daniel Kirchhofer (Genentech, Inc.) for providing Ab25. They thank Dr. Gerard Evan for providing pBabe-puro MycERtm and pBabe-puro MycΔERtm vectors. Vector pWzl blast h-TERT was a kind gift from Dr. Martin McMahon and Robert A. Weinberg. Vectors pWzl hygro (empty backbone) and pWzl-hygro HrasV12 were a kind gift from Dr. Scott Lowe. The authors thank Dr. Kristian Helin and Dr. Karin Helin for providing the vectors pBabe-puro (HAERE2F-1) E2F-1ERtm and pBabe-puro (HAERE132) E2F-1ΔERtm. Vectors pBabe-puro K-RasV12 and RasV12 (HrasV12) in pENTR1A (w99-1) were kind gifts from Dr. William Hahn and Dr. Eric Campeau, respectively. They also thank Dr. Outi Monni for providing MPE 600 cell line. The authors are grateful to Biomedicum Functional Genomics Unit/Libraries, Biomedicum Virus Core Unit, Genome Biology Unit and Biomedicum Imaging Unit, and Laboratory Animal Center [all from Helsinki Institute of Life Science Infrastructures (HiLIFE), University of Helsinki and Biocenter Finland] for their services. They thank Dr. Jeroen Pouwels and other Klefström laboratory personnel for discussions and critical comments on the article. The authors thank T. Raatikainen, M. Merisalo-Soikkeli, K. Karjalainan, and T. Välimäki for providing technical assistance. This work was funded by grants from The Academy of Finland, Business Finland, EU H2020 RESCUER, Finnish Cancer Organizations, Sigrid Juselius Foundation and iCAN Digital Precision Cancer Medicine Flagship, and TEKES. T.A. Tervonen was funded by the HiLIFE of University of Helsinki.