Rhbdf2 mutations increase its protein stability and drive EGFR hyperactivation through enhanced secretion of amphiregulin

Significance

Epidermal growth factor receptor (EGFR) signal transduction plays a major role in growth, proliferation, and differentiation of mammalian cells. Although inactive rhomboids (iRhoms) are cardinal regulators of EGFR signaling in Drosophila melanogaster, their physiological role in regulating EGFR signaling and their substrates in mammals remain unclear. Here, we show that iRhoms are short-lived proteins, but dominant mutations increase their protein stability and stimulate secretion of specific EGF family ligand amphiregulin independent of metalloprotease activity. This study demonstrates the significance of mammalian iRhoms in regulating an EGFR signaling event that promotes accelerated wound healing and triggers tumorigenesis. Given their ability to regulate EGFR signaling in parallel with metalloproteases, iRhoms can be potential therapeutic targets in impaired wound healing and cancer.

Abstract

The rhomboid 5 homolog 2 (Rhbdf2) gene encodes an inactive rhomboid (iRhom) protease, iRhom2, one of a family of enzymes containing a long cytosolic N terminus and a dormant peptidase domain of unknown function. iRhom2 has been implicated in epithelial regeneration and cancer growth through constitutive activation of epidermal growth factor receptor (EGFR) signaling. However, little is known about the physiological substrates for iRhom2 or the molecular mechanisms underlying these functions. We show that iRhom2 is a short-lived protein whose stability can be increased by select mutations in the N-terminal domain. In turn, these stable variants function to augment the secretion of EGF family ligands, including amphiregulin, independent of metalloprotease a disintegrin and metalloproteinase 17 (ADAM17) activity. In vivo, N-terminal iRhom2 mutations induce accelerated wound healing as well as accelerated tumorigenesis, but they do not drive spontaneous tumor development. This work underscores the physiological prominence of iRhom2 in controlling EGFR signaling events involved in wound healing and neoplastic growth, and yields insight into the function of key iRhom2 domains.

Inactive rhomboids (iRhoms) are highly conserved but proteolytically inactive intramembrane proteins (1). iRhoms are characterized by a long cytosolic N-terminal domain, a conserved cysteine-rich iRhom homology domain (IRHD), and a dormant proteolytic site lacking an active-site serine residue within the peptidase domain (1). Recently, Greenblatt et al. (2) reported that Derlin-1 belongs to the rhomboid family and suggested that a dual role exists for the cytosolic and peptidase domains of this novel rhomboid pseudoprotease. They showed that although the cytosolic domain of Derlin-1 is essential in mammalian cells for clearance of misfolded proteins from the endoplasmic reticulum, the transmembrane domain is required to interact with its substrates (2), suggesting that these two domains have distinctive functions. The physiological significance of these domains is unclear.

Despite their lack of proteolytic activity, iRhoms participate in a diverse range of functions in a variety of species, including regulation of epidermal growth factor receptor (EGFR) signaling in Drosophila melanogaster (3), survival of human squamous epithelial cancer cells (4, 5), misfolded protein clearance from endoplasmic reticulum membranes in mammalian cell lines (2), induction of migration in primary mouse keratinocytes (6), secretion of soluble TNF-α in mice (7, 8), and regulation of substrate selectivity of stimulated a disintegrin and metalloproteinase 17 (ADAM17)-mediated metalloprotease shedding in mouse embryonic fibroblasts (MEFs) (6, 9). Although the physiological targets of iRhoms are largely unknown, mounting literature suggests that EGF-like ligands are potential substrates. For example, studies in D. melanogaster and in mammalian cell lines showed that iRhoms could negatively regulate EGFR signaling by breaking down EGF-like substrates through endoplasmic reticulum-associated degradation (ERAD) (3).

Recent data also point to a role for iRhoms in EGFR-mediated human neoplastic growth. siRNA-mediated silencing of iRhom1 inhibits tumor growth by inducing apoptosis (4). Likewise, missense mutations in RHBDF2 (p.I186T, p.P189L, and p.D188N), the gene encoding iRhom2, cause tylosis with human esophageal cancer, which is characterized by palmoplantar and oral hyperkeratosis (10). Other studies have revealed that iRhom1 promotes the survival of epithelial tumors through EGFR transactivation (5), whereas somatic mutations in iRhom2 are strongly linked to gain of EGFR signaling (11). Although the mechanisms underlying the pathogenesis of cancer are evolving, these studies further strengthen the link between iRhoms and EGFR signaling.

Rhomboids have been extensively characterized in Drosophila, in which changes in EGFR signaling can be detected by studying the wing phenotype (12, 13), whereas in mice, such alterations can be analyzed by examining hair follicle development (14), response to wound healing (15), and tumorigenesis (16). Here, we show that a spontaneous deletion within the rhomboid 5 homolog 2 (Rhbdf2) gene in mice underlies the curly-bare (cub) mutation, in which loss of the cytosolic N-terminal domain of iRhom2 causes subsequent effects on hair follicle development, wound healing, and tumorigenesis. We find that iRhom2 is a short-lived protein but that gain-of-function mutations in the N terminus (tylosis) or loss of the N terminus (cub mutation) increase mutant protein stability, leading to metalloprotease-independent secretion of the EGFR ligand amphiregulin (AREG). Using a genetic modifier of the cub phenotype (Mcub), we demonstrate that AREG is a physiological target of iRhom2. We also identify key amino acids in the peptidase domain of iRhom2 that are necessary for AREG secretion, suggesting that the peptidase domain of this pseudoenzyme might be functional despite lacking a serine residue in the putative active site. This study therefore yields insight into the function of key iRhom domains and establishes a framework for understanding the relationship between iRhoms, EGFR signaling, and the biological processes involved in wound healing and tumorigenesis.

Results

cub Mutation Leads to Hyperactivation of the EGFR Signaling Pathway.

We previously described a recessive mouse mutation named cub, which is characterized by a hairless phenotype (17). We previously mapped the cub mutation to a locus on distal chromosome (Chr) 11 and mapped a dominant Mcub to a 10-cM interval on Chr 5 (17). A single copy of the dominant Mcub allele in combination with the cub/cub genotype results in a full, wavy coat rather than the hairless coat of cub/cub mcub/mcub mice (17). Because of the potential relationship between the hair coat phenotypes and EGFR signaling (18, 19), we investigated whether the cub/cub mcub/mcub genotype might also be associated with other aspects of altered EGFR signaling, such as cell proliferation and cell migration.

We performed proliferation, cell migration (scratch-wound healing), and immunoblot assays on mouse MEFs isolated from cub/cub mcub/mcub and control (^+/+ mcub/mcub) mice. The cub/cub mcub/mcub MEFs had significantly higher rates of proliferation and migration relative to control MEFs (Fig. 1 A and B). These changes were associated with significant increases in the phosphorylation of canonical signal transduction proteins of the EGFR pathway, including, Akt, S6, mTOR, and p38 (Fig. 1C). Further, we observed a significant reduction in cell-surface EGFR levels in cub/cub mcub/mcub MEFs (Fig. 1C), indicating internalization and constitutive activation of EGFR signaling (20–22).

Fig. 1.

cub mutation accelerates EGFR-related cell proliferation and migration, as well as cutaneous healing. (A) Quantitation of MEFs using a proliferation assay (SI Materials and Methods). Cells were seeded at the numbers indicated on the x axis and incubated for 24 h. The extent of fluorescence is proportional to the amount of total cellular DNA. (B) Quantitation of migration of cub/cub mcub/mcub and ^+/+ mcub/mcub MEFs using a scratch-wound assay (Materials and Methods). The width of the scratch wound was measured at time 0 (100% open), and the increase in wound closure at each time point was calculated as a percentage of the original width. (C) Immunoblot analysis of cub/cub mcub/mcub or ^+/+ mcub/mcub MEFs for various markers of EGFR signaling. Cell lysates were run in duplicate. Actin served as a loading control. (D) Representative images of regenerating ear tissue in 6- to 40-wk-old female cub/cub mcub/mcub and ^+/+ mcub/mcub mice (n = 3 per group) at 0, 7, 14, 21, and 28 d postwounding. (Magnification: 4×.) (Scale bar: 1mm.) (E) Quantification of ear hole closures shown in D. (F) Cross-section of ears from cub/cub mcub/mcub and ^+/+ mcub/mcub mice at 0, 7, and 14 d postwounding stained with H&E. (Magnification: 10×.) (Scale bars: 100 μm.) Notice the undifferentiated and thickened epidermis (E; 10–12 nucleated layers) and the extensive degree of proliferation (M) in the ears of cub/cub mcub/mcub mice. The dotted line indicates the site of excision. Data in A, B, and E are shown as mean ± SD.

Because activated EGFR signaling mediates epithelial regeneration (23–25), we examined whether this increase in EGFR signaling corresponds to changes in epithelial proliferation. We conducted wound-healing assays in which we punched 2-mm through-and-through holes into the ears of 6- to 40-wk-old cub mice and monitored the rate of closure over subsequent days (Fig. 1 D–F). Within 14 d, cub/cub mcub/mcub mice showed accelerated closure compared with control ^+/+ mcub/mcub littermates. The ⁺/cub mcub/mcub mice (which have a normal coat) also showed faster healing 14 d postinjury relative to control mice (Fig. S1 A and B), suggesting that a single mutant cub allele can trigger increased EGFR signaling, although not at levels high enough to block hair follicle induction (17, 26). These data strongly indicate that the cub/cub mcub/mcub genotype results in a hyperactive EGFR phenotype.

cub Is a Mutation of the Rhbdf2 Gene, Encoding iRhom2.

Next, we examined the iRhom2 gene Rhbdf2 as a candidate for the cub mutation because of its coincident position with the cub mutation on Chr 11 and the established relationship between rhomboid proteases and EGFR-mediated signaling (3, 6). DNA sequencing of cub/cub mice identified a 12,681-bp deletion in the Rhbdf2 gene, which results in loss of exons 2–6 (Fig. 2A). To test whether the deletion produces an aberrant Rhbdf2 transcript, we performed reverse transcriptase (RT) PCR on RNA derived from WT and cub/cub mcub/mcub MEFs using primers designed to amplify exons 2, 5, 12, and 19. As expected from the extent of the genomic deletion, the cub transcript contained exons 12 and 19 but lacked exons 2 and 5 (Fig. 2B). Further, quantitative RT-PCR (qPCR) on RNA from cub/cub mcub/mcub and ^+/+ mcub/mcub skin amplified cub transcripts in which exons 2–6 were deleted but the remaining exons were expressed, suggesting that the rest of the cub gene is transcribed in cub mice (Fig. 2C). The structure of the cub genomic deletion suggests that cub transcripts would result in splicing of exons 1–7. To test this interpretation, we performed qPCR with a probe specifically designed to amplify only transcripts containing exon 1 spliced to exon 7. Using this probe, we detected a PCR product in cub/cub mcub/mcub mice but not in ^+/+ mcub/mcub mice (Fig. 2D). These findings confirm that cub is a mutation of the Rhbdf2 gene, and it will henceforth be referred to as Rhbdf2^cub.

Fig. 2.

cub is a mutation of the Rhbdf2 gene. (A) cub mutation is a 12,681-bp deletion in the mouse Rhbdf2 gene. The deletion starts midway between exons 1 and 2, and encompasses exons 2–6, ending shortly after exon 6. (B) RT-PCR on MEFs from ^+/+ mcub/mcub, ⁺/cub mcub/mcub, and cub/cub mcub/mcub mice using primers against exons 2, 5, 12, and 19. ntc, no template control. (C) qPCR on cDNA extracted from skin tissues using TaqMan gene expression assays against the indicated exon boundaries. Actin served as an endogenous control. Data are normalized to ^+/+ mcub/mcub actin levels. Samples were run in triplicate with four biological replicates. (D) qPCR with a custom TaqMan gene expression assay to detect transcripts with an exon 1/exon 7 boundary. Samples were run in triplicate with three biological replicates (each color bar represents an individual mouse) Data in C and D are shown as mean ± SD. (E) Anti-Flag immunoblot of HEK 293 cells transiently expressing Flag-tagged HuWt (human full-length RHBDF2 cDNA) and HuCub (human version of cub cDNA). Actin served as a loading control. (F) Representative images of Flag-tagged HuWt- and HuCub-expressing B6 primary MEFs stained using a Flag-specific antibody (brown). DAPI was used to counterstain the nucleus (blue). Arrowheads represent cytoplasmic expression of both full-length and mutant proteins. (Magnification: 40×.) (Scale bars: 25 μm.)

Given that the normal translation initiation site in exon 3 is missing in Rhbdf2^cub transcripts, we next tested whether these mutant transcripts could produce a mutant iRhom2 protein. Sequence analysis of Rhbdf2^cub DNA revealed that the next in-frame translation initiation site (ATG) was in exon 8, which would result in an ∼63.5-kDa protein. Because we lacked an antibody to iRhom2, we tested whether Rhbdf2^cub transcripts could produce a protein product in vitro by cloning both full-length WT human RHBDF2 cDNA (HuWt) and a version that mimicked the mutant Rhbdf2^cub transcript (HuCub) into a C-terminal Flag-tagged expression vector. First, using immunoblotting, we determined that the HuWt clone generated an ∼100-kDa protein product, whereas the HuCub mutant construct generated an ∼67-kDa product, consistent with the expected molecular mass of the tagged proteins (Fig. 2E). Furthermore, there was no evidence of any shortened protein products in the HuWt clone. Second, using immunostaining, we examined the localization patterns of both forms of protein after transfection of B6 MEFs (Fig. 2F and Fig. S2 A and B) with either the HuWt or HuCub clone. Both forms were expressed in the endoplasmic reticulum, and no staining was observed in either the Golgi or nucleus (Fig. S2 A and B), suggesting that the mutation does not lead to altered protein localization.

Genetic Noncomplementation Confirms That cub Is a Mutant Allele of the Rhbdf2 Gene and a Gain-of-Function Mutation.

To examine whether Rhbdf2^cub is a gain-of-function mutation rather than a null mutation, we generated Rhbdf2 KO (Rhbdf2^−/−) mice using ES cells from the Knockout Mouse Project (KOMP) repository, in which lacZ expression is under control of the endogenous Rhbdf2 promoter (Fig. 3A). Rhbdf2 promoter-driven lacZ expression was predominantly observed in the epidermis and the inner and outer sheath layers of hair follicles in the skin (Fig. 3B). However, the wound healing (Fig. 3C) and loss-of-hair (Fig. 3D) phenotypes observed in the Rhbdf2^cub/cub mice were not seen in Rhbdf2^−/− mice, which appeared otherwise normal. An allele test mating between Rhbdf2^cub/cub and Rhbdf2^−/− mice yielded compound mutant (Rhbdf2^−/cub) mice with a sparse hair coat (Fig. 3E), indicating genetic noncomplementation and confirming that cub is a mutant allele of the Rhbdf2 gene. Additionally, the level of hair growth in the Rhbdf2^−/cub compound mutant mice was intermediate between that of Rhbdf2^cub/cub and Rhbdf2^+/+ or Rhbdf2^−/− mice, confirming that expression of the mutant Rhbdf2^cub protein product, rather than Rhbdf2 deficiency, causes the Rhbdf2^cub/cub phenotype.

Fig. 3.

Rhbdf2^cub is a gain-of-function mutation rather than a null mutation. (A) PCR for WT product in KOMP Rhbdf2 KO mice. The expected product is 2,181 bp. L, New England Biolabs 1-kb DNA ladder. (B) Reporter gene analysis of Rhbdf2 expression. A whole-mount X-gal–stained embryonic day 18.5 Rhbdf2^−/− embryo shows strong expression of β-gal in the epidermis (arrowhead), and X-gal–stained 2-wk-old female Rhbdf2^−/− skin shows β-gal positivity in the inner and outer sheath layers of hair follicles (arrow). (Magnification: 2.5× and 5×, respectively.) No staining was observed in the hair shaft. (Scale bars, 1 mm.) (C) Mice with a null mutation of Rhbdf2 lack a regenerative phenotype as demonstrated by significantly delayed wound closure in the ears of 6- to 8-wk-old female Rhbdf2^−/− mice compared with those of Rhbdf2^cub/cub mice. (D) Mice with a null mutation of Rhbdf2 have normal skin and hair morphology. H&E-stained sections of adult skin from Rhbdf2^cub/cub and Rhbdf2^−/− mice were taken at the indicated times after wounding. The adult Rhbdf2^cub/cub skin displays a thin hypodermal fat layer (F), abnormal hair follicles (H), thick epidermis (E), enlarged sebaceous glands (SG), dense interlacing bundles of collagen fibers (C), and no full differentiation of hair follicles and hair bulb (HF). The adult Rhbdf2^−/− skin shows normal epidermis and hair follicles. (Scale bars, 100 μm.) (Magnification: 10× and 40×, respectively.) (E) Rhbdf2^−/− mice develop a normal hair coat (arrowhead), whereas compound mutant Rhbdf2^−/cub mice exhibit a sparse hair coat (arrow).

Genetic Mcub Is a Loss-of-Function Mutation of the Areg Gene.

We next sought to map the mutation underlying the Rhbdf2^cub modifier gene (Mcub). The map position (17) coincides with that of a cluster of four EGFR ligand-encoding genes on Chr 5: Epng, Ereg, Areg, and Btc. We considered each of these four genes as candidate loci for the Mcub mutation. By sequencing the exons and flanking regions of each gene in Mcub/Mcub Rhbdf2^cub/cub mice (Table S1), we found that Mcub is a loss-of-function mutation in Areg (Fig. 4A), a gene that encodes the autocrine keratinocyte growth factor AREG (27).

Fig. 4.

Mcub is a mutation of the Areg gene. (A) Mcub is a T-to-G point mutation that disrupts the normal donor splice site (exon 1) in the Areg gene, causing the use of an alternative downstream splice site. (B) Healing of ear holes in 6- to 8-wk-old female Rhbdf2^+/+ Areg^+/+, Rhbdf2^cub/cub Areg^+/+, and Rhbdf2^cub/cub Areg^Mcub/Mcub mice (n = 3 mice per group) over a period of 28 d. Data are shown as mean ± SD. (C) H&E-stained sections show postexcision healing of ear holes of Rhbdf2^cub/cub Areg^Mcub/Mcub mice. The dotted line indicates the site of excision. (Magnification: 10×.) (Scale bars: 100 μm.) E, epidermis; M, proliferation. Compare with Fig. 1F. (D) Serum AREG levels in age-matched Rhbdf2^+/+ Areg^+/+, Rhbdf2^cub/cub Areg^+/+, and Rhbdf2^cub/cub Areg^Mcub/Mcub female mice. AREG was not detected (nd) in the serum of Rhbdf2^cub/cub Areg^Mcub/Mcub mice. ***P < 0.001. (E) ELISA quantitation of AREG levels in the supernatants of cultured mouse epidermal keratinocytes (MEKs) isolated from Rhbdf2^+/+ Areg^+/+, Rhbdf2^cub/cub Areg^+/+, and Rhbdf2^cub/cub Areg^Mcub/Mcub mice. AREG was not detected in MEKs from Rhbdf2^cub/cub Areg^Mcub/Mcub mice. ***P < 0.001. (F) qPCR of EGFR ligands: cDNA was extracted from skin tissues of Rhbdf2^cub/cub Areg^+/+ and Rhbdf2^+/+ Areg^+/+ mice using TaqMan gene expression assays. Actin served as an endogenous control. Data are normalized to Rhbdf2^+/+ Areg^+/+ actin levels. *P < 0.05. Data in D–F are shown as mean ± SD of three independent experiments.

Mcub is a T-to-G point mutation that destroys the canonical donor splice site of exon 1 and leads to the exclusive use of an alternative downstream splice site that adds 22 extra nucleotides to the Areg transcript; this addition of extra nucleotides disrupts the coding frame and introduces a premature stop codon (Fig. S3). This mutation will henceforth be referred to as Areg^Mcub. Notably, the hyperactive EGFR signaling (Fig. S4A) and the rapid wound closure capability of Rhbdf2^cub/cub mice are significantly reduced (Fig. 4 B and C), and the loss-of-hair phenotype is prevented when a single copy of the dominant Areg^Mcub allele is present (17). The dominant Areg^Mcub mutation does not confer a normal hair coat to the Rhbdf2^cub/cub mice but, rather, a wavy hair phenotype (17), suggesting remaining abnormalities in the EGFR pathway (28).

We next measured serum levels of AREG in Rhbdf2^cub/cub Areg^+/+, Rhbdf2^cub/cub Areg^Mcub/Mcub, and Rhbdf2^+/+ Areg^+/+ mice. We found no detectable AREG in the serum of Rhbdf2^cub/cub Areg^Mcub/Mcub mice, but we did observe a dramatic increase in serum AREG levels in Rhbdf2^cub/cub Areg^+/+ mice compared with Rhbdf2^+/+ Areg^+/+ mice (Fig. 4D). Measurements of supernatant AREG levels from cultured mouse epidermal keratinocytes of Rhbdf2^cub/cub Areg^+/+, Rhbdf2^cub/cub Areg^Mcub/Mcub, and Rhbdf2^+/+ Areg^+/+ mice yielded similar results (Fig. 4E). AREG is abundantly expressed in normal skin (29); therefore, we next performed qPCR on skin samples from Rhbdf2^cub/cub Areg^+/+ and Rhbdf2^+/+ Areg^+/+ mice to measure transcript levels of Areg, as well as six other genes known to encode EGFR ligands (Egf, Tgfα, Btc, Epgn, Ereg, and Hbegf). Compared with controls, we observed a fourfold increase in Areg, and subtle but statistically significant increases in Epgn and Hbegf mRNAs, in Rhbdf2^cub/cub Areg^+/+ mice. Also, there was a statistically significant decrease in Btc and Egf transcript levels (Fig. 4F).

Lastly, to examine whether AREG mediates the hyperactive EGFR phenotype, we silenced Areg expression in Rhbdf2^+/+ and Rhbdf2^cub/cub MEFs (Fig. S4B) using lentiviral shRNA and performed proliferation assays. Whereas silencing of Areg had a subtle effect on proliferation of Rhbdf2^+/+ MEFs (Fig. S4C), proliferation rates of Rhbdf2^cub/cub MEFs were significantly reduced (Fig. S4D). Taken together, these data suggest that enhanced AREG levels mediate the mutant phenotype of the Rhbdf2^cub/cub mice via a gain of EGFR signaling. Our findings are consistent with previous reports showing that exposure to high concentrations of AREG influences hair follicle development (26, 27) and tissue regeneration (30, 31). Furthermore, our data reveal that the Rhbdf2^cub phenotype is modified by Areg^Mcub expression.

N-terminal–Truncated iRhom2 Induces Substrate-Specific Secretion of EGFR Ligands Independent of ADAM17.

Nakagawa et al. (32) found that N-terminal–truncated but not full-length iRhom1 induces heparin-binding (HB) EGF secretion in Drosophila. Moreover, recent evidence suggests that active rhomboid proteases, with short or nonexistent N-terminal domains, cleave membrane-tethered EGF independent of metalloprotease activity (33). Thus, we asked whether Rhbdf2^cub, with its short N-terminal domain, could induce secretion of AREG independent of metalloprotease activity. We performed in vitro cleavage assays in the presence or absence of marimastat (MM), a potent broad-spectrum metalloprotease inhibitor that can block both ADAM17- and ADAM10-dependent shedding of substrates (34, 35). Because of the significant homology between the mouse and human RHBDF2 genes (3, 32, 33, 36), we transfected 293T cells with human AREG either alone or with the HuWt or HuCub RHBDF2 gene (Fig. 5A) and measured AREG levels in conditioned medium. In the absence of MM, we observed no difference in AREG levels between AREG-expressing and AREG/HuCub-coexpressing cells, whereas coexpression of HuWt and AREG reduced AREG levels by ∼60% (Fig. 5B). Intriguingly, in the presence of MM, AREG levels were approximately twofold higher in HuCub/AREG-cotransfected cells compared with cells transfected with HuWt and AREG or AREG alone (Fig. 5B). These data suggest that N-terminal–truncated iRhom2 can enhance AREG secretion, which only becomes apparent when metalloprotease activity is diminished.

Fig. 5.

Rhbdf2^cub regulates substrate-specific production of EGFR ligands. (A) Schematic representation of the HuWt gene and the HuCub gene. (B) ELISA quantitation of cleaved AREG after coexpression of HuWt or HuCub with the AREG gene in 293T cells. At 24 h posttransfection, cells were incubated with either DMSO or 10 μM MM for 24 h and AREG levels were analyzed in the conditioned medium. Transfections were performed in duplicate, and the conditioned medium was diluted fivefold. (C) Quantitation of serum TNF-α levels by ELISA, 3 h after LPS injection of 8- to 12-wk-old female mice of the indicated genotypes. TNF-α was not detected in the serum of Rhbdf2^+/+ mice with no LPS injection. *P < 0.05. (D) ELISA quantitation of cleaved AREG/HB-EGF/EGF after coexpression of the human RHBDL2 (HuRHBDL2) gene with the AREG, HB-EGF, or EGF gene. At 24 h posttransfection, cells were incubated with 10 μM MM for 24 h and AREG/HB-EGF/EGF levels were analyzed in the conditioned medium. Whereas HB-EGF and EGF conditioned media were undiluted, AREG medium was diluted fivefold. Data represent mean ± SD of three independent experiments. ***P < 0.001. (E) ELISA quantitation of cleaved HB-EGF after coexpression of HuWt or HuCub with the HB-EGF gene in 293T cells. At 24 h posttransfection, cells were incubated with 10 μM MM for 24 h and HB-EGF levels were analyzed in the conditioned medium. ***P < 0.001. (F) ELISA quantitation of cleaved EGF after coexpression of HuWt or HuCub with the EGF gene in 293T cells. At 24 h posttransfection, cells were incubated with either DMSO or 10 μM MM for 24 h and EGF levels were analyzed in the conditioned medium. **P < 0.01. (G) ELISA quantitation of cleaved AREG after coexpression of HuCub or HuCub without the peptidase domain (HuCubΔPD) with the AREG gene in 293T cells. At 24 h posttransfection, cells were incubated with 10 μM MM for 24 h and AREG levels were analyzed in the conditioned medium. ***P < 0.001. (H) ELISA quantitation of cleaved AREG after coexpression of HuCub or HuCub with individual alanine mutations with the AREG gene in 293T cells. At 24 h posttransfection, cells were incubated with 10 μM MM for 24 h and AREG levels were analyzed in the conditioned medium. ***P < 0.001. (I) ELISA quantitation of cleaved EGF after coexpression of HuWt, HuCub, or HuCub with individual alanine mutations with the EGF gene in 293T cells. At 24 h posttransfection, cells were incubated with 10 μM MM for 24 h and EGF levels were analyzed in the conditioned medium. **P < 0.001. (J) Membrane topology of Rhbdf2^cub. The amino acids shown are critical for regulation of EGFR ligand production by Rhbdf2^cub.

iRhom2 has been shown to regulate maturation of ADAM17 and, in turn, ADAM17-dependent shedding (7, 8, 37). Because Rhbdf2^cub is a gain-of-function mutation rather than a null mutation, there is a possibility that Rhbdf2^cub could enhance ADAM17 activity, and thereby increase AREG secretion. ADAM17 activity can be measured in mice by examining TNF-α secretion after stimulation with bacterial endotoxin LPS (7, 8). We therefore used this approach to test changes in ADAM17 activity in Rhbdf2^cub mice by injecting LPS into Rhbdf2^cub/cub, Rhbdf2^−/−, and Rhbdf2^+/+ mice and measuring serum TNF-α levels. We found that both LPS-injected Rhbdf2^cub/cub and Rhbdf2^−/− mice had a markedly lower induction of TNF-α relative to LPS-injected Rhbdf2^+/+ mice (Fig. 5C). These results suggest that ADAM17 activity is significantly attenuated in Rhbdf2^cub/cub and Rhbdf2^−/−, and they also implicate a role for the N-terminal domain of iRhom2 in regulation of ADAM17-dependent TNF-α release. Notably, serum TNF-α levels in LPS-stimulated Rhbdf2^cub/cub mice were not completely abrogated, suggesting that ADAM17 activity is attenuated but not eliminated (38). The observation that ADAM17 activity is attenuated in Rhbdf2^cub/cub mice explains the nonlethal or noninflammatory cutaneous phenotype in Rhbdf2^−/− and Rhbdf2^cub/cub mice compared with Adam17^−/− mice (39, 40) or with transgenic mice overexpressing AREG (41).

We next sought to assess the protease selectivity of the active rhomboid proteases for other EGF-like substrates, and whether these differed from HuCub. In accordance with a previous study (33), we found that RHBDL2 selectively increased secretion of EGF but not AREG or HB-EGF compared with empty vector-transfected controls (Fig. 5D). In contrast, HuCub selectively increased secretion of AREG and HB-EGF but not EGF compared with empty vector-transfected cells (Fig. 5 E–G). These results suggest that the rhomboid peptidase domain confers substrate selectivity. The ability of N-terminal–truncated iRhom1 to induce AREG but not EGF secretion is consistent with this assertion (Fig. S5 A and B).

To determine how the iRhom2 peptidase domain helps to regulate the secretion of EGF-like substrates, we initially aligned sequences for amino acids of the peptidase domains of human and mouse iRhoms. We found a significant difference in protein sequence homology between the peptidase domains of human iRhom2 and RHBDL2 (Fig. S5C). By contrast, the peptidase domains of iRhom1 and iRhom2 show 96% homology. Together, these data suggest that the residues that form the active site or play a role in substrate recognition differ significantly between active rhomboids and iRhoms. In addition, we found that deletion of the HuCub peptidase domain significantly diminished AREG secretion compared with native HuCub (Fig. 5G), suggesting that the peptidase domain is essential for secretion of EGF-like substrates. To identify the critical residues, we performed site-directed mutagenesis such that key residues in the HuCub peptidase domain were mutated to alanines. We found that glutamine-426 and cysteine-C432 in transmembrane domain 4, and histidine-366 and histidine-475 in transmembrane domains 2 and 6, respectively, were critical not only for mediating enhanced secretion of AREG/HB-EGF but also for suppression of EGF (Fig. 5 H and I). However, mutations in serines S362A, 402A, and 425A; glutamic acid E436A; and glutamine Q439A did not alter AREG secretion (Fig. S5D). Together, these results suggest that key residues (Fig. 5J and Fig. S5E) in the peptidase domain of N-terminal–truncated iRhom regulate the secretion of EGF-like substrates independent of metalloprotease activity.

N-Terminal–Truncated iRhom2 Increases Susceptibility to Epithelial Cancers.

Two recent studies indicated that missense mutations in RHBDF2 (p.I186T, p.P189L, and p.D188N) underlie a familial tylosis with esophageal cancer syndrome in families in the United States, United Kingdom, Germany, and Finland (11, 42). Based on our results on Rhbdf2^cub mutant mice, we predicted that increased AREG secretion due to dominant N-terminal mutations in RHBDF2 might drive these human pathological changes. We first tested this hypothesis in vitro by coexpressing AREG with a HuWt clone containing the human missense mutation (RHBDF2 p.I186T) (11), HuCub, or HuWt in 293T cells. As expected, expression of the RHBDF2 p.I186T or HuCub resulted in greater levels of AREG in conditioned medium and lower intracellular levels compared with HuWt (Fig. 6 A and B). Further, RHBDF2 p.I186T produced AREG levels comparable to those produced by HuCub, suggesting that loss of, or dominant mutations in, the iRhom2 N terminus lead to increased AREG secretion. Additionally, we found that loss of at least one of four critical residues (H, C, Q, and H) in the peptidase domain of the RHBDF2 p.I186T mutant resulted in significantly decreased AREG secretion (Fig. 6C).

Fig. 6.

Heterozygosity of Rhbdf2^cub results in increased adenoma formation and decreased survival in Apc^Min/+ mice. (A) Quantitation of cleaved AREG by ELISA from conditioned medium of HEK 293 cells transfected with AREG and the human RHBDF2 p.I186T mutant, HuWt, or HuCub in the presence of 10 μM MM. **P < 0.01. (B) Western blot of HEK 293 cells cotransfected with AREG and HuCub, HuWt, or p.I186T; incubated with DMSO or MM; and immunoblotted for HA-AREG. Coexpression of AREG and either HuCub or p.I186T significantly reduces the intracellular levels of pro-AREG compared with those of AREG/HuWt or AREG alone, even in the presence of 10 μM MM. (C) AREG levels in cells expressing various RHBDF2 P.I186T point mutants and cotransfected with the AREG gene or empty vector. Each residue was mutated to alanine. Cells were incubated with 10 μM MM for 24 h and assayed for AREG levels in conditioned medium. (D) X-gal–stained 2-wk-old female Rhbdf2^−/− intestine reveals reporter gene β-gal expression in the middle and upper villous regions. (Magnification: 2.5×.) (Scale bar: 1mm.) (E) Kaplan–Meier survival curves of Apc^Min/+ Rhbdf2^+/+ (n = 28) and Apc^Min/+ Rhbdf2^+/cub (n = 23) mice. The median survival of Apc^Min/+ Rhbdf2^+/+ mice was 172 d, compared with 135 d for Apc^Min/+ Rhbdf2^+/cub mice. (F) H&E-stained sections of intestinal tissue from mice of the indicated genotypes at 3 mo of age. (Magnification: 2.5×.) (Scale bars: 1mm.) Number (G) and size (H) of polyps per mouse of the indicated genotypes of mice at 3 mo of age are shown. The mean number of polyps in Apc^Min/+ Rhbdf2^+/+ mice was 12, whereas in Apc^Min/+ Rhbdf2^+/cub mice, the mean number of polyps was 21. The mean tumor sizes were 1.8 mm² and 3.7 mm² for Apc^Min/+ Rhbdf2^+/+ and Apc^Min/+ Rhbdf2^+/cub mice, respectively.

To determine whether the Rhbdf2^cub allele increases tumor susceptibility, we investigated how its expression would affect adenoma formation in Apc^Min/+ mice, a mouse model of human familial adenomatous polyposis. In Apc^Min/+ mice, spontaneous loss of one WT Apc allele induces intestinal epithelial adenoma formation and premature death at a median age of 169 d (43). Notably, Rhbdf2 (Fig. 6D) and Areg (31) expression is observed in the small intestine, suggesting a potential functional relationship. We generated and observed Apc^Min/+ Rhbdf2^+/cub and Apc^Min/+ Rhbdf2^+/+ mice, but because of increased lethality, we could not generate enough Apc^Min/+ Rhbdf2^cub/cub mice for meaningful comparison. Nonetheless, we did observe a significant difference in the median survival age of Apc^Min/+ Rhbdf2^+/+ (172 d) and Apc^Min/+ Rhbdf2^+/cub (135 d) mice (Fig. 6E). Necropsy of Apc^Min/+ Rhbdf2^+/+ and Apc^Min/+ Rhbdf2^+/cub mice at 3 mo of age revealed that the presence of a single Rhbdf2^cub allele significantly increased the number of polyps (Fig. 6 F and G) and adenoma size (Fig. 6 F and H) in Apc^Min/+ mice, suggesting that the cub mutation increases the growth of epithelial tumors. However, there was no spontaneous incidence of cancer in Rhbdf2^cub/cub mice aged up to 2 y, suggesting that Rhbdf2^cub mutation creates a conducive environment for, but alone does not drive, tumor development.

Loss of the Cytosolic N Terminus or Dominant Mutations in the N Terminus of the RHBDF2 Gene Increase Its Protein Stability.

iRhom2 negatively regulates EGFR signaling by promoting degradation of EGF-like ligands through the proteasomal pathway (3). Moreover, we observed that iRhoms induce secretion of AREG/HB-EGF when the cytosolic N terminus is lacking (Fig. 5). Thus, we asked whether gain-of-function mutations in the amino terminus of iRhom2 interfere with proteasomal processing, and thereby increase its stability. We initially examined whether the tylotic RHBDF2 mutant p.I186T has an ability to interact with AREG. We found that similar to HuWt and HuCub, p.I186T forms physical complexes with AREG (Fig. 7A). We then compared the protein expression levels of HuWt, HuCub, and p.I186T in 293T and COS7 cells by immunocytochemistry and flow cytometry. We observed that HuWt protein expression was significantly lower compared with both HuCub and p.I186T expression (Fig. 7 B and C). Further, when we subjected COS7 cells to a cycloheximide (a protein synthesis inhibitor) chase for the indicated times, within 1 h, we observed an ∼50% reduction in immunoreactivity for HuWt compared with either HuCub or the p.I186T (Fig. 7 D and E). Endoplasmic reticulum-localized rhomboid proteases interact with the ubiquitin proteasome system to promote ERAD (2, 3, 36). Because iRhom2 participates in ERAD, it is possible that it could be a target of proteasomal degradation. We determined the protein t_1/2 of HuWt and HuCub in the presence of a potent proteasomal inhibitor, MG-132. Expectedly, the protein t_1/2 of HuWt, but not HuCub, was significantly increased (Fig. 7F), suggesting that the proteasomal degradation of iRhom2 might be affecting its protein stability. We conclude that missense mutations in the amino terminus of iRhom2, similar to the Rhbdf2^cub mutation, increase its stability and contribute to enhanced AREG secretion independent of metalloprotease activity.

Fig. 7.

Mammalian iRHOM2 is a short-lived protease. (A) Coimmunoprecipitation (IP) of human iRhom2–AREG complex. Lysates from COS7 cells cotransfected with Flag-tagged HuWt or Flag-tagged HuCub or with Flag-tagged p.I186T and HA-tagged AREG were immunoprecipitated with anti-Flag magnetic beads and probed with anti-Flag and anti-HA antibodies. (B) Flow cytometry results from 293T or COS7 (Fig. S6) cells transfected with the indicated Flag (FL)-tagged genes and immunolabeled using a Flag-specific PE-labeled antibody. Cells were immunolabeled 48 h posttransfection. ddk, also known as Flag. (C) Quantification of the data obtained in B. (D) Transiently transfected COS7 cells were subjected to a chase with 150 μg/mL cycloheximide (the protein synthesis inhibitor) for the indicated times and evaluated for protein expression using Flag-specific PE-labeled antibody. Data are representative of one of a total of three experiments. (E) Quantification of the data obtained in D. (F) Transiently transfected COS7 cells were preincubated for 4 h with 10 μM MG-132, a cell-permeable protease inhibitor, followed by a chase with 150 μg/mL cycloheximide in the presence of MG-132 for the indicated times. Protein expression was determined as described in D. ***P < 0.001.

Discussion

The EGFR signal transduction pathway plays an essential role in growth, proliferation, and differentiation of mammalian cells. Canonical EGFR ligands, including EGF, AREG, and HB-EGF, exist as proproteins expressed at the cell surface that, to bind EGFRs, must be shed into the extracellular compartment. Different classes of proteases cleave membrane-tethered EGFR proligands to regulate a broad range of biological activities during various stages of development. Here, we report that iRhom2, a member of a family of rhomboid proteases well known as regulators of EGFR signaling in Drosophila, has an ability to regulate EGFR signaling during cutaneous healing and tumor development. We show that iRhom2 is a short-lived protein whose stability can be increased by select mutations in the N-terminal domain. In turn, these stable variants function to enhance AREG secretion independent of metalloprotease activity. We identify an important role for iRhoms in EGFR-dependent cell proliferation and wound healing, and show how iRhom2 mutations that increase EGFR signaling, under the right circumstances, can promote cancer development.

N-Terminal and Peptidase Domains Have Separate Functions in Regulating EGFR Signaling.

iRhoms are complex multidomain enzymes that contain a long cytosolic N terminus, a dormant peptidase domain, and a conserved IRHD; the function of these domains remains unknown. Under normal circumstances, iRhoms negatively regulate EGFR signaling by promoting the degradation of EGF-like substrates (3). However, the Rhbdf2^cub mutation is unlikely to be simply a loss-of-function mutation. Rhbdf2^−/− mice failed to recapitulate the Rhbdf2^cub phenotype. In addition, we demonstrate that similar to the Rhbdf2^cub mutation, dominant missense mutations in the N terminus of iRhom2 induce secretion of AREG and HB-EGF in a manner mediated by key amino acids in transmembrane helices 2, 4, and 6 of the peptidase domain. Thus, our results suggest that the cytosolic N terminus of iRhom2 negatively regulates EGFR signaling by suppressing the peptidase domain and, consequently, secretion of AREG/HB-EGF (Fig. 8). These findings are consistent with a recent study suggesting that in mammalian cells, whereas the cytosolic domain of Derlin-1, a novel rhomboid pseudoprotease, is essential for clearance of misfolded proteins from the endoplasmic reticulum, the transmembrane domain is required to interact with its substrates (2). Our findings also reveal more subtle regulatory functions for iRhom2 that are unmasked in N-terminal mutations, such as Rhbdf2^cub.

Fig. 8.

Working model for our study shows that WT iRhom2 is a short-lived protein, whereas loss of its N terminus or mutations in its N-terminal domain, including those that underlie epithelial cancers, increase its protein stability. Increased protein stability, in turn, augments the secretion of selective EGF family ligands, including AREG. Inhibition of ADAM17 has no effect on AREG secretion, whereas loss of amino acids H366, Q426, C432, and H475 in the peptidase domain of iRhom2 abrogates AREG secretion. Enhanced secretion of AREG leads to hyperactivation of EGFR signaling, and thereby increased cell proliferation and migration. ER, endoplasmic reticulum; ERAD, endoplasmic reticulum associated degradation.

Instead, the Rhbdf2^cub may be considered a gain-of-function mutation. This conclusion is supported by several pieces of evidence. First, the negative regulatory role of iRhom2 seems to be minimal because Rhbdf2^−/− mice do not present an overt “EGFR hyperactive” phenotype except when combined with the Rhbdf2^cub mutation. Second, cotransfection of HuCub and AREG results in approximately two- to threefold greater levels of AREG compared with transfections of either HuWt and AREG or AREG alone. Third, expression of HuCub induced secretion of membrane-anchored AREG and HB-EGF independent of metalloprotease activity. Fourth, mutant iRhom2 alleles fail to induce secretion of AREG/HB-EGF in the absence of the peptidase domain. Consistent with these observations, transgenic expression of the N-terminal–truncated but not full-length RHBDF1 induces a strong EGFR signaling-related wing phenotype in Drosophila. Additionally, coexpression of truncated RHBDF1 with HB-EGF intensifies the altered wing phenotype in Drosophila, indicating that the truncated iRhom1 might induce secretion of HB-EGF (32), and thereby activate EGFR signaling. These results validate the concept that the cytosolic N terminus contributes to iRhom-elicited ubiquitin processing of EGF family ligands, whereas the iRhom peptidase domain stimulates EGFR signaling when not suppressed by the N terminus.

It has been hypothesized that mutant iRhom2 alleles can enhance maturation of ADAM17 (6), leading to constitutive activation of EGFR signaling. However, we found that mutant iRhom2 alleles induce secretion of AREG and HB-EGF in the presence of saturating concentrations of MM, a potent broad-spectrum metalloprotease inhibitor. Maturation of other members of the ADAM family is unaffected by the deficiency of both iRhoms (9). Moreover, we demonstrate that ADAM17 activity is attenuated in Rhbdf2^cub mice. Thus, we conclude that Rhbdf2^cub and tylotic mutations selectively induce secretion of AREG and HB-EGF independent of metalloprotease activity, and that the transmembrane peptidase domain is necessary for this function.

Hyperactive EGFR Pathway Underlies Accelerated Cutaneous Healing in Rhbdf2^cub Mice.

The biology of wound healing is complex. In human adults, wounds are vulnerable to nonfunctional fibrotic tissue formation, whereas wounds occurring during the prenatal period accomplish complete regeneration, resembling scar-free healing in vertebrates, such as axolotls and planarians. Even though the highly orchestrated and rapid events that occur following injury are well documented, the tissue regeneration and remodeling field is still in its infancy. The literature suggests that a strong link exists between rhomboid proteases and cutaneous wound healing. Active rhomboid RHBDL2 has been demonstrated to induce cell proliferation and migration by specifically cleaving EGF, thrombomodulin, and EphrinB3 (33, 44, 45). In vitro wound-healing assays performed in a human keratinocyte cell line indicate that the expression of RHBDL2 is significantly up-regulated after wounding compared with unwounded controls. Also, increased RHBLD2 expression correlates with increased shedding of membrane-bound thrombomodulin (46). Although these results suggest that targeting rhomboid proteases might have a therapeutic benefit in impaired cutaneous healing, the physiological substrates of rhomboid proteases are still unclear.

In several mouse models of wound healing, including the “superhealing” Murphy–Roths–Large (MRL/MpJ) strain, studies have shown that excessive inflammation delays healing, whereas rapid reepithelialization and reduced inflammation lead to accelerated and scar-free healing (15, 47–51). The ability of Rhbdf2^cub mice to heal wounds rapidly without significant scar formation might be due to a combination of decreased TNF-α secretion due to attenuated ADAM17 activity and rapid reepithelialization induced by augmented AREG production/EGFR hyperactivation. We propose that although decreased TNF-α contributes to a lesser degree of inflammation, increased AREG production facilitates accelerated proliferation and migration of keratinocytes to the wound site in Rhbdf2^cub mice. Moreover, iRhom2 is predominantly expressed in the skin, making it a potential therapeutic target in impaired cutaneous wound healing.

iRhom2-AREG-EGFR Pathway Is Constitutively Active in Some Epithelial Cancers.

Several studies have implicated iRhom mutations in cancer. For example, overexpression of the RHBDF1 gene, which encodes iRhom1, is crucial in sustaining growth signaling in epithelial cancer cells (4, 5). Significant elevation of RHBDF1 transcript levels is observed in breast cancer clinical specimens. Although iRhom1 suppresses secretion of EGF family ligands and negatively regulates EGFR signaling (3), a recent study suggests that iRhom1 might have additional physiological roles (9). For instance, it has been implicated in G protein-coupled receptor transactivation of the EGFR signaling pathway (5). Moreover, unlike iRhom2 KO mice, iRhom1 KO mice have a more severe phenotype and survive for about 6 wk (9).

In humans, dominant mutations in the N terminus of the RHBDF2 gene are associated with hyperkeratosis and esophageal cancer (11, 42). Although the mechanisms underlying the pathogenesis of hyperkeratosis and cancer are unclear, analyses of skin biopsies of patients suggest constitutive EGFR activation. Specifically, EGFR levels were shown to be significantly lower in tylotic keratinocytes compared with control keratinocytes, implicating excessive or prolonged activation-induced down-regulation of EGFR (11). Interestingly, because the RHBDF2 mutations described to date arose in the N terminus, dominant mutations in the N terminus of iRhom2 could trigger esophageal cancer via an EGFR-dependent signaling event. In the present study, we find that N-terminal–truncated iRhom2 promotes increased AREG production independent of ADAM17 activity, and thereby induces EGFR activation. In particular, the phenotype of Apc^Min/+ Rhbdf2^+/cub mice recapitulates the increased susceptibility to epithelial cancers seen in patients with dominant RHBDF2 mutations (11, 42). However, Rhbdf2^cub/cub mice did not spontaneously develop tumors, suggesting that the Rhbdf2^cub mutation might not drive cancer development but, instead, might promote tumor growth and progression by creating a conducive environment.

Materials and Methods

Full details are available in SI Materials and Methods.

Mice were obtained, bred, and maintained under modified barrier conditions at The Jackson Laboratory. All genotypes, including cub and Mcub, were maintained on the C57BL/6J (B6) genetic background. To generate Rhbdf2^−/− mice, ES cell clones (EPD0208_1_A09) obtained from the KOMP repository were injected into B6-Tyr^c (B6 albino) blastocysts. Males displaying >50% chimerism were mated to B6 albino females; black offspring were genotyped by PCR. Heterozygotes were mated with each other to produce homozygotes, or with Rhbdf2^cub/cub mice to produce Rhbdf2^−/cub mice. The Animal Care and Use Committee at The Jackson Laboratory approved all of the experimental procedures.