Analysis of the function of ADAM17 in iRhom2 curly-bare and tylosis with esophageal cancer mutant mice

iRhom2 (or iR2, encoded by Rhbdf2) and the related iRhom1 (or iR1, encoded by Rhbdf1) are membrane-spanning inactive rhomboid-like proteins with seven transmembrane domains that regulate the maturation and function of the cell surface metalloprotease ADAM17 and its role in activating the TNFα (or TNF), IL-6R and EGF receptor pathways (Adrain et al., 2012; McIlwain et al., 2012; Siggs et al., 2012; Maretzky et al., 2013; Li et al., 2017, 2015; Christova et al., 2013). ADAM17 controls the release of the extracellular domain of substrates such as TNFα, IL-6R and several EGFR ligands [e.g. TGFα (or TGF), heparin-binding EGF (HB-EGF) and amphiregulin (AREG)] from their membrane anchor in a process known as ectodomain shedding (Moss et al., 1997; Black et al., 1997; Horiuchi et al., 2007; Blobel, 2005; Peschon et al., 1998; Jackson et al., 2003; Sternlicht et al., 2005). Adam17^−/− mice resemble mice that lack EGFR, in that they die perinatally with open eyes at birth (Blobel, 2005; Peschon et al., 1998; Jackson et al., 2003; Sternlicht et al., 2005). However, iR2^−/− mice, which lack mature and functional ADAM17 in myeloid cells but have mature ADAM17 supported by iR1 in most other cells and tissues (Li et al., 2015; Christova et al., 2013), have no spontaneous pathological phenotypes (Adrain et al., 2012; McIlwain et al., 2012; Issuree et al., 2013). iR2^−/− mice are protected from the deleterious effects of the soluble pro-inflammatory cytokine TNFα in models for septic shock and inflammatory arthritis, just like conditional knockout mice lacking ADAM17 in myeloid cells (McIlwain et al., 2012; Horiuchi et al., 2007; Issuree et al., 2013). iR2^−/− mice are also protected from lupus nephritis (Qing et al., 2018), hemophilia arthropathy (Haxaire et al., 2018) and inflammation of adipose tissue (Badenes et al., 2020; Skurski et al., 2020).

The cytoplasmic tail of iR2 is the site of several distinct gain-of-function point mutations in patients suffering from the familial Howel–Evans syndrome or tylosis with esophageal cancer (TOC) (Brooke et al., 2014). These patients have palmoplantar keratoderma, oral leukokeratosis and a predisposition for developing esophageal cancer. Knock-in mice carrying a specific mutation in iR2 (P159L), which is orthologous to the P189L TOC mutation in humans, exhibit mild to moderate alopecia with wavy fur (Hosur et al., 2017). This phenotype is similar, but less severe than that of mice carrying the curly-bare (cub) mutation in iR2, which leads to a curly coat or a hairless bare skin, depending on the presence of absence of a modifier allele Mcub (Johnson et al., 2003), later identified as the EGFR-ligand AREG (Hosur et al., 2014; Siggs et al., 2014). The cub mutation deletes the highly conserved N-terminal cytoplasmic domain of iR2, including the sequences corresponding to those mutated in TOC patients (Hosur et al., 2014; Siggs et al., 2014). This suggests that cub and toc mutant mice could help to improve our understanding of how the cytoplasmic domain of iR2 regulates its function and that of its essential binding partner ADAM17.

As iR2^−/− mice appear healthy and normal, but iR2^cub/cub mice with two wild-type (WT) alleles of Areg have bare skin and iR2^toc/toc mice exhibit alopecia with wavy fur, the iR2^cub and iR2^toc mutations are considered gain-of-function mutations (Siggs et al., 2014; Hosur et al., 2014). The iR2^cub and iR2^toc mutations both accelerate cutaneous wound healing and EGFR-dependent cell migration. An iR2-lacZ reporter highlighted high expression of iR2 in the hair follicles, which are the most strongly affected structures in the mutant animals (Hosur et al., 2014). iR2^cub/cub mice also have increased tumor formation in the Apc^Min mouse model for intestinal cancer, providing additional evidence for a gain-of-function phenotype (Hosur et al., 2014). Interestingly, the modifier of the cub mutation, Mcub (Areg) is an iR2-regulated-ADAM17-dependent EGFR ligand (Maretzky et al., 2013; Li et al., 2015; Siggs et al., 2014; Hosur et al., 2014). Whereas Areg^−/− mice have no evident phenotypic abnormalities, loss of one or both alleles of Areg was sufficient to largely rescue the iR2^cub/cub and iR2^toc/toc phenotypes, suggesting that both are caused by a gene-dosage-dependent gain of function that depends on the presence of AREG (Siggs et al., 2014; Hosur et al., 2014; Johnson et al., 2003).

Hosur et al. (2014) reported enhanced levels of AREG in the serum of iR2^cub/cub mice and presented data that were interpreted to suggest that iR2 and iR2^cub have inherent catalytic activity independent of ADAM17. In a second study, Siggs et al. (2014) reported that the iR2^cub mutation causes a loss of iR2-regulated-ADAM17-dependent shedding of AREG. Importantly, Siggs et al. (2014) presented compelling arguments against the notion that iR2^cub itself has catalytic activity. Specific amino acid residues required for the catalytic activity of the related rhomboid intramembrane proteinases are not present in iR2, and a proline residue in the sequences corresponding to the active site in rhomboids is thought to prevent catalytic activity in iR2 (Adrain and Freeman, 2012; Dulloo et al., 2019). Previous studies have shown that overexpression of iR1 or iR2 carrying an N-terminal deletion that is similar to the cub mutation activates ADAM17-dependent shedding (Maney et al., 2015). Increased ADAM17 activity was also reported in keratinocytes isolated from tylosis patients (Blaydon et al., 2012; Brooke et al., 2014). When iR2^cub/cub mice lacking ADAM17 in keratinocytes were generated, this largely rescued the skin phenotype, suggesting that the function of iR2^cub depends on ADAM17 (Hosur et al., 2018a), although this conclusion was later called back into question (Burzenski et al., 2021). Thus, no clear picture of how the endogenous iR2^cub or iR2^toc mutations affect the maturation and function of ADAM17 has emerged from these previous studies.

The goal of this study was to learn more about the functional consequences of the iR2^cub and iR2^toc mutations by analyzing the phenotypes of iR2^cub/cub and iR2^toc/toc mice that also lack iR1. Moreover, we characterized the maturation and sheddase activity of ADAM17 in cells carrying the iR2^cub and iR2^toc mutations in the absence of iR1. Finally, as iR2^cub was initially proposed to have catalytic activity and to support AREG secretion independently of ADAM17, we tested how the systemic inactivation of ADAM17 affected the iR2^cub and iR2^toc phenotypes in mice.

To understand how the iR2^cub mutation affects ADAM17, we performed a western blot on bone marrow-derived macrophages (BMDMs), in which the function of ADAM17 depends on iR2 (McIlwain et al., 2012; Adrain et al., 2012; Issuree et al., 2013). Mature ADAM17 could be cell-surface biotinylated on WT BMDMs, whereas no biotin-labeled mature ADAM17 was detectable on the surface of iR2^cub/cub or iR2^−/− BMDMs (Fig. 1A, top row) (Adrain et al., 2012; McIlwain et al., 2012). In addition, there was no detectable mature ADAM17 in a western blot of iR2^cub/cub BMDMs, just like in iR2^−/− BMDMs, even though mature ADAM17 was present in WT controls (Fig. 1A, top row). ERK1/2 (encoded by Mapk3 and Mapk1) served as a negative control for cell surface biotinylation, and as a loading control for the western blots (Fig. 1A, second row). The mature form of ADAM10 was used as a positive control for surface biotinylation in all three mouse embryonic fibroblast (mEF) lines, with western blots shown on the right (Fig. 1A, bottom row). Flow cytometric analysis of cell surface ADAM17 showed normal levels in WT BMDMs and comparably low levels in iR2^cub/cub and iR2^−/− BMDMs (Fig. 1B). Moreover, there was no detectable lipopolysaccharide (LPS)-stimulated TNFα release from iR2^cub/cub BMDMs, as reported previously (Hosur et al., 2014), similar to iR2^−/− BMDMs (Fig. 1C) (Adrain et al., 2012; McIlwain et al., 2012). When iR2^cub/cub BMDMs were stimulated with phorbol 12-myristate 13-acetate (PMA), which activates ADAM17, the iR2-regulated-ADAM17 substrate macrophage colony-stimulating factor 1 receptor (MCSFR or CSF1R) was not downregulated compared to untreated controls, similar to iR2^−/− BMDMs (Fig. 1D,E) (Qing et al., 2016). Under these conditions, PMA stimulation led to strong downregulation of MCSFR on WT BMDMs (Fig. 1D,E). Taken together, these results demonstrate that iR2^cub/cub BMDMs lack detectable mature and functional ADAM17.

In order to analyze the function of the iR2^cub mutant in mouse development in the absence of the related iR1, we crossed iR2^cub mice with a strain of iR1^−/− mice that have no evident spontaneous pathological phenotypes but lack mature and functional ADAM17 when iR2 is also inactivated (Li et al., 2015). Therefore, iR1^−/− iR2^−/− double-knockout mice, generated with this strain of iR1^−/− mice, resemble Adam17^−/− mice in that they die shortly after birth with open eyes, heart valve defects and enlarged long bone growth plates owing to loss of ADAM17 activity (Li et al., 2015). We found that iR1^−/− iR2^cub/cub double-mutant mice survived into adulthood and resembled iR2^cub/cub mice in terms of their bare skin (Fig. 2A,B). Western blot analyses of ADAM17 in tissue extracts from iR1^−/− iR2^cub/cub mice showed unprocessed pro-ADAM17, but no detectable processed mature ADAM17 in the tissues analyzed here (heart, lung, spleen and kidney), although small amounts of a non-specific band that migrated between the position of mature and pro-ADAM17 were present in the spleen samples [Fig. 2C, WT, iR2^−/− and iR2^cub/cub samples are shown for comparison; see Fig. S1 for a quantitative PCR (qPCR) analysis of Adam17 and iR2 in heart tissue from each genotype].

As Adam17^−/− mice and iR1^−/− iR2^−/− mice, generated with the iR1^−/− strain used here, die shortly after birth (Peschon et al., 1998; Horiuchi et al., 2007; Li et al., 2015), we performed a histopathological analysis on newborn iR2^cub/cub and iR1^−/− iR2^cub/cub mice (Fig. S2). iR2^cub/cub mice had normally closed eyes at birth and normal growth plates and heart valves. One out of three newborn iR1^−/− iR2^cub/cub pups had one open eye at birth and all three had enlarged aortic and pulmonic valves and enlarged long bone growth plates, similar to Adam17^−/− and iR1^−/− iR2^−/− mice (Fig. S2). These developmental defects are consistent with at least a partial loss of ADAM17-dependent paracrine functions of proteolytically released EGFR-ligands in iR1^−/− iR2^cub/cub mutant mice (Peschon et al., 1998; Hall et al., 2013; Saito et al., 2013; Usmani et al., 2012; Jackson et al., 2003; Mine et al., 2005). However, the postnatal viability of the iR1^−/− iR2^cub/cub mice supported the conclusion that iR2^cub is a gain-of-function mutation (Siggs et al., 2014; Hosur et al., 2014).

To further explore how iR2^cub affects the function of ADAM17, we isolated mEFs from iR1^−/− iR2^cub/cub mice for substrate-shedding assays. We found that the PMA-stimulated shedding of the iR1- and iR2-regulated-ADAM17-dependent substrate TGFα and of the iR2-regulated-ADAM17-selective substrates AREG and HB-EGF was comparable in WT and iR1^−/− mEFs, but was abrogated in iR1^−/− iR2^cub/cub mEFs, as in iR1^−/− iR2^−/− mEFs (Fig. 2D) (Maretzky et al., 2013; Li et al., 2015; Sanderson et al., 2005; Sahin et al., 2004). The constitutive shedding of these substrates was also strongly reduced in iR1^−/− iR2^cub/cub mEFs, just as in iR1^−/− and iR1^−/− iR2^−/− mEFs, consistent with previous reports that iR2 has little, if any constitutive activity (Zhao et al., 2022). However, shedding of the ADAM10-substrate betacellulin (BTC) could be stimulated by ionomycin in all four mEF lines tested here (Fig. 2D). The characterization of iR1^−/− iR2^cub/cub in mEFs thus corroborated that iR2^cub does not support ADAM17 activity, at least towards the ADAM17 substrates tested here.

Previous studies have shown that the cytoplasmic binding partner of iR2, FRMD8, prevents the degradation of mature ADAM17 in the lysosome (Kunzel et al., 2018; Oikonomidi et al., 2018). Therefore, there was no detectable mature ADAM17 in iR1^−/− iR2^−/− mEFs expressing a mutant form of iR2 lacking the FRMD8-binding site (Kunzel et al., 2018). However, degradation of mature ADAM17 under these conditions could be prevented by the lysosomal degradation inhibitor bafilomycin (Kunzel et al., 2018; Oikonomidi et al., 2018). When we added bafilomycin to WT, iR1^−/− or iR1^−/− iR2^−/− mEFs, it had no detectable effect on the levels of mature ADAM17 compared to those in untreated controls (Fig. S3A). However, after incubation of iR1^−/− iR2^cub/cub mEFs with bafilomycin, a mature form of ADAM17 was detectable that was not present in untreated controls (Fig. S3A). Treatment with bafilomycin was not able to rescue PMA-stimulated TGFα shedding from iR1^−/− iR2^cub/cub mEFs, just like in iR1^−/− iR2^−/− mEFs, with bafilomycin-treated iR1^−/− mEFs serving as a positive control (Fig. S3B). Thus, the presence of mature ADAM17 in bafilomycin-treated iR1^−/− iR2^cub/cub mEFs as detected by western blotting was not sufficient to rescue its function as a TGFα sheddase.

A previous study suggested that the iR2^cub mutant can support AREG shedding independently of ADAM17 (Hosur et al., 2014). To test this notion experimentally, we generated iR2^cub/cub Adam17^+/− and iR2^cub/cub Adam17^−/− mice. The loss of one allele of ADAM17 in iR2^cub/cub Adam17^+/− mice restored the normal appearance of the fur and whiskers (Fig. 3A; Fig. S4A). Crosses of iR2^cub/cub Adam17^+/− mice yielded offspring with a Mendelian distribution of the targeted Adam17 allele at birth (Fig. 3B). The iR2^cub/cub Adam17^−/− mice died shortly after birth and had open eyes, enlarged aortic and pulmonic valves and enlarged long bone growth plates, similar to Adam17^−/− mice (Fig. S4B). The newborn iR2^cub/cub pups with both WT alleles of Adam17 had no abnormal histopathological phenotypes at birth (Fig. S4B) but displayed the iR2^cub/cub phenotypes as adults (Fig. 3A). Western blots of organ lysates from newborn mice showed mature ADAM17 in the iR2^cub/cub Adam17^+/+ and iR2^cub/cub Adam17^+/− samples, but no detectable ADAM17 in organs of newborn iR2^cub/cub Adam17^−/− mice (Fig. 3C).

Substrate-shedding assays in mEFs showed slightly reduced PMA-stimulated shedding of TGFα from iR2^cub/cub Adam17^+/+ and iR2^cub/cub Adam17^+/− mEFs compared to that in WT controls and only background constitutive and PMA-stimulated TGFα shedding in iR2^cub/cub Adam17^−/− mEFs (Fig. 3D). Moreover, only background constitutive and PMA-stimulated shedding of the iR2-regulated-ADAM17-selective substrates AREG and HB-EGF was seen in iR2^cub/cubAdam17^−/− mEFs, whereas there was some constitutive and PMA-stimulated AREG and HB-EGF shedding in iR2^cub/cub Adam17^+/+ and iR2^cub/cub Adam17^+/− mEFs (Fig. 3D). Moreover, ionomycin stimulated the shedding of the ADAM10 substrate BTC in all four mEF lines tested here (Fig. 3D). Finally, shedding experiments performed in the presence or absence of the metalloprotease inhibitor marimastat reduced the release of AREG to a similar degree in mEFs from WT, iR2^−/−, iR2^cub/cub, iR1^−/− or iR1^−/− iR2^cub/cub mice (Fig. S5), arguing against a functionally significant increase in marimastat-insensitive shedding of AREG by iR2^cub. Taken together, the shedding of AREG and the other ADAM17 substrates tested here was not supported by iR2^cub independently of ADAM17, contrary to what was previously proposed (Hosur et al., 2014).

Mice carrying the P159L iR2^toc knock-in mutation in the cytoplasmic domain of iR2 have wavy and sparse fur, and thus a similar, albeit less severe, phenotype to that of the iR2^cub/cub mutant mice (Hosur et al., 2017). Similar to the iR2^cub/cub mice, the skin and hair phenotype of the iR2^toc/toc mice can be rescued by inactivation of both alleles of Areg (Hosur et al., 2017). However, mature ADAM17 was labeled on iR2^toc/toc BMDMs by cell surface biotinylation and its levels were comparable to those in WT BMDMs, with iR2^−/− BMDMs serving as a negative control (Fig. 4A, top row). Moreover, western blots of iR2^toc/toc BMDMs confirmed the presence of mature ADAM17, similar to in WT BMDMs and unlike BMDMs from iR2^−/− (Fig. 4A, top row) or iR2^cub/cub (Fig. 1A) mice. There was no labeling of the cytoplasmic proteins ERK1/2 under these conditions in the three cell types shown (Fig. 4A, second row), with ERK1/2 serving as a loading control for the western blot analysis (Fig. 4A, second row). The mature form of the cell surface metalloprotease ADAM10 could be biotinylated in all samples shown and was also detected by western blotting (Fig. 4A, bottom row). Flow analysis with anti-ADAM17 antibodies corroborated that the levels of ADAM17 on iR2^toc/toc BMDMs were comparable to those on WT BMDMs (Fig. 4B). The LPS-stimulated release of TNFα from iR2^toc/toc BMDMs was somewhat reduced compared to that from WT BMDMs, but higher than from iR2^−/− BMDMs (Fig. 4C). Finally, the levels of the iR2-regulated-ADAM17 substrate MCSFR were substantially lower in iR2^toc/toc BMDMs than in WT BMDMs under unstimulated conditions and after PMA stimulation (Fig. 4D,E, iR2^−/− BMDMs included as a negative control). Thus, the iR2^toc mutant supported the shedding of TNFα and increased constitutive and stimulated downregulation of MCSFR by mature ADAM17 in BMDMs.

To determine how inactivation of iR1 affects the survival of iR2^toc/toc mice, we mated iR1^−/− iR2^toc/+ parents, which produced offspring at the expected Mendelian ratio that survived into adulthood (Fig. 5A,B). The skin phenotype of iR1^−/− iR2^toc/toc mice appeared more severe than in iR2^toc/toc mice (Fig. S6). Moreover, iR1^−/− iR2^toc/toc mice had similar levels of mature ADAM17 in western blots of heart, lung and spleen samples compared to those in WT controls, but lower levels in the kidney, presumably caused by the lack of iR1 in this tissue (Fig. 5C) (see also Li et al., 2015). A qPCR analysis of mRNA expression of Adam17 and iR2 showed no significant difference in heart tissue from iR2^toc/toc or iR1^−/− iR2^toc/toc mice compared to WT mice (Fig. S1). In shedding experiments, we found that PMA stimulated the release of the iR1- and iR2-regulated-ADAM17 substrates TGFα, ICAM-1 (hereafter ICAM) and CD62L (L-selectin, encoded by Sell) in iR1^−/− iR2^toc/toc mEFs, although at somewhat reduced levels compared to in WT or iR1^−/− controls, whereas no stimulation was seen in iR1^−/− iR2^−/− mEFs (Fig. 6A, top panels) (Li et al., 2015). However, there was no significant increase in PMA-stimulated shedding of the iR2-selective substrates AREG, HB-EGF and EREG in iR1^−/− iR2^toc/toc mEFs, similar to in iR1^−/− iR2^−/− mEFs (Fig. 6A, lower panels) (Li et al., 2015), although the shedding of these substrates was significantly increased in WT and iR1^−/− mEFs following PMA stimulation. These results demonstrated that the iR2^toc mutation has distinct effects on different ADAM17 substrates, with no significant PMA-stimulated shedding of the iR2-selective substrates AREG, HB-EGF and EREG, somewhat reduced PMA-stimulated shedding of TGFα, ICAM, CD62L and TNFα, and increased MCSFR shedding, compared to those in WT controls.

A histopathological analysis of newborn iR1^−/− iR2^toc/toc mice showed that all three pups had normal growth plates and closed eyelids, which depend on processing of TGFα by ADAM17 (Peschon et al., 1998; Hall et al., 2013; Saito et al., 2013), just like iR1^−/− mice (Fig. 6B) (Li et al., 2015). However, two out of three iR1^−/− iR2^toc/toc pups had thickened aortic and pulmonic heart valves (Fig. 6B), which is a well-established developmental defect caused by lack of HB-EGF processing by ADAM17 (Jackson et al., 2003) and was not found in iR1^−/− mice (Fig. 6B) (see also Li et al., 2015).

To assess the role of ADAM17 in maintaining the skin phenotype in iR2^toc/toc mice, we generated iR2^toc/toc Adam17^+/− mice. The trunk skin and hair of these animals were phenotypically normal compared to those of their iR2^toc/toc Adam17^+/+ littermates (Fig. 7A; histopathological analysis shown in Fig. S7). However, all homozygous double-mutant iR2^toc/toc Adam17^−/− mice died perinatally with open eyes, thickened heart valves and enlarged growth plates, just like Adam17^−/− and iR2^cub/cub Adam17^−/− pups (Fig. 7B; Fig. S7B).

The main goal of this study was to explore how two distinct mutations in the cytoplasmic domain of iR2 affect the function of its obligate binding partner, ADAM17. The iR2^toc knock-in P159L mutation (Hosur et al., 2017) recapitulates one of the point mutations in human iR2 that cause palmoplantar hyperkeratosis and TOC (Brooke et al., 2014; Qu et al., 2019; Mokoena et al., 2018). The mouse iR2 curly-bare (cub) mutant lacks the first 268 cytoplasmic residues of iR2, including the conserved sequence that is mutated in TOC. In both mutant mouse strains, a full wavy coat of hair is restored if one or both alleles of Areg are inactivated (Hosur et al., 2018b, 2014, 2017; Siggs et al., 2014). This suggests that both phenotypes are caused by a gene-dosage-dependent gain of function of the EGFR ligand AREG, which presumably leads to increased EGFR signaling. Here, we show that the pathological skin and hair phenotypes of both strains also effectively disappear upon inactivation of one allele of Adam17. Thus, both alleles of three genes (iR2^cub or iR2^toc, Adam17 and Areg) are required to produce the full pathological skin and hair phenotypes.

To explain the dependency of these two phenotypes on ADAM17, we propose that the iR2^cub and iR2^toc mutant proteins both form a complex with ADAM17, similar to WT iR2 (Adrain et al., 2012; Kahveci-Türköz et al., 2022). Presumably, the restoration of the normal appearance of the skin and hair in iR2^cub/cub and iR2^toc/toc mice upon inactivation of one allele of Adam17 can be explained by reduced levels of the mutant iR2–ADAM17 complexes, which would no longer sufficiently exert their gain-of-function effects together with AREG. However, although inactivation of both alleles of Areg also rescues the skin and hair defects in iR2^cub/cub and iR2^toc/toc mice, a homozygous knockout of Adam17 overrides the gain-of-function phenotypes and leads to perinatal lethality with open eyes at birth, just like in Adam17^−/− mice. This is consistent with the finding that iR2^toc is not detectable in the absence of ADAM17, as previously reported for WT iR2 (Weskamp et al., 2020). Presumably, the same is also the case for iR2^cub, although this could not be tested because the anti-iR2 cytoplasmic tail (or cytotail) antibody used here binds to cytoplasmic sequences that are deleted in iR2^cub. The recently identified role of the signal peptidase complex in processing iR2, which can be suppressed by increased levels of ADAM17, lends additional support to this explanation (Zanotti et al., 2022).

Despite the similarities in the genetic dependency of iR2^cub/cub and iR2^toc/toc mice on AREG and ADAM17, unexpected differences emerged upon further analysis. Very low levels, if any, of mature ADAM17 were present in iR2^cub/cub BMDMs, leading to strongly reduced TNFα and MCSFR shedding, like in iR2^−/− BMDMs and consistent with previous studies (Siggs et al., 2014; Hosur et al., 2014). However, there were similar levels of mature ADAM17 in iR2^toc/toc BMDMs as in WT controls. iR2^toc/toc BMDMs had enhanced MCSFR processing, consistent with the increased activity of iR2^toc reported previously (Brooke et al., 2014; Maney et al., 2015; Sieber et al., 2022) although its TNFα sheddase activity was somewhat reduced. The inability of iR2^cub to support detectable amounts of the mature form of ADAM17 was further corroborated in iR1^−/− iR2^cub/cub double-mutant mice and mEFs. When we analyzed newborn iR1^−/− iR2^cub/cub mice, we found some of the phenotypes caused by lack of ADAM17-dependent paracrine EGFR-ligand signaling (enlarged heart valves and growth plates and, in one case, open eyes at birth). In cell-based assays, iR1^−/− iR2^cub/cub mEFs showed comparably low levels of constitutive or stimulated shedding of ADAM17 substrates such as TGFα, AREG and HB-EGF, as in iR1^−/−iR2^−/− mEFs, which resemble Adam17^−/− mEFs (Li et al., 2015; Sahin et al., 2004). Finally, there was no evident mature ADAM17 in the tissues analyzed in adult iR1^−/− iR2^cub/cub mice. Taken together, these findings suggest that the iR2^cub skin and hair phenotype depends on a non-physiological gain-of-function effect caused by a protein complex comprising iR2^cub, inactive ADAM17 and AREG. The iR2^cub mutation removes the known binding site for FRMD8 (also referred to as iTAP) (Kunzel et al., 2018; Oikonomidi et al., 2018), which is required for stabilizing iR2–ADAM17 on the cell surface and preventing their degradation in the lysosome (Badenes et al., 2023). This point is further supported by our finding that treatment of iR1^−/− iR2^cub/cub mEFs with the lysosomal acidification inhibitor bafilomycin restores the presence of mature ADAM17, although not its function as a TGFα sheddase. Taken together, the deletion of the FRMD8-binding site in iR2^cub provides a plausible explanation for the inability to detect mature ADAM17 in cells and tissues from iR1^−/− iR2^cub/cub mice.

As the iR2^toc/toc and iR2^cub/cub skin and hair phenotypes both require two alleles of Adam17 and Areg, but iR2^toc has an intact FRMD8-binding site, this suggests that mutations in the iR2^toc site, but not the FRMD8-binding site, cause the bare-skin phenotypes in iR2^cub/cub mice (Fig. 8). Thus, iR2^cub is presumably defective in at least two distinct functions of iR2, an essential role in stabilizing mature ADAM17 and preventing its degradation in the lysosome, and a separate gain-of-function effect caused by the loss of the sequences surrounding the iR2^toc mutation. Perhaps iR2^cub causes a more severe phenotype than iR2^toc because the entire highly conserved sequence that includes the iR2^toc site is deleted, which could have a stronger effect than a single point mutation in iR2^toc (P159L) (Fig. 8). In addition, the N-terminal deletion in iR2^cub could affect other functions of the cytoplasmic domain of iR2.

How could AREG signal through the EGFR without being processed by ADAM17 in iR2^cub/cub mice? A characteristic property of ADAM17 is that it can be rapidly and post-translationally activated by numerous stimuli, such as TNFα, G protein-coupled receptors or the phorbol ester PMA (Sahin et al., 2004; Le Gall et al., 2010). This is thought to trigger an allosteric conformational change that depends on the interaction between iR2 or iR1 and ADAM17 and enhances the catalytic activity of ADAM17 (Li et al., 2017; Tang et al., 2020; Doedens et al., 2003). Interestingly, when a soluble form of EGF is expressed without its membrane anchor, it activates EGFR signaling in an intracrine manner in the absence of proteolytic processing, which cannot be blocked by addition of an anti-EGFR antibody (Wiley et al., 1998). This showed that it is, in principle, possible to activate EGFR in the secretory pathway via intracrine signaling. However, normally the membrane anchors of EGFR ligands are thought to prevent inappropriate intracrine EGFR activation in the endoplasmic reticulum, thereby ensuring spatially and temporally regulated ADAM17-dependent EGFR signaling (Wiley et al., 1998; Jackson et al., 2003; Sternlicht et al., 2005; Peschon et al., 1998; Sahin et al., 2004; Sunnarborg et al., 2002; Blobel, 2005; Borrell-Pagès et al., 2003). Thus, the survival of iR2^cub/cub mice could conceivably be explained by a non-physiological and non-proteolytic activation of AREG by iR2^cub bound to inactive pro-ADAM17. However, we cannot rule out that processing of AREG by minute amounts of residual mature ADAM17, perhaps in a compartment of the secretory pathway where this does not normally occur, causes an intracrine gain of function.

The analysis of the maturation and function of iR2^toc-regulated ADAM17 provided additional clues as to the possible mechanism underlying the gain-of-function phenotypes in iR2^cub/cub and iR2^toc/toc mice. iR2^toc/toc BMDMs had increased activity towards the MCSFR, a commonly used substrate to monitor iR2-regulated ADAM17 activity in BMDMs (Li et al., 2017; Lora et al., 2021; Qing et al., 2016). In addition, iR2^toc-regulated ADAM17 supported reduced but nevertheless substantial amounts of shedding of TNFα, TGFα, ICAM and CD62L. However, endogenous iR2^toc–ADAM17 in iR1^−/− iR2^toc/toc mEFs behaved differently from overexpressed iR2^toc in that it had strongly reduced or almost abolished activity towards the iR2-selective ADAM17 substrate AREG (Sieber et al., 2022) as well as towards EREG and HB-EGF.

A recent study has suggested that the substrate selectivity of iR2-regulated ADAM17 towards EREG versus TGFα is determined, at least in part, by the positioning of the substrate cleavage site relative to the catalytic site of ADAM17 (Zhao et al., 2022). Moreover, the substrate selectivity of the related ADAM10 and its binding partners, the C8-tetraspanins (Tspan-C8), is thought to require the proper positioning of the catalytic site with the substrate cleavage site (Lipper et al., 2022 preprint). If we assume that the iR2^cub–ADAM17 complex engages AREG such that it changes the presentation of its EGF-like repeat, this could conceivably allow uncleaved pro-AREG to trigger intracrine EGFR signaling. In this scenario, both iR2^cub and iR2^toc would somehow alter the binding and positioning of AREG relative to AREG binding with WT iR2. An altered alignment or positioning of the catalytic site of ADAM17–iR2^toc versus ADAM17–iR2^WT could also explain the unusual functional properties observed for iR2^toc, including the somewhat reduced, but still robust, stimulated processing of TGFα, ICAM and CD62L in iR1^−/− iR2^toc/toc mEFs, the increased processing of MCSFR in iR2^toc/toc BMDMs and the loss of stimulated shedding of the iR2-selective substrates HB-EGF, AREG and EREG, compared to the shedding observed for WT iR2 in iR1^−/− cells. Genetic support for a defect in HB-EGF shedding by iR2^toc was provided by the thickened aortic and pulmonic heart valves in two out of three iR1^−/− iR2^toc/toc pups, but not in iR1^−/− controls with WT iR2 (see also Li et al., 2015), as normal development of these valves in mice depends on HB-EGF processing by ADAM17 (Jackson et al., 2003; Blobel, 2005; Yamazaki et al., 2003). However, the normal growth plates and closed eyelids in iR1^−/− iR2^toc/toc pups are consistent with sufficient processing of TGFα by iR2^toc-regulated ADAM17 in vivo (Peschon et al., 1998; Hall et al., 2013; Saito et al., 2013). This is the first evidence, to our knowledge, for a substrate-selective developmental defect caused by a point mutation in endogenous iR2, albeit with only partial penetrance.

In summary, this study provides new insights into the effects of mutations in the cytoplasmic domain of iR2 on the function of ADAM17 in cells and in vivo. The iR2^cub deletion removes the FRMD8-binding site, which is required for stabilization of the iR2–ADAM17 complex, thus explaining the lack of mature ADAM17 in iR1^−/− iR2^cub/cub mutant mice. As the only common feature of iR2^toc and iR2^cub is an altered iR2^toc site, this is most likely the principal cause of the hair and skin phenotypes. We propose that the gain-of-function phenotypes caused by both mutations are triggered by a change in the binding or presentation of AREG, which causes increased and most likely proteolysis-independent EGFR signaling by iR2^cub- and iR2^toc-regulated ADAM17, although other interpretations could also be compatible with the data. Further studies will be necessary to determine whether the human iR2^toc mutations cause TOC by a similar or different mechanism. We hope that a better mechanistic understanding of the effect of these mutations in iR2 on ADAM17 and AREG in mice could potentially help devise new treatment option for patients with TOC, for example, by screening for compounds that restore the proper alignment and function of AREG and ADAM17 in the presence of the iR2^toc mutation.

PMA was used at a concentration of 25 ng/ml (Sigma-Aldrich, St Louis, MO, USA; P8139), ionomycin at 2.5 µM (Sigma-Aldrich, I9657), the metalloprotease inhibitor marimastat at 5 µM (Tocris Biosciences, Bristol, UK; 2631), LPS at 10 ng/ml (Sigma-Aldrich, L4391), and bafilomycin A1 at 1 µM (Cell Signaling Technology, Danvers, MA, USA; 54645). The rabbit polyclonal anti-ADAM17 cytotail antibodies used for western blotting were described previously (Schlöndorff et al., 2000; Lum et al., 1998) and used at 1:2000. The ADAM10 antibody was from Abcam, Cambridge, UK (ab124695, used at 1:2000). The rat monoclonal antibody against the cytoplasmic domain of iR2, used at 1:100, has been previously described (Weskamp et al., 2020) and was kindly provided by Dr Stefan Lichtenthaler, Technical University Munich, Munich, Germany. The LC3B antibody was from Novus Biologicals, Centennial, CO, USA (NB100-2220, lot EX, used at 1:1000). The ERK1/2 antibody was from Sigma-Aldrich (M5670, lot 059M4758V, used at 1:5000). The anti-α-tubulin rabbit polyclonal antibody was from Cell Signaling Technology (2144S, lot 7, used at 1:2000), the anti-GAPDH antibody was from Abclonal (AC002, lot 3500033014, used at 1:10,000). The secondary anti-rabbit IgG HRP conjugate was from Promega, Madison, WI, USA (W401B, lot 0000527210, used at 1:5000) The secondary anti-mouse IgG HRP conjugate was also from Promega (W402B, lot 0000149014, used at 1:5000). The secondary anti-rat IgG peroxidase antibody was from Sigma-Aldrich (A5795, used at 1:5000). For cell surface biotinylation, EZ Link Sulfo-NHS-LC-Biotin (Thermo Fisher Scientific, Waltham, MA, USA; 21335) was used, followed by streptavidin-sepharose pulldown (Thermo Fisher Scientific, 434341) for Figs 1A and 4A. Concanavalin-A sepharose beads (GE Healthcare, Uppsala, Sweden; 17-0440-01) were used for ADAM17 enrichment for western blot analysis (Schlöndorff et al., 2000; McIlwain et al., 2012) for Figs 2C, 3C, 5C and 7C. For flow cytometry to detect cell-surface MCSFR on BMDMs, IgG2a Fc block (BioLegend, San Diego, CA, USA; 101302, lot B264872) was used, followed by either PE mouse IgG2a isotype control (1:200, BD Biosciences, Franklin Lakes, NJ, USA; 558595, lot 9072688) or PE anti-mouse CD115/MCSFR (1:200, Invitrogen, Waltham, MA, USA; 12-1152-81, lot 2481371). For flow cytometry to detect cell-surface ADAM17 on BMDMs, a rabbit-anti ADAM17 ectodomain antibody (anti-ADAM17ecto) was used at 1:500 (Lora et al., 2021) and detected with PE-donkey anti-Rabbit IgG (1:200, Jackson ImmunoResearch, 711-116-152, lot 152858). Normal rabbit serum (1:200, Jackson ImmunoResearch, 011-000-001) served as the control serum. 4′6-diamidino-2-phenylindole, dilactate (DAPI) (Invitrogen, D3571) was employed as a live/dead stain for flow cytometry. For all alkaline phosphatase (AP)-based assays, we used the AP substrate 4-nitrophenyl phosphate (Thermo Fisher Scientific, 34045) at a concentration of 1 mg/ml.

iR2^cub/cub mice and iR2^toc/toc mice in a C57BL/6J background were kindly provided by Dr Vishnu Hosur and Dr Leonard D. Shultz at the Jackson Laboratories in Bar Harbor, ME, USA (Hosur et al., 2014, 2017). iR2^−/− mice were obtained from the Knockout Mouse Project (KOMP) Repository at University of California, Davis, and correspond to the Rhbdf2^{tm1b(KOMP)Wtsi} option available at KOMP (C57BL/6N). These animals were crossed with 129X1Sv/J mice to generate mixed-background 129X1Sv/J/C57BL/6N mice. iR1^−/− and Adam17^−/+ mice were previously described (Li et al., 2015; Horiuchi et al., 2007). Specifically, we used the iR1^−/− mice generated from embryonic stem cells from the European Conditional Mouse Mutagenesis program (EUCOMM) in this study because these particular iR1^−/− mutant mice have no spontaneous pathological phenotypes, even though this mutation abolishes the iR1-dependent maturation and function of ADAM17 (see Li et al., 2015 for details). Therefore, double-knockout mice lacking iR1 (EUCOMM) and iR2 resemble mice lacking Adam17 (Li et al., 2015), providing a good comparison for the iR1^−/− iR2^cub/cub and iR1^−/− iR2^toc/toc mice described here. It is important to note that two additional strains of iR1^−/− mice have more severe phenotypes than the iR1^−/− (EUCOMM) mice used in this study, presumably because they also lack exons 2 and 3 of iR1 (Christova et al., 2013; Hosur et al., 2020), which are still present in the iR1^−/− (EUCOMM) mice. One possible explanation for the normal phenotype of iR1^−/− (EUCOMM) mice is that the N-terminal 82 amino acid residues encoded by exons 2 and 3 of iR1 that are still present in these animals have additional essential functions not related to the regulation of ADAM17. Therefore, use of another iR1^−/− strain would complicate any interpretations with respect to how iR2^cub and iR2^toc affect the function of ADAM17 in the absence of iR1-dependent functions of ADAM17. The more severe phenotype of iR1^−/− mice carrying deletions that include exons 1–3 could be related to the import of the N-terminal sequence encoded by exons 2 and 3 into the nucleus and the resulting transcriptional regulation, as has been proposed for the cytoplasmic domain of the related iR2 (Dulloo et al., 2022 preprint), although the cause of the more severe phenotypes remains to be determined. We note that the iR1^−/− (EUCOMM) mice used here have also been referred to as mice carrying a gain-of-function mutation (Rhbdf1^v2/v2 mice in Hosur et al., 2020). In our view, a more accurate description is that iR1^−/− (EUCOMM) mice have a complete loss-of-function phenotype with respect to the iR1-dependent functions of ADAM17, but do not have the additional more severe and presumably non-ADAM17-dependent phenotypes, the mechanistic cause of which remains to be determined.

iR2^cub/cub mice and iR2^toc/toc mice of the C57BL/6J background were crossed with WT mice of the C57BL/6J/129X1SV/J mixed background to generate mixed-background heterozygotes, which were then bred with each other to generate mixed-background homozygous mutant populations for experiments. BMDMs were isolated from WT C57BL/6J/129X1SV/J mixed-background animals, or homozygous iR2^cub/cub or iR2^toc/toc mixed-background mice for the experiments presented in Figs 1 and 4, with separately bred iR2^−/− mice (129X1Sv/J/C57BL/6N) serving as controls. The same mice were used to isolate tissues for Figs 2 and 5 and Fig. S1. iR1^−/− (EUCOMM) mice (referred to as iR1^−/− mice in the text) of the C57BL/6N/129X1Sv/J mixed genetic background were used to generate iR1^−/− iR2^cub/cub and iR1^−/− iR2^toc/toc double-mutant mice of the C57BL/6J/C57BL/6N/129X1Sv/J mixed genetic background for Figs 2 and 5 and Figs S2 and S6. Embryos isolated from crosses of these double-mutant mice were used to generate mEFs for Figs 2 and 6 and Figs S3 and S5.

Mixed-background iR2^cub/cub and iR2^toc/toc mice (C57BL/6J/129X1SV/J) were also crossed with Adam17^+/− mice of the C57BL/6J/129X1SV/J mixed genetic background. Then iR2^cub/cub Adam17^+/− parents were mated to generate iR2^cub/cub and iR2^cub/cub Adam17^+/− adult littermates and iR2^cub/cub Adam17^−/− newborn pups for Fig. 3A and Fig. S4A, and newborn iR2^cub/cub, iR2^cub/cub Adam17^+/− and iR2^cub/cub Adam17^−/− pups for Fig. 3C and Fig. S4B. Embryos from the cross of iR2^cub/cub Adam17^+/− mice were isolated to generate mEFs for Fig. 3 and Fig. S5. Similarly, iR2^toc/toc Adam17^+/− parents were mated to generate iR2^toc/toc and iR2^toc/toc Adam17^+/− littermates for Fig. 7 and Fig. S7A and newborn iR2^toc/toc and iR2^toc/toc Adam17^−/− pups for Fig. S7B. Embryos from the cross iR2^toc/toc Adam17^+/− mice were isolated to generate mEFs for Fig. 6. All genotypes were confirmed by PCR and the iR2^toc point mutation was confirmed by Sanger sequencing (Azenta Life Sciences, Chelmsford, MA, USA). The adult animals were between 2 and 8 months of age, and both male and female mice were included in comparable numbers. All animal experiments were approved by the Institutional Animal Care and Use Committee (IACUC) of the Hospital for Special Surgery at Weill Cornell Medicine.

BMDMs were isolated from adult mice older than 8 weeks (Weskamp et al., 2020). All mice were genotyped before weaning, and the genotype was confirmed by PCR using a tail sample from the euthanized mice. Briefly, bone marrow was flushed from femurs and tibiae of adult mice with Hanks' balanced salt solution (HBSS) (Corning, Corning, NY, USA; 21-022-CV), centrifuged at 400 g for 10 min to collect the marrow and cultured in Dulbecco's modified Eagle medium (DMEM) (Corning; 10-014-CV) containing 10% fetal bovine serum (FBS) (R&D Systems; S1155OH, lot G22149), 1% penicillin-streptomycin and 10 ng/ml murine macrophage colony-stimulating factor (Peprotech, Rocky Hill, NJ, USA; 315-02) on Petri dishes. After 7 days in culture, mature macrophages were collected using Accutase cell dissociation reagent (Innovative Cell Technologies, San Diego, CA, USA; AT-104), counted and plated for experiments. Primary mEFs were isolated as previously described (Sahin et al., 2006). Briefly, mouse embryos were collected from iR2^cub/cub, iR2^cub/cub iR1^−/−, iR2^cub/cub Adam17^+/−, iR2^toc/toc, iR2^toc/toc iR1^−/− or iR2^toc/toc Adam17^+/− dams at embryonic day 14, the organs were removed, and the remaining tissue was digested in trypsin at 37°C to generate a single-cell suspension. Genotypes of the isolated embryos were determined by PCR on genomic DNA. Cells were cultured in DMEM containing 10% FBS (R&D Systems, Flowery Branch, GA, USA; S11550H) containing 1% penicillin-streptomycin (Corning, 30-0020Cl). After one passage, the mEFs were immortalized by transduction with a retroviral plasmid encoding the SV40 large T antigen. Retroviruses were generated using the GP2-293 packaging cell line (TakaraBio, Kusatsu, Shiga, Japan; 631458). iR2^cub/cub, iR2^cub/cub iR1^−/−, iR2^cub/cub Adam17^+/−, iR2^cub/cub Adam17^−/−, iR2^toc/toc, iR2^toc/toc iR1^−/− and iR2^toc/toc Adam17^−/− cells lines were generated and immortalized for the studies shown in this paper. The WT, iR1^−/−, iR2^−/−, iR1^−/− iR2^−/− and Adam17^−/− cell lines have been previously described (Maretzky et al., 2013; Li et al., 2015; Horiuchi et al., 2007). Immortalized cell lines used for shedding experiments were tested for mycoplasma contamination by PCR at least every 6 months.

AP-tagged expression constructs for TGFα, ICAM, CD62L, BTC, AREG, HB-EGF and EREG have been previously described (Sahin et al., 2004; Maretzky et al., 2013). SV40 large T cell antigen was expressed from the pMSCVzeo retroviral expression vector (Clontech, Mountain View, CA, USA, 631461) and used for immortalization of mEF lines with different genotypes (see above) in conjunction with the pEco envelope vector (Clontech, 631457).

BMDMs were plated in 96-well round-bottomed plates at 3×10⁵ cells/well in OptiMEM (Gibco, Grand Island, NY, USA; 51985091). Cells were either sham treated or treated with 25 ng/ml PMA for 1 h and the plates were then placed on ice. The live and non-permeabilized cells were blocked with Fc-block (Biolegend; 101302, lot B264872) for 15 min and then stained with anti-CD115 at 1:200 for surface MCSFR detection and DAPI for live/dead differentiation. For flow cytometry of cell-surface ADAM17, 3×10⁵ live, non-permeabilized BMDMs of different genotypes were stained with the anti-ADAM17ecto antibody (Lora et al., 2021) at 1:500 and bound primary antibodies were detected with a PE-anti-rabbit IgG antibody (1:200).

BMDM were cultured for 7 days, then collected using Accutase cell dissociation reagent and plated in 24-well tissue culture plates at 1×10⁵ cells/well in 0.4 ml complete DMEM overnight. The following day, the cells were either sham treated or treated with 10 ng/ml LPS for 1 h to stimulate TNFα production and shedding. After 1 h, the supernatant was collected to measure the concentration of the released TNFα using the mouse TNFα DuoSet ELISA kit (Bio-Techne, R&D Systems, Minneapolis, MN, USA; DY410) following the manufacturer's protocol.

Adult mice were euthanized with CO₂. Gross post-mortem examination of all organs was performed, followed by fixation in 10% neutral buffered formalin and decalcification of bones in a formic acid solution (Surgipath Decalcifier I, Leica Biosystems, Wetzlar, Germany). Tissues were then processed in ethanol and xylene and embedded in paraffin. Five-micrometer sections were prepared using a HistoCore Autocut Automated Rotary microtome (Leica Biosystems), stained with Hematoxylin and Eosin, and examined by a board-certified veterinary pathologist (S.M.). Newborn mice were euthanized, and fixation, decalcification, slide preparation and examination were performed as described for adults, with the addition of serial sections performed to evaluate heart valves and the proximal humeral and tibial and distal femoral growth plates.

Fibroblasts were transfected with AP-tagged substrates using Lipofectamine 3000 (Invitrogen, L3000015) following the manufacturer's protocol. Shedding experiments were performed 1 day post transfection. Cells were washed, serum starved in OptiMEM for 1 h, and then the medium was changed to fresh OptiMEM with the indicated inhibitors or stimuli and incubated for 1 h. Supernatants and cell lysates were collected separately and the AP activity in the respective fractions was measured at an optical density of 405 nm after incubation with the AP substrate 4-nitrophenyl phosphate (Thermo Fisher Scientific, 34045) overnight at room temperature. No AP activity was detected in non-transfected control cells. Three identical wells were prepared for each fraction. The fold change of AP activity in the untreated versus stimulated supernatant was calculated for normalization for Fig. S5. Fold change represents the ratio of the AP signal of the PMA-treated supernatant over the untreated supernatant. The AP ratio shown in all other figures represents the AP signal from the supernatant divided by the total supernatant plus lysate signal for the same sample multiplied by 100. Each experiment was performed at least three times.

Cells and organs were lysed on ice in standard lysis buffer composed of PBS containing Triton X-100 (1%), 1,10-phenanthroline (10 mM; Sigma Aldrich; P9375-259, lot SLCD2111), marimistat (5 µM) and protease inhibitor cocktail (1:500, Roche, Basel, Switzerland), and then incubated with concanavalin A beads to enrich for glycoproteins (Schlöndorff et al., 2000; McIlwain et al., 2012), except for panels Fig. 1A, Fig. 4A and Fig. S3A, where the western blots were performed on cell lysates directly, without concanavalin A enrichment. Lysates containing comparable amounts of glycoproteins were separated on NuPAGE 10% Bis-Tris Polyacrylamide gels (Invitrogen, NP0315BOX) using NuPAGE MOPS SDS Running Buffer (Invitrogen, NP0001) and transferred onto BioTrace NT nitrocellulose (Pall Corporation, Port Washington, NY, USA; 66485). Membranes were blocked with 5% skim milk in PBS, incubated with primary antibodies and washed three times in 0.05% Tween 20 in PBS, and bound primary antibodies were detected with peroxidase-conjugated goat anti-rabbit antibody (Promega, W401B) using the ECL detection system (Cytiva, Marlborough, MA, USA, RPN2106, and Millipore, Burlington, MA, USA, WBULS0100) and a Chemidoc image analyzer (Bio-Rad, Hercules, CA, USA). See Fig. S8 for blot transparency images of the western blots used for this study.

mEFs were seeded in 10 cm tissue culture dishes at 6×10⁵ cells per plate in full medium (DMEM containing 10% FBS and 1% penicillin-streptomycin). The following evening, cells were washed and OptiMEM containing either 1 µM bafilomycin A1 or an equivalent volume of DMSO was added to the cells, which were incubated for 16 h at 37°C in 5% CO₂. After 16 h, the cells were washed with PBS and lysed in standard lysis buffer (see above). Western blot analysis for LC3B accumulation was used to determine the efficacy of lysosome inhibition.

Fibroblasts were transfected as described above. On the following evening, cells were washed and serum starved in OptiMEM containing bafilomycin A1 at a concentration of 1 µM for 16 h. After 16 h, the cells were either sham treated or treated with 25 ng/ml PMA for 1 h in the continued presence of bafilomycin A1 at 1 µM. Supernatants and lysates were collected and analyzed as described above.

BMDMs were treated with LPS (10 ng/ml for 15 h) prior to labelling. Cell surface proteins of BMDMs were labeled with membrane impermeable EZ-Link-Sulfo-NHS-LC-Biotin at a concentration of 1 mg/ml following the manufacturer's protocol. Cells were lysed at 4°C in standard cell lysis buffer composed of PBS, Triton X-100 (1%), 1,10-phenanthroline (10 mM), marimastat (5 µM) and protease inhibitor cocktail (1:500). Biotinylated proteins were enriched using streptavidin-sepharose 4B beads overnight at 4°C. The beads were washed four times for at least 1 h at 4°C with cell lysis buffer plus 0.1% SDS and the bound material was eluted in sample loading buffer (0.33% SDS, 1.6 mM EDTA, 110 mM Tris, 16% glycerol, 0.05% Bromophenol Blue and 100 mM dithiothreitol in water) and boiled for 10 min at 100°C. The samples were analyzed by western blotting as described above.

Adult mice of 2–7 months of age or newborn postnatal day 0 (P0) pups were euthanized and dissected to isolate individual organs as indicated, which were collected into ice-cold PBS to rinse and then moved to standard cell lysis buffer (see above) on ice. The tissues were mechanically disrupted by 30-s bursts with a polytron homogenizer (Polytron-Aggregate model PT 10-35 GT, probe PT-DA 07/2 EG-B101, Kinematica, Switzerland). Tissue suspensions were then incubated on ice for a minimum of 15 min. Samples were transferred to 50 ml Nalgene Round Centrifuge Tubes (Thermo Fisher Scienctific Nalgene Products, Rochester, NY, USA; 3117-9500) and centrifuged at 16,000 g for 30 min at 4°C in a Sorvall SS-34 Rotor in a Sorvall Evolution RC Centrifuge (Thermo Fisher Scientific). Lysate supernatants were then incubated with concanavalin A beads to enrich for glycoproteins, including ADAM17 (Schlöndorff et al., 2000; McIlwain et al., 2012). The beads were washed twice with cell lysis buffer and bound material was eluted in sample loading buffer by boiling for 10 min at 100°C. Samples were analyzed by western blotting as described above.

Adult mice between 2 and 7 months of age were euthanized and dissected to isolate hearts as a representative tissue for qPCR analysis. A portion of the heart (∼30 mg) was placed into RNAlater solution (Invitrogen, AM7020), incubated on ice for 1 h and then moved to −80°C. In preparation for qPCR analysis, the samples were thawed on ice and lysed using the polytron homogenizer and the RNA was isolated using RNeasy Plus Mini Kit (QIAGEN, Hilden, Germany; 74134) following the manufacturer's protocol. RNA quantity was determined using a NanoDrop spectrophotometer (Thermo Fisher Scientific; Nanodrop one). Reverse-transcription PCR was performed using M-MulV Reverse Transcriptase (New England Biolabs, Ipswich, MA, USA; M0253) following the manufacturer's protocol. qPCR was performed using Maxima SYBR Green/ROX qPCR master mix (2×) (Thermo Fisher Scientific, K0222) following the manufacturer's protocol. qPCR was run on a Quant Studio 6 Pro (Applied Biosystems, Waltham, MA, USA). Data were exported from the Design and Analysis software to Microsoft Excel for analysis. Adam17 qPCR primers were purchased from Eurofins Genomics with the following sequences: F, 5′-GATGCTGAAGATGACACTGTG-3′, and R, 5′-GAGTTGTCAGTGTCAACGC-3′. Gapdh qPCR primers were from Quantitech Primer Assays (QIAGEN, QT01658692 Mm_GAPDH_3_SG). iR2 (Rhbdf2) qPCR primers were from Quantitech Primer Assays (QIAGEN, QT00132657 Mm_Rhbdf2_1_SG).

Candidate sequences for iR1 and iR2 in different species were identified by BLASTP (https://blast.ncbi.nlm.nih.gov/Blast.cgi?PAGE=Proteins) using the mouse iR1 or iR2 protein sequence as a query and the following species to identify orthologous sequences: Xenopus laevis (African clawed frog), Gallus Gallus (chicken), Lagenorhynchus obliquidens (dolphin), Homo sapiens (human) and Mus musculus (mouse). The aligned sequences of the conserved TOC locus in Fig. 8B were identified by pairwise alignment of the mouse sequence as the query and the candidate orthologue as the target sequence using BLASTP. The accession numbers for the orthologues used for Fig. 8B are: X. laevis iR1, XP_018094554.1; X. laevis iR2, XP_018090848.1; G. Gallus iR1, XP_040503413.1; G. Gallus iR2, XP_004946173.2; L. obliquidens iR1, XP_026973861.1, L. obliquidens iR2, XP_026936583.1; H. sapiens iR1, NP_071895.3; H. sapiens iR2, NP_078875.4; M. musculus iR1, XP_011241960.1; and M. musculus iR2, NP_001161152.1.

All graphs are presented as mean±s.d. and were analyzed using GraphPad Prism 9.3.1. The unpaired two-tailed Welch's unequal variance t-test was used to compare treated to untreated conditions for a given genotype, as indicated. In Fig. S1, the comparisons were between WT and one mutant sample each, as indicated by individual bars above the graphs in panels A and B. A P-value of <0.05 was considered statistically significant.

We thank Alejandra Vela Moreno from the Blobel laboratory, and Iveta Simanska, Kim McBride and Jacqueline Candelier from the Tri-Institutional Laboratory of Comparative Pathology, Hospital for Special Surgery, Memorial Sloan Kettering Cancer Center, The Rockefeller University, Weill Cornell Medicine, New York, USA, for excellent technical assistance, Jose Lora for helpful discussions, and Leonard D. Shultz and Vishnu Hosur from the Jackson Laboratory in Bar Harbor, ME, USA for providing iR2^cub and iR2^toc mice.