Integral Membrane Protein 2A Is a Negative Regulator of Canonical and Non-Canonical Hedgehog Signalling

The Hedgehog (Hh) receptor PTCH1 and the integral membrane protein 2A (ITM2A) inhibit autophagy by reducing autolysosome formation. In this study, we demonstrate that ITM2A physically interacts with PTCH1; however, the two proteins inhibit autophagic flux independently, since silencing of ITM2A did not prevent the accumulation of LC3BII and p62 in PTCH1-overexpressing cells, suggesting that they provide alternative modes to limit autophagy. Knockdown of ITM2A potentiated PTCH1-induced autophagic flux blockade and increased PTCH1 expression, while ITM2A overexpression reduced PTCH1 protein levels, indicating that it is a negative regulator of PTCH1 non-canonical signalling. Our study also revealed that endogenous ITM2A is necessary for timely induction of myogenic differentiation markers in C2C12 cells since partial knockdown delays the timing of differentiation. We also found that basal autophagic flux decreases during myogenic differentiation at the same time that ITM2A expression increases. Given that canonical Hh signalling prevents myogenic differentiation, we investigated the effect of ITM2A on canonical Hh signalling using GLI-luciferase assays. Our findings demonstrate that ITM2A is a strong negative regulator of GLI transcriptional activity and of GLI1 stability. In summary, ITM2A negatively regulates canonical and non-canonical Hh signalling.


Introduction
The Hedgehog (Hh) signalling pathway has essential functions in embryonic development, tissue homeostasis and cancer [1]. PTCH1 is the most important Hh receptor and a well-established tumour suppressor [2]. The canonical signalling pathway is initiated by binding of one of the Hh ligands (Shh, Ihh, or Dhh) to the 12-transmembrane receptor Patched1 (PTCH1). In the absence of Hh ligands, PTCH1 represses the activation of the G protein-coupled receptor Smoothened (SMO), resulting in processing of the GLI2 and GLI3 transcription factors into repressors. Binding of a Hh protein to PTCH1 inhibits its intrinsic activity, believed to be the transport of cholesterol across the plasma membrane to generate a localized gradient, leading to derepression of SMO and its accumulation at the primary cilium. Ciliary SMO induces activation of GLI2 and GLI3 and prevents their processing, initiating a GLI-dependent transcriptional response that results in upregulation of the short-lived strong activator GLI1 and of the Hh protein-sequestering membrane proteins PTCH1, PTCH2 and HHIP, among other cell type-specific targets.
More recent studies have revealed that PTCH1 has additional functions independently of SMO/GLI, known as "Type I non-canonical Hh signalling" to differentiate it from SMO-dependent, GLI-independent processes [2,3]. The C-terminal domain (CTD) of PTCH1 is dispensable for canonical Hh signalling [4,5], although one study suggests that the CTD is necessary for PTCH1 localization to the primary cilium [6]. The first identified role of the CTD was a pro-apoptotic function, mediated by interaction with the DRAL/TUCAN/Caspase9 complex and independent of SMO [7][8][9][10]. With the goal of identifying additional proteins that interact with the CTD of PTCH1, we previously performed a yeast-2-hybrid study, in which we identified the autophagy-related protein ATG101, a subunit of the ULK complex that mediates initiation of autophagy [11]. Our study established that the cytosolic CTD of PTCH1 inhibits autophagic flux by impairment of autophagosome-lysosome fusion or acidification and that its interaction with ATG101 was important for that function. Paradoxically, ATG101 regulates autophagy initiation and expansion of phagophore membranes to form the autophagosome and no role in the terminal step of autophagy has been discovered. This led us to hypothesize that PTCH1 might block autophagy completion with the aid of additional interacting proteins. In this study, we focus on the potential regulatory role of integral membrane protein 2A (ITM2A), which was identified in a large proportion of our positive clones in the yeast-2hybrid screen. Remarkably, out of 63 sequenced and validated positive hits, ITM2A was recovered in 21 of them (33%). A novel function of ITM2A in blocking autophagic flux by an inhibitory interaction with v-ATPase was reported a few years ago [12], suggesting a potential functional interaction with PTCH1. However, the role of ITM2A in autophagy remains controversial since it was proposed to induce autophagy in breast cancer cells [13].
ITM2A is a type II single-pass membrane protein with intracellular N-terminal domain and an extracellular BRICHOS domain [14]. Two alternatively spliced forms of ITM2A were described: the long isoform of 263 aa (30 kDa) and a shorter isoform of 219 aa lacking the transmembrane domain (25 kDa). Although the function of ITM2A is not completely elucidated, it has been linked to differentiation of skeletal muscle, cartilage tissue and T cell development [15,16]. Early reports indicate that ITM2A levels increase during skeletal muscle differentiation and the formation of myotubes in vivo and in C2C12 cells, a model of in vitro myogenesis [15,17,18]. Overexpression of ITM2A can accelerate myotube formation; however, it is not essential for muscle development, as indicated by in vitro knockdown studies and in conditional knockout mice [17,18]. Canonical Hh signalling prevents myogenic differentiation by keeping C2C12 and primary satellite muscle cells in the proliferative state [19]. In particular, Gli1 and Gli2 were shown to repress MyoDdependent transcription, preventing terminal myotube differentiation [20]. This suggests that ITM2A function could be inversely correlated to canonical Hh signalling activity.
Given the involvement of ITM2A in biological processes that are regulated themselves by different types of Hh signalling and the potential physical interaction with PTCH1, we decided to investigate their crosstalk.

Reagents
The antibodies used in this study (name, catalog number, vendor and dilution used in Western blotting) is described in Supplementary Table S1 To generate  shRNA stable cells, cells were transduced with lentiviral particles, encoding scrambled or  ITM2A targeting shRNA sequences (Table 1), produced in HEK293T cells as previously described [21]. Table 1. shRNA sequences used to silence ITM2A in human and mouse cell lines.

Name
Sequence (5 -3 ) C2C12 cells were cultured in DMEM containing 10% fetal bovine serum and penicillin/streptomycin in a humidified 5% CO 2 incubator. In order to keep the cells in the undifferentiated state, they were split when reaching 70% confluency. Differentiation was induced by changing the medium of 100% confluent cultures by DMEM containing 2% horse serum (Sigma cat. H1270) and penicillin/streptomycin. The day the differentiation medium was added was called day 0, and the cells were kept for 7 days with media refresh every other day. The generation of stable shRNA ITM2A and control C2C12 cells was performed by transduction of undifferentiated cells with lentiviruses encoding hairpin RNAs (Table 1), followed by selection with puromycin.

Quantitative PCR
Total RNA was isolated from each of the three cell lines with RNeasy Mini Kit (QI-AGEN, Manchester, UK; cat 74104) and quantified by A260nm/A280nm in a Nanodrop. Synthesis of cDNA was performed using 1 µg RNA with iScript cDNA Synthesis Kit (BioRad) using hexarandom primers following the manufacturer's instructions. Realtime quantitative PCR was performed using 1-5 µL of cDNA and target-specific primers ( Table 2) with the SsoFast EvaGreen Supermix (BioRad) with a CFX Connect Real-Time PCR Detection System (BioRad) thermocycler. Amplification was quantified and expressed as fold using the ∆∆Ct method. reagent and 8 µg total DNA. For co-immunoprecipitation (co-IP) assays, two different plasmids were used in the same reaction (4 µg of each plasmid). Twenty-four hours after transfection, cells were washed with 2 mL of ice-cold 1× PBS and scraped in 700 µL of co-IP lysis buffer (50 mM Tris HCl pH 7.5, 150 mM NaCl, 1% NP-40, 0.05% Sodium deoxycholate, 1 mM EDTA, 2.5 Mm MgCl 2 ) supplemented immediately before use with 1× Proteoloc protease inhibitor cocktail (Gentaur, London, UK; cat. 44204), 0.2 mM PMSF and 1 mM DTT. Lysates were incubated at 4 • C for 30 min with rotation and then centrifuged at 13,000 RPM for 15 min at 4 • C. The supernatant was separated and 200 µL set apart as whole cell lysate (WCL) and mixed with 40 µL of 6× Laemmli buffer. The remaining supernatant was incubated with 4 µg primary antibody (Table 3) at 4 • C with rotation for 1.5 h, followed by the addition of 30 µL Dynabeads Protein G (Thermo Fisher; cat. 10003D) for an additional 1 h incubation at 4 • C with rotation. A magnetic rack was used to collect and wash the beads three times with 1 mL of co-IP lysis buffer. Immunoprecipitates were extracted from the bead by the addition of 18 µL of 2× Laemmli buffer. Both beads and WCL were incubated on heat block at 45 • C for 25 min and stored at −80 • C for Western blotting.

Gli-Luciferase Assay
NIH3T3 cells and Ptch1 −/− mouse embryonic fibroblasts (MEFs) were used for this assay. Cells were grown in a 10 cm culture dish to an approximate confluence of 90%, trypsinised, and seeded in 24-well plates at a 1:5 dilution in growth media. Cells were transfected 24 h after plating using different transfection reagents for each cell line: FuGENE ® HD (Promega, Southampton, UK; cat. E2311) for NIH3T3 and TransIT-X2 ® (Mirus Bio, Madison, WI, USA; cat MIR6003) for MEFs.
Assays were always conducted in triplicate. For each well, a fixed amount of p8xGBS-Luc (135 ng), pRL-SV40 or pRL-TK (15 ng/well) and testing plasmid DNA (375 ng, containing a single plasmid or a combination of plasmids) and 1.5 µL of the transfection reagent were used, following the manufacturer's instructions. After 24 h, transfected cells were incubated with serum starvation media (DMEM, 0.5% FBS, 1% GlutaMAX) for an additional 48 h. Cells were washed with 0.5 mL of PBS and lysed with 100 µL passive lysis buffer with vigorous shaking at room temperature for 15 min. Dual-luciferase reporter assay system (Promega; cat. E1910) was used to measure the activity of Firefly-luciferase and Renilla-luciferase using a Promega GloMax 20/20 luminometer (cat. E5311). Values of the individual measurements were generated as relative luciferase units (RLUs) and as Firefly-luciferase/ Renilla-luciferase ratio.

Mass Spectrometry
To prepare the samples for proteomic analysis, 3 × 10 5 HEK293 cells were seeded in 10 cm culture dishes and transiently transfected 24 h later with HA-ITM2A or empty pcDNA3.1 plasmid, using 8 µg of DNA and 20 µL of Lipofectamine TM 2000. Twenty-four hours after transfection, cells were washed with 2 mL of ice-cold 1x PBS and processed for IP as described above. Samples were submitted in solution form to the mass spectrometry facility of the University of Leeds for the identification of immunoprecipitation-interacting protein partners and for the detection of protein post-translational modifications. Protein digestion of samples was carried out with trypsin. Data was processed by the facility using the software Peaks (Bioinformatics Solutions, Waterloo, ON, Canada).

STRING Protein Association Network
An interactome map was constructed using STRING for the ITM2A-specific identified proteins by mass spectrometry analysis. The interactome was constructed with the STRING database using default settings; the confidence score was set to medium (0.4) with network edges showing the confidence of an interaction.

Statistical Analysis
GraphPad Prism 9 was used for the statistical analysis and for the generation of graphs. Unless otherwise specified, three biological replicates were performed. Error bars are shown as standard error of the mean (SEM). Two-tailed paired Student's t-test was used to analyse the significant difference between two groups. One-way ANOVA analysis was selected to analyse at least three different groups when the samples showed normal distribution and equal variance.

ITM2A Physically Interacts with the Hedgehog Receptor PTCH1
Following the interaction studies in yeast cells, we tested if human PTCH1 interacts with ITM2A in mammalian cells by co-immunoprecipitation. First, we determined that fulllength PTCH1-HA was capable of interacting with myc-tagged ITM2A when co-expressed in HEK293 cells. Reciprocal pulldowns using anti-myc or anti-HA antibodies confirmed the interaction in mammalian cells ( Figure 1A). Unexpectedly, the CTD alone (aa 1176-1447) in soluble form was unable to interact strongly with ITM2A ( Figure 1A). Since the yeast-2-hybrid experiment that initially indicated an interaction of ITM2A with PTCH1 used only the CTD as bait, we considered the possibility that for interaction of the proteins in mammalian cells, co-localisation at the plasma membrane was required. Therefore, we expressed a membrane-associated CTD (through fusion with a myristoylated GFP (CAAX-eGFP-CTD) [10] and compared the interaction to ITM2A with full-length PTCH1-eGFP or empty myr-eGFP as a negative control. Immunoprecipiation of myc-ITM2A resulted in detection of full length PTCH1 and the CTD fragment, indicating that membrane-associated CTD was sufficient for their interaction ( Figure 1B).
To determine which domains of PTCH1 are necessary for PTCH1-ITM2A physical interaction, we generated PTCH1 mutants with deletions of the two largest intracellular domains, the middle loop (ML) and the CTD [5]. Removal of the ML alone (PTCH1∆ML) or both ML and CTD (PTCH1∆ML∆CTD) did not prevent ITM2A interaction, indicating that, in addition to the CTD, the transmembrane and/or extracellular loops of PTCH1 independently interact with presumably different regions of the transmembrane protein ITM2A ( Figure 1C).
PTCH1 has been recently shown to form stable dimers even in the absence of Hh ligands [22]. We reasoned that ITM2A binding to the CTD and additional domains could cluster together two molecules of ITM2A. Thus, we tested if ITM2A is competent to form self-interactions, as such could be permissive for the formation of higher order magnitude complexes with PTCH1 dimers. As shown in Figure 1D, ITM2A forms strong interactions with itself, supporting a stoichiometric model of 2 PTCH1: 2 ITM2A molecules in the complex.
PTCH1 has been recently shown to form stable dimers even in the absence of Hh ligands [22]. We reasoned that ITM2A binding to the CTD and additional domains could cluster together two molecules of ITM2A. Thus, we tested if ITM2A is competent to form self-interactions, as such could be permissive for the formation of higher order magnitude complexes with PTCH1 dimers. As shown in Figure 1D, ITM2A forms strong interactions with itself, supporting a stoichiometric model of 2 PTCH1: 2 ITM2A molecules in the complex.

ITM2A Does Not Mediate PTCH1 Autophagic Flux Inhibition
Since ITM2A was also reported to reduce autophagic flux via interaction with the vacuolar vATPase [12] and we confirmed that ITM2A is a bona fide PTCH1 CTD interactor, we hypothesised that ITM2A could be a mediator of non-canonical PTCH1 signalling in autophagy. Transient transfection of PTCH1 or ITM2A alone increased the levels of LC3BII, the lipidated form of LC3B, a marker of autophagosome formation. However, addition of Bafilomycin A1 during the last 4 h did not increase LC3BII in cells expressing PTCH1 or ITM2A, unlike in cells transfected with empty vector, suggesting that it is the result of LC3BII accumulation as a consequence of an autophagic flux blockade at the last step of the autophagosome-lysosome fusion (Figure 2A,B). Co-expression of PTCH1 and ITM2A had a very similar effect (Figure 2A,B), although we observed a reproducible reduction of PTCH1 expression level when co-expressed with ITM2A (Figure 2A,D), which was not associated with changes in PTCH1 ubiquitylation ( Figure S2). Densitometry-based analysis of the magnitude of the Bafilomycin A1 effect revealed that autophagic flux was reduced by more than 80% in all groups ( Figure 2C).

ITM2A Does Not Mediate PTCH1 Autophagic Flux Inhibition
Since ITM2A was also reported to reduce autophagic flux via interaction with the vacuolar vATPase [12] and we confirmed that ITM2A is a bona fide PTCH1 CTD interactor, we hypothesised that ITM2A could be a mediator of non-canonical PTCH1 signalling in autophagy. Transient transfection of PTCH1 or ITM2A alone increased the levels of LC3BII, the lipidated form of LC3B, a marker of autophagosome formation. However, addition of Bafilomycin A1 during the last 4 h did not increase LC3BII in cells expressing PTCH1 or ITM2A, unlike in cells transfected with empty vector, suggesting that it is the result of LC3BII accumulation as a consequence of an autophagic flux blockade at the last step of the autophagosome-lysosome fusion (Figure 2A,B). Co-expression of PTCH1 and ITM2A had a very similar effect (Figure 2A,B), although we observed a reproducible reduction of PTCH1 expression level when co-expressed with ITM2A (Figure 2A,D), which was not associated with changes in PTCH1 ubiquitylation ( Figure S2). Densitometrybased analysis of the magnitude of the Bafilomycin A1 effect revealed that autophagic flux was reduced by more than 80% in all groups ( Figure 2C).  Since the lack of an additive effect of PTCH1 and ITM2A could be the result of maximal inhibition of autophagic flux by overexpression of either of the two proteins, we decided to study the effect of silencing endogenous ITM2A using shRNA. HEK293 cells were transfected with shITM2A and selected with puromycin to generate multiclonal cells; however, ITM2A expression was restored after several passages (data not shown). This might be because as a heterogeneous population, cells expressing higher levels of ITM2A had a growth advantage and their proportion increased during passaging. Given this problem, we isolated shITM2A single cell-derived clones using the limiting dilution method. Three different shRNA sequences targeting ITM2A and a control sequence (shScrambled) were used and clones with the highest knockdown efficiency that was maintained for several passages were used for the following experiments. Clone shITM2A-C maintained a minimum of 65% knockdown ( Figure 3A).
Since the lack of an additive effect of PTCH1 and ITM2A could be the result of maximal inhibition of autophagic flux by overexpression of either of the two proteins, we decided to study the effect of silencing endogenous ITM2A using shRNA. HEK293 cells were transfected with shITM2A and selected with puromycin to generate multiclonal cells; however, ITM2A expression was restored after several passages (data not shown). This might be because as a heterogeneous population, cells expressing higher levels of ITM2A had a growth advantage and their proportion increased during passaging. Given this problem, we isolated shITM2A single cell-derived clones using the limiting dilution method. Three different shRNA sequences targeting ITM2A and a control sequence (shScrambled) were used and clones with the highest knockdown efficiency that was maintained for several passages were used for the following experiments. Clone shITM2A-C maintained a minimum of 65% knockdown ( Figure 3A). Transient expression of PTCH1 in shITM2A-A HEK293 cells showed a small but significant increase in LC3BII and p62 (an autophagy scaffolding protein that is degraded along with the cargo) accumulation by PTCH1 compared to control shScramble cells ( Figure 3B,C) and a larger inhibition of autophagic flux ( Figure 3D). The enhancing effect of ITM2A depletion on PTCH1-dependent autophagy regulation was also observed in HeLa cells (Supplementary Figure S1A,B). Interestingly, silencing of ITM2A resulted in a small increase of PTCH1 expression levels (Figure 2A and Supplementary Figure S1A). Altogether, these findings suggest that ITM2A is not necessary for autophagy regulation by PTCH1, and that it might actually reduce non-canonical PTCH1 signalling by modulating its turnover.

Increase in ITM2A Levels Accompanies Reduction of Autophagic Flux during Skeletal Muscle Differentiation
Given the involvement of Hh signalling and ITM2A in myotube differentiation of the C2C12 myoblastic cell line, and that ITM2A has been reported to be a marker of skeletal muscle differentiation, we investigated their potential interplay during in vitro differentiation of C2C12 myoblasts into myotubes ( Figure 4A). As C2C12 cells differentiate at high density, they upregulate myosin expression from day 3 onwards ( Figure 4B). As previously reported [15], ITM2A levels increase significantly during differentiation ( Figure 4B). To investigate the regulation of autophagy by endogenous ITM2A and endogenous PTCH1, we first investigated if there were any changes in basal autophagic flux during the 7 days of differentiation of C2C12 cells. The level of p62 and LC3BII increased during differentiation, becoming gradually more insensitive to Bafilomycin A1 treatment ( Figure 4C,D). This observation suggests that autophagic flux is reduced in differentiated myotubes compared to proliferating myoblasts. nificant increase in LC3BII and p62 (an autophagy scaffolding protein that is degraded along with the cargo) accumulation by PTCH1 compared to control shScramble cells (Figure 3B,C) and a larger inhibition of autophagic flux ( Figure 3D). The enhancing effect of ITM2A depletion on PTCH1-dependent autophagy regulation was also observed in HeLa cells (Supplementary Figure S1A,B). Interestingly, silencing of ITM2A resulted in a small increase of PTCH1 expression levels (Figure 2A and Supplementary Figure S1A). Altogether, these findings suggest that ITM2A is not necessary for autophagy regulation by PTCH1, and that it might actually reduce non-canonical PTCH1 signalling by modulating its turnover.

Increase in ITM2A Levels Accompanies Reduction of Autophagic Flux during Skeletal Muscle Differentiation
Given the involvement of Hh signalling and ITM2A in myotube differentiation of the C2C12 myoblastic cell line, and that ITM2A has been reported to be a marker of skeletal muscle differentiation, we investigated their potential interplay during in vitro differentiation of C2C12 myoblasts into myotubes ( Figure 4A). As C2C12 cells differentiate at high density, they upregulate myosin expression from day 3 onwards ( Figure 4B). As previously reported [15], ITM2A levels increase significantly during differentiation ( Figure 4B). To investigate the regulation of autophagy by endogenous ITM2A and endogenous PTCH1, we first investigated if there were any changes in basal autophagic flux during the 7 days of differentiation of C2C12 cells. The level of p62 and LC3BII increased during differentiation, becoming gradually more insensitive to Bafilomycin A1 treatment ( Figure  4C,D). This observation suggests that autophagic flux is reduced in differentiated myotubes compared to proliferating myoblasts. We next generated stable C2C12 myoblastic lines with reduced ITM2A expression using three different shRNA sequences. Puromycin-resistant shITM2A cells were analysed for knockdown of ITM2A. Cells stably expressing the shITM2A-B and -C sequences showed 21 and 55% reduction of ITM2A by qPCR, respectively, while shITM2A-A was ineffective ( Figure 5B). We continued our study with the shITM2A-C cells, which showed the best knockdown efficiency. C2C12 shITM2A-C cells presented a morphological delay in my-otube differentiation ( Figure 5A), which was confirmed by delayed myosin upregulation by Western blot, compared to parental and control shScramble C2C12 cells ( Figure 5A). Analysis of key cell cycle regulators in shITM2A-C C2C12 cells showed an early increase of p21 at day 3 and a lack of or delayed upregulation of CDK2, CDK4, CDK6, cyclin D1 and cyclin D3 compared to control shCONTROL cells ( Figure 5C). We next generated stable C2C12 myoblastic lines with reduced ITM2A expression using three different shRNA sequences. Puromycin-resistant shITM2A cells were analysed for knockdown of ITM2A. Cells stably expressing the shITM2A-B and -C sequences showed 21 and 55% reduction of ITM2A by qPCR, respectively, while shITM2A-A was ineffective ( Figure 5B). We continued our study with the shITM2A-C cells, which showed the best knockdown efficiency. C2C12 shITM2A-C cells presented a morphological delay in myotube differentiation ( Figure 5A), which was confirmed by delayed myosin upregulation by Western blot, compared to parental and control shScramble C2C12 cells ( Figure  5A). Analysis of key cell cycle regulators in shITM2A-C C2C12 cells showed an early increase of p21 at day 3 and a lack of or delayed upregulation of CDK2, CDK4, CDK6, cyclin D1 and cyclin D3 compared to control shCONTROL cells ( Figure 5C).

ITM2A Is as a Negative Regulator of Canonical Hh Signalling
Given that we observed a regulatory effect of ITM2A on PTCH1 expression level and non-canonical signalling in autophagy in epithelial cell lines, we next sought to investigate if ITM2A also affects canonical Hh signalling using a well-established GLI-luciferase activity assay in the Hh-responsive NIH 3T3 cells. Remarkably, the overexpression of ITM2A inhibited Gli-luciferase activation by Shh, an oncogenic SMO mutant (SMO-M2), Gli1 and Gli2 ( Figure 6A). The same effect was observed in Ptch1 −/− mouse embryonic fibroblasts (MEFs), which display high constitutive canonical Hh signalling activity due to loss of endogenous Ptch1, when overexpressing mouse Gli1 or Gli2 ( Figure 6B). Furthermore, ITM2A strongly reduced myc-GLI1 levels when expressed in NIH3T3 cells ( Figure 6C), or in HEK293 cells ( Figure 6D), a non-ciliated cell type in which GLI-dependent transcription can be stimulated in response to GLI1 overexpression. These findings suggest that ITM2A acts as a negative regulator of Gli stability and transcriptional activity, independently of its interaction with PTCH1.
non-canonical signalling in autophagy in epithelial cell lines, we next sought to investigate if ITM2A also affects canonical Hh signalling using a well-established GLI-luciferase activity assay in the Hh-responsive NIH 3T3 cells. Remarkably, the overexpression of ITM2A inhibited Gli-luciferase activation by Shh, an oncogenic SMO mutant (SMO-M2), Gli1 and Gli2 ( Figure 6A). The same effect was observed in Ptch1 −/− mouse embryonic fibroblasts (MEFs), which display high constitutive canonical Hh signalling activity due to loss of endogenous Ptch1, when overexpressing mouse Gli1 or Gli2 ( Figure 6B). Furthermore, ITM2A strongly reduced myc-GLI1 levels when expressed in NIH3T3 cells ( Figure  6C), or in HEK293 cells ( Figure 6D), a non-ciliated cell type in which GLI-dependent transcription can be stimulated in response to GLI1 overexpression. These findings suggest that ITM2A acts as a negative regulator of Gli stability and transcriptional activity, independently of its interaction with PTCH1.

Identification of ITM2A-Interacting Proteins by Mass Spectrometry
In order to better understand the role of ITM2A in Hh signalling and skeletal muscle differentiation, we overexpressed a tagged ITM2A protein in HEK293 cells and performed immunoprecipitation followed by mass spectrometry to identify potential interacting proteins. Cells expressing empty vector were used as a negative control to subtract non-specific contaminating proteins. ITM2A immunoprecipitates contained a number of proteins listed in Table 4.
STRING analysis reveals a highly significant clustering (p = 1 × 10 −16 ) of over 90% of the proteins identified (Figure 7), with the existence of sub-clusters related to ER protein quality control, mitochondria membrane protein translocation, nuclear import, and stress granules. ITM2A interaction with NUP93 and CSE1L could be involved in negative regulation of Gli-dependent transcription by the impairment of GLI1 and GLI2 nuclear import. Some other proteins identified play important roles in the recognition of defective mitochondria for clearance by mitophagy (prohibitin and prohibitin-2, ADACT3, HSPA1B). This snapshot of the ITM2A interactome suggests that it may function at ER-mitochondria contacts to regulate mitophagy, which could contribute to mitochondrial network remodeling during myogenic differentiation.

Discussion
In this study, we present evidence that the single-pass membrane protein ITM2A is a negative regulator of canonical and non-canonical Hh signalling. On the one hand, ITM2A physically interacts with PTCH1 and reduces its stability and biological function as an endogenous inhibitor of autophagy. On the other hand, ITM2A reduces GLI1 stability and inhibits GLI-transcriptional activity at a step downstream of PTCH1, impairing

Discussion
In this study, we present evidence that the single-pass membrane protein ITM2A is a negative regulator of canonical and non-canonical Hh signalling. On the one hand, ITM2A physically interacts with PTCH1 and reduces its stability and biological function as an endogenous inhibitor of autophagy. On the other hand, ITM2A reduces GLI1 stability and inhibits GLI-transcriptional activity at a step downstream of PTCH1, impairing canonical Hh signalling. Our findings refute our original hypothesis that ITM2A could be a mediator of PTCH1-induced autophagic flux reduction, suggesting instead that ITM2A's effect on autophagy are independent of PTCH1. The identification of ITM2A-interacting proteins by mass spectrometry provides the basis for testable hypotheses of the mechanism by which ITM2A acts as a negative regulator of Hh signalling and how it regulates myogenic differentiation, which will be explored in the future.
In the first part of our study, we confirm that ITM2A interacts with the CTD of PTCH1, and at least with another domain, since a PTCH1 mutant with deletions of the CTD and cytosolic middle loop, was able to co-immunoprecipitate ITM2A. This suggests that ITM2A likely interacts with PTCH1 through its intracellular 53 aa domain and the extracellular BRICHOS domain.
ITM2A and PTCH1 exert the same autophagy blocking phenotype characterised by an accumulation of autophagosomes [11,12]. However, our findings demonstrate that their interaction is not necessary for the autophagic flux blockade induced by PTCH1 in HEK293 or HeLa cells. In agreement with the lack of cooperative effect, the expression of ITM2A causes a direct or indirect degradation of PTCH1, while silencing of endogenous ITM2A results in a concomitant increase in PTCH1 levels. Even when the change in the expression of PTCH1 in the ITM2A knockdown cells was not statistically significant, the slight increase in PTCH1 expression was enough to enhance its inhibitory effect on autophagy. It is possible that the effects on the knockdown are not as marked as the effects of the overexpression because the endogenous levels of ITM2A in HEK293 and HeLa are low, and also because the efficiency of the knockdown was not 100%. This is supported by the observation that, when comparing the effects in the two cells lines, the results are statistically significant in HeLa cells, which presented a higher knockdown efficiency. In summary, our study suggests that ITM2A is not a necessary mediator of the autophagic flux blockade regulated by PTCH1, but instead reduces its activity, while still maintaining its capacity to block autophagy independently of their interaction.
While the effect of ITM2A over PTCH1 protein levels suggested that it could stimulate canonical Hh signalling, GLI-luciferase experiments show a clear inhibitory effect when the pathway is activated by upregulation of ligand (Shh), by an oncogenic SMO mutant, or by the overexpression of the main GLI family transcriptional activators, GLI1 and GLI2. This indicates that ITM2A exerts an inhibitory effect at the level of the GLI transcription factors. While the inhibitory mechanism is unknown, we speculate that interaction of ITM2A with NUP93 (a nucleoporin) and CSE1L (exportin) might regulate GLI2 ciliary trafficking and/or nuclear accumulation. NUP93 has been shown to localise at the ciliary base and regulate the permeability barrier [23]. CSE1L mediates the re-export of importin-α from the nucleus after importin substrates, which include GLI2, are released in the nucleoplasm [24,25].
The negative regulation of ITM2A on Hh signalling could play an important role during myogenic differentiation. It is well known that canonical Shh signalling maintains myoblasts and satellite cells in the proliferative state, when ITM2A levels are lower. Induction of myogenic terminal differentiation is accompanied by upregulation of ITM2A and a simultaneous decrease in autophagic flux. While our interactome analysis in HEK293 cannot be directly extrapolated to other cell types, it showed a subset of proteins that participate in the recognition of defective mitochondria for clearance by mitophagy (prohibitin and prohibitin-2, ADACT3, HSPA1B). This selective type of autophagy has been shown to play an important role during mitochondrial network remodelling in myogenic differentiation and could explain the regulatory role of ITM2A in C2C12 cells.
The process of myogenic differentiation requires the induction of muscle-specific genes, such as myosin, and irreversible cell cycle withdrawal [26,27]. Previous studies showed that the upregulation of cyclin D3, as opposed to cyclin D1 and cyclin D2, plays a key role in cell cycle withdrawal during myogenic differentiation [28,29]. Our results show that knockdown of ITM2A expression in C2C12 cells results in a delay in differentiation together with reduced levels of cyclin D3 expression. Therefore, it is possible that ITM2A regulates a process needed for upregulation of cyclin D3 during differentiation of C2C12 cells. Furthermore, ITM2A could regulate cell cycle progression through interaction with CDKN2A (also known as p16 INK4 ), as suggested by our mass spectrometry analysis. Given the positive role of ITM2A in C2C12 differentiation, it is possible that it can stimulate the function of this CDK4/6 inhibitor to accelerate cell cycle exit. Future studies will investigate if ITM2A is involved in stabilization of the main complexes involved in the maintenance of the cell cycle withdrawal required for the differentiation of myoblasts into myotubes. The potential physical interaction with CDKN2 suggested by the proteomics analysis is an attractive starting point to further investigate this process.
In conclusion, the results presented here support a negative role of ITM2A on Hh signalling and confirm its requirement during terminal differentiation of skeletal muscle.

Institutional Review Board Statement: Not applicable.
Informed Consent Statement: Not applicable.

Data Availability Statement:
All data related to this study is presented in this manuscript; however, any reasonable request can be directed to N.A.R.-D.G.