Arih2 regulates Hedgehog signaling through smoothened ubiquitylation and ER-associated degradation

The Hedgehog pathway is an evolutionarily conserved signaling cascade that functions in embryonic development and tissue homeostasis. Malfunction of the pathway causes a variety of developmental syndromes and cancers. In vertebrates, this pathway is mediated by cilia, hair-like organelles found on nearly all eukaryotic cells. In brief, in the basal state without sonic hedgehog (SHH) ligand present, the SHH receptor Ptch1 localizes to cilia and inhibits the ciliary accumulation and activation of Smo. Upon binding of the Hedgehog ligand, Ptch1 stops inhibiting Smo, and Ptch1 exits cilia. Uninhibited Smo accumulates in cilia, becomes activated and promotes the activation of the Gli transcription factors, which move to the nucleus to promote gene expression. Our previous work showed that ciliary Smo levels are regulated by ubiquitylation. At the basal state, the ubiquitin E3 ligase Wwp1 localizes to cilia by binding Ptch1. This promotes the ubiquitylation of Smo, which promotes the interaction of Smo with the intraflagellar transport (IFT) system for removal from cilia (Desai et al., 2020; Lv et al., 2021).

The post-translational modification of proteins by ubiquitylation plays pivotal roles in a wide range of signaling processes (Otten et al., 2021). The covalent attachment of the ubiquitin peptide to a target can change the stability, localization, trafficking, activity and protein–protein interactions of the target. Ubiquitylation requires the activation of ubiquitin by E1 ubiquitin-activating enzymes, the transfer of the ubiquitin to an E2 ubiquitin-conjugating enzyme followed by the ligation of the ubiquitin onto the target protein by an E3 ubiquitin ligase or a complex of E2 and E3 enzymes. There are two E1 activating enzymes, ∼40 E2 conjugating enzymes and more than 600 E3 ligases encoded in the human genome.

In this work, we examine the role of Arih2 in regulating Smo. Our prior work showed that loss of Arih2 elevated ciliary Smo levels at the basal state and increased the level of Smo in cells (Lv et al., 2021). Arih2, also known as TRIAD1 is a RING-in-between-RING (RBR) type E3 ubiquitin ligase, has been mostly studied in the context of cancer where it has anti-proliferative effects on myeloid progenitor cells and its expression is reduced in acute myeloid leukemia (Marteijn et al., 2005; Wang et al., 2011). The gene is widely distributed across metazoans. It is found in organisms like mouse and humans, which utilize ciliary Hedgehog signaling, Drosophila melanogaster, which has non-ciliary Hedgehog signaling, and organisms like Caenorhabditis elegans and Arabidopsis thaliana, which do not have Hedgehog signaling. This distribution suggests that Arih2 regulates other pathways besides Hedgehog (Aguilera et al., 2000; Marín and Ferrús, 2002; Mladek et al., 2003). Most mice lacking Arih2 die perinatally and the few that survive past the first week of life are runty and show signs of excessive inflammation (Lin et al., 2013). The authors of that work did not report phenotypes associated with defective Hedgehog signaling; however, unpublished work from the International Mouse Phenotyping Consortium documents structural birth defects of the heart, kidney, and skin in Arih2 heterozygotes that might be Hedgehog related (https://www.mousephenotype.org/data/genes/MGI:1344361). Vertebrate Arih2 undergoes extensive alternative splicing, producing nine protein variants in human and two in mouse. We find that in mouse, the isoforms are differentially localized, with Arih2α localized to the nucleus and cytoplasm and Arih2β localized to the endoplasmic reticulum (ER). Only Arih2β rescues the Smo phenotypes in Arih2^−/− cells suggesting the Arih2β controls the biosynthesis of Smo possibly through ER-associated degradation (ERAD).

In a CRISPR-based screen to identify ubiquitin-related genes regulating Hedgehog signaling, we identified the E3 ligase Arih2 as a negative regulator of the pathway, whose loss increased cellular Smo levels and increased Smo ciliary localization (Lv et al., 2021). In our current work, we seek to understand how Arih2 regulates the Hedgehog pathway. To ensure that the effect of Arih2 loss was not limited to our reporter, we assessed the activity of the endogenous pathway by measuring Gli1 gene expression (Fig. 1A), which is elevated by pathway activation (Lee et al., 1997). This analysis reproduced our finding that the loss of Arih2 elevated basal, but not SHH-induced Hedgehog signaling (Lv et al., 2021). As previously shown, loss of Arih2 elevated ciliary Smo levels at the basal state (Fig. 1B,C) and increased total cellular Smo levels as detected by western blotting (Fig. 1Da). Loss of Arih2 did not affect ciliogenesis (Fig. S1). The excess Smo appeared to be largely in intracellular pools, as the amount exposed to the surface in Arih2 mutant cells was similar to the amount that was surface exposed in control cells stimulated with SHH (Fig. 1D). Smo mRNA levels were similar in control and Arih2^−/− cells, suggesting that the increased Smo results from post-transcriptional mechanisms (Fig. 1E).

NCBI Gene (https://www.ncbi.nlm.nih.gov/gene) describes four mouse Arih2 transcript variants that differ at the N-terminal coding region (Fig. 1F). There are also two Arih2 pseudogenes in the mouse genome, but these are highly mutated and do not appear to express proteins (Fig. 1F). The four transcripts encode isoforms of 492 and 421 residues, which we named Arih2α and Arih2β (Fig. 1F). mRNA representing both isoforms is readily detected in fibroblasts (Fig. 1G) although Arih2α mRNA levels appear to be ∼10-fold higher than Arih2β (Fig. 1H). Arih2α and Arih2β are similar with shared ring, ring between ring, and Ariadne domains, but differ at their N-termini. The N-terminus of Arih2α carries a cullin-5-binding site that is missing from Arih2β, whereas the N-terminus of Arih2β is predicted to encode a signal peptide with a low probability cleavage site (at the alanine residue in position 19) (Fig. 1I) (Kelsall et al., 2013). Even though Arih2α is expressed at higher levels than Arih2β, measuring the levels of a Flag-Avi-tag knocked into the endogenous Arih2 gene just prior to the stop codon suggests that both isoforms are present in the cell at approximately equal levels (Fig. 1J,K).

To ensure that the phenotype in the Arih2 mutant cells was due to the loss of Arih2, we rescued Arih2 mutant cells with constructs that express Myc-tagged Arih2α or Arih2β, and Myc-tagged Arih2α^C309A or Arih2β^C238A in which the active sites are mutated (Fig. 2A). Interestingly expression of Arih2β returned ciliary Smo (Fig. 2B,C) and total Smo (Fig. 2D) levels back to normal, whereas expression of Arih2α was not effective at doing this. Arih2β^C238A was not functional indicating that the E3 ubiquitin ligase activity of Arih2β is required for rescue (Fig. 2B–D).

To determine whether Arih2 is capable of ubiquitylating Smo, we transfected HEK293 cells with Smo–Flag and Ty1–ubiquitin along with Arih1, Arih2α, Arih2α^C309A, Arih2β or Arih2β^C238A. Immunoprecipitating Smo–Flag and measuring the incorporation of Ty1–ubiquitin indicated that cells expressing Arih2β incorporated more ubiquitin onto Smo than cells expressing the Arih2β active site mutation or the Arih2α and Arih1 isoforms (Fig. 2H).

To explore physical interactions between Smo and Arih2, we used the recently developed fluorescent protein–protein interaction (Fluoppi) approach (Watanabe et al., 2017). In this method, one protein is tagged with an Azami-Green (AG) fluorescent tag and the other protein is tagged with the PB1 homodimerization domain from the p62 autophagy protein (Fig. 2E). If the bait and prey proteins do not interact, the AG fluorescence will reflect the distribution of its fusion partner. If the bait and prey interact, the AG fusion protein will be drawn to the PB1 clusters, and the fluorescence will appear as puncta in the cytoplasm. Arih2α–AG expressed alone or co-expressed with Smo–PB1 showed the same localization pattern in the cytoplasm and nucleus indicating no interaction between Arih2α and Smo (Fig. 2F). Arih2β–AG expressed alone showed cytoplasmic localization, which was redistributed to strong puncta when co-expressed with Smo–PB1, indicating a physical interaction (Fig. 2F). Deletion of the Ariadne domain of Arih2β or the C-terminal tail of Smo abolished the interaction, whereas the interaction was maintained when the Triad domain was deleted (Fig. 2G).

Previously, we reported that Arih2 was localized predominately in the nucleus with a lower amount in the cytoplasm and none detected in the cilium (Lv et al., 2021). However, this work only examined the 492-residue Arih2α isoform. Since Arih2β is relevant to Hedgehog signaling, we repeated this work and examined both isoforms. We could not detect ciliary localization with either isoform (Fig. 3A). As previously shown, Arih2α localized predominately to the nucleus with some cytoplasmic localization in mouse embryonic fibroblasts (MEFs) (Fig. 3A,B), which is similar to the distribution of its paralog Arih1 (Fig. S2A). Arih2α in the cytoplasm did not appear to associate with any cytoskeletal or vesicular structures. In contrast to the prominent nuclear localization of Arih2α, Arih2β localizes to the cytoplasm with no localization in the nucleus (Fig. 3A,B). Within the cytoplasm, Arih2β was concentrated near the nucleus where it associated with tubular and vesicular structures, and colocalized with cell body Smo (Fig. 3B). A similar localization of both isoforms was seen in IMCD3, hTERT RPE-1, HEK 293T and NIH/3T3 cells (Fig. S2B). No association was seen between Arih2β and endosomes or the Golgi complex (Fig. S2C,D), and the localization of Arih2β was not affected by brefeldin A (BFA) treatment (Fig. S2D). However, extensive colocalization was observed between Arih2β and an ER-targeted GFP construct and the ER-localized ubiquitin E3 ligase Syvn1 (Fig. 3C) (Kaneko et al., 2002).

Arih2α and Arih2β differ at their N-termini. The N-terminus of Arih2β is critical for localizing it to the ER, as a form with a C-terminal GFP fusion is retained at the ER, whereas that with an N-terminal GFP fusion is dispersed throughout the cell (Fig. 3D). To identify the ER localization signal of Arih2β, we generated a series of deletions where the N-terminal helical domains were progressively removed starting at the RING1 domain and moving back toward the N-terminus. Immunofluorescence results showed that the first 22 residues, which are evolutionally conserved in mammals, are necessary for the ER localization of Arih2β (Fig. 3E). Inserting GFP after the 22nd residue did not disrupt ER localization. However, GFP fusions that only carried the first 22 Arih2β residues are not as highly enriched at the ER, suggesting that other portions of Arih2β also contribute to ER localization (Fig. 3F). Fluoppi analysis showed that the Ariadne domain is needed for interaction with Smo (Fig. 2G), indicating that Smo binding might enhance ER enrichment of Arih2β.

The signal sequence at the N-terminus of Arih2β could function to tether Arih2β to the outer surface of the ER or it could direct the translocation of Arih2β into the lumen of the ER. There is currently no evidence for ubiquitylation activity in the lumen of the ER, suggesting that the signal sequence is more likely to function to tether the protein to the cytoplasmic surface of the ER. However, to distinguish these possibilities, we made Arih2β fusions to the Ca²⁺-measuring organelle-entrapped protein indicators (CEPIA) (Suzuki et al., 2014) (Fig. 4A,B). CEPIA is a Ca²⁺indicator that is fluorescent when exposed to high Ca²⁺ levels as found in the ER lumen but is non-fluorescent in the lower Ca²⁺ environment of the cytoplasm (Fig. 4C). Targeting CEPIA to the ER lumen by fusion with a modified ER signal sequence from the mouse immunoglobulin heavy chain variable region yielded a high fluorescent signal. However, fusion of CEPIA to C-terminus of Arih2β or inserting it after the signal sequence at the N-terminus correctly localized the indicator to the ER but produced little signal (Fig. 4C). The lack of signal indicates that the CEPIA domain in these constructs is in the cytoplasm supporting a model where the signal sequence of Arih2β tethers the protein onto the cytoplasmic face of the ER.

Prior to our finding that Arih2β localizes to the ER, more than 30 E3 ligases had been localized to this organelle where they are thought to function primarily in protein quality control, but they also regulate Ca²⁺ release and cholesterol biosynthesis (Rusilowicz-Jones et al., 2021). Our finding that cellular Smo levels are increased by the loss of Arih2 suggests that this E3 is regulating the destruction of excess or misfolded Smo. Transmembrane proteins in the ER typically undergo proteasomal degradation via ER-associated degradation (ERAD). During ERAD, cytoplasmic domains of misfolded membrane proteins become ubiquitylated, targeting them for extraction from the membrane by Vcp (also known as p97 and Cdc48) and degradation by the proteosome (Bodnar and Rapoport, 2017). Treatment of control cells with eeyarestatin I (EerI), which blocks ERAD by inhibiting Vcp (Wang et al., 2008), elevated levels of the ER form of Smo and promoted the production of high molecular mass forms of Smo (Fig. 5A; Fig. S3). These higher molecular mass forms are not seen in the parental cell line, indicating that they are not simply a cross reactive product (Fig. S4). It is likely that these larger forms are polyubiquitylated Smo forms, which under normal conditions would be degraded by ERAD but accumulate when ERAD is blocked. If Arih2 is responsible for targeting misfolded Smo for degradation, we would expect that the amount of ubiquitylated Smo that accumulates upon EerI treatment would be reduced in Arih2 mutants. Supporting this idea, blocking ERAD with EerI without blocking translation caused a greater accumulation of ubiquitylated Smo in control cells compared to in Arih2 mutant cells (Fig. 5A–C).

The ER accumulation of Smo and other potential Arih2β substrates in Arih2 mutants is likely to cause ER stress and the unfolded protein response. Consistent with this idea, protein aggregates as detected with thioflavin T (Beriault and Werstuck, 2013) were as abundant in untreated Arih2 mutant cells as they are in control cells treated with the ERAD inhibitor EerI (Fig. 5D). As a more direct test, we measured expression of unfolded protein response target genes (Oslowski and Urano, 2011). Changes were consistent with an increased unfolded protein response, with Atf4, Hspa5, Ddit3, Hsp90b1 and spliced Xbp1 increased, and unspliced Xbp1 decreased in Arih2 mutant cells (Fig. 5E).

Our findings show that Arih2 loss elevates basal expression of Hedgehog-responsive genes, increases the total cellular level of Smo and causes Smo to accumulate in cilia at the basal state. Arih2, also known as Triad1, is a ring-between-ring E3 ligase that functions with the E2 conjugating enzyme Ube2l3 to ubiquitylate substrates. Consistent with Ube2l3 being a functional E2 for Arih2, we found that Ube2l3 loss also elevated cellular levels of Smo (Lv et al., 2021). In mouse, Arih2 encodes two major isoforms that differ at their N-termini. Arih2α has a longer N-terminus that includes a cullin-5-binding site and localizes to the nucleus, whereas the N-terminus of Arih2β has a hydrophobic helix that anchors it to the cytoplasmic face of the ER. Expression of the ER-localized Arih2β isoform fully rescues the Smo phenotypes caused by loss of Arih2, whereas the nuclear-localized form does not. Most work on Arih2 focuses on the role of Arih2α as part of the cullin-5 complex (Hüttenhain et al., 2019; Kelsall et al., 2013; Kostrhon et al., 2021). Complete loss of Arih2 in mouse leads to embryonic lethality and increased inflammatory responses (Lin et al., 2013). Heterozygotes show a variety of structural birth defects in kidney, skin, bone, and heart (https://www.mousephenotype.org/data/genes/MGI:1344361; Bult et al., 2019; Dickinson et al., 2016), which are similar to phenotypes caused by cilia dysfunction.

The localization of Arih2β at the ER and our finding that loss of Arih2 elevates total Smo levels suggest that Arih2 regulates the production of Smo, possibly through a quality control mechanism (Fig. 5F). Misfolded proteins in the ER are targeted for degradation by the ERAD system. During ERAD, misfolded proteins are targeted to a dislocation complex in the ER membrane where the misfolded protein is polyubiquitylated on cytoplasmic lysine residues. The polyubiquitylated protein is removed from the membrane by the Vcp complex and sent to the proteosome for degradation. The major E3 ligases involved in ERAD are thought to be Syvn1 and Amfr. However, more than a dozen other E3s have been implicated in ERAD of specific substrates (Olzmann et al., 2013), and more than 30 E3s localize to the ER (Fenech et al., 2020; Rusilowicz-Jones et al., 2021). Our finding that loss of Arih2 reduces the incorporation of ubiquitin onto Smo when ERAD is blocked by the Vcp inhibitor EerI supports a role for Arih2β in ERAD. However, Arih2 is thought to catalyze only the initial mono-ubiquitylation (Hüttenhain et al., 2019; Kelsall et al., 2013), indicating that additional E3 ligases are needed to extend the chain. Syvn1 is an interesting possibility as its loss also elevates expression of Hedgehog-responsive genes, but we did not observe increased ciliary Smo levels in the knockouts (Lv et al., 2021).

The loss of Arih2 elevates the total level of Smo in the cell. The majority of the extra Smo is not exposed on the cell surface. However, exposed Smo is increased in the mutant to about the level seen in activated control cells. Smo is thought to constantly diffuse into cilia with regulated removal dictating the ciliary level (Milenkovic et al., 2009). Increased plasma membrane Smo would increase the amount that diffuses into the cilium. Overexpression of Smo is sufficient to saturate the retrieval process (Corbit et al., 2005), and so it is likely that this is the reason for elevated ciliary Smo in Arih2 mutants. However, ciliary localization of Smo is not sufficient to activate the pathway, raising the question of why the loss of Arih2 elevates basal expression of Hedgehog responsive genes. It is likely that Arih2 is involved in the production of Smo and is not directly involved in the Hedgehog signal transduction cascade. Activation of Smo is driven by phosphorylation and sterol binding (Deshpande et al., 2019; Jia et al., 2004; Nedelcu et al., 2013; Zhang et al., 2004) but point mutations, such as the SmoM2 W539L mutation, activate the pathway independent of upstream signals. This mutation is thought to shift helixes changing the protein into an active conformation (Huang et al., 2018). It is possible that some misfolded protein that escapes the ER in the absence of Arih2 is in an active conformation analogous to the SmoM2 state. Another possibility is that if the loss of Arih2 causes Smo to remain in the ER for an extended time, the high Ca²⁺ environment of the ER lumen might promote the esterification of Smo with cholesterol and drive it towards an active conformation (Hu et al., 2022).

In summary, our work identifies a new regulatory mechanism in the ER that controls the cellular levels of Smo. When this mechanism is defective, ciliary Smo levels are elevated and basal expression through the pathway is increased.

Plasmids were assembled by TEDA method (Xia et al., 2019) into the pHAGE lentiviral backbone (Wilson et al., 2008). All inserts are derived from mouse unless otherwise stated. Mutations were generated by PCR amplification with mutated primers and the corresponding amplicons were TEDA assembled. All inserts were fully sequenced and matched the Ensembl reference sequence, NCBI reference sequence or expected mutant forms. Plasmids are listed in Table S1 and SnapGene files will be provided upon request.

Wild-type mouse embryonic fibroblasts (MEFs) were derived from E14 embryos and immortalized with SV40 Large T antigen. These cells were cultured in 95% DMEM (4.5 g/l glucose), 5% fetal bovine serum (FBS), 100 U/ml penicillin, and 100 µg/ml streptomycin. IMCD3 cells (obtained from Jagesh Shah, Harvard Medical School, Boston, USA) were cultured in 42.5% DMEM (4.5 g/l glucose), 42.5% F12, 5% FBS, 100 U/ml penicillin and 100 µg/ml streptomycin. NIH/3T3 cells (obtained from Stephen Doxsey, UMass Medical School, Worcester, USA) and HEK 293T cells (obtained from Julie Jonassen, UMass Medical School, Worcester, USA) were cultured in 90% DMEM (4.5 g/l glucose), 10% FBS, 100 U/ml penicillin and 100 µg/ml streptomycin. hTert RPE-1 cells (obtained from Stephen Doxsey) were cultured in 90% DMEM (1 g/l glucose), 10% FBS, 100 U/ml penicillin and 100 µg/ml streptomycin (all from Gibco-Invitrogen). Cells were confirmed to be of mouse or human origin as expected and monitored for mycoplasma contamination by PCR (Tang et al., 2000) and DAPI staining.

For smoothened agonist (SAG) experiments, MEFs were plated at near-confluent densities and serum starved (same culture medium described above but with 0.25% FBS) for 24 h prior to treatment to allow ciliation. SAG (Calbiochem) was used at 400 nM.

Sonic hedgehog (SHH) conditioned medium was generated from HEK 293T cells expressing plasmids encoding HsSHH (BL243; Table S1), XtScube2 (BL244; Table S1) and MmDisp1 (BL323; Table S1). Cells stably secreting SHH were grown to confluency in 90% DMEM (4.5 g/l glucose), 10% FBS, 100 U/ml penicillin and 100 µg/ml streptomycin; the medium was then replaced with low-serum medium (0.25% FBS) and cells were cultured for a further 24 h. Medium was collected, filtered sterilized with 0.45 µm filter (Millipore) and titered for the ability to cause relocation of Smo to cilia. Dilutions similar in effect to 400 nM SAG were used for experiments.

The chemicals used in this study include protein transport from the ER to the Golgi complex inhibitor brefeldin A (BFA) (50 µg/ml), protein synthesis inhibitor cycloheximide (CHX) (150 μg/ml), the ERAD inhibitor eeyarestatin I (EerI) (50 μM), and the proteasomal degradation inhibitor MG132 (1 μM).

Lentiviral packaged pHAGE-derived plasmids (Wilson et al., 2008) were used for transfection. These vectors are packaged by a third-generation system comprising four distinct packaging vectors (Tat, Rev, Gag/Pol, VSV-g or MLV-env) using HEK 293T cells as the host. DNA (plasmid of interest, 5 µg; Tat, 0.5 µg; Rev, 0.5 µg; Gag/Pol, 0.5 µg; VSV-g/MLV-env, 1 µg) was delivered to the HEK 393T cells as calcium phosphate precipitates. After 48 h, the supernatant was harvested, filtered through a 0.45 µm filter (Millipore), and added to subconfluent cells. After 24 h, cells were selected with corresponding antibiotics [nourseothricin (Nat, 50 µg/ml), puromycin (Puro, 1 µg/ml), zeocin (Zeo, 500 µg/ml) or blasticidin (Bsd, 60 µg/ml)].

For flow sorting, pelleted cells (2000 g for 3 min) were resuspended in the corresponding media and sorted into tubes or 96-well plates containing medium with 10% fetal bovine serum by a BD FACS C-Aria II Cell Sorter (BSL-2+/BSC).

Guide RNAs were selected from the Brie library (Doench et al., 2016) or designed using CHOPCHOP (Labun et al., 2019). Corresponding oligonucleotides were cloned into lentiCRISPR v2 Puro (Addgene plasmid #52961, deposited by Feng Zhang; Sanjana et al., 2014) or lentiCRISPR v2 Puro^P93S (BL245) and screened by sequencing. lentiCRISPR v2 Puro^P93S is like its parent except for a proline to serine mutation in the puromycin N-acetyl-transferase gene, which increases its resistance to Puro (https://www.addgene.org). The vectors were packaged into lentiviral particles and transfected into MEF cells. After selection, the pools were analyzed by flow cytometry. Individual cells were sorted into 96-well plates by flow cytometry or dilution cloning. Single mutant clones were identified with Sanger sequencing, GENEWIZ Amplicon-EZ sequencing, immunofluorescence or immunoblotting. Sequencing results were analyzed with GEAR Indigo (https://www.gear-genomics.com), Poly peak parser and the SWS method (Hill et al., 2014; Jie et al., 2017; Rausch et al., 2020).

Flag-Avi knock-in was achieved by using CRISPR-Cas9 genome editing. Cas9 and sgRNA were expressed from lentiCRISPR v2 Puro (BL1139; Table S1). The template for homology-directed repair was designed in Benchling (https://www.benchling.com/). Each homology arm was about 600 bp. The main components of the donor sequence are left arm-3xFlag-Avi (containing a stop codon)-loxP-IRES2-Puro-loxP-right arm. The donor (BL1137; Table S1) and guide RNA vectors (BL1139) were transfected into cells with Qiagen Effectene Transfection Reagent according to the manufacturer's protocol. Transfected cells were drug-selected with Puro and then tested in population.

PCR products were analyzed by Amplicon-EZ paired-end sequencing (Azenta Genewiz). Using bwa-mem2 (https://github.com/bwa-mem2) with default parameters, paired-end sequencing reads from each replicate were first aligned to Arih2 pseudogenes (Gm12263 and Gm49867), which contain many substitutions compared to the parental Arih2 gene; unmapped reads were then extracted and mapped onto the enriched region of Arih2. Because the cassette exon is specifically spliced out in Arih2β but not Arih2α, sequencing read pairs were deemed to belong to Arih2β if they cover the cassette exon with an overhang of at least 5 nucleotides, otherwise they were considered as reads from Arih2α.

Cells were fixed with 2% paraformaldehyde for 15 min, permeabilized with 0.1% Triton-X-100 for 2 min and stained as described previously (Follit et al., 2006). In some cases, fixed cells were treated with 0.05% SDS for 5 min before prehybridization to retrieve antigens. The primary and secondary antibodies are described in Table S2. For the thioflavin T fluorescence assay, cells were incubated with 5 μM thioflavin T (Millipore Sigma) for 10 min before fixation (Beriault and Werstuck, 2013).

For live-cell imaging, cells were seeded in 35 mm glass bottom collagen-coated dishes (MatTek Corporation) and cultured for at least 24 h before imaging.

Confocal images were taken with a LSM910 equipped with a 63× objective and converted into a maximum projection with ZEN 3.1 blue edition (Zeiss).

For western blots, cells were pelleted (2000 g for 3 min) and lysed directly with denaturing gel loading buffer [Tris-HCl 125 mM pH 6.8, 20% (v/v) glycerol, 4% (v/v) SDS, 10% (v/v) β-mercaptoethanol, and 0.04% Bromophenol Blue]. The primary and secondary antibodies are described in Table S2. Western blots were developed by chemiluminescence (Super Signal West Dura, Pierce Thermo) and imaged using an Amersham Imager 600 imager (GE Healthcare Life Sciences). Bands were quantified with Gel-Pro Analyzer 4 (Meyer Instruments).

For immunoprecipitations, cells were serum starved for 48 h and proteins were extracted with lysis buffer (20 mM HEPES pH 7.5, 50 mM KCl, 1 mM MgCl₂) with 0.5% digitonin and protease inhibitor (cOmplete EDTA-Free, Roche). Insoluble components were removed by centrifugation at 20,000 g. Primary antibodies pre-adsorbed to protein G–Sepharose beads (GE Healthcare) were added to the cell extract and the mixture incubated for 2 h at 4°C. After centrifugation (200 g for 1 min), beads were washed with lysis buffer supplemented with 0.1% digitonin before elution in denaturing gel loading buffer for SDS-PAGE electrophoresis and western blotting analysis.

Isolation of mRNA and quantitative mRNA analysis was performed as previously described (Jonassen et al., 2008) using the primers tabulated in Table S3.

Cell surface proteins were biotinylated by a non-cell permeable EZ-Link Sulfo-NHS-SS-Biotin (Thermo Fisher Scientific) as described previously (Pusapati et al., 2018). Cells with surface proteins biotinylated were lysed with CelLytic M solution containing cOmplete, EDTA-free Protease Inhibitor Cocktail (Millipore Sigma). The lysate was clarified by centrifugation at 4°C, 18,000 g for 10 min. Biotinylated proteins were captured and purified from the supernatant by Pierce High Capacity NeutrAvidin agarose (Thermo Fisher Scientific). After washing six times with lysis buffer, the beads were extracted with denaturing gel loading buffer containing 100 mM DTT at 37°C for 1 h to release biotinylated proteins.

The Smo ubiquitylation assay was performed as described previously (Lv et al., 2021). In brief, HEK 293T cells were plated at 60% confluent density onto a 10 cm plate. After 24 h, the cells were transfected with 5 µg of each vector using calcium phosphate transfection. At 24 h after transfection, cells were treated with 50 μM Eeyarestatin I and 1 µM MG132 for 4 h to block ERAD and proteasomal degradation. Cells were lysed and Smo was captured with anti-Flag M2 affinity gel (Millipore Sigma) as described in the immunoprecipitation method in Protein and mRNA analysis section.

Statistical results were obtained from at least three independent experiments. Statistical differences between groups were tested by one-way ANOVA, two-way ANOVA, or repeated measures ANOVA with Sidak or Tukey post-hoc tests  in GraphPad Prism 7.04. Differences between groups were considered statistically significant if P<0.05. Otherwise, results were labeled as non-significant (n.s.). Statistical significance is denoted with asterisks (*P<0.05; **P<0.01; ***P<0.001, ****P<0.0001). Error bars indicate standard deviation (s.d.).

We thank Dr Carol E. Schrader and the staff of the University of Massachusetts Medical School Flow Cytometry Core for assistance during this project. Flow Cytometry Resources were supported by National Institutes of Health S10 1S10OD028576 to Carol E. Schrader. We thank Drs Vadim Arshavsky and William Spencer (Duke University) for comments on this work.

The peer review history> is available online at https://journals.biologists.com/jcs/lookup/doi/10.1242/jcs.260299.reviewer-comments.pdf.