Ubiquitin chains earmark GPCRs for BBSome-mediated removal from cilia

The regulated trafficking of signaling receptors in and out of cilia is a central regulatory mechanism of many cilia-based signaling pathways (Nachury and Mick, 2019; Anvarian et al., 2019; Mykytyn and Askwith, 2017; Gigante and Caspary, 2020). For example, upon activation of the Hedgehog signaling pathway, the Hedgehog receptor Patched 1 and the G protein–coupled receptor (GPCR) GPR161 disappear from cilia, while the GPCR Smoothened (SMO) accumulates in cilia (Rohatgi et al., 2007; Corbit et al., 2005; Mukhopadhyay et al., 2013). In all three cases, regulated exit from cilia plays a major role in the on-demand redistribution of signaling molecules (Nachury and Mick, 2019). Signal-dependent exit is likely to be a general characteristic of ciliary GPCRs as the somatostatin receptor 3 (SSTR3), the dopamine receptor 1 (D1R), the melanocortin concentrating hormone receptor 1 (MCHR1), and the neuropeptide receptor 2 (NPY2R) all disappear from cilia upon exposure to agonist (Domire et al., 2011; Loktev and Jackson, 2013; Nager et al., 2017; Green et al., 2015). A major question is how GPCRs are selected for removal from cilia in an activity-dependent manner.

The major trafficking entities mediating signal-dependent exit from cilia are the BBSome and β-arrestin 2. The BBSome is an evolutionarily conserved complex of eight Bardet–Biedl syndrome (BBS) proteins that directly recognizes intracellular determinants in GPCRs and ferries them out of cilia (Nachury, 2018; Wingfield et al., 2018). The BBSome associates with the intraflagellar transport (IFT) machinery and has been proposed to act as adaptor between the motor-driven IFT trains and the membrane proteins that are to be removed from cilia. Because the BBSome is not known to have the ability to discriminate active from inactive GPCRs, there must exist another layer of regulation that commits activated GPCRs for exit. β-Arrestin 2 is a well-established molecular sensor of the activation state of GPCRs that is required for the signal-dependent exit of GPR161 and SSTR3 (Pal et al., 2016; Green et al., 2015; Nager et al., 2017). No association of β-arrestin 2 with ciliary trafficking complexes has been reported to date, and it remains unclear how β-arrestin 2 relays information regarding the state of activation of ciliary GPCRs to the ciliary exit machinery.

An emerging player in ciliary exit is ubiquitination (Shearer and Saunders, 2016). Ubiquitin (Ub), a 76–amino acid polypeptide, becomes conjugated to acceptor lysine residues on substrate proteins by Ub ligases and tags substrates for degradation or other regulatory fates (Yau and Rape, 2016; Swatek and Komander, 2016). A role for Ub in promoting ciliary exit is suggested by multiple lines of evidence. First, interfering with Patched1 ubiquitination blocks its signal-dependent exit from cilia (Kim et al., 2015; Yue et al., 2014). Second, fusing Ub to PKD-2, the olfactory receptor ODR-10, or the transient receptor potential channel OSM-9 at their cytoplasmic end results in the disappearance of these proteins from cilia (Hu et al., 2007; Xu et al., 2015). The accumulation of these Ub fusions inside cilia when BBSome function is compromised suggests that the BBSome might sort ubiquitinated signaling receptors out of cilia, in line with a reported association of BBSome with ubiquitinated proteins in trypanosomes (Langousis et al., 2016). Interestingly, TRIM32 is a Ub ligase mutated in BBS patients (Chiang et al., 2006). Third, the Ub ligase Cbl is recruited to cilia upon activation of the ciliary tyrosine kinase receptor platelet-derived growth factor receptor αα (PDGFRαα), and Cbl is required for termination of PDGFRαα signaling (Schmid et al., 2018). Last, a dramatic rise in ubiquitination of the ciliary proteome is observed when Chlamydomonasrheinardtii cilia disassemble (Huang et al., 2009). In particular, α-tubulin becomes ubiquitinated on K304 upon cilia disassembly, and expression of α-tubulin[K304R] slows down cilia disassembly (Wang et al., 2019).

After conjugation onto substrates, Ub itself often serves as a substrate for ubiquitination. The resulting Ub chains are characterized by the acceptor lysine residue on Ub with lysine 48–linked Ub (UbK48) chains targeting soluble proteins for proteasomal degradation, while lysine 63–linked Ub (UbK63) chains assembled onto membrane proteins constitute the major driver for sorting from the limiting membrane of late endosomes into intralumenal vesicles and ultimately lysosomal degradation (Piper et al., 2014; Yau and Rape, 2016; Swatek and Komander, 2016). Intriguingly, in shortening C. rheinardtii cilia, UbK63 linkages predominate at least fivefold over K48 linkages, and only UbK63 linkages are detected on α-tubulin. A role for UbK63 chains in ciliary trafficking remains to be determined.

Together, these data led us to investigate the interplay between ubiquitination, β-arrestin 2, and the ciliary exit machinery. Using a combination of cellular imaging, biochemical studies, and cilia-specific manipulations, we find that β-arrestin 2 directs the activation-dependent addition of Ub chains built at least in part with K63 linkages onto ciliary GPCRs that are then selected by the BBSome for removal from cilia.

As a first step in investigating the interplay between ubiquitination and the ciliary exit machinery, we sought to determine whether activation of ciliary GPCRs leads to their ubiquitination inside cilia before retrieval into the cell. We blocked BBSome-dependent exit by deleting its cognate GTPase ARL6/BBS3 in mouse inner medulla collecting duct (IMCD3) cells. IMCD3 cells represent a validated cell culture system to study ciliary trafficking, and the IMCD3 Arl6^−/− line has been previously characterized (Liew et al., 2014; Ye et al., 2018). As in previous studies (Ye et al., 2018; Nager et al., 2017), all GPCRs were expressed at near-endogenous levels by driving expression with extremely weak promoters. The fluorescent protein mNeonGreen (NG) fused to the cytoplasm-facing C terminus of GPCRs allowed direct visualization. A biotinylation acceptor peptide (AP) fused to the extracellular N terminus biotinylated by an ER-localized biotin ligase (BirA-ER) enabled pulse labeling of surface-exposed molecules and denaturing purifications. When SSTR3 was expressed in Arl6^−/− IMCD3 cells, addition of its agonist somatostatin (sst) led to a drastic increase in the ciliary levels of Ub observed via immunostaining with the well-characterized FK2 (Fig. 1, A and B) and FK1 (Fig. S1, A and B) monoclonal antibodies (Fujimuro et al., 1994; Emmerich and Cohen, 2015; Haglund et al., 2003). All further experiments were conducted with the FK2 antibody. Arl6^−/− cells accumulated Ub signals inside cilia in the absence of SSTR3 expression (Fig. 1, A and B), demonstrating that some endogenous proteins become ubiquitinated inside the cilia of Arl6^−/− IMCD3 cells. No Ub signal was detected inside cilia when BBSome function was intact (Fig. 1, A and B; and Fig. S1, A–D), indicating that the BBSome efficiently removes ubiquitinated proteins from cilia. Because the sst-dependent increase in ciliary Ub signal depends upon SSTR3 expression (Fig. 1, A and B), these results suggest that either SSTR3 itself or a downstream effector of SSTR3 become ubiquitinated inside cilia upon activation.

To determine whether SSTR3 becomes ubiquitinated in response to sst, we biochemically isolated ^APSSTR3^NG under denaturing conditions via the biotinylated AP tag and probed for Ub via an HA tag on transfected HA-Ub. SSTR3 migrated as a broad band—as typical for a glycosylated protein—centered around 100 kD (Fig. 1 C). The amount of SSTR3 recovered did not change appreciably between WT and Arl6^−/− cells or upon stimulation with sst (Fig. 1 C). Remarkably, while only very faint Ub signals were detected associated with SSTR3 in WT cells, a Ub smear extending from 100 kD upward was detected in the SSTR3 pull-downs from Arl6^−/− cells. Furthermore, stimulation with sst resulted in a modest but reproducible increase in SSTR3 ubiquitination in Arl6^−/− cells (Fig. 1 D). We conclude that the BBSome/ARL6 system is required for the degradation of ubiquitinated SSTR3 and that SSTR3 becomes ubiquitinated in response to stimulation with sst.

We next sought to test whether signaling receptors endogenous to IMCD3 cells are ubiquitinated inside cilia in an activity-dependent manner. One signaling pathway that is natively expressed in IMCD3 cells is the Hedgehog pathway. Prior experiments have detected normal trafficking dynamics of SMO and GPR161 in IMCD3 cells (Ye et al., 2018; Mukhopadhyay et al., 2013; see Fig. 4 D). As GPR161 and SMO both accumulate in cilia of Arl6^−/− cells without hedgehog (Hh) pathway stimulation (Liew et al., 2014; Zhang et al., 2011), the Ub signal detected in Arl6^−/− cilia may correspond to Ub conjugated to SMO or GPR161. We detected a significant elevation of the ciliary Ub levels in Arl6^−/− cells when treated with the SMO agonist SAG (Fig. 1, E and F). Because SMO ubiquitination was previously shown to decrease upon its activation (Xia et al., 2012; Jiang et al., 2019), it is likely that the elevated ciliary Ub signal detected in SAG-treated cells is associated with GPR161. As activation of SMO triggers exit of GPR161 but this exit is frustrated when BBSome function goes awry, these data suggest that GPR161 becomes ubiquitinated before its BBSome-dependent exit. Collectively, these data suggest that GPCRs are ubiquitinated in a regulated fashion inside cilia and subsequently retrieved into the cell by the BBSome.

To test the functional importance of GPCR ubiquitination in BBSome-mediated exit from cilia, we removed all ubiquitination sites from SSTR3 and GPR161 by mutating all cytoplasm-exposed lysine residues to arginine (cK0 variants). Exit kinetics were precisely monitored by real-time tracking of individual cilia in live cells. In both cases, the cK0 variant underwent markedly slower signal-dependent exit from cilia than the WT allele (Fig. 2, A and B; and Fig. S2, A and B). Measurement of exit rates revealed that exit of SSTR3cK0 was approximately twofold slower than SSTR3 and that exit of GPR161cK0 was ∼2.5-fold slower than GPR161. The residual exit rates of the cK0 mutants indicate the existence of alternative mechanisms of exit that complement GPCR ubiquitination or bypass BBSome-dependent retrieval (see Discussion).

To rule out that the observed ciliary exit defect of the cK0 mutants was an indirect consequence of defective endocytosis, for example, because of clogging of the ciliary exit path, we directly assessed endocytosis of SSTR3 and SSTR3cK0. The surface-exposed pool of ^APSSTR3 was pulse labeled with fluorescently labeled monovalent streptavidin (mSA), and cells were stimulated with sst. The faint hazy signal corresponding to the plasma membrane pool of SSTR3 disappeared after 10 min in the presence of sst in a similar fashion for both WT and cK0 variants (Fig. 2 C). Concurrently, the WT and cK0 variants appeared in cytoplasmic foci corresponding to endocytic vesicles (Fig. 2 C). Counting cytoplasmic foci revealed that endocytosis proceeded normally irrespective of the ubiquitination competence of SSTR3 (Fig. 2 D). SSTR3 thus behaves similarly to nearly every GPCR tested to date, in that it does not require ubiquitination for its signal-dependent internalization. Indeed, while ubiquitination of plasma membrane proteins is a major driver of internalization in yeast, ubiquitination of signaling receptors is generally dispensable for signal-dependent internalization in mammalian cells (Dores and Trejo, 2019; Piper et al., 2014; Skieterska et al., 2017).

These results strongly support a direct role for GPCR ubiquitination in promoting signal-dependent exit from cilia.

We next sought to characterize the type of Ub conjugates that are attached to ciliary proteins. The FK1 and FK2 antibodies used in our immunofluorescent studies recognize Ub either as single adduct or in chain but not free Ub (Fujimuro et al., 1994; Emmerich and Cohen, 2015). The availability of antibodies specific for UbK48 and UbK63 linkages (Newton et al., 2008) enabled us to determine the types of Ub conjugates that are attached to ciliary GPCRs. While we found no detectable signal for UbK48 in cilia, the signal for UbK63 linkages mirrored the ciliary response observed with the FK1 and FK2 antibody in IMCD3-[SSTR3] cells subjected to agonist treatment (Fig. 3, A and B; and Fig. S1, C and D). Given the high specificity and selectivity of the K48 and K63 linkage antibodies, these data strongly suggest that UbK63 chains are added onto SSTR3 inside cilia upon activation.

To confirm that K63 of Ub is the main linkage used for the elongation of Ub chains on SSTR3, we transfected variants of Ub with lysines mutated to arginines into Arl6^−/− cells stably expressing SSTR3. Signal-dependent ubiquitination of SSTR3 was abrogated when all seven lysines of Ub were mutated to arginines (Fig. 3, C and D). However, when K63 was the only lysine left intact on Ub, signal-dependent ubiquitination of SSTR3 was restored to the same extent as with WT Ub (Fig. 3, C and D). We conclude that the Ub chains assembled onto SSTR3 in response to SSTR3 activation inside cilia comprise K63 linkages. While UbK48 linkages were not detected inside cilia, it remains possible that mixed chains containing K63 linkages as well as other linkages besides K48 are assembled onto SSTR3 inside cilia.

To determine the functional importance of UbK63 linkages in tagging GPCRs for exit from cilia, we sought to specifically interfere with UbK63 linkages inside cilia (Fig. 4 A). Structural and biochemical studies have demonstrated that the deubiquitinase AMSH (associated molecule with the SH3 domain of STAM) possesses a near-absolute specificity for UbK63 linkages and does not cleave other Ub chain linkages or Ub-substrates linkages (Sato et al., 2008; McCullough et al., 2004). AMSH normally functions on the surface of late endosomes in concert with the endosomal sorting complex required for transport (ESCRT) protein STAM (McCullough et al., 2006). AMSH was previously fused to the epidermal growth factor receptor (EGFR) to demonstrate that UbK63 chains assembled on EGFR are dispensable for internalization but required for sorting into lysosomes (Huang et al., 2013). We targeted AMSH to cilia using the well-validated ciliary targeting signal from the ciliopathy protein NPHP3 (Nakata et al., 2012; Wright et al., 2011) and analyzed the exit of three distinct GPCRs: SMO, GPR161, and SSTR3. To alleviate concerns related to the interaction of AMSH with the STAM, we deleted the STAM-interacting domain of AMSH and only expressed the catalytic domain of AMSH. The signal-dependent exit of SSTR3 was assessed by treating IMCD3-[SSTR3] cells with sst for 2 h. In cells expressing cilia-AMSH (a fusion of NPHP3[1–200] to GFP and the catalytic domain of AMSH), sst no longer triggered a significant reduction in the ciliary levels of SSTR3 (Fig. 4, B and C). Care was taken to only analyze cells that expressed modest levels of cilia-AMSH as judged by GFP fluorescence (Fig. S3). In contrast, SSTR3 exit proceeded normally in cells that expressed NPHP3[1–200]-GFP alone or a catalytically dead variant of cilia-AMSH that we named cilia-ASMH^† (Fig. 4, B and C). These controls strongly suggest that cilia-AMSH exerts its effects on SSTR3 exit by cleaving UbK63 chains inside cilia, a conclusion borne out by imaging ciliary UbK63 in cells transfected with cilia-AMSH (Fig. S4). Similarly, the exit of endogenous GPR161 from retinal pigmented epithelium (RPE) cell cilia upon activation of the Hedgehog pathway was abolished in cells expressing moderate levels of cilia-AMSH (Fig. 4, D and E). Neither NPHP3[1–200]-GFP nor cilia-AMSH^† blocked the signal-dependent exit of GPR161 from cilia (Fig. 4, D and E).

Unlike GPR161 and other ciliary GPCRs such as SSTR3 that are removed from cilia by the BBSome only when they become activated, SMO undergoes constitutive BBSome-dependent exit from cilia in the absence of Hedgehog pathway stimulation (Nachury, 2018; Ye et al., 2018; Ocbina and Anderson, 2008; Zhang et al., 2011). Signal-dependent accumulation of SMO in cilia is achieved, at least in part, by suppression of its exit (Milenkovic et al., 2015; Nachury and Mick, 2019; Ye et al., 2018). In cells expressing cilia-AMSH, endogenous SMO spontaneously accumulated inside cilia in the absence of Hedgehog pathway activation (Fig. 4, F and G). Again, neither NPHP3[1–200]-GFP nor cilia-AMSH^† influenced the ciliary accumulation of SMO (Fig. 4, F and G).

Together, these results indicate that K63-linked Ub chains inside cilia, presumably built onto each ciliary GPCR under a specific signaling stimulus, are required for GPCR removal from cilia.

As β-arrestin 2 is required for signal-dependent exit of GPR161 and SSTR3 from cilia (Green et al., 2015; Pal et al., 2016; Ye et al., 2018), we sought to determine the functional relationship between β-arrestin 2 and the BBSome. As previously described (Green et al., 2015), β-arrestin 2 is initially undetectable inside cilia and becomes recruited to cilia within minutes of SSTR3 agonist addition (Fig. 5, A–C). The ciliary β-arrestin 2 signal reaches a plateau after ∼10 min, and the t_1/2 for ciliary accumulation is <5 min (Fig. 5 C). Meanwhile, the t_1/2 for BBSome recruitment to the tip of cilia was 10 min, and the first signs of detectable SSTR3 exit were seen between 10 and 15 min (Fig. 5 C). These kinetics suggested that β-arrestin 2 is first recruited to activated ciliary SSTR3 before the BBSome ferries activated SSTR3 out of cilia. When ARL6 was depleted, the basal levels of ciliary β-arrestin 2 were elevated but the kinetics of β-arrestin 2 accumulation in cilia upon SSTR3 activation were indistinguishable from control-depleted cells (Fig. 5, A and B). These data suggest that a small fraction of SSTR3 that is tonically active fails to exit cilia in ARL6-depleted cells and recruits β-arrestin 2. From these data, we conclude that β-arrestin 2 functions either upstream of, or in parallel with, the BBSome in signal-dependent retrieval of GPCRs. Similarly, the kinetics of GPR161 exit and β-arrestin 2 ciliary recruitment upon activation of the Hh pathway indicated that β-arrestin 2 reaches its maximal ciliary level before the first signs of GPR161 exit become evident (Fig. 5 D).

We previously proposed that β-arrestin 2 may bridge activated GPCRs to the BBSome (Ye et al., 2018; Nachury, 2018). Yet we have thus far failed to detect a biochemical interaction between β-arrestin 2 and the BBSome. Furthermore, the continuous distribution patterns of β-arrestin 2 and SSTR3 inside cilia are similar to one another and distinct from the discontinuous pattern of the BBSome (Fig. 5 A), arguing against a direct association of β-arrestin 2 with BBSome/IFT trains inside cilia. We considered the alternative hypothesis that β-arrestin 2 functions upstream of the BBSome. Given that β-arrestin 2 functions in the signal-dependent targeting of GPCRs from the plasma membrane to the lysosome by recruiting Ub ligases to activated GPCRs (Henry et al., 2012; Bhandari et al., 2007; Shenoy et al., 2008; Martin et al., 2003), we posited that β-arrestin 2 recognizes activated GPCRs inside cilia to direct their ubiquitination and subsequent selection by the BBSome for removal from cilia. A central prediction of this model is that β-arrestin 2 is required for the ubiquitination of ciliary GPCRs in response to stimulation. To test this prediction, we deleted the β-arrestin 2 gene Arrb2 in Arl6 knockout IMCD3 cells (Nager et al., 2017). While the signal-dependent exit of SSTR3 from cilia failed in both Arl6^−/− and Arl6^−/−Arrb2^−/− cells (Fig. 1, A and B; and Fig. 6, A and B), Ub staining in cilia and by inference signal-dependent ubiquitination of SSTR3 was only observed in Arl6^−/− cells (Fig. 6, A and B). In Arl6^−/−Arrb2^−/− cells, the signal-dependent increase in ciliary Ub signal was no longer observed (Fig. 6, A and B). Similarly, the increase in ciliary Ub signal seen in Arl6^−/− cells treated with the Hedgehog pathway agonist SAG was no longer observed in the absence of β-arrestin 2 (Fig. 6, C and D). These data indicate that, in the absence of BBSome function, ciliary GPCRs become ubiquitinated in response to activation in a β-arrestin 2–dependent manner. We sought to confirm our imaging-based finding with a biochemical analysis of SSTR3 ubiquitination. While SSTR3 ubiquitination was readily detected in Arl6^−/− cells and measurably increased upon stimulation with sst, the deletion of β-arrestin 2 in Arl6^−/− cells drastically reduced SSTR3 ubiquitination levels, and no sst-dependent increase of SSTR3 ubiquitination was detected in Arl6^−/−/Arrb2^−/− cells (Fig. 6, E and F).

Together, these data suggest the following order of action: GPCR activation, β-arrestin 2 engagement, Ub ligase recruitment, assembly of UbK63 chains on the GPCR, and finally selection of ubiquitinated GPCRs for BBSome-mediated retrieval (Fig. 7 E).

Besides removing GPCRs from cilia in a regulated fashion, the BBSome also clears cilia of unwanted proteins that accidentally enter cilia. In the single-cell flagellated organism C. reinhardtii, mutations in the BBSome subunit Bbs4 cause a constitutive accumulation of several proteins including phospholipase D in cilia (Lechtreck et al., 2009). When stained for Ub, Bbs4 C. rheinardtii showed a twofold enrichment in ciliary signal compared with WT (Fig. 7, A and B). Isolation of cilia revealed the accumulation of Ub conjugates above 80 kD in Bbs4 mutant C. rheinardtii cilia (Fig. 7 C). These results are consistent with the hypothesis that ciliary proteins subject to constitutive retrieval are ubiquitinated before their removal from cilia by the BBSome.

In the mammalian retina, proteomics studies have found >100 proteins accumulating in the outer segment (the equivalent of the cilium) of Bbs photoreceptors compared with WT photoreceptors (Datta et al., 2015). When we stained sections of mouse retina for Ub, we found a drastic accumulation of Ub signal in the outer segment of Bbs4^−/− photoreceptors compared with WT photoreceptors (Fig. 7 D). We note that the nature of Ub conjugates remains to be determined in the case of constitutive cargoes. These results suggest that a variety of proteins that accidentally flow into the photoreceptors outer segments, possibly caught on the tremendous flux of rhodopsin trafficking to the outer segment (Baker et al., 2008), are recognized as foreign to the cilium, ubiquitinated on site, and selected for removal by the BBSome (Fig. 7 E).

Our results indicate that Ub chains built at least in part with K63 linkages mark activated GPCRs and likely other unwanted proteins for BBSome-mediated removal from cilia. Prior findings in trypanosomes suggest that the BBSome may directly recognize Ub (Langousis et al., 2016), although direct biochemical evidence and information regarding Ub chain specificity is still missing. It is also known that the BBSome directly recognizes cytoplasmic determinants in the GPCRs it ferries out of cilia (Jin et al., 2010; Klink et al., 2017; Ye et al., 2018; Seo et al., 2011; Domire et al., 2011). We propose that the summation of weak molecular interactions between the BBSome and Ub chains on the one hand and the BBSome and GPCRs cytoplasmic loops on the other hand increases the BBSome–cargo interaction to enable sorting of ubiquitinated GPCRs out of cilia. In support of this bivalent recognition of ubiquitinated cargoes by the BBSome, the affinity of the BBSome for cytoplasmic determinants on GPCRs is in the micromolar range (Klink et al., 2017), and affinities of Ub-binding proteins for Ub typically reside in the sub-millimolar range (Husnjak and Dikic, 2012). The nature of the biochemical entity that directly recognizes Ub, and in particular UbK63 linkages, remains to be determined: while the BBSome was pulled down on Ub-agarose resin from trypanosome lysates (Langousis et al., 2016), no known Ub-binding domain is found in the BBSome polypeptides. Furthermore, the IFT-A subunit IFT-139 associates with ubiquitinated α-tubulin in ciliary extracts from C. rheinardtii, suggesting a possible recognition of Ub by IFT139 (Wang et al., 2019). Again, the absence of a known Ub-binding domain in the sequence of IFT139 presents a challenge to this hypothesis.

Wang et al. (2019) found that α-tubulin is modified by UbK63 chains and not UbK48 chains upon ciliary disassembly. The Ub ligases that have been associated with ciliary clearance of Patched 1 and of PDGFRαα specifically assemble UbK63 chains (Kim et al., 2015; Yue et al., 2014; Schmid et al., 2018). The suppression of GPR161, SSTR3, and SMO exit by cilia-AMSH (Fig. 4) demonstrates that UbK63 chains are recognized inside cilia to direct trafficking out of cilia. The milder exit defect observed when acceptor lysines are mutated on ciliary GPCRs (Fig. 2) than when cilia-AMSH is expressed (Fig. 4) suggests that UbK63 linkages added onto other ciliary proteins besides GPCRs may participate in ciliary exit. A role for ubiquitination of the ciliary transport machinery in exit is in line with the functional importance of ubiquitination of the endolysosomal machinery in GPCR trafficking (Dores and Trejo, 2019). β-Arrestin ubiquitination promotes internalization of the β₂-adrenergic receptor (β₂AR; Shenoy et al., 2001, 2007; Shenoy and Lefkowitz, 2003), ubiquitination of the ESCRT components Hrs and STAM participates in sorting of the chemokine receptor CXCR4 into the multivesicular body (MVB; Malik and Marchese, 2010; Marchese et al., 2003), and ubiquitination of the ESCRT-III recruitment factor ALIX enhances MVB sorting of the protease-activated receptor 1 (Dores et al., 2015).

The requirement for UbK63 chains in signal-dependent exit from the cilium contrasts with the more limited role of Ub and UbK63 chains in signal-dependent endocytosis of signaling receptors at the plasma membrane in mammals. Although ubiquitination is a major driver of endocytosis in yeast (Stringer and Piper, 2011; Piper et al., 2014), foundational studies on EGFR trafficking found that signal-dependent receptor ubiquitination is not essential for internalization because it is redundant with other internalization mechanisms (Goh et al., 2010; Huang et al., 2013; Fortian et al., 2015). Studies of an EGFR-AMSH fusion found that UbK63 chains attached onto EGFR become increasingly necessary as EGFR progresses along the degradative route, and a strict requirement for K63-linked Ub chains is observed at the terminal step of EGFR sorting into the lumen of MVB (Huang et al., 2013). A strict requirement for receptor ubiquitination in degradative sorting but not in internalization can be generalized to nearly all GPCRs (Skieterska et al., 2017), namely β₂AR (Shenoy et al., 2001), CXCR4 (Marchese and Benovic, 2001), µ opioid receptor (Hislop et al., 2011), κ opioid receptor (Li et al., 2008), δ opioid receptor (Henry et al., 2011), neurokinin-1 receptor (Cottrell et al., 2006), protease-activated receptor 2 (Jacob et al., 2005), and vasopressin receptor 2 (Martin et al., 2003). Similarly, we find that SSTR3 endocytosis proceeds normally in the absence of SSTR3 ubiquitination (Fig. 2, C and D). Studying the importance of UbK63 linkages in mammalian cells via genetics remains challenging because of the multiple genes encoding Ub, but in yeast strains expressing UbK63R as the sole source of Ub, the Gap1 permease is internalized normally and yet fails to get sorted into the MVB (Lauwers et al., 2009).

The parallels between sorting at the late endosome and ciliary trafficking extend to β-arrestin–directed ubiquitination. While β-arrestin functions inside cilia to direct the addition of Ub onto activated SSTR3 and GPR161 (Fig. 5 and Fig. 6), β-arrestin–mediated ubiquitination selectively affects sorting at the level of endosomes rather that at the plasma membrane for activated CXCR4 (Bhandari et al., 2007), β₂AR (Shenoy et al., 2008), and vasopressin receptor 2 (Martin et al., 2003), and β-arrestin–mediated ubiquitination of CXCR4 was shown to take place on endosomes (Malik and Marchese, 2010). In this context, the recognition of K63-linked Ub chain inside cilia may befit the endosomal origin of primary cilia (Sorokin, 1962; Westlake et al., 2011; Mitchell, 2017).

The parallels between endosomal and ciliary trafficking lead us to speculate that some ubiquitinated proteins might be accidentally sorted to cilia instead of late endosomes. These proteins would then need to be retrieved from cilia by the BBSome. Thus, the Ub signal detected in photoreceptor outer segment and in C. rheinardtii flagella might result from the accidental import of ubiquitinated proteins into cilia rather than in situ ubiquitination of unwanted proteins inside cilia.

An interesting consideration is the relationship between ciliary exit and endolysosomal sorting. It has been proposed that endocytosis of signaling receptors is intimately coupled to their exit from cilia (Pedersen et al., 2016). However, single molecule imaging of GPR161 exiting cilia suggests that activated GPR161 diffuses into the plasma membrane after it exits from the ciliary compartment (Ye et al., 2018). Prior findings that worm mutants for BBSome or the ESCRT machinery both fail to remove ciliary proteins fused to Ub from cilia (Hu et al., 2007; Xu et al., 2015) suggest the provocative possibility that ciliary exit may be tightly coupled to degradative sorting. Nonetheless, the hypothesis that interfering with UbK63 chain formation on ciliary GPCRs indirectly blocks ciliary exit because of a primary defect in endocytosis can be rejected because ubiquitination is not a major determinant of SSTR3 endocytosis and because cilia-AMSH potently and specifically blocks the signal-dependent exit of ciliary GPCRs.

In this context, the contrast between the dramatic increase in ubiquitinated SSTR3 levels in Arl6^−/− cells compared with WT cells even in the absence of stimulation (Fig. 1, C and D) and the modest increase in ciliary Ub levels in Arl6^−/− cells compared with WT cells in the absence of stimulation (Fig. 1, A and B) suggest that a considerable amount of ubiquitinated SSTR3 accumulates outside of cilia in Arl6^−/− cells. One interpretation consistent with a direct coupling between ciliary exit and degradation is that a small fraction of ubiquitinated SSTR3 escapes cilia in Arl6^−/− cells but fails to reach the correct degradative route. In this interpretation, the BBSome would deliver its cargoes to the ESCRT machinery for routing into the lysosomal degradative pathway.

A role for ubiquitination in selecting unwanted proteins for removal from cilia may unify the functions of the BBSome in signal-dependent retrieval of GPCRs and the constitutive clearance of proteins that accidentally enter cilia. In both of these cases, unwanted ciliary proteins need to be recognized as “nonself” by a ciliary surveillance machinery. For GPCRs, β-arrestin orchestrates the recognition of activated GPCRs as unwanted by directing their ubiquitination. A fascinating question for future investigations is how the >100 proteins that accumulate in outer segments of Bbs photoreceptors are recognized as nonself and tagged with Ub.

The mouse IMCD3 cell lines used in the study were generated from a parental IMCD3-FlpIn cell line (gift from P.K. Jackson, Stanford University, Stanford, CA). IMCD3-FlpIn cells were cultured in DMEM/F12 (11330–057; Gibco) supplemented with 10% FBS (100–106; Gemini Bio-products), 100 U/ml penicillin-streptomycin (400–109; Gemini Bio-products), and 2 mM L-glutamine (400–106; Gemini Bio-products). The RPE1-hTERT cell line (CRL-4000; ATCC) was cultured in DMEM/F12 supplemented with 10% FBS, 100 U/ml penicillin-streptomycin, 2 mM L-glutamine, and 0.26% sodium bicarbonate (25080; Gibco).

Ciliation was induced by serum starvation in media containing 0.2% FBS for 16–24 h.

Stable isogenic IMCD3 cell lines were generated using the Flp-In system (Thermo Fisher Scientific). Low-expression promoters and additional expression cassettes were introduced into pEF5/FRT (flippase recognition target) plasmid as described (Nager et al., 2017; Ye et al., 2018). To reduce expression levels, the EF1α promoter was attenuated by mutating the TATA-box (pEF1α^Δ) or replaced with a minimal chicken lens δ-crystallin promoter (pCrys).

Coding sequences were amplified from plasmids encoding mouse GPR161 (BC028163; Mammalian Gene Collection; GE Healthcare), mouse SSTR3 (gift from Kirk Mykytyn, Ohio State University, Columbus, OH), and human BBS5 (gift from V. Sheffield, University of Iowa, Iowa City, IA), BirA-ER (gift from Alice Ting, Stanford University). β-Arrestin 2 (a gift from Mark Scott, Institut Cochin, Paris, France) was stably expressed from a cytomegalovirus promoter-based plasmid (pEGFP-N). SSTR3 and SMO expression were driven by pEF1α^Δ, GPR161 expression by pCrys, and BBS5 expression by pEF1α. Coding sequences were fused to GFP, NG (Shaner et al., 2013), or an acceptor peptide for the biotin ligase BirA (AP; Howarth and Ting, 2008). Multiple rounds of site-directed mutagenesis were performed to generate SSTR3-cK0, a variant with all the cytoplasm-facing lysine residues (K233, K330, K356, K407, and K421) mutated to arginine. GPR161-cK0, a variant with 18 cytoplasm-exposed lysine residues (K83, K84, K93, K167, K247, K250, K269, K296, K358, K362, K455, K469, K473, K481, K486, K497, K504, and K548) mutated to arginine, was gene synthesized (GenScript). Cilia-AMSH was generated by fusing the catalytic domain of mouse AMSH (gift from David Komander, Walter and Eliza Hall Institute of Medical Research, Melbourne, Australia; plasmid no. 66712; Addgene; Michel et al., 2015) with NPHP3[1–200] and GFP to create NPHP3[1–200]-GFP-AMSH[243–424]. A catalytically dead version of AMSH was generated by mutating the water-activating Glu280 residue to Ala (Sato et al., 2008; Huang et al., 2013). pRK5-HA-Ub WT, K0 (all seven acceptor lysine residues, K6, K11, K27, K29, K33, K48, and K63, are mutated to arginine), and K63 only (all lysines except K63 mutated to arginines; gift from Ted Dawson, Johns Hopkins, Baltimore, MD; plasmid nos. 17608, 17603, and 17606, respectively; Addgene).

CRISPR-edited Arl6^−/− and Arl6^−/−/Arrb2^−/− cell lines were described previously (Liew et al., 2014; Nager et al., 2017). The genotypes are Arl6: NM_001347244.1:c.10_25del; c.3_6del; and Arrb2: NM_001271358.1:c.1 12del;c112_113del.

For generation of all stable cell lines except for the ones expressing β-arrestin 2^GFP, FRT-based plasmids were reverse-transfected using XtremeGene9 (Roche) into IMCD3 Flp-In (flippase integration) cells along with a plasmid encoding the Flp recombinase (pOG44), and stable transformants were selected by blasticidin resistance (4 µg/ml). The transfection mixture was then added to the cells. pEGFP-N• β-arrestin 2 was transfected into IMCD3 cells using Lipofectamine 2000, and clones were selected using neomycin resistance.

The following monoclonal antibodies were used for immunofluorescence: anti-acetylated tubulin (mouse; clone 6-11B-1; Sigma-Aldrich; 1:500), anti-Ub clone FK2 used in all experiments except where indicated (mouse; D058-3; Medical and Biological Laboratories; 1:500), anti-Ub FK1 in Fig. S1, A and B (mouse; D071-3; Medical and Biological Laboratories; 1:500), anti-UbK48 linkage (rabbit; clone Apu2; 05–1307; Millipore-Genentech; 1:500), and anti-UbK63 linkage (human; clone Apu3; a gift from Genentech; 1:500). The following monoclonal antibodies were used for immunoblotting: anti-Ub (mouse; clone P4D1; 3936; Cell Signaling; 1:1,000), anti–α-tubulin (mouse; clone DM1A; MS-581-P1ABX; Thermo Fisher Scientific; 1:1,000), anti-HA (mouse; clone 16B12; 901501; Biolegend; 1:500), and streptavidin-HRP (Pierce-21140; Thermo Fisher Scientific; 1:10,000). The following polyclonal antibodies were used for immunofluorescence: anti-GPR161 (rabbit; 13398–1-AP; Proteintech; 1:100), anti-SMO (rabbit; a gift from Kathryn Anderson, Memorial Sloan-Kettering Cancer Center, New York, NY; 1:500), and anti–cone-arrestin (rabbit; AB15282; Millipore; 1:500). Biotinylated SSTR3 and GPR161 were detected using Alexa Fluor 647–labeled mSA (mSA647; Ye et al., 2018). The following reagents were used at the indicated concentrations: 200 nM SAG and 10 µM sst-14. Somatostatin 14 stocks were made in DMEM/F12 media, Hepes, no phenol red (11039–021; GIBCO). SAG was dissolved in DMSO (276855; Sigma-Aldrich).

Cells were imaged either on a DeltaVision system (Applied Precision) equipped with a PlanApo 60×/1.40 objective lens (Olympus), CoolSNAP HQ2 camera (Photometrics), and solid-state illumination module (InsightSSI), or on a confocal LSM 700 (Zeiss) microscope equipped with 40× Plan-Apochromat 1.3 DIC oil objective. Z stacks were acquired at 0.5-µm intervals (LSM 700; Fig. 1, Fig. 3, Fig. 7, and Fig. S1) or 0.2-µm intervals (DeltaVision; Fig. 4, Fig. 6, Fig. S3, and Fig. S4).

For fixed imaging, 60,000 cells were seeded on acid-washed coverglasses (12 mm no. 1.5; 12–545-81; Thermo Fisher Scientific). Cells were grown for 24 h and then starved for 16–24 h in 0.2% FBS media before experimental treatment. After treatment, cells were fixed with 4% PFA in PBS for 15 min at 37°C and extracted in −20°C methanol for 5 min. Cells were then permeabilized in immunofluorescence (IF) buffer (PBS supplemented with 0.1% Triton X-100, 5% normal donkey serum [017–000-121; Jackson ImmunoResearch Laboratories], and 3% BSA [BP1605-100; Thermo Fisher Scientific]), incubated at room temperature for 1 h with primary antibodies diluted in IF buffer, washed three times in IF buffer, incubated with secondary antibodies (Jackson ImmunoResearch Laboratories), diluted in IF buffer for 30 min, and washed three times with IF buffer. DNA was stained with Hoechst 33258 (H1398; Molecular Probes), cells were washed twice more with PBS, and coverglasses were mounted on slides using fluoromount-G (17984–25; Electron Microscopy Sciences).

For live-cell imaging, 300,000 cells were seeded on acid-washed 25-mm coverglasses (Electron Microscopy Sciences). After 24 h of growth, cells were serum starved for 16 h and transferred to the DeltaVision stage for imaging at 37°C within an environmental chamber that encloses the microscope and the stage. Cells were imaged in DMEM/F12 media, Hepes, no phenol red (11039–021; GIBCO). To measure SSTR3 exit, cells expressing ^APSSTR3 were first washed three times with PBS and then pulse labeled with 2 µg/ml mSA647 for 5 min at 37°C. Cells were then imaged after addition of sst (for SSTR3) or SAG (for GPR161) for 2 h with an acquisition every 10 min. Each acquisition consisted of a three-plane Z-stack with each plane separated by 0.5 µm. Transmittance was set to 5% and exposure time to 500 ms.

WT or Bbs4^−/− mice (Mykytyn et al., 2004) eyes were enucleated at P21. Whole eyes were fixed in 4% PFA/PBS overnight at 4°C and then washed three times with PBS. The eyes were infiltrated in 30% sucrose/PBS at 4°C overnight. Eyes were placed in O.C.T. compound embedding medium (4583; Tissue-Tek), frozen on dry ice, and stored at −80°C. 14-µm sections were cut on a cryostat (CM 1850; Leica). Sections were processed for immunofluorescence as follows. Sections were blocked in blocking solution (50 mM Tris, pH 7.4, 100 mM NaCl, 0.1% Triton X-100, 3% normal donkey serum, and 0.1% BSA) for 1 h at room temperature and then incubated with primary antibodies overnight at 4°C, followed by three washes in wash solution (50 mM Tris, pH 7.4, 100 mM NaCl, and 0.1% Triton X-100). After rinsing in wash solution, sections were incubated with secondary antibodies for 1 h at room temperature. DNA was counterstained using Hoechst 33258 dye, and sections washed twice more with PBS and mounted on slides using Fluoromount-G. Sections were imaged on a Confocal LSM 700 (Zeiss) microscope.

Endocytosed SSTR3 foci were revealed by contrast enhancement and counted with the ImageJ particle analysis tool.

C. rheinardtii WT-g1 (nit1, agg1, mating type [mt⁺]; gift from George B. Witman, University of Massachusetts Medical School, Worcester, MA) and Bbs4⁻ CC-4377 ptx5-1/bbs4-1:: NIT1 agg1 mt⁺ (Chlamydomonas Resource Center) strains were cultured asynchronously under light in Tris-acetate-phosphate media (Gorman and Levine, 1965) for 72 h. Flagellar fractions were prepared from 2 liters of cultures of WT or Bbs4 cells as described (Craige et al., 2013). Briefly, cells were harvested by centrifugation for 5 min at 2,000 rpm (1,100 g) at room temperature. Cells were resuspended in 10 mM Hepes, pH 7.4, and centrifuged for 5 min at 2,000 rpm (1,100 g). Next, the cell pellet was gently resuspended in ice-cold HMDS (10 mM Hepes, pH 7.4, 5 mM MgSO₄, 1 mM DTT, and 4% [wt/vol] sucrose). Deflagellation was achieved by addition of 5 mM dibucaine to each tube of cells and by pipetting up and down ∼10 times using a 10-ml plastic serological pipette. The suspension was spun for 5 min at 1,800 g at 4°C. The supernatant containing flagella was underlaid with 9 ml ice-cold HMDS-25% sucrose (10 mM Hepes, pH 7.4, 5 mM MgSO₄, 1 mM DTT, and 25% [wt/vol] sucrose), followed by a centrifugation step for 10 min at 2,800 rpm (2,400 g), 4°C. The supernatant above the sucrose interface was collected using a 25-ml serological pipette, and flagella were pelleted by centrifugation for 20 min at 30,000 g (16,000 rpm), 4°C. The flagellar pellet was resuspended in 100 µl HMDS buffer. Protein concentrations were measured by Bradford assays, and 25 μg were resolved by SDS-PAGE for immunoblotting.

C. rheinardtii immunofluorescence was performed as follows. Cells were fixed with 4% PFA in MT buffer (30 mM Hepes, pH 7, 5 mM EGTA, 5 mM MgSO₄, and 25 mM KCl) for 20 min in suspension. The cells were centrifuged at 1,000 rpm for 5 min, resuspended in 100 µl of fixative, and transferred onto slides coated with 1 mg/ml poly-L-lysines. After 5 min, the unadhered cells were washed off by rinsing with PBS. Cells were permeabilized in 0.5% Triton X-100 for 20 min followed by blocking for 1 h in blocking buffer (3% fish skin gelatin, 1% BSA, and 0.1% Tween 20 in PBS) at room temperature. Cells were incubated at 4°C overnight with primary antibodies diluted in blocking buffer, washed five times in PBS, and incubated with secondary antibodies for 2 h at room temperature. After five washes in PBS, coverglasses were mounted on slides using Fluoromount-G. Cells were imaged on the LSM700 confocal microscope.

IMCD3 cells stably expressing pEF1α^Δ−APSSTR3^NG and BirA-ER were transiently transfected with HA-tagged Ub plasmids. Cells were reverse transfected using XtremeGene9 and plated onto 15-cm plates. Cells were moved to medium containing 0.2% serum and supplemented with 10 µM biotin 24 h after transfection to promote ciliation and maximize biotinylation of SSTR3. 18 h after the medium change, sst was added for 0, 10, or 20 min. The assay was stopped by washing cells twice with ice-cold 1× PBS and scraping cells off on ice into 1 ml of ice-cold radioimmunoprecipitation assay buffer (50 mM Tris/HCl, pH 8, 150 mM NaCl, 1% Nonidet P-40, 0.5% sodium deoxycholate, 0.1% SDS, 100 mM iodoacetamide, and 10 mM EDTA) supplemented with protease inhibitors (1 mM 4-(2-aminoethyl)benzenesulfonyl fluoride hydrochloride, 0.8 mM aprotinin, 15 mM E-64, 10 mg/ml bestatin, 10 mg/ml pepstatin A, and 10 mg/ml leupeptin). After gently rocking at 4°C for 5 min, lysates were clarified by spinning in a tabletop Eppendorf centrifuge at 15,000 rpm and 4°C for 20 min. Each clarified lysate was incubated with 50 µl streptavidin Sepharose beads (GE Healthcare) for 1 h at 4°C to capture biotinylated SSTR3. Beads were washed four times with 500 µl radioimmunoprecipitation assay buffer, and bound proteins were eluted by heating the beads in 1× lauryl dodecyl sulfate sample loading buffer containing 2 mM biotin at 70°C for 10 min. After SDS-PAGE and transfer, polyvinylidene fluoride membranes were probed with anti-HA antibody and streptavidin-HRP.

Signals were acquired with a ChemiDoc imager (Bio-Rad) and analyzed with ImageLab 6.0.1 (Bio-Rad). The integrated densities of the SSTR3 bands in the streptavidin-HRP blot were normalized to the integrated densities of the endogenous biotinylated proteins to correct for variations in the recovery of biotinylated proteins on streptavidin Sepharose. The resulting value is SSTR3_bio. Second, the integrated densities of ubiquitinated SSTR3 were collected by measuring the area from 100 kD to the top of the gel on the HA blot. The values were background-corrected by subtracting the value of the equivalent area in the IMCD3 control lane. The resulting value is Ub-SSTR3_raw. The relative amount of SSTR3 that is ubiquitinated, Ub-SSTR3_frac = Ub-SSTR3_raw/SSTR3_bio, was plotted on the graphs using PlotsOfData (Postma and Goedhart, 2019).

Fig. S1 shows staining of WT and Arl6^−/− IMCD3 cells expressing ^APSSTR3^NG with the FK1 anti-Ub antibody and with the UbK63 antibody. Fig. S2 presents the measurement of the ciliary levels of the cytoplasmic lysineless mutants of SSTR3 and GPR161. Fig. S3 shows a micrograph representative of the NPHP3^CTS-AMSH expression levels. Fig. S4 contains data supporting the removal of ciliary UbK63 chains by NPHP3^CTS-AMSH.

We thank Drs. Kirk Mykytyn, Val Sheffield, Alice Ting, and Mark Scott for the gifts of cDNAs; Kathryn Anderson, Vishva Dixit (Genentech, San Francisco, CA), and Felice Dunn (University of California, San Francisco [UCSF], San Francisco, CA) for the gifts of antibodies; Val Sheffield for the gift of the Bbs4^−/− mice; George B. Witman and the Chlamydomonas Resource Center for the gift of Chlamydomonas strains; Yien-Ming Kuo for help with cryosections of mouse retina and core support; Dhivya Kumar and Tyler Picariello for the assistance with Chlamydomonas cultures; Nicholas Morante (UCSF) and Jeremy Reiter (UCSF) for providing Bbs4^−/− mouse eyes; Fan Ye (UCSF) for generating some of the cell lines used in the study; Irene Ojeda Naharros for help with single cilia tracking; the Nachury and von Zastrow laboratories for helpful discussions; and the Ogden and Caspary laboratories for comments on the preprint.

This work was funded by the National Institute of General Medical Sciences (R01-GM089933 to M.V. Nachury). This work was made possible, in part, by a National Eye Institute Core Grant for Vision Research (EY002162) and by the Research to Prevent Blindness Unrestricted Grant (M.V. Nachury).

The authors declare no competing financial interests.

Author contributions: S.R. Shinde conducted all experimental work except for the experiments in Fig. 5, which were conducted by A.R. Nager. M.V. Nachury supervised research. S.R. Shinde and M.V. Nachury wrote the manuscript.