Protein quality control machinery supports primary ciliogenesis by eliminating GDP-bound Rab8-family GTPases

Summary The small GTPase Rab8 plays a vital role in the vesicular trafficking of cargo proteins from the trans-Golgi network to target membranes. Upon reaching its target destination, Rab8 is released from the vesicular membrane into the cytoplasm via guanosine triphosphate (GTP) hydrolysis. The fate of GDP-bound Rab8 released from the destination membranes, however, has not been investigated adequately. In this study, we found that GDP-bound Rab8 subfamily proteins are targeted for immediate degradation, and the pre-emptive quality control machinery is responsible for eliminating these proteins in a nucleotide-specific manner. We provide evidence that components of this quality control machinery have a critical role in vesicular trafficking events, including the formation of primary cilia, a process regulated by the Rab8 subfamily. These results suggest that the protein degradation machinery plays a critical role in the integrity of membrane trafficking by limiting the excessive accumulation of GDP-bound Rab8 subfamily proteins.


INTRODUCTION
Small monomeric GTPases belonging to the Ras superfamily are molecular switches that cycle between their active guanosine triphosphate (GTP)-bound and inactive GDP-bound states. 1,2 In the case of cell-membraneanchored Ras protein, GDP/GTP exchange is stimulated by guanine nucleotide exchange factor (GEF) to transduce proliferative signals in response to extracellular growth stimuli. [3][4][5] In the absence of growth factors, Ras-bound GTP nucleotide is hydrolyzed by the intrinsic GTPase activity of Ras, resulting in its rapid inactivation. 2 Thus, the activity of Ras is controlled by its reversible nucleotide exchange cycle on the cell membrane.
In contrast to Ras, the Rab family of small GTPases is responsible for membrane vesicular trafficking between cytoplasmic organelles. [6][7][8][9][10][11] Rab8a, for example, plays an important role in the long-distance vesicular trafficking of cargo proteins from the trans-Golgi network to target destinations with specific cytoskeletal-filament interactions. [12][13][14][15][16][17][18][19][20] At the end of its journey, GTP hydrolysis facilitates the dissociation of Rab8a from the vesicular membrane to the cytoplasm. It has long been assumed that GDP-bound Rab8a released from the destination membrane is retroverted through cytoplasmic transport (or simple diffusion) to the departure organelles where it is reactivated by a specific GEF. 21,22 The cytoplasmic machinery responsible for transporting GDP-bound Rab8a back to the departure membranes, if it exists, is not adequately understood.
The cytoplasmic accumulation of GDP-bound Rab8a has been reported to have deleterious effects on cell vesicular trafficking due to its dominant-negative and aggregation-prone nature. 17,[23][24][25][26][27][28][29][30] Indeed, the forced expression of GDP-bound Rab8a inhibits several Rab8a-mediated processes, such as primary ciliogenesis, 17,31,32 insulin-stimulated GLUT4 translocation, 30,33 the transformation of macropinosomes into tubules, 17 and the distribution of RABIN8, a GEF for Rab8a. 34 These observations suggest that the amount of GDP-bound Rab8a must be maintained at a minimum level. Therefore, it is necessary to understand the fate of GDP-bound Rab8a after its cytoplasmic retrieval. However, the mechanism responsible for maintaining minimum amounts of inactive Rab8a remains to be elucidated.
In a previous study, we proposed that cytoplasmic Rab8a in its GDP-bound form is recognized by BAG6 and targeted for degradation. 35 In contrast, the vast majority of active GTP-bound Rab8a is resistant to iScience Article degradation and does not have detectable affinity for BAG6. It is not known how BAG6 selectively determines the stability of GDP-bound Rab8a. BAG6 is a multifunctional protein localized in the cytoplasm and nucleus, and one of its cytoplasmic functions is pre-emptive protein quality control. [36][37][38][39][40][41][42] This quality control machinery is essential for degrading mislocalized transmembrane domain proteins that fail to be incorporated into the target membrane and thus expose their hydrophobic stretches (such as the transmembrane domain) in the cytoplasm. 36,38 Indeed, BAG6 possesses a preference for such exposed hydrophobicity. 37,38,43,44 The pre-emptive quality control machinery is composed of not only BAG6 but also the RING finger-type ubiquitin ligase RNF126, 45,46 and ubiquitin receptor UBQLN4. 47,48 Therefore, it is necessary to investigate whether these components of the pre-emptive quality control system are essential for membrane trafficking events by controlling Rab8 subfamily proteins, such as Rab8a and Rab10, in a GDP-bound form-specific manner.
In this study, we focused on this issue by investigating the possible interplay between Rab8 subfamily proteins and pre-emptive quality control components. We found that UBQLN4 predominantly recognizes the GDP-bound forms of Rab8a and Rab10, which are destined for immediate degradation, while their stable GTP-bound counterparts are scarcely recognized. Furthermore, major components of the pre-emptive protein quality control machinery, including RNF126 and BAG6, are all associated with Rab8 subfamily proteins in a nucleotide-dependent manner. UBQLN4 and RNF126 are both necessary for the instability of GDP-bound Rab8a, since their depletion results in the stabilization of GDP-bound species. We also found that the degradation of GDP-bound Rab10 is inhibited by the depletion of BAG6 or RNF126. Furthermore, we demonstrated that deficiencies in the pre-emptive quality control machinery induce defects in primary ciliogenesis, the vesicular trafficking of which is regulated in a redundant manner by Rab8 and Rab10. 49 Collectively, these results suggest that the ubiquitination machinery responsible for pre-emptive protein quality control prevents the excessive accumulation of GDP-bound Rab8 subfamily proteins, which must be maintained at a minimum level to ensure proper membrane vesicle sorting in the cytoplasm.

GDP-bound Rab8 subfamily proteins are unstable in HeLa cells
Rab10, a paralog of Rab8a, was previously shown to bind to BAG6, 35 and functions redundantly with Rab8 in primary ciliogenesis. 49 Cycloheximide (CHX) chase experiments suggested that wild-type Rab10 and Rab8a were stable proteins ( Figures 1A, 1B, and 1D). However, we found that the GDP-bound Rab10-T23N mutant has a very short half-life in HeLa cells ( Figures 1A and 1D), similar to the GDP-bound Rab8a-T22N mutant ( Figures 1B and 1D). In contrast, constitutively active GTP-bound Rab10-Q68L and Rab8a-Q67L mutant proteins, which have compromised GTPase activity, were highly stable ( Figures 1A,  1B and 1D). These observations suggest that the stability of the Rab8 subfamily of small GTPases is controlled in a nucleotide-specific manner.
In contrast to Rab8 subfamily proteins, Rab7a was identified as a small GTPase with a non-detectable affinity for BAG6. 35 We found that Rab7a was highly stable in HeLa cells, even in its GDP-bound T22N mutant form ( Figures 1C and 1D). These results suggest that the instability of GDP-bound Rab proteins is subfamily specific and apparently correlates with their affinity for BAG6.

UBQLN4 is required for Rab8a degradation in a nucleotide-specific manner
Because UBQLN4 is a critical ubiquitin receptor for pre-emptive quality control, 47,48 we examined whether UBQLN4 has an affinity for Rab8 subfamily proteins. We found that GDP-bound Rab8a-T22N co-immunoprecipitated with UBQLN4 in HeLa cell extracts ( Figure 1E). In contrast, constitutively active Rab8a-Q67L Figure 1. Continued even in its GDP-bound T22N mutant form (C). Tubulin was used as a loading control. Note that GDP-bound Rab10-T23N is difficult to express at a level equivalent to that of stable WT Rab10 due to its extreme instability.
(D) Graphs indicate the quantified signal intensities of Rab10 (A), Rab8a (B), and Rab7a (C) blots relative to loading controls at the indicated time points after CHX addition. Signals at time zero were defined as 100%.
(E and F) Co-immunoprecipitation analysis shows that UBQLN4 efficiently co-precipitates with Rab8a-T22N, but scarcely with Rab8a-Q67L (E). The case was similar for the GDP-locked Rab10-T23N and nucleotide hydrolysis-deficient Rab10-Q68L mutants (F). MG-132 (10 mM) was added to the cell cultures 4 h before harvesting the cells.
(G) UBQLN4 co-precipitates with Rab8a-T22N but not Rab7a-T22N. T7-tagged Rab8a and Rab7a in their GDP-bound forms were co-expressed with Flag-UBQLN4 in HeLa cells. The cells were treated with 10 mM MG-132 for 4 h. Flag-precipitates were blotted with anti-T7 and anti-Flag antibodies, respectively. iScience Article did not co-precipitate with UBQLN4 under identical conditions ( Figure 1E). Similarly, GDP-bound Rab10-T23N co-immunoprecipitated with UBQLN4, whereas constitutively active Rab10-Q68L did not ( Figure 1F), even though Rab10-Q68L accumulates in much larger amounts than Rab10-T23N due to their difference in stability. Notably, the highly stable Rab7a protein showed little affinity for UBQLN4 ( Figure 1G), even in its GDP-bound form. These observations suggest a clear preference of UBQLN4 for unstable Rab proteins, and their interactions with UBQLN4 are nucleotide-specific.
A previous study suggested that defective model transmembrane protein substrates for pre-emptive quality control are predominantly recognized by UBQLN4 over UBQLN1, which is a UBQLN4-related ubiquitin-like (UBL) and ubiquitin-associated (UBA) domain protein. 47,50 To determine the specificity of UBQLN family proteins for GDP-bound Rab8a, we compared the binding of UBQLN4 and UBQLN1 to Rab8a-T22N. As shown in Figure 2A, UBQLN4 co-immunoprecipitated with Rab8a-T22N more effectively than UBQLN1, suggesting that UBQLN4 possesses a higher affinity for inactive species of Rab8a. In accordance with this observation, the STI-II region, which determines the substrate specificity of UBQLN4, 47 is necessary for the recognition of Rab8a-T22N (Figures S1A and S1B).
Because UBQLN4 was found to be associated with Rab8a-T22N, we examined the possible role of UBQLN4 in the instability of GDP-bound Rab8a. We found that UBQLN4 knockdown effectively blocked the degradation of Rab8a-T22N ( Figure 2B), suggesting that endogenous UBQLN4 is essential for the degradation of GDP-bound Rab8a.
UBQLN4 recognizes the exposed hydrophobic residues of client proteins and targets them for degradation. 47 Therefore, we suspected that UBQLN4 could recognize the exposed hydrophobic residues of GDP-bound Rab8a upon its release from membranes. Previous studies suggested that hydrophobic residues in the Switch I loop are exposed to the cytoplasmic surface when Rab8a is bound to guanosine diphosphate (GDP) ( Figure 2C, upper panel). 51,52 To examine whether this exposed hydrophobicity in the GTPase domain is important for UBQLN4 recognition, we used the 3IS mutant, in which three hydrophobic residues in the Switch I region were substituted with hydrophilic serine residues ( Figure 2C, lower panel). We found that this highly stable Rab8a-T22N-3IS mutant 35 failed to co-immunoprecipitate with UBQLN4 ( Figure 2D). The interaction between UBQLN4 and Rab8a also depends on ubiquitination events, given that treatment with TAK-243 (MLN7243), a small molecule inhibitor of the ubiquitin-activating enzyme E1, weakened the co-precipitation of Rab8a-T22N with UBQLN4 ( Figure S1C). Collectively, these findings indicate that UBQLN4 targets Rab8a for degradation in a nucleotide-and ubiquitination-dependent manner by recognizing exposed hydrophobic stretches in the GTPase domain.
The GEF substrate portion of wild-type Rab8a is unstable for proteasomal degradation To examine whether the nucleotide exchange of wild-type Rab8a is indeed critical for UBQLN4 recognition, we depleted the endogenous GEF for Rab8a, RABIN8/RAB3IP. 52,53 This depletion should theoretically lock the GDP/GTP exchange cycle of wild-type Rab8a in the GDP-bound state, thereby causing the cytoplasmic accumulation of GDP-bound species. In accordance with this idea, the co-precipitation of UBQLN4 with wild-type Rab8a was stimulated by RABIN8 knockdown ( Figure 2E). Importantly, the amount of wildtype Rab8a that co-precipitated with UBQLN4 under GEF suppression was greatly augmented after treatment with MG-132, a proteasome inhibitor ( Figure 2F), even though the total amount of wild-type Rab8a was insensitive to MG-132. These observations suggest that a cryptic portion of the GEF substrate (a GDP-bound species of wild-type Rab8a) is an unstable protein targeted by UBQLN4 for immediate degradation.

GDP-bound cryptic species of Rab8a is polyubiquitinated
Given that UBQLN4 is known to function as a ubiquitin receptor for proteasome-dependent protein degradation, 47,48 we speculated that GDP-bound Rab8a, a UBQLN4 client, might be polyubiquitinated. We co-expressed T7-tagged ubiquitin with Rab8a-T22N mutant protein in HeLa cells to examine this possibility. Strikingly, Rab8a-T22N massively co-immunoprecipitated with T7-tagged polyubiquitin, while highly stable GTP-bound Rab8a-Q67L did not ( Figure 3A). Polyubiquitination of Rab8a-T22N was enhanced following MG-132 treatment ( Figure 3B), suggesting that polyubiquitinated Rab8a-T22N is a proteasomal substrate for degradation. In accordance with this idea, SDS-mediated denaturation before immunoprecipitation did not abolish the co-precipitation of endogenous polyubiquitin chains with Rab8a-T22N (   Rab8a polyubiquitination was observed not only in the T22N GDP-bound mutant but also in its wild-type form following Rab8a-specific GEF knockdown. Although only a trace amount of polyubiquitin signals co-precipitated with wild-type Rab8a ( Figure 3E, lane 1; Figure 3F, lane 1), Rab8a-GEF depletion augmented the signals of covalent modifications of wild-type Rab8a that co-precipitated with UBQLN4 ( Figure 3E, smear signals in lane 2). In accordance with this observation, co-precipitation of T7-tagged ubiquitin with wild-type Rab8a was stimulated with RABIN8 knockdown ( Figure 3F, upper panel; compare lanes 2 and 3). These signals of wild-type Rab8a were not detected in the absence of MG-132 ( Figure 3F; compare lanes 3 and 4), suggesting that the polyubiquitinated species induced by RABIN8 depletion are proteasomal substrates. Intriguingly, RABIN8 depletion also stimulates mono-ubiquitination of Rab8a WT at 35-kDa irrespective of MG-132 addition ( Figure 3F, second panel, indicated by a white arrowhead). These observations collectively support the notion that native Rab8a can be ubiquitinated in a GDP-bound species-specific manner.

RNF126 ubiquitin ligase is essential for the degradation of GDP-bound Rab8a
The identification of the ubiquitination machinery responsible for the targeted degradation of Rab8a proteins in a nucleotide-dependent manner is crucial for understanding the novel regulatory mechanism of this small GTPase protein. RNF126 was identified as a BAG6-associated RING finger E3 ligase that is critical for the pre-emptive quality control of mislocalized prion proteins. 45 Therefore, we investigated whether RNF126 has a physical affinity for GDP-bound Rab8a targeted for degradation.
We found that RNF126 co-precipitated with GDP-bound Rab8a-T22N more efficiently than its GTP-bound Q67L counterpart ( Figure 4A). This was also observed for RNF115, an RNF126-related RING finger family protein ( Figure S2A) that is also a BAG6-associated E3 ubiquitin ligase. 45 Collectively, these results show that the major components of the pre-emptive quality control machinery, including RNF126 family E3 ligases, UBQLN4, and BAG6, are all associated with GDP-bound inactive Rab8a in the cytoplasm.
Similar to UBQLN4, the co-immunoprecipitation of RNF126 with wild-type Rab8a was enhanced following the depletion of RABIN8, an endogenous GEF ( Figure 4B). Furthermore, the amount of wild-type Rab8a that co-precipitated with RNF126 was augmented by MG-132 treatment with GEF suppression ( Figure 4B). These results correspond to the observation that RABIN8 depletion stimulated the polyubiquitination of (C) Endogenous ubiquitin covalently modifies GDP-bound Rab8a. Before immunoprecipitations, cell extracts were boiled with 1% SDS to denature Rab8a-T22N, and then diluted. Flag-tagged Rab8a-T22N was immunoprecipitated and probed with an anti-polyubiquitin FK2 antibody to detect covalent ubiquitin conjugation on Rab8a-T22N. UBQLN4 was used as a negative control for direct ubiquitination.
(D) Flag-tagged BAG6 fragments and T7-tagged Rab8a-T22N were co-immunoprecipitated by anti-Flag antibody and probed with an anti-T7 antibody to detect co-precipitated Rab8a-T22N. Prolonged exposure of T7 blots with N400 or N468 precipitates shows multiple ladder-like signals at approximately 8 kDa intervals (indicated by white arrowheads). N219, N200, and N468 stand for N-terminal 219, 200, and 468 residues fragments of BAG6, respectively, and N400 stands for a tandemly fused fragment of N200. ''Mock'' indicates empty vector transfection as a negative control. . Rab8a co-precipitates with RNF126 in a nucleotide-specific manner (A) T7-tagged RNF126 was preferentially co-precipitated with Flag-tagged Rab8a-T22N, and less efficiently coprecipitated with Rab8a-Q67L, a constitutively active mutant. MG-132 (10 mM) was added to the cell culture 4 h before harvesting. Note that the T22N mutant Rab8a was expressed at lower levels than either WT or Q67L (see Flag-blot panel). Graph shows the relative signal intensities of co-precipitated T7-RNF126 with Flag-Rab8a (signal intensities of T7 blots were divided by that of Flag blots). The signal of WT was defined as 1.0. Data are presented as means G standard deviation (SD) from five independent experiments. *p < 0.05 (Student's t test). See also Figure S2A.
(B) Enhanced physical interaction was observed between wild-type (WT) Rab8a and RNF126 following RABIN8 depletion (top panel), even though the total amount of WT Rab8a did not change (second panel iScience Article wild-type Rab8a ( Figures 3E and 3F), suggesting that proteasomes degrade a cryptic (GDP-bound) portion of wild-type Rab8a in a nucleotide-dependent manner.
When flag-tagged RNF126 was immunoprecipitated, we noticed that co-precipitated Rab8a-T22N showed multiple ladder-like signals at approximately 8-kDa intervals ( Figures 4C and 4D, indicated by white arrowheads and an arrow, see also S2B). Prolonged exposure of the Rab8a blot revealed high-molecular-weight smear signals of Rab8a ( Figure 4D, a right lane, indicated by a red line), which were not visible in the control precipitates ( Figure 4D, a left lane). These multiple ladder-like signals were abolished by treatment with TAK-243 ( Figures 4E and S2C, indicated by white arrowheads and an arrow). These observations suggest that RNF126 is associated with covalently ubiquitinated Rab8a species. Although depletions of BAG6, UBQLN4, and RNF126 all resulted in stabilization of GDP-bound Rab8-family proteins, the polyubiquitination state for Rab8a-T22N seems to be different. As shown in Figure 5D, polyubiquitin modifications of Rab8a-T22N were suppressed by depletion of endogenous BAG6, whereas UBQLN4 depletion enhanced the accumulation of polyubiquitinated Rab8a-T22N. These observations suggest that BAG6 is required for the recognition of Rab8a-T22N before its polyubiquitination ( Figure 5F), and that UBQLN4 is critical for proteasomal targeting of polyubiquitinated Rab8a-T22N ( Figure 5F). The interaction between RNF126 and GDP-bound Rab8a is dependent on BAG6 ( Figure 5E), which might explain why BAG6 depletion blocked the polyubiquitination of Rab8a-T22N. We observed that RNF126 depletion did not completely suppress polyubiquitination of Rab8a-T22N, suggesting that one or more other E3 ligases may collaboratively participate in this process.
Accumulation of GDP-bound species of wild-type Rab8a in RNF126-depleted cells To examine whether cryptic GDP-bound species of wild-type Rab8a truly accumulate in RNF126-depleted cells, we need a specific probe for GDP-bound Rab8a. To solve this issue, we took advantage of an exogenously expressed BAG6 fragment as a GDP-form-specific probe, given that BAG6 possesses superior binding affinity to GDP-bound Rab8a. 35 This GDP-form-specific probe enabled us to detect the enhanced accumulation of inactive species of wild-type Rab8a in the cytoplasm of RNF126-depleted cells ( Figures 6A  and 6B, upper panels). We confirmed that the total amount of wild-type Rab8a was unchanged with RNF126 depletion (Figures 6A and 6B, second panels). Collectively, the endogenous ubiquitin ligase of the pre-emptive quality control pathway is necessary to selectively limit the excessive accumulation of Rab8 subfamily proteins in their cryptic GDP-bound form.
Deficiency in RNF126 induces the defective distribution of Patched 1 Because the RNF126 was suggested to play a role in the instability of GDP-bound Rab8a ( Figures 5A and  5C), and RNF126 deficiency was found to induce mis-accumulation of GDP-bound Rab8a ( Figures 6A and  6B), we were interested in examining whether this ubiquitin ligase is linked with the vesicular trafficking of Rab8 client proteins. Patched 1 (Ptc1) is a sonic hedgehog receptor that predominantly localizes in cytoplasmic endosomes 55 and its vesicular localization is controlled by Rab8a. 35 Therefore, we examined the endosomal distribution of Ptc1 in RNF126-depleted cells. In control siRNA-treated cells, the majority of Ptc1 was detected in cytoplasmic vesicular structures ( Figures 6C-a), as reported previously. 35,55 In contrast, Ptc1 was present mostly in congregated perinuclear vesicular compartments and reduced cytoplasmic endosomal structures in RNF126-defective cells (Figures 6C-b and S4). Double depletion of iScience Article two related E3 ligases (RNF126 and RNF115) also altered Ptc1 distribution (Figures 6C-c). Note that the expression level of Ptc1-Flag protein was not affected by RNF126 siRNA (Figure 6D), and no overt difference in the distribution of early endosome marker EEA1 was observed in RNF126-depleted cells ( Figure S4).
These results indicate that endogenous RNF126-family ubiquitin ligases play a role in controlling the localization of Rab8a client membrane proteins.
Pre-emptive quality control machinery is necessary for primary cilium formation The primary cilium is a microtubule-based sensory organelle that protrudes from the surface of almost all vertebrate cells. 56 The formation of primary cilia (i.e., ciliogenesis) requires polarized vesicular trafficking events. 56,57 Importantly, Rab10 depletion in Rab8a/b double knockout cells has been reported to lead to a significant reduction of ciliation, 49 suggesting that these paralogs have a redundant but critical role in ciliogenesis. 31,32,49,58,59 Given the present finding that the pre-emptive quality control system is essential for eliminating GDP-bound Rab8 and Rab10, this quality control machinery may be necessary for ciliogenesis. We therefore, tested this hypothesis by depleting several components of the pre-emptive quality control machinery.
To  Figures 7A-d, and h). We measured the percentage of ciliated cells by scoring ARL13B-positive structures greater than 1 mm in length. We found that 23.6% of cells (255/1082 cells) were ciliated in the presence of control siRNA, whereas only 6.1% of BAG6 siRNA-treated cells (21/344 cells) were ciliated ( Figure 7C). Depletion of UBQLN4 by siRNA also resulted in shorter primary cilia (7.1%, 37/522 cells). Although RNF126 knockdown did not significantly reduce the rate of ciliated cells having cilia longer than 1 mm, we found that the average length of primary cilia was significantly shorter in RNF126 siRNA-treated cells than in control cells ( Figure 7D; average length: 1.6 mm in RNF126-depleted cells, 2.3 mm in control cells).
The cilium is assembled from the distal end of the mother centriole or basal body, which serves to nucleate growth of the axoneme, which is a microtubule doublet. To examine the formation and position of the axoneme on the basal body, we double-stained serum-starved RPE1 cells with antibodies against anti-acetyltubulin (a cilia axoneme marker) and anti-pericentrin (a basal body marker). After 48 h of serum starvation, the mother centrioles were associated with the ciliary pockets with extended axonemal shafts ( Figures 7B and S5A). Depletion experiments of pre-emptive quality control components, including UBQLN4, RNF126, and BAG6, confirmed a reduction in the number of elongated axonemes on the normal basal body, as determined by acetylated tubulin and pericentrin staining ( Figures 7B and S5A-S5D). These observations further support the critical function of these proteins as degrading machinery for ciliogenesis. We confirmed that endogenous Rab8a protein in RPE1 cells can be physically recognized by BAG6 and that their interaction was stimulated by both MG-132 treatment and RABIN8 siRNA ( Figure S5E). These observations suggest that the proteasome is responsible for the elimination of endogenous GDP-bound Rab8a in the RPE1 cells.
Cliogenesis and cell division are mutually exclusive events, and primary cilia are known to be formed specifically during the G0/G1 phase. To examine whether the phenotypes observed in this study are not the results of secondary cell-cycle defects, we performed flow cytometric analysis upon depletion of the pre-emptive quality control machinery. Knockdown of UBQLN4 or RNF126 did not lead to any decrease     Figure S6) compared with the control (95.4% G0/G1 cells). Although BAG6 depletion showed a slight reduction in the G0/G1 population (89.8% G0/G1 cells in BAG6 depletion) as previously reported (94% of G0/G1 cells in control cells, 87% in BAG6 depleted cells), 60 the observed effect was very small. We concluded that the reduced G0/G1-phase population was not a major cause of the overt defects in cilia formation in UBQLN4-, RNF126-, and BAG6-defective cells. Taken together, these findings suggest that the pre-emptive quality control machinery have an indispensable role in ciliogenesis.

DISCUSSION
The Rab family of small GTPases has long been thought to be stable proteins; indeed, the half-lives of Rab8a and Rab10 in their GTP-bound form are very long. In contrast to this generally believed notion, our study suggested that the cryptic portion of Rab8, that is a GDP-bound form, is extremely unstable. The instability of GDP-bound Rab8a seems not to be caused by unexpected structural defects resulting from the T22N mutation, given that the amount of wild-type Rab8a that co-precipitated with UBQLN4 under GEF suppression was also augmented after proteasomal inhibition ( Figure 2F). Similarly, polyubiquitination of wild-type Rab8a was stimulated following GEF depletion and proteasomal inhibition ( Figures 3E and 3F). Although we can't exclude the possibility that T22N mutation may alter the protein stability in ways unrelated to the nucleotide-bound state, we think this is largely unlikely. Wild-type Rab8a looks highly stable as a whole because the equilibrium between GTP-and GDP-binding to Rab8a is largely in favor of GTP.
The chaperone protein BAG6 plays a role in the instability of GDP-bound Rab8 subfamily proteins. BAG6 is a multifunctional protein that participates in diverse processes, such as protein biogenesis and degradation, with many distinct regulatory subunits dedicated to these respective processes. For example, BAG6 associates with the transmembrane recognition complex (TRC) to facilitate the biogenesis of tailanchored membrane proteins. 42,61,62 It is also a component of the pre-emptive protein quality control machinery. 38,45,47 Therefore, distinguishing the machinery required for BAG6-dependent Rab8a regulation is important. In this study, we provide the first evidence that pre-emptive quality control components, including RNF126 and UBQLN4, are responsible for the instability of GDP-bound Rab8 subfamily proteins ( Figures 2B, 5A, 5C, and 5D). The pre-emptive quality control system is known to selectively target aggregation-prone hydrophobic clients to BAG6-associated ubiquitin ligases such as RNF126 and the ubiquitin receptor UBQLN4. 38,39,45,47,63,64 Because Rab8a exposes its hydrophobic stretch in a GDP-bound form-specific manner 51 and it has been reported that GDP-bound Rab8a is aggregation-prone in the cytoplasm, 29 GDP-bound Rab8 can be a preferential target for these components of the ubiquitination machinery.
Since Rab10 protein is post-translationally modified by highly lipophilic geranylgeranyl groups (consisting of 20-carbon isoprenoid groups) at the C-terminal Cys199 and Cys200 residues, we suspected that these iScience Article hydrophobicities might contribute to its instability. To examine whether these geranylgeranylations are essential for the association of Rab10-T23N with BAG6, we mutated cysteine residues to serine (designated C199S/C200S mutant Rab10, Figure S7A). We found that C199S/C200S mutations of Rab10 did not reduce its BAG6 binding as the GDP-bound cytosolic form ( Figure S7B). In accordance with this, C199S/C200S mutant Rab10 remains highly unstable ( Figure S7C), suggesting that C-terminal geranylgeranyl groups are not responsible for rapid degradation of GDP-bound Rab10. These observations are consistent with our previous report that the exposed hydrophobicity resulting from isoprenylation of the C-terminus is not involved in the instability of GDP-bound Rab8a. 35 This view is further supported by our observation that the overexpression of GDP dissociation inhibitor 2 (GDI2), a known cytoplasmic chaperone dedicated for isoprenylated Rab proteins in its GDP-bound form ( Figure S7A), did not stabilize Rab8a-T22N (Figure S7D) or Rab10-T23N ( Figure S7E) in HeLa cells. Therefore, a shortage of GDI is not sufficient to explain the observed instability of GDP-bound Rab8a subfamily proteins.
Rab8-family proteins are responsible for vesicular trafficking between cytoplasmic organelles. Unlike Ras ( Figure S8), GTP hydrolysis and the subsequent retrieval of Rab8a from membranes occur at locations far from the departure organelles. It remains unclear how GDP-bound Rab8a generated at the destination membranes returns to its departure organelles through long-distance transportation. We suggest that most, if not all, GDP-bound Rab8a and Rab10 must be targeted immediately for degradation after cytoplasmic release from the destination membranes ( Figure 8). In relation to this, it has been reported that the accumulation of GDP-bound Rab proteins in the cytoplasm is detrimental to intracellular membrane Figure 8. Pre-emptive quality control machinery targets GDP-bound Rab8a/10 for ubiquitin-dependent degradation GDP-bound Rab8a/10 released from the destination membranes is thought to return to its departure organelles where it is reactivated by specific GEFs. In this study, we show that a non-negligible portion of GDP-bound Rab8a/10 is targeted for degradation by the pre-emptive quality control machinery. The exposed hydrophobic region of GDP-bound Rab8a in Switch I is critical for its ubiquitin-mediated degradation, thereby preventing the excessive accumulation of inactive Rab species during GDP-GTP cycling. The association of pre-emptive quality control machinery components with wild-type Rab8a was greatly enhanced under RABIN8 (GEF)-suppressed conditions. Furthermore, depletions of BAG6, UBQLN4, and RNF126 resulted in the stabilization of GDP-bound Rab8a/10, thereby preventing ciliogenesis. These results suggest that the protein degradation machinery plays a critical role in maintaining the integrity of membrane trafficking events by limiting the excessive accumulation of GDP-bound forms of Rab8a/10 small GTPases. See also Figures  iScience Article trafficking. [23][24][25][26][27][28][29][30] The inadequate elimination of GDP-bound Rab8a might cause the unregulated accumulation of aggregation-prone Rab8a ( Figures 6A and 6B), which might impair vesicular trafficking. 17,31,32,65 Therefore, pre-emptive quality control-mediated degradation of Rab8-subfamily proteins during GDP/ GTP cycling are necessary for selectively limiting the excessive accumulation of Rab8 in its GDP-bound form to maintain the integrity of vesicular trafficking (Figure 8).
In this context, it is interesting to recall previous reports that RNF126 plays a role in some membrane trafficking events via an unknown mechanism. For example, RNF126 is involved in the cytoplasmic incorporation of cell surface receptors 66 and plays a role in the efficiency of endosomal sorting. 67 Additionally, the golgi apparatus is mildly disrupted in RNF126-depleted cells, 67 similar to the case in BAG6 depletion. 35 Although the substrates of the RNF126 ubiquitin ligase in this context were poorly investigated, the data in the present study suggest that this ubiquitin ligase is necessary for preventing the accumulation of GDP-bound Rab family small GTPases.
Recently, it was reported that RNF115, an RNF126 paralog, interacts with Rab1A and Rab13 predominantly in their GDP-bound forms, thereby inducing K11-linked polyubiquitin modifications. 68 RNF115 modulates the vesicular trafficking of toll-like receptor from the ER/golgi apparatus to the cell surface. 68 Since residues identified as ubiquitination sites of Rab1A (Lys49 and Lys61) and Rab13 (Lys46 and Lys58) were conserved in multiple Rab proteins, including Rab8a and Rab10, 68 it is conceivable that Rab8-family proteins were also polyubiquitinated on these residues. Although proteasomal degradation of Rab1A and Rab13 has not been reported, our preliminary study suggests that GDP-bound Rab13 is highly unstable in HeLa cells, such as in the case of Rab8a-T22N and Rab10-T23N mutants.
A notable finding in this study was that primary cilia formation was disrupted in cells with depleted components of the pre-emptive quality control machinery (Figure 7 and S5). Depletion of these components also disrupted the intracellular localization of the sonic hedgehog receptor Ptc1 ( Figure 6C). The sonic hedgehog morphogen requires Rab8 subfamily proteins for transduction of its intermediate signaling at the tip of the primary cilium. 69 Accordingly, it has been reported that the forced accumulation of GDPlocked variants of Rab8a or depletion of RABIN8 abolishes ciliogenesis in hTERT-RPE1 cells. 31,32 In addition, the Exocyst-Rab8-RABIN8 complex has a crucial role in the docking of exocytic vesicles to the plasma membrane and is necessary for primary ciliogenesis. 59,70 Given that defects in primary cilia are associated with the defective transduction of sonic hedgehog signals 71 and with a large class of inherited disorders collectively known as ciliopathies, 56,71,72 the pre-emptive quality control system might possess huge importance in the pathomechanism of ciliary function-associated disorders. This should be explored further in future study.
We recently found that BAG6 is necessary for stress fiber formation via supporting RhoA stability. 73 Since blocking stress fiber was reported to facilitate ciliogenesis and results in abnormally long cilia, 74 observed defects in cilial elongation as a result of BAG6 depletion are consequence of the predominant suppression of cilia over the effects of defective stress fiber formation. In addition, neither UBQLN4 nor RNF126 is involved in the formation of stress fiber, 73 suggesting that the observed phenotypes of UBQLN4 and RNF126 depletion in ciliogenesis are independent of stress fiber formation.
An interesting issue that remains to be elucidated is whether the pre-emptive quality control machinery is responsible for other Rab family proteins. The human genome encodes more than 60 Rab genes. 9,75 The instability of GDP-bound Rab proteins seems to be subfamily specific ( Figures 1A-1D) and partially correlates with their affinity for BAG6. Many mammalian Rab family proteins show high binding affinity with BAG6, 35 thus, these BAG6-phile Rab proteins might be potential clients for pre-emptive quality control in a nucleotide-specific manner. Therefore, the effects of defective pre-emptive quality control on membrane trafficking might not solely rely on the dysfunction of Rab8a and Rab10. RNF126 has also been reported to play a role in retrograde sorting of the cation-independent mannose 6-phosphate receptor. 66 Considering that vesicular trafficking of the mannose 6-phosphate receptor is mediated by Rab9a and that Rab9a has the highest affinity for BAG6, 35 Rab9a might be an excellent candidate client protein for pre-emptive quality control. We are investigating this possibility in parallel studies.
In summary, we found evidence that the pre-emptive quality control machinery is essential for proper membrane trafficking via the irreversible degradation of Rab8 subfamily proteins in a nucleotide-specific iScience Article manner ( Figure 8). This finding implies that the simple nucleotide exchange cycle (as shown in the case of Ras, Figure S8) might not be sufficient to explain the regulation of Rab8 subfamily small GTPases, contrary to the currently accepted paradigm. We speculate that the cytoplasmic transport of GDP-bound Rab8a back to the departure organelle may be, at least in part, dispensable. Instead, excessive amounts of GDP-bound Rab8 subfamily proteins must be degraded immediately (Figure 8). The importance of the degradation of other Rab family small GTPases should be explored further to determine the fate of these mobile proteins after GTP hydrolysis at the end of their journey.

Limitations of the study
This study showed Rab8a and Rab10 in their GDP-bound forms are highly unstable, whereas Rab7a is stable in HeLa cells. The instabilities of other GDP-bound Rab-family proteins remain an important area for future investigation. Most of the experiments in this study were performed in cells ectopically expressing the Rab GTPases. Although, we demonstrated that co-precipitation of endogenous Rab8a with BAG6-probe was stimulated by GEF suppression and MG-132 treatment, direct estimation of the endogenous level of GDP-bound Rab8a should be necessary. Finally, a detailed mechanistic understanding of how RNF126 cooperates with BAG6 to promote polyubiquitination of GDP-bound Rab8a will require biochemical reconstitution from purified components. Such studies will be crucial for elucidating the mechanism by which pre-emptive protein quality control machinery selectively regulates specific species of Rab-family proteins in a nucleotide-dependent manner.

STAR+METHODS
Detailed methods are provided in the online version of this paper and include the following:    Data and code availability d All data reported in this paper will be shared by the lead contact upon request.

AUTHOR CONTRIBUTIONS
d This paper does not report original code.
d Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

Plasmid construction
Full-length cDNAs for UBQLN4, RNF126, RNF115, BAG6, Ptc1, GDI2 and Rab-family proteins were amplified by PCR from cDNA library derived from HEK293 cells. The PCR fragments were cloned into the pCI-neo expression vector (Promega, Madison, WI) for expression in cultured cells. It should be noted that expression vectors encode three repeats of a Flag-, T7-, or single S-tags at the N-terminus of their products. The truncated and mutated versions of the products were constructed by PCR. These expression vectors were used for experiments after verification of the sequence of inserted DNA.

Cell culture and transfection
HeLa cell was cultured in Dulbecco's modified Eagle's medium (Sigma Chemical Co.) supplemented with 10% heat-inactivated calf serum at 37 C under a 5% CO 2 atmosphere. Transfection in HeLa cells with pCIneo expression vectors was performed by using PEI MAX (Polysciences, Inc, PA, USA) or HilyMax (Dojindo, Kyoto, Japan) according to the protocols supplied by the manufacturers. At 24 h after plasmid transfection, the cells were harvested and subjected to immunological analysis, unless otherwise noted.

RNA interference
UBQLN4 depletion in human cells was performed with duplex siRNAs covering the targeted sequences; 5 0 -CUCUUCAGAUGCUGGCAGUTT-3 0 (UBQLN4 siRNA#3) RABIN8 depletion in human cells was performed with duplex siRNAs covering the targeted sequences;

Immunocytochemical observations of Ptc1
For the immunostaining of Ptc1 in HeLa cells, we used pCI-neo-based modified expression plasmid with largely compromised promoter activity (by deleting promoter region partially) to keep the expression of Flag-tagged Ptc1 protein (Ptc1-Flag) at nearly physiological levels. 35 This was due to preclude the possibility of inappropriate aggregation of this polytopic TMD protein in the cytosol. Transfected HeLa cells were grown on micro coverglass (Matsunami, Osaka, Japan), fixed by incubating in 4% paraformaldehyde for 10 min at room temperature, and permeabilized with 0.1% Triton X-100 for 3 min at room temperature. All cells after fixation and permeabilization were blocked with 3% calf serum solution in PBS for 30 min at room temperature, reacted with anti-Flag M2 primary antibody at 4 C for overnight, and were subsequently reacted with secondary antibody, Alexa Fluor R 488 goat anti-mouse IgG. To observe the nucleus, cells were treated with 2.5 mg/mL Hoechst 33258. Immunofluorescent images were obtained by BIOREVO BZ9000 fluorescence microscope (Keyence, Osaka, Japan). One day after plating of hTERT-RPE1 cells, siRNAs (5 nM) were transfected into the cells by using Lipofectamine RNAi MAX (Thermo Fisher Scientific) according to the manufacturer's instructions, and cultured for 72 h in complete medium to approximately 60-70% confluence. To induce ciliogenesis, the complete medium was replaced with serum-free medium, and the cells were cultured for a further 24 h. For immunocytochemical observations of primary cilia, hTERT-RPE1 cells were fixed in 4% paraformaldehyde in PBS for 15 min at room temperature, and after two washes with PBS, the cells were permeabilized in ice-cold methanol for 5 min. Then, cells were blocked in 5% FBS-PBS for 30 min at room temperature. Cilia were analyzed by immunostaining of endogenous ARL13B, a marker for primary cilia, with anti-ARL13B antibody or antiacetylated a-tubulin as primary antibodies, and were subsequently reacted with a secondary antibody, Alexa Fluor R 488 goat anti-mouse IgG. To evaluate the length and frequency of primary cilia, ARL13B-positive immunosignals were measured using ImageJ software under the BZ-9000 fluorescence microscope (Keyence). Data for at least 100 cilia per treatment were obtained from at least three independent biological replicates, and the values are presented as means G S.D.

Flow cytometry for cell cycle analysis
For quantification of cell cycle stage distribution, the hTERT-RPE1 cells were washed twice with PBS, and harvested by trypsinization from culture plates. The cells were re-suspended in 500 mL PBS, and 1 mL of ethanol was subsequently added to the cell suspension. After incubation at 4 C for at least 1 h, the cells were washed 3 times with PBS. The cells were suspended again in PBS and incubated at 4 C for 30 min. After incubation, RPE1 cells were centrifuged and the supernatant was removed. Next, 0.5 mL of 250 U/mL RNase A in PBS was added to cell, and the cells were incubated at room temperature for 20 min. After incubation, 50 mg/mL propidium iodide (PI) was added. The cell cycle profile was analyzed using a flow cytometer (model BD Accuri C6, BD Bioscience) by 488 nm excitation. 76 FCS files from cell cycle assay were extracted and analyzed using FCS Express 7.

QUANTIFICATION AND STATISTICAL ANALYSIS
Statistical analysis was performed using R (4.1.3) or Microsoft Excel 2016. Statistical details of experiments are stated in the legends of figures displaying the respective data, including the statistical tests used, the number of replicates and number of investigated cells, and measures of precision. For two-sample comparisons, students' t-test or the Mann-Whitney U test was used. For multiple comparisons, statistical significance was tested by one-way ANOVA and Dunnett's test. p value <0.05 was considered statistically significant (p < 0.05*, 0.01**, 0.001***). Note that independent experiments were repeated at least three times to ensure the reproducibility of the data.

ll
OPEN ACCESS