Rab34 small GTPase is required for Hedgehog signaling and an early step of ciliary vesicle formation in mouse

The primary cilium is a microtubule-based protrusion from the cell surface on most vertebrate cells. It is a sensory organelle and also a location for the transduction of many signaling molecules, including Hedgehog (Hh), Wnt and PDGF. Defects in ciliary structure and function are associated with a broad spectrum of developmental abnormalities and metabolic diseases, i.e. ciliopathies, such as open brain, holoprosencephaly, polydactyly, microphthalmia, cystic kidney disease and retinal degeneration, and can also result in obesity (Bangs and Anderson, 2017; Gerdes et al., 2009; Reiter and Leroux, 2017).

Most of developmental defects associated to ciliary gene mutation are attributed to impairment of Hh signaling between Smoothened (Smo) – a seven transmembrane G-protein-coupled receptor – and the transcription regulators Gli2 and Gli3 (Bangs and Anderson, 2017). Proteolytic cleavage of the full-length Gli2 and Gli3 transcriptional activators (Gli2^FL and Gli3^FL, respectively) into their C-terminally truncated transcriptional repressor forms (Gli2^Rep and Gli3^Rep, respectively) is attenuated and, although activities of Gli2^FL and Gli3^FL are reduced, their levels are increased. Moreover, because the patterning of the central nervous system and limb in vertebrates is dependent on the correct ratio of Gli2^FL/Gli3^FL activators to Gli2^Rep/Gli^Rep, most of ciliary gene mutations – unsurprisingly – cause abnormalities in the central nervous system and limb development.

Ciliogenesis starts with the docking of the mother centriole (also known as basal body) to a ciliary vesicle (Sorokin, 1962), which results from the successive fusion of many small preciliary vesicles near the basal body, a process that requires Ehd1 (Lu et al., 2015). Following docking, the capping protein Cp110 (officially known as CCP110) is removed from the distal end of the mother centriole (according to observations using cultured cells; Schmidt et al., 2009; Spektor et al., 2007), and the basal body extends an axoneme that invaginates the vesicle (Sánchez and Dynlacht, 2016). During and after the process of vesicle docking and axoneme extension, the basal body migrates close to the plasma membrane from the perinuclear area. The vesicle then fuses with the plasma membrane to expose the cilia to the extracellular environment.

The cilia assembly requires vesicular trafficking that brings lipids and membrane proteins to the cilia membrane. The BBSome, an octameric protein complex that comprises seven Bardet–Biedl syndrome (BBS) proteins as well as BBIP1, functions as a coat complex required for sorting of specific membrane proteins to the primary cilia and is required for ciliogenesis (Jin et al., 2010). This ciliogenic function is thought to be mediated, at least in part, by its binding to the guanine nucleotide exchange factor (GEF) for Rab8 – Rabin8 (Nachury et al., 2007). Studies in cultured cells by using overexpression of Rab proteins and RNA interference (RNAi) of Rab RNA expression have demonstrated that the vesicular trafficking is dependent on Rab small GTPases, in particular the Rab11–Rab8 cascade (Knodler et al., 2010; Nachury et al., 2007; Westlake et al., 2010; Yoshimura et al., 2007). In this cascade, Rabin8 (which is enriched at the mother centriole) binds to Rab11 (also enriched at the mother centriole) and, subsequently, activates Rab8, which localizes to cilia and the basal body, to promote ciliary assembly. Surprisingly, however, Rab8 is dispensable for ciliogenesis in vivo, as Rab8a-Rab8b double-knockout mice display normal cilium morphology, length and number. Furthermore, the cilia in these double-mutant mice appear to be functionally normal, as no ciliopathies are detected (Sato et al., 2014). Unlike Rab8, mutations in Rab23 do result in ligand-independent activation of the Hh pathway (Eggenschwiler et al., 2001). This is because Rab23 regulates the transport of Kif17 (a kinesin that binds Gli2 and Gli3, and negatively regulates their activity) into cilia (Lim and Tang, 2015). Rab23 also maintains ciliary steady state of Smo (Boehlke et al., 2010). Nevertheless, cilia form normally in Rab23 mutant mice (Fuller et al., 2014) and, so far, no Rab proteins have been linked to ciliogenesis in vivo. In addition, no Rab GTPases have been shown to regulate the fusion of preciliary vesicles to generate ciliary vesicles – even in cultured cells.

Rab34 is a Golgi-associated small GTPase (Wang and Hong, 2002). Loss of Rab34 has been recently shown to result in polydactyly in mice (Dickinson et al., 2016). While this manuscript was being reviewed, a large-scale CRISPR screen study reported that Rab34 is associated with ciliogenesis and Hh signaling in cultured cells (Pusapati et al., 2018). However, no in vivo study and mechanism were presented. We show in this study that Rab34 localized to cilia and that loss of Rab34 resulted in a significant decrease in ciliogenesis in mice. By using transmission electron microscopy (TEM), our study reveals that Rab34 is required for the successive fusion of small preciliary vesicles to form ciliary vesicles and for the migration of the mother centriole from the perinuclear area to the plasma membrane. Consistent with the defect in ciliogenesis, we also found Hh signaling impaired and Gli3 processing reduced in Rab34 mutant mice. This altered the ratio of Gli3^FL activator to Gli3 repressor (Gli3^Rep) and, consequently, caused polydactyly in Rab34 mutants. In support of the impairment of Hh signaling, Rab34 mutant mice exhibited cleft lip and cleft palate. Therefore, Rab34 is required for ciliogenesis and Hh signaling in vivo, and the fusion of preciliary vesicles to form ciliary vesicles.

A recent report has shown that a Rab34 mutation results in polydactyly in mice (Dickinson et al., 2016). Given that many known mutations of ciliary genes affect limb patterning due to reduced levels of the Gli3 repressor (Bangs and Anderson, 2017), we explored the possibility that Rab34 is involved in ciliogenesis. First, the Rab34 gene was mutated in NIH3T3 cells by using CRISPR gene editing with two independent single-guide (sg) RNAs specifically targeting the Rab34 gene (Rab34 sgRNA1 and Rab34 sgRNA2; see Materials and Methods) and a control sgRNA for green fluorescent protein (GFP). Heterogeneous populations of NIH3T3 cells transduced with lentivirus expressing either of the two sgRNAs together with Cas9 were then subjected to immunostaining for the ciliary marker Arl13b. The results showed that ∼90% of GFP sgRNA NIH3T3 cells formed cilia, whereas only ∼30% of the Rab34 sgRNA1 cells and <20% of Rab34 sgRNA2 cells developed cilia. Analysis of clonal Rab34 sgRNA2 NIH3T3 cells, which carried a large deletion in the gene (Fig. S1), showed a similar number of ciliated cells (Fig. 1A). Similar results were also obtained using C3H10T1/2 cells (Fig. S2).

To confirm the reduced ciliogenesis was specifically due to mutations in Rab34, a rescue experiment was performed. For this purpose, a FLAG- and streptavidin (FS)-tagged Rab34 cDNA mutant expression construct (FS-Rab34CR) was created. This mutant construct should be resistant to targeting by Rab34 sgRNA2 because nucleotide mutations created in the region corresponding sgRNA2 destroyed the sgRNA2 PAM sequence – amino acid residues, however, were unchanged. Stable overexpression of FS-Rab34CR in clonal Rab34 sgRNA2 NIH3T3 cells restored ciliogenesis of the mutant cells (Fig. 1A), indicating that reduced ciliogenesis in Rab34 sgRNA2 NIH3T3 cells is specifically due to mutations in the Rab34 gene.

To determine whether Rab34 is required for ciliogenesis in vivo, a Rab34 mutant mouse line was generated by deleting exons 2–6 in a targeted gene knockout approach. The deletion was confirmed by both Southern blotting (Fig. S3A,B) and RT-PCR (Fig. S3C); it removed the Rab34 genomic sequence [amino acid (aa) residues 19–171] from the 256 aa full-length protein and caused a reading frame shift, if exon 1 were spliced to exon 8 (Fig. S3A). Therefore, the mutant allele is most likely null. To determine whether the Rab34 mutant mice form cilia, the neural tube and limb bud sections of both wild-type (WT) and mutant embryos at embryonic day 10.5 (E10.5) were immunostained for ciliary markers Arl13b and acetylated tubulin. The results showed that cilium density in both tissues was significantly reduced. Only 5% of mutant primary mouse embryonic fibroblasts (pMEFs) formed cilia, as compared to >60% in WT (Fig. 1B). Thus, Rab34 is required for ciliogenesis in vivo.

Mice heterozygous for Rab34 mutant did not exhibit any noticeable abnormal phenotypes. Approximately 25% of E18.5 embryos examined were homozygous, and they were all alive (Table 1). However, no newborn homozygotes were alive, indicating that Rab34 mutant mice died around birth. Rab34 mutant embryos exhibited polydactyly and cleft lip and palate (Fig. 2A and Table 1) and occasionally exencephaly (n=1/21). This is consistent with defects in Hh signaling. However, heart, lung and liver of Rab34 mutant embryos appeared to be morphologically normal (data not shown), and heart looping was also normal (Table S1). Thus, the lethality of Rab34 mutant newborns was most likely due to functional defects of a vital organ(s). Since ciliary gene mutations often disrupt Hh signaling, we wanted to know in more detail whether Hh signaling is affected in the Rab34 mutant.

First, neural tube patterning along the anterior and posterior axis of embryos was assessed by examining the specification and patterning of neuronal markers that are regulated by Sonic Hedgehog (Shh) expressed in the notochord and floor plate (Briscoe et al., 2000). Foxa2, Nkx2.2 and Hb9 mark the floor plate, V3 and motoneuron progenitors, respectively, and are specified and patterned by Shh signaling. In contrast, Pax6 is a dorsal marker for Hh signaling and neural tube patterning, and is ventrally restricted by low levels of Shh signaling. Surprisingly, all these markers were specified and appeared to be patterned normally in Rab34 mutant embryos (Fig. 2B).

Second, since most ciliary gene mutations affect the processing of Gli3^FL to Gli3^Rep and of Gli2^FL levels, Gli3^FL and Gli3^Rep were examined by western blotting, showing that levels of both Gli2^FL and Gli3^FL were increased. However, levels of Gli3^Rep were decreased in mutant compared to the WT Rab34 mice (Fig. 2C), indicating that Gli3^Fl to Gli3^Rep processing is reduced in the mutant.

Third, to more directly determine whether loss of Rab34 affects Hh signaling, the RNA levels of Ptch1 and Gli1, two direct Hh targets (Goodrich et al., 1996; Marigo et al., 1996), were examined by RT-qPCR before and after WT and Rab34 mutant pMEFs were stimulated with SAG, a synthetic Smo agonist (Chen et al., 2002). The results showed that, although stimulation with SAG upregulated Ptch1 and Gli1 RNA levels ∼6- and 13-fold in WT cells, it did not so in the mutant cells (Fig. 2D), indicating that Hh signaling was impaired.

Smo and Gli2 accumulate in cilia upon Hh signaling (Chen et al., 2009; Corbit et al., 2005; Haycraft et al., 2005; Rohatgi et al., 2007; Wen et al., 2010). Since a small number of mutant pMEFs still form cilia, we were curious whether stimulation with SAG leads to the accumulation of Smo and Gli2 in cilia. Surprisingly, the percentage of Smo- or Gli2-positive cilia increased in both WT and Rab34 mutant cells in a dose-dependent manner, although it appeared to reach the maximum after stimulation with 100 nM SAG (Fig. 3A-C). The percentages for mutant cells varied substantially because of the small number of ciliated cells that could be found. Thus, the difference between WT and mutant following stimulation with a certain dose of SAG seemed to be statistically significant, whereas it did not with another dose. Nevertheless, the results indicate that these ciliated mutant cells are still capable of responding to Hh signaling, at least based upon accumulation of Smo and Gli2 in cilia. Similar results were observed for Smo in clonal Rab34 sgRNA2 NIH3T3 cells (Fig. 3D,E). Taken together, these results suggest that Hh signaling is impaired in nonciliated, but not in ciliated, Rab34 mutant cells and that Hh signaling is not reduced enough to impact the neural tube patterning, presumably because some of neuroepithelial cells in the neural tube still develop cilia.

Rab34 has been previously shown to localize to the Golgi complex (Wang and Hong, 2002). Given its role in ciliogenesis, we wondered whether Rab34 also localized to cilia and/or the basal body. To that end, NIH3T3 cells that stably expressed LAP-Rab34WT fusion protein (GFP- and S-tagged Rab34WT) or LAP alone were created and subjected to co-staining for GFP and acetylated tubulin. The LAP-Rab34WT fusion protein localized to cilia in ∼25% LAP-Rab34WT-expressing cells, whereas LAP alone localized diffusely in the cytoplasm and nucleus (Fig. 4A,B).

Rab proteins in the cell exist in an either GDP-bound (inactive) or GTP-bound (active) form. We, therefore, created the two mutant constructs LAP-Rab34T66N and LAP-Rab34Q111L that mimic the inactive and active form, respectively. Interestingly, only active LAP-Rab34Q111L but not the dominant-negative inactive LAP-Rab34T66N localized to cilia (Fig. 4A,B). This suggests that Rab34 is dependent on GTP-binding to either enter or reside in cilia. Given the role of Rab34 in ciliogenesis, we also wondered whether overexpression of Rab34 or of its mutants would promote ciliogenesis. Surprisingly, unlike the Rab8–Rab11 cascade (Knodler et al., 2010; Nachury et al., 2007; Yoshimura et al., 2007), it neither promoted nor inhibited ciliogenesis in NIH3T3 cells (Fig. 4A,B). It is worth noting that LAP-Rab34T66N was less stable than LAP-Rab34WT (Fig. 4C). Thus, its inability to inhibit ciliogenesis is probably partially due to its low expression level.

Studies with cultured cells using overexpression and RNAi showed that the Rab8–Rab11 cascade is required for ciliogenesis (Knodler et, al., 2010; Nachury et al., 2007; Westlake et al., 2010; Yoshimura et al., 2007). Consistent with this, Rab8 localizes to cilia and the centrosome, and Rab11 is enriched in the centrosome (Knodler et al., 2010; Nachury et al., 2007; Westlake et al., 2010; Yoshimura et al., 2007). Given the role of Rab34 in ciliogenesis, we next wanted to know whether loss of Rab34 affects subcellular localization of Rab8, Rab11 and Rabin8. To this end, LAP-tagged Rab8, Rab11 and Rabin8 were stably overexpressed in either parental or Rab34 mutant NIH3T3 cells, and then subjected to immunostaining for GFP and γ-tubulin (marking centrioles), or for GFP and acetylated α-tubulin (marking cilia). The results showed that the centrosomal and ciliary localizations of none of these proteins were altered in the mutant cells (Fig. 5).

BBS4 is a part of the BBSome and localizes to centrioles and its satellite matrix; the BBSome functions as a coat complex and is required for sorting of specific membrane proteins to the primary cilia (Jin et al., 2010; Nachury et al., 2007). IFT20 is a subunit of anterograde IFT-B complex and localizes to both centrioles and Golgi network (Follit et al., 2006). Since Rab proteins are essential for membrane vesicle sorting, we wondered whether the subcellular localization of the BBSome and IFT20 are affected by loss of Rab34. However, immunostaining results showed that their subcellular localization in Rab34 mutant cells was similar to those in WT cells and, thus, independent of Rab34 (Fig. 5).

Ehd1 localizes to the ciliary pocket and regulates the fusion of preciliary vesicles in order to form ciliary vesicles (Lu et al., 2015). As expected, overexpressed LAP-Ehd1 was found in parts of cilia near the basal body (presumably the cilium pocket) and in the cytoplasm of WT pMEFs. However, it localized diffusely to the cytoplasm without accumulation at the basal body in Rab34 mutant Rab34 pMEFs (Fig. 5), suggesting that Ehd1 is not recruited to the mother centrioles in mutant cells.

Overexpression and RNAi studies using cultured mammalian cells have shown that Cp110 is a capping protein at the distal end of centrioles and is removed from the mother centrioles during an early step of ciliogenesis (Schmidt et al., 2009; Spektor et al., 2007). Our staining results indicated that Cp110 was properly removed from the mother centrioles in Rab34 mutant cells, as evidenced by the presence of Cp110 signals only at daughter centrioles in most mutant cells (Fig. 5). Thus, loss of Rab34 does not affect the removal of Cp110 from the mother centriole.

Ciliogenesis begins with successive fusion of preciliary vesicles to form ciliary vesicles; these then dock onto distal appendages of the mother centrioles, followed by assembly of the transition zone and axoneme and fusion of ciliary vesicles with the plasma membrane (Sánchez and Dynlacht, 2016). During and after the process of axoneme initiation and elongation, mother centrioles need to migrate from the perinuclear region to near the plasma membrane so that, once ciliary vesicles fuse with the plasma membrane, the mother centriole is tethered to the plasma membrane.

To better understand which step in ciliogenesis was affected by loss of Rab34, the ultrafine structure of cilia on the neuroepithelial progenitors in the neural tube of both WT and mutant embryos was examined by using TEM. Micrographs were divided into five categories according to the state of mother centrioles: (1) mother centrioles not anchored to plasma membrane but surrounded by preciliary vesicles, (2) mother centrioles anchored to plasma membrane and surrounded by preciliary vesicles, (3) mother centrioles docked to ciliary vesicles, (4) mother centrioles with elongated axoneme and, (5) mother centrioles alone. In mutant cells 43.2% of mother centrioles were not anchored to the plasma membrane but surrounded by many preciliary vesicles, whereas, in WT cells only 5.7% of mother centrioles fell into this category. Similarly, 11.1% of mother centrioles in mutant cells showed elongated axonemes, whereas in WT 32% did. This is consistent with our observations of reduced ciliogenesis using immunostaining (Fig. 1B). In addition, we also observed that the percentage of mother centrioles anchored to the plasma membrane in WT embryos was markedly higher than that in the mutant (11.3% vs 1.2%) (Fig. 6A and B). Together, these results suggest that Rab34 regulates the successive fusion of preciliary vesicles to form ciliary vesicles and the migration of the mother centrioles from perinuclear area to plasma membrane.

Here we show that loss of Rab34 results in a significant decrease in ciliogenesis in both cultured pMEFs and mouse tissues (Fig. 1). The decreased ciliogenesis consequently disrupts Hh signaling and causes Hh-related phenotypes (Fig. 2). Thus, Rab34 is required for both ciliogenesis and Hh signaling.

There are two pieces of evidence that Hh signaling is impaired in Rab34 mutants. First, the levels of Gli2^FL protein are increased and Gli3^FL processing is reduced in Rab34 mutant (Fig. 2C). The latter alters the ratio of Gli3^FL to Gli3^Rep, which explains why Rab34 mutant displays polydactyly (Fig. 2A). Second, RT-qPCR results demonstrate that the expression of Ptch1 and Gli1 RNAs, two direct Hh targets, fails to respond to Smo activation following treatment with SAG in Rab34 mutant pMEFs (Fig. 2D). These observations resemble those in other known ciliary gene mutants (Bangs and Anderson, 2017).

However, it was unexpected that the neural tube patterning appears to be generally normal in the Rab34 mutant (Fig. 2B). A similar phenotype has also recently reported in a Dzip1l ciliary gene mutant (Lu et al., 2017; Wang et al., 2018). Since neural tube patterning is largely dependent on Gli2^FL activator activity, this finding suggests that the Gli2^FL activator function is not significantly affected in the neural tube of Rab34 mutants.

It is even more surprising that a small number of Rab34 mutant pMEFs and NIH3T3 cells both still develop cilia and that these ciliated cells appear to be able to respond to stimulation with SAG on the basis of the accumulation of both Smo and Gli2 in cilia (Fig. 3). This suggests that Rab34 is dispensable for ciliogenesis of these cells because either these cells do not normally express Rab34 or, if they do, there are another Rab protein(s) that compensate the loss of Rab34 function. Given that a small number of Rab34 mutant NIH3T3 cells, which are presumably homogenous, remain ciliated, the latter is more likely. While this manuscript was being revised, a large-scale CRISPR mutagenesis study uncovered that Rab34 mutations affect ciliogenesis of NIH3T3 cells. However, the study showed that Smo fails to be enriched in the cilia of Rab34 mutant cells upon stimulation with SAG (Pusapati et al., 2018). The discrepancy between that study and ours is probably due to the sensitivity of different Smo antibodies or a variation of NIH3T3 cells used in two studies.

A genome-wide chromosomal immunoprecipitation (ChIP) study identified Rab34 to be a potential transcriptional target of Gli transcription factors (Vokes et al., 2007). However, our RT-qPCR analysis showed that Rab34 RNA expression in pMEFs was not significantly upregulated upon stimulation with SAG (Fig. S4). Thus, additional studies are needed to determine whether Rab34 is a true Hh target or not.

Both Rab34wt and Rab34Q111L localize to cilia only in a small fraction of overexpressed cells (Fig. 4). However, Rab34T66N does not. This suggests that GTP binding is required for Rab34 to be localized in cilia. Since both Rab34wt and Rab34Q111L exhibit the similar ability to be localized in cilia, GTP binding is probably required only for the protein to enter but not reside in cilia.

Studies in cultured cells by using overexpression and RNAi approaches have linked Rab8-Rab11 cascade to ciliogenesis (Knodler et al., 2010; Nachury et al., 2007; Westlake et al., 2010; Yoshimura et al., 2007). However, an in vivo study has demonstrated that Rab8 is dispensable for ciliogenesis, as cilia are formed normally in Rab8a and Rab8b double mutant mice (Sato et al., 2014). Similarly, cell culture studies have suggested that Rab23 is involved in ciliogenesis (Lim and Tang, 2015; Yoshimura et al., 2007), but the in vivo evidence does not support that (Eggenschwiler et al., 2001; Fuller et al., 2014). Thus, to our knowledge, no Rab proteins have been so far linked to ciliogenesis in vivo and Rab34 is the first one shown to be required for ciliogenesis in vivo.

Ciliogenesis begins with the successive fusion of preciliary vesicles to form ciliary vesicles, which dock onto the distal appendages of the mother centrioles, followed by the formation of the transition zone and initiation and elongation of axoneme (Sánchez and Dynlacht, 2016). During and after the initiation and elongation of axoneme, the mother centrioles must migrate close to plasma membrane from the perinuclear area so that ciliary vesicles can fuse with plasma membrane to expose cilia to extracellular space. A study that used TEM to investigate RNAi of Ehd1 in RPE cells showed that endocytic recycling protein Ehd1 is required for successive fusion of preciliary vesicles to form ciliary vesicles (Lu et al., 2015). In our present study, we show that Rab34 also appears to regulate this process, as there was a marked accumulation of preciliary vesicles in Rab34 mutant (43.2% mother centrioles surrounded by preciliary vesicles in the mutant versus 5.7% in WT). More interestingly, there were fewer basal bodies anchored to the plasma membrane and surrounded by preciliary vesicles in the Rab34 mutant compared with those in WT embryos (1.2% vs 11.3%) (Fig. 6A and B). Thus, Rab34 appears to regulate the making of ciliary vesicles from preciliary vesicles and the migration of the basal bodies to cell cortex.

It is currently unclear how loss of Rab34 results in preciliary vesicle accumulation. Rab34 does not seem to function through Rab8-Rab11 cascade, as the ciliary and/or centriolar localizations of Rab8, Rab11 and Rabin8 are independent of Rab34 (Fig. 5). It does not also appear to act via BBS4 and IFT20, since their subcellular localization is unaffected in the absence of Rab34 (Fig. 5). The relationship between Rab34 and Ehd1 with respect of ciliary vesicle formation is also not clear. Because the vast majority of Rab34 mutant NIH3T3 cells and pMEFs cannot form cilia, and because we could not obtain an antibody that recognizes endogenous Ehd1, we were unable to determine whether Ehd1 ciliary localization is dependent on Rab34 (Fig. 5). Nevertheless, Ehd1 siRNA knockdown appears to impair the recruitment of IFT20 to the mother centrioles and the removal of Cp110 from the mother centrioles (Lu et al., 2015), whereas the subcellular localization of both proteins appears to be normal in Rab34 mutant cells (Fig. 5). This suggests that Rab34 and Ehd1 are likely to act in different steps to regulate the fusion of preciliary vesicles to generate ciliary vesicles. Additional studies are necessary to understand the relationship between Rab34 and Ehd1 in ciliogenesis.

Institutional Animal Care and Use Committee at Weill Cornell Medical College approved this research, including the use of mice and mouse embryonic fibroblasts.

A BAC clone containing C57BL/6J mouse Rab34 genomic DNA sequences was purchased from the BACPAC Resources Center (Oakland, CA) and used to create a Rab34-targeting construct. The construct was engineered by replacing exons 2–6 of the Rab34 gene with the neomycin cassette flanked by loxP sites (Fig. S3A). The linearized construct was electroporated into W4 mouse embryonic stem (ES) cells (129S6/SvEvTac), and targeted ES cell clones were identified by digestion of genomic DNA with Acc65I, followed by Southern blot analysis of ES cell DNA using a probe as indicated (Fig. S3B). Two Rab34-targeted ES cell clones were injected into C57BL/6 blastocysts to generate chimeric founders, which were then bred with Swiss Webster (SW) outbred mice to establish F1 heterozygotes and maintain the mouse colony. PCR analysis was used for routine genotyping with the following primers: forward primer BW1901F1, 5′-CTTCCAGATAGCCTTGTCATC-3′ and reverse primer BW1901R, 5′-CTCTTAGACTTCCTGCATTCC-3′ for WT allele, which produced a 260 bp fragment; and forward primer BW294, 5′-ATTGGGAAGACAATAGCAGGC-3′ and BW1901R for the targeted Rab34 allele, which produced a 230 bp fragment. The primers used for RT-PCR to confirm Rab34 mutant allele were BW1645F, 5′-CTGTGGGGAAGACCTGTCTCATTAAT-3′ on exon 4, and BW1646R, 5′-TGAACCTTTCCTGACCAGCCGTGTCC-3′ on exon 5.

WT and mutant Rab34 primary mouse embryonic fibroblasts (pMEFs) were prepared from E13.5 or E14.5 mouse embryos. pMEFs in the first passage were used for this study. HEK293, C3H10T1/2 and pMEFs were cultured in DMEM supplemented with 10% fetal bovine serum (FBS), penicillin and streptomycin. NIH3T3 and its derived cells were cultured in DMEM supplemented with 10% calf serum, penicillin and streptomycin. The stable cell lines were established by transducing NIH3T3 cells with lentivirus carrying the designated constructs and then selecting clones resistant to puromycin (3 µg/ml) (Santa Cruz Biotechnology).

DNA oligonucleotides used to express single-guide (sg) RNAs specific for GFP and Rab34 were: sgRNA GFP, forward 5′-caccGAGCTGGACGGCGACGTAAA-3′ and reverse 5′-aaacTTTACGTCGCCGTCCAGCTC-3′; Rab34 sgRNA1, forward 5′-caccGCCTGCCAGGAGCACCGGAC-3′, reverse aaacGTCCGGTGCTCCTGGCAGGC; and Rab34 sgRNA2, forward 5′-caccGCGGAGGGACCGCGTCCTGG-3′ and reverse 5′-aaacCCAGGACGCGGTCCCTCCGC-3′. The nucleotides in lower case letters were used for cloning the double oligonucleotides into BsmBI site of pLenti-Crispr-v2 vector, Addgene plasmid #52961 (deposited by Feng Zhang; Sanjana et al., 2014). To generate lentivirus, each of the resulting constructs was transfected into HEK293 cells together with packaging plasmids: pMD2.G and psPAX2, Addgene plasmids #12259 and #12260 (deposited by Didier Trono) using the calcium phosphate precipitation method (Wang et al., 2013). NIH3T3 and C3H10T1/2 cells were transduced with lentivirus by using medium containing lentivirus and polybrene (4 µg/ml) overnight, followed by selection with puromycin (3 µg/ml). Single Rab34 sgRNA2 clones were selected by a limited dilution of puromycin-resistant clones in 96-well plates. The mutations were identified by nested PCR using primers BW1648F 5′-CGGATCAGGCCTCTGCGCCTTC-3′, BW1648R1 5′-GCGGCCTCTTTCTTCAGGCAC-3′, BW1648R2 5′-CTGAGAGGAACATTGTGAGAAT-3′, followed by digestion with PflF1 and DNA sequencing. The mutant clone used for the study contains a 51 bp deletion at the junction of the first exon and intron, which destroys RNA splicing (Fig. S1).

Rab34, Rab8, Rab11, Rabin8 and BBS4 cDNAs were amplified by PCR using a mouse cDNA library and specific primers for these cDNAs. Rab34T66N and Rab34Q111L were generated by overlap extension PCR using primers carrying respective point mutations. pENTR-LAP and pENTR-FS were engineered by inserting either a localization and affinity purification (LAP) tag (GFP and S double tags) or FS (FLAG and streptavidin double tags) into the pENTR4(686-1) vector. All LAP-tagged fusion constructs were created by inserting the indicated cDNAs into pENTR-LAP and then moving the LAP-cDNAs into pLenti-CMV-puro DEST vector by using LR-Clonase II (Thermo Scientific). pLenti-FS-Rab34 construct was created by inserting Rab34 into the pENTR-FS vector followed by moving it into pLenti-CMV-puro DEST vector using LR-Clonase II. The CRISPR-resistant FS-Rab34 construct FS-Rab34CR was created by a PCR-based mutagenesis method using pENTR-FS-Rab34 as a template and primers BW1970F (5′-GACCGCGTaCTGGCaGAGCTGCCC-3′) and BW1970R (5′-GGGCAGCTCtGCCAGtACGCGGTC-3′), in which the nucleotides in lower case indicate mutations. This avoids targeting by sgRNA2 without altering the amino acid sequence of Rab34. All constructs created by PCR were sequenced. pENTR4(686-1) and pLenti-CMV-Puro DEST(w118-1) were Addgene plasmids #17424 and #17452 (deposited by Eic Campeau and Paul Kaufman; Campeau et al., 2009). The LAP-Ehd1 construct was a gift from Christopher Westlake (Lu et al., 2015). The generation and transduction of lentivirus is described above.

For immunofluorescence of neural tube sections, mouse embryos at 10.5 days post coitus (E10.5) were dissected, fixed in 4% paraformaldehyde (PFA)/PBS for 1 h at 4°C, equilibrated in 30% sucrose/PBS overnight at 4°C and embedded in OCT. The frozen embryos were transversely cryosectioned (10 µm/section). Tissue sections around fore- and hind-limb areas were immunostained using antibodies against Foxa2 (concentrated), Nkx2.2, Hb9 [Developmental Study Hybridoma Bank (DSHB), Iowa] and Pax6 [Biolegend, #PRB-278P), or against Arl13b together with anti-acetylated tubulin antibody as described (Pan et al., 2009).

For cell ciliation studies, cells were plated on coverslips that had been coated with 0.1% gelatin for at least overnight and serum-starved with 0.1% FBS for 20–24 h to arrest the cells. For centrosome staining, cells were fixed with cold (−20°C) methanol for 5 min. For cytoplasmic and cilium staining, cells were fixed in 4% PFA/PBS for 15 min. For both centrosome and cilium staining, cells were fixed in 4% PFA for 2 min and then cold methanol for 5 min. After a wash (3×) with PBS, the cells were incubated with blocking solution (PBS, 0.2% Triton X-100, 2% heat-inactivated calf serum) for 20 min. The cells were then incubated with primary antibodies in blocking solution for 1 h at room temperature. In some cases, the fixed cells were incubated with PBS/0.2% Triton-X-100 for 5 min and then immediately with primary antibodies. The cells were washed with PBS and incubated with secondary antibodies in blocking solution for 1 h at room temperature. After washes in PBS (3×), the coverslips were mounted to glass slides using Vectashield mounting fluid supplemented with DAPI (Vector Labs). Staining was visualized using a Zeiss Axiovert fluorescent microscope.

Anti-Cp110 antibodies were generated by immunizing rabbits with a His-tagged Cp110 fragment (1-149 aa). Other antibodies include: anti-Gli2, anti-Gli3, anti-Smo, anti-Arl13b and anti-GFP (all 1:1000) (Wang et al., 2000, 2013; Wu et al., 2017, 2014), as well as antibodies against acetylated tubulin (1:2000), γ-tubulin (1:4000) (Sigma) and IFT20 [1:500; a gift from Gregory J. Pazour (Pazour et al., 2002)]. Secondary antibodies Alexa Fluor 488-conjugated goat anti-rabbit IgG and Cy3-conjugated goat anti-mouse IgG were purchased from Jackson Immunoresearch, Inc.

E10.5 mouse embryos used to detect Gli2 or Gli3 were lysed in RIPA buffer (50 mM Tris-HCl pH 7.4, 150 mM NaCl, 1 mM EDTA, 1%Triton X-100, 1% sodium deoxycholate, 0.1% SDS, protease inhibitors). Stable NIH3T3 cells were lysed in lysis buffer (50 mM HEPES pH 7.4, 150 mM NaCl, 1 mM EDTA, 1% NP40, 10% glycerol, protease inhibitors). Immunoblotting was performed as described (Wang et al., 2000).

E10.5 embryos were fixed with 2% paraformaldehyde/2.5% glutaraldehyde in 0.1 M sodium cacodylate buffer pH 7.4 (EM quality) overnight at 4°C. After washes with sodium cacodylate buffer (3×) and then water (3×), the fixed embryos were transversely cut into 5–6 segments. The segments were incubated with 1% OsO4 for 1 h in dark. After washing 3× with water, segments were incubated in 1% uranyl acetate for 2 h at room temperature, washed 3× with water and dehydrated with a series of graded ethanol. The embryo segments were then infiltrated and embedded with Embed812 mixed solution according to the manufacturer's instruction (Electron Microscopy Sciences, Inc.). Transverse ultrathin sections (65 nm) were cut with a Leica UC6 ultramicrotome (Leica Microsystems), stained with uranyl acetate and lead nitrate and imaged with a JEOL1400 transmission electron microscope.

Confluent WT and Rab34-mutant pMEFs in 60 mm plates were incubated with growth medium containing SAG (200 nM; Cayman Chemical) or vehicle control overnight. Total RNA was isolated from the pMEFs by using TRIzol solution according to the manufacturer's instruction (Thermo Fisher Scientific). Of the isolated total RNAs, 5μg of each were used to synthesize first stranded cDNA using SuperScript II reverse transcriptase according to the manufacturer's instruction (Thermo Fisher Scientific). Reverse transcription quantitative PCR (RT-qPCR) was performed using a qPCR kit (MasterMix-R, Applied Biological Materials, Inc., Canada) and primers described previously (Wang et al., 2017). Primers for Rab34 were: BW1916F 5′-CCAAGAAGGACCTGAGTACTCCT-3′ and BW1916R 5′-TTCCCGGACATTCTCACCAGTGA-3′. For a 20 µl reaction, 0.15 µl of the first strand cDNA was used.

We thank Dr Gregory Pazour for the anti-IFT20 antibody, Dr Christopher J. Westlake for the LAP-Ehd1 construct and Ms Leona Cohen-Gould for help with TEM. Monoclonal antibodies against Foxa2, Nkx2.2 and Hb9 were purchased from the Developmental Studies Hybridoma Bank maintained by the University of Iowa, Department of Biological Sciences, Iowa City, under contract NO1-HD-7-3263 from the NICHD.