Disease-associated mutations in WDR34 lead to diverse impacts on the assembly and function of dynein-2

The primary cilium is a microtubule-based structure that protrudes from the cell surface, with its membrane continuous with the plasma membrane. The organelle acts as a cellular antenna, detecting stimuli from the surrounding environment and transmitting the information into the cell via receptors on its surface to mediate the appropriate response. Primary cilia have a major role in Hedgehog (Hh) signalling, directing proliferation, migration and differentiation of cells with a particular importance during skeletal development (Goetz and Anderson, 2010). Defects in cilia assembly and function lead to a wide range of ciliopathies (Reiter and Leroux, 2017).

Assembly and ongoing function of the primary cilium is mediated by intraflagellar transport (IFT) of cargoes (Kozminski et al., 1993; Pigino, 2021; Webb et al., 2020). IFT is bidirectional and utilizes the organelle-specific motors kinesin-2 and cytoplasmic dynein-2 (hereafter referred to as dynein-2) to transit along microtubule doublets arranged along the central axis of the cilium (Pigino, 2021; Vuolo et al., 2020). Kinesin-2 works with IFT-B trains (Funabashi et al., 2018) to drive anterograde transport of cargoes along B-tubules of the microtubule doublets (Stepanek and Pigino, 2016) towards the tip of the cilium. Anterograde IFT trains also include IFT-A proteins and inactive dynein-2 (Jordan et al., 2018; Toropova et al., 2017; Toropova et al., 2019; Webb et al., 2020). At the tip, and likely at other sites along the axoneme (Nievergelt et al., 2022), reassembly of proteins and a change in conformation of dynein-2 leads to its activation, allowing retrograde trafficking of proteins by dynein-2 and IFT-A trains along the A-tubules of the microtubule doublets (Stepanek and Pigino, 2016). A ciliary necklace (Gilula and Satir, 1972) connects to the transition zone (TZ), distal to the mother centriole, to form a diffusion barrier at the ciliary base from which proteins enter and exit the organelle. IFT is required both to maintain the TZ (Jensen et al., 2018; Vuolo et al., 2018) as well as enable crossing of this barrier (De-Castro et al., 2022; Park and Leroux, 2022).

Metazoan dynein-2 is formed of multiple subunits (Asante et al., 2014; Vuolo et al., 2020) built around two copies of the dynein heavy chains (encoded by DYNC2H1) (Pazour et al., 1999; Porter et al., 1999; Signor et al., 1999). The motor complex is asymmetric, containing two different dynein-2 intermediate chains (Asante et al., 2014; Toropova et al., 2019) – DYNC2I1 [originally identified as FAP163 in Chlamydomonas (Patel-King et al., 2013) and commonly called WDR60 in mammals] and DYNC2I2 [originally identified as FAP133 in Chlamydomonas (Rompolas et al., 2007) and commonly called WDR34 in mammals]. They each contain N-terminal regions that bind dynein light chains (Pazour et al., 1998) and a C-terminal WD-propellor structure, each of which binds one dynein-2 heavy chain (Toropova et al., 2019). A dimeric dynein-2-specific light intermediate chain, commonly called LIC3 (encoded by DYNC2LI1; Grissom et al., 2002; Hou et al., 2004), in turn binds to and stabilizes the individual heavy chains. Light chains DYNLL1 and DYNLL2 (from the LC8 family), DYNLRB1 and DYNLRB2 (from the roadblock family), DYNLT1 and DYNLT2 (from the Tctex family) are shared with dynein-1; DYNLT2B, encoding TCTEX1D2, is unique to dynein-2 (Asante et al., 2014). We use the more common names throughout for ease of reading.

Mutations in each of the genes encoding dynein-2 subunits lead to ciliopathies, specifically, those associated with skeletal abnormalities (Huber and Cormier-Daire, 2012). In particular, multiple studies have identified WDR34 mutations in individuals with Jeune syndrome or Short-rib polydactyly syndrome (SRPS) (Huber et al., 2013; Schmidts et al., 2013; You et al., 2017) with more continuing to be identified (Yin et al., 2020). Clinical phenotypes differ between individuals (Schmidts et al., 2013) but all have skeletal defects in common. Table S1 shows the diversity of clinical outcomes for individuals with disease-causing variants in WDR34 emphasising the deficit in correlation between genotype and phenotype (Huber et al., 2013; Schmidts et al., 2013; You et al., 2017). Table S2 shows predictions of the functional impact of mutations in WDR34, again highlighting the lack of clear correlation between genotype and phenotype.

Although these mutations are often homozygous, they are not always found in isolation (Schmidts et al., 2013). This is also true of other ciliopathies, for example, where mutations in TTC21B (which encodes IFT139) can cause both isolated nephronophthisis and syndromic Jeune asphyxiating thoracic dystrophy as well as showing extensive genetic interactions with other ciliopathy loci (Davis et al., 2011). It has been suggested that mutations could act in trans, and digenic inheritance has been proposed in some cases, complicating the explanation of clinical presentations. For example, patients with p.A22V disease-causing variants in WDR34 are also heterozygous for other mutations in WDR19 and IFT140 that are predicted to be deleterious to function (Schmidts et al., 2013; Table S2).

Knockdown and knockout (KO) models of WDR34 have been studied in cells and model organisms (Huber et al., 2013; Schmidts et al., 2013; Wu et al., 2017). A WDR34 ‘KO’ cell culture model containing only the first 24 or 44 residues of WDR34 resulted in longer cilia, mis-localization of the IFT-B protein IFT88 and impaired Hh signalling (Tsurumi et al., 2019). Similar but more pronounced phenotypes were observed in WDR60 KO cells (Hamada et al., 2018; Tsurumi et al., 2019; Vuolo et al., 2018). Different WDR34 ‘KO’ cell lines (Tsurumi et al., 2019; Vuolo et al., 2018) and a ‘KO’ mouse model (Wu et al., 2017) have been reported to have very short primary cilia; notably these retain expression of a longer N-terminal WDR34 region of 114–146 amino acids. These deletion allele models, show changes in IFT88 and IFT140 distribution (Vuolo et al., 2018; Wu et al., 2017), impaired Hh signalling (Wu et al., 2017) and reduced assembly of the dynein-2 holocomplex (Vuolo et al., 2018).

In this study, we set out to determine the impact solely of the mutations in WDR34 to gain a better understanding of the disease phenotypes as well as the role of this key dynein-2 subunit. Here, we have examined the ability of clinically identified point mutations in WDR34 to restore normal ciliary structure and function to WDR34 KO cells. In order to achieve this, we have reconciled previous work using different WDR34 KO models and then expressed WDR34 mutants in a null background. Using proteomics, biochemistry and cellular imaging, our data show that WDR34 mutations have diverse impacts on dynein-2 integrity, ciliogenesis, cilia length control and signalling. Notably, many mutations can restore significant aspects of cilia function and selective defects are seen with each mutation.

Several different WDR34 KO models have been generated with some key differences in the observed phenotypes. Experiments in mouse embryonic fibroblasts and retinal pigment epithelial (RPE1) cells introduced frameshift mutations into the WDR34 gene (Fig. 1A) (Vuolo et al., 2018; Wu et al., 2017) that caused a severe reduction in both the frequency of ciliated cells and cilia length. These mutants retain expression of an N-terminal polypeptide of WDR34. In contrast, other WDR34 KO cells that remove expression completely have longer cilia than the wild-type (WT) RPE1 line (Tsurumi et al., 2019). To reconcile these findings, Tsurumi et al. (2019) provided evidence that the residual N-terminal WDR34 fragment might exert a dominant negative effect on cilia formation by binding to and sequestering the dynein light chains DYNLL1 and DYNLL2 (depicted in Fig. 1A). WDR34 KO cell lines generated with p.Val119Alafs*4 or p.Thr151Thrfs*41 mutations and encoding similar N-terminal regions also assemble short cilia (Tsurumi et al., 2019). We considered that the same could also be true for WDR34 for the disease-associated mutation p.Q158*.

To validate these findings, we generated a new RPE1 WDR34 KO cell line in which the start codon of WDR34 was deleted (termed here WDR34 KO D10) (Fig. S1) and compared this to the RPE1 WDR34 KO 1-5 cell line generated by Tsurumi et al. (2019). Cells were serum starved to allow analysis of cilia. In both cases, the WDR34 KO cells were able to form cilia with a frequency comparable to control cells, with a minor increase in cilia length (Fig. 1B,C).

The above data support the notion that N-terminal WDR34 fragments can cause a more severe ciliary phenotype than when the entire WDR34 protein is absent (Tsurumi et al., 2019). To test whether the p.Q158* mutant could exert an inhibitory effect on cilia formation, we expressed either GFP-tagged WDR34 or GFP-tagged WDR34-p.Q158* in WDR34 1-5 KO RPE1 cells. Whereas cilia formation was not affected by expression of full-length WDR34, expression of WDR34 p.Q158* in WDR34 KO cells resulted in a severe decrease in both the fraction of ciliated cells and cilia length (Fig. 1D,E; Fig. S1C). Notably, this was not the case when p.Q158* was expressed in a WT background (Fig. S1C). Proteomics experiments demonstrated that WDR34 p.Q158* retained the ability to bind to dynein light chain DYNLL1 and other dynein-2 subunits, consistent with a sequestering effect (Table S3).

At the base of the cilium, the TZ acts as a gatekeeper between the ciliary compartment and the cytoplasm, helping to determine the protein composition of the cilium (Shi et al., 2017). Previous studies demonstrated that loss of WDR60 is associated with mis-localisation of the TZ marker RPGRIP1L (Vuolo et al., 2018). We therefore tested the localisation of RPGRIP1L in RPE1 WDR34 KO cells. Whereas RPGRIP1L was restricted to a region near the ciliary base in control cells, in both WDR34 KO D10 and WDR34 KO 1-5 cell lines, RPGRIP1L was mis-localised and extended into the cilium in a significant fraction of cells (Fig. 1F,G). We conclude that although WDR34 is not critical for cilium formation in RPE1 cells, consistent with the findings of Tsurumi et al., (2019), but is required for proper organisation of the TZ as is also true for WDR60 (Vuolo et al., 2018).

Twelve point mutations and one truncation mutation in WDR34 were investigated (Fig. 2A). Eleven of the residues are well conserved among species (summarized in Fig. S3), suggesting an important role in protein function. Fig. 2B shows a schematic of the location of individual subunits including WDR34 within the dynein-2 complex. The location of each mutant relative to its position within the dynein-2 complex is shown in Fig. S4. We include predictions of the potential functional impact of these mutations from PolyPhen2 (Adzhubei et al., 2010), SIFT (Ng and Henikoff, 2003) and PROVEAN (Choi et al., 2012) in Table S2. To study these WDR34 mutations, we used the WDR34 KO cell line 1-5 (Tsurumi et al., 2019) and transduced them with lentiviruses to generate polyclonal cell lines stably expressing WT GFP–WDR34, specific WDR34 mutants or GFP only. All constructs were expressed as intact proteins at the expected molecular mass, as shown by immunoblotting of cell lysates, but at different levels (Fig. 2C; Fig. S2A). Notably, GFP–WDR34-p.Q158* was expressed at a low level. Varying phenotypes in cells expressing WDR34 mutations might result from changes in protein function but might also be due to the different protein levels. Therefore, we also analysed the relative stability of each mutant using cycloheximide (to inhibit further protein synthesis) and MG132 (to inhibit proteasomal degradation). Fig. 2C shows that p.A22V, pQ158*, p.R206C and p.T354M have similar stability to WT WDR34. In contrast, p.C148F, p.R182W, p.P390L, p.G393S, p.S410I, p.R447Q and p.R447W all appear less stable (enhanced degradation with cycloheximide). WDR34 p.A341V is expressed at low levels making interpretation difficult. Notably, p.K436R shows enhanced stability compared to both WT and other WDR34 mutants. p.R447W appears to be more susceptible to degradation than p.R447Q, even in the presence of MG132. Full gels are shown in Fig. S2B along with tubulin as a control.

Fig. 2B shows a schematic of the location of dynein-2 subunits colour-coded as in the subsequent plots. We utilized 16-plex multiplex tandem-mass-tag (TMT)-labelling to relatively quantify the interaction of WDR34 point mutant proteins with other dynein-2 components. To prepare cell samples for proteomics, clarified protein lysates were incubated with GFP–Trap beads to identify proteins interacting with GFP–WDR34. Purified proteins were labelled with tandem labelled tags prior to mass spectrometry (MS) analysis to allow the relative quantification of proteins in each sample. Fig. 2D shows that we do not efficiently detect any proteins in GFP–nanotrap pulldowns from cells expressing GFP only relative to those expressing GFP–WDR34-WT. These data are normalized to the abundance of GFP. Fig. 2E,F show data from cells expressing GFP–WDR34 mutants relative to GFP–WDR34-WT; these data are normalized to the abundance of WDR34 to counter differences in expression levels and stability of each mutant. Dynein-2 complex components DYNC2H1, WDR60, LIC3, DYNLL1, DYNLL2, DYNLRB1, DYNLT1 and DYNLT2B were identified as interacting partners across all samples in all three experimental repeats (Fig. 2E,F). DYNLRB2 and DYNLT3 were not identified in these experiments. These data are summarized graphically in Fig. S5 where the central circle represents the abundance relative to GFP–WDR34-WT. Those subunits shown within the circle are identified more readily, suggesting tighter interactions; those outside the circle indicate less abundance and likely weaker interactions.

Dynein-2 can be split into three subcomplexes (Hamada et al., 2018), DYNC2H1–LIC3, WDR60–TCTEX1D2–DYNLT1, and WDR34–DYNLL1/DYNLL2–DYNLRB1, where DYNLL1/DYNLL2 means either DYNLL1 or DYNLL2. Association of WDR34 A22V, R447Q and R447W with DYNC2H1–LIC3 and WDR60–TCTEX1D2–DYNLT1 was similar to WDR34 WT. Mutations that resulted in a >50% reduction in interaction were C148F, T354M, G393S and S410I, whereas mutations that resulted in a 2-fold increased interaction with DYNC2H1–LIC3 were p.Q158*, p.R182W, p.R206C, p.A341V and p.K436R. These WDR34 mutants also resulted in a modestly increased interaction with WDR60–DYNLT2B–DYNLT1 (Fig. 2E,F).

Formation of the third dynein-2 subcomplex, WDR34–DYNLL1/DYNLL2–DYNLRB1 involves binding of the light chains to the N-terminal region of WDR34 (Toropova et al., 2019). Interestingly, apart from for WDR34-K436R, interaction with DYNLRB1 was reduced by more than half in all WDR34 point mutants located in a WD domain, suggesting that defective assembly of DYNLRB1 in dynein-2 might be a common phenotype in WDR34-related disorders. The association of DYNLL2 with WDR34 was only significantly reduced in cells expressing WDR34 p.A341V, whereas DYNLL1 was not markedly affected. The differences in light chain binding to WDR34 mutants suggest they bind independently of one another. These results show that WDR34 mutations impact dynein-2 assembly by either enhancing or decreasing association depending on the mutation.

All GFP–WDR34 proteins showed a significant cytosolic pool (Fig. 3A) with some minor localization to centrioles. Consistent with the immunoblotting results above, GFP–WDR34-p.Q158* was expressed at a lower level and was more prominently observed at the ciliary base. To examine the effect of WDR34 mutations on the regulation of ciliogenesis, cells were immunolabelled to detect the cilia membrane marker ARL13B (Fig. 3A). We scored the percentage of ciliated cells and cilia length, testing all cell lines against the parental WDR34 KO cells expressing GFP only; experiments included WT RPE1 cells for comparison.

Expression of some mutants in the WDR34 KO line (p.A22V, p.R182W, p.A341V, p.P390L, p.K436R, p.R447Q and p.R447W) led to an increase in the percentage of ciliated cells (Fig. 3B). The truncated version, p.Q158*, however, led to a significant reduction in the ability of cells to form cilia, consistent with a dominant-negative impact on ciliogenesis (see Fig. 1 and Tsurumi et al., 2019). The increase in cilia length in WDR34 KO cells is reversed in cells expressing WT WDR34 and almost all mutant proteins (Fig. 3C). The p.C148F, p.G393S and p.S410I mutations were not able to rescue cilia length. The p.Q158* mutation led to further reduction in cilia length, again suggesting the fragment acquired functions that are restrictive to cilia assembly (Tsurumi et al., 2019).

To summarize the proteomics and ARL13B data, WDR34 mutants that significantly increase the proportion of ciliated cells either showed no change or enhanced dynein-2 assembly. Conversely, mutations unable to rescue cilia length also showed reduced dynein-2 assembly.

The TZ at the ciliary base mediates the entry and exit of proteins from the cilium. This diffusion barrier has been found to be disrupted in WDR60 KO cells (Vuolo et al., 2018). We immunolabelled ciliated cells with an antibody against RPGRIP1L, a TZ protein, to investigate whether this defect also arises in WDR34 KO cells and if so whether it could be ‘rescued’ by any of the disease-associated mutations in WDR34 (Fig. 4A). As expected in RPE1 WT cells, RPGRIP1L was localized at the ciliary base as compact circular dots, adjacent to the basal body. In the WDR34KO–GFP cells, a proportion of the cells had RPGRIP1L staining that stretched and extended towards the cilium, similar to what is seen in WDR60 KO cells (Vuolo et al., 2018). This was significantly reduced upon expression of WT WDR34. RPGRIP1L localization was similarly rescued by nearly all mutations, including the p.Q158* truncation (Fig. 4B), the only exception being the p.R447Q mutant. This supports a role for WDR34 in maintaining TZ integrity. Notably, of p.R447Q and p.R447W, only p.R447Q fails to restore the localization of RPGRIP1L. Our proteomics study showed that the interaction of WDR34-p.R447Q, but not p.R447W, with other dynein-2 proteins is slightly decreased.

To analyse the effects of the WDR34 mutations on IFT-A proteins, IFT140 was immunolabelled alongside ARL13B as a ciliary marker (Fig. 5). In WDR34 KO cells, IFT140 localizes to the ciliary base, as it does in WT RPE1 cells. The specific lot of IFT140 antibody used also stains non-specific nuclear dots, as observed by others (Hamada et al., 2018; Hirano et al., 2017; Takahara et al., 2019), and described by the manufacturer. This complicates analysis and so we have not applied automated quantification to these data but instead scored for the robust presence of IFT140 by eye. Specific IFT140 staining was observed at the ciliary base in WDR34 KO cells expressing GFP only and nearly all WDR34 point mutations. This shows that WDR34 KO cells retain sufficient retrograde IFT function to maintain IFT140, and by inference the IFT-A complex, at the base. The only exception to this is seen with p.Q158* cells where IFT140 was seen throughout the length of the cilium, again suggesting that this truncated WDR34 protein results in dominant-negative functions within the cilium, here in the assembly or localization of IFT-A.

We next investigated the localization of the IFT-B protein, IFT88, as a well-characterized reporter of IFT function as it accumulates at ciliary tips when retrograde defects occur. IFT88 is localized at the ciliary base in WT cells, but at the tips in WDR60 and WDR34 KO cells (Vuolo et al., 2018). We classified the distribution of IFT88 (Fig. 6A) into groups: base only, base and tip, base and throughout, and uniformly throughout (Fig. 6B). In RPE1 cells, IFT88 was mostly observed at the ciliary base and tip, or at the base only. In WDR34KO–GFP cells, the ‘base and throughout’ localization became the most dominant distribution, whereas WDR34 WT rescued IFT88 distribution. The p.Q158* truncation resulted in a uniform distribution of IFT88 throughout the cilia. WDR34-p.A341V- and p.T354M-expressing cells also had a small increased proportion of cilia with IFT88 along the cilium. This change is concurrent with both a decrease in proportion of cells with IFT88 at the base only or at the base and tip, suggesting both localizations reflect normal IFT-B function.

The studies shown above provide insight to the assembly of the dynein-2 and IFT complexes. It is, however, the functional changes resulting from defects in IFT, such as trafficking of receptors and signalling proteins, that translate these mutations into disease phenotypes. Given that WDR34 mutations lead to skeletal ciliopathies, Hh signalling is of key interest owing to its well-established roles in skeletal development. In cell culture, Hh signalling can be artificially stimulated with smoothened agonist (SAG). Under basal conditions, the GPCR family member smoothened (SMO) is typically excluded from cilia at the basal state, and SAG stimulation promotes its entry into and accumulation within cilia.

As expected in basal conditions, SMO is only observed in the cilium of a few cells in RPE1 WT cells (Fig. 7A). As shown previously for the cognate dynein-2 intermediate chain WDR60 (Vuolo et al., 2018), WDR34 KO cells show SMO localization to the cilium under basal conditions in a greater proportion of cells (Fig. 7A, quantified in 7B). Expression of GFP–WDR34-WT was able to reduce basal levels of intraciliary SMO, although not to the same low level as RPE1 WT cells and not to an extent that is statistically detected as significant. p.Q158* shows a large increase in basal SMO to a level greater than in WDR34 KO cells.

SAG stimulation led to SMO entry and accumulation in cilia in WT RPE1 cells (Fig. 7A; quantified in Fig. 7B). SAG treatment of cells expressing GFP resulted in only a small increase in SMO within the cilium. This contrasts with our previous work (Tsurumi et al., 2019) and (Vuolo et al., 2018) using an alternative WDR34 KO cell line and likely results from either further time in culture or the stable expression of GFP in these cells. We could clearly define that cells expressing WDR34 WT and WDR34 mutants were responsive to SAG, as shown by an increased proportion of cells with SMO in the axoneme. Interestingly, p.C148F, p.G393S, p.T354M and p.S410I mutations resulted in the least SMO accumulation and are also the mutations that led to weakened assembly of the core dynein-2 proteins. This supports a key role for dynein-2 in regulating the entry of SMO as well its exit from cilia.

Skeletal ciliopathies caused by mutations in dynein-2 are well known for their poor correlation of genotypes and phenotypes (Huber and Cormier-Daire, 2012). Only limited work has been done on cells from these patients, in part hampered by limited availability of material. Our data provide for the opportunity to analyse the impact of many mutations in large numbers of isogenic cells. Many of the residues that are mutated are highly conserved through evolution (Fig. S3) and many are in regions likely to impact the folding or stability, either of WDR34 or of the entire dynein-2 complex (Fig. S4). We explored this using proteomics and then sought to correlate the findings with functional data from microscopy assays. We were unable to define any clear common outcomes in relation to complex integrity, cilia length and number, IFT or SAG responses. These data are summarized in Fig. 8. Major impacts depend, at least in part, on how effectively each WDR34 mutant is associated with other dynein-2 proteins and also likely relate to the stability of the mutant protein. Given that we were unable to cluster these phenotypes, we discuss each mutant individually.

From our data, we can conclude that cilia formation can proceed in the absence of WDR34 in RPE1 cells.­­­ Our work reconciles previously apparently conflicting findings from different KO models. We show that these changes relate to retained expression of N-terminal fragments of WDR34. In the control background, p.Q158* does not impair ciliogenesis, whereas in WDR34 KO background, p.Q158* significantly perturbs cilium formation. Our finding that the WDR34 N-terminal fragment impairs ciliogenesis in the WDR34 KO background is in good agreement with results from Tsurumi et al. (2019). Our finding that the WDR34 N-terminal fragment does not impair ciliogenesis when expressed in a WT background (i.e. in the presence of full-length WDR34) differs from Tsurumi et al. (2019), who found that WDR34 N-terminal fragments exerted a dominant-negative effect on cilia formation in control cells. This difference might relate to differences in expression level of the WDR34 N-terminal fragment; stable versus transient expression of the WDR34 N-terminal fragment (stable in our study; a mixture of stable and transient in Tsurumi et al., 2019), and the precise N-terminal fragment used (WDR34 1–157 in this study; WDR34 1–146 in Tsurumi et al., 2019). Compound heterozygous WDR34 variants p.Q158* and p.K436R were previously reported in a patient with Jeune syndrome (Schmidts et al., 2013). That this patient presented with Jeune syndrome rather than the more severe SRPS phenotype can be explained by our observations that WDR34 p.Q158* does not disrupt cilium formation in a background where residual WDR34 function exists.

p.A22V is at the N-terminal end of the protein. It has some effects on dynein-2 complex integrity, notably with regard to association with the WDR60–TCTEX module. Its expression effectively restores length and cilia number even above re-expression of WT GFP–WDR34 when expressed in WDR34 KO cells. It also fully restores the localization of RPGRIP1L and is most similar to the WT protein in terms of localization of IFT-A and IFT-B and SMO signalling. This relatively minor impact is consistent with functional predictions (Table S2).

p.C148F shows defects in dynein-2 complex assembly or stability. It cannot restore the cilia length defect seen in WDR34 KO cells. It partially restores the steady-state distributions of RPGRIP1L and IFT-B. SMO localization in response to SAG is dampened, indicative of a defect in entry into cilia.

p.R182W results in an enhanced association with most dynein-2 subunits except for DYNLRB2. It restores cilia number even above re-expression of WT GFP–WDR34 and effectively restores cilia length, RPGRIP1L and IFT88 localization. This mutant has high levels of SMO in cilia at basal levels but retains a robust response to SAG stimulation. The SAG response remaining robust indicates that the basal levels of SMO in cilia might underpin p.R182W-related defects.

p.R206C shows a similar proteomic profile to p.R182W but shows other key differences. It restores the cilia length defect of WDR34 KO cells. It very effectively reverses the TZ defect shown with RPGRIP1L and IFT88 localization. A key difference to p.R182W is seen with Hh signalling where basal levels are reduced compared to WT, whereas the SAG response also remains robust.

p.A341V also shows a similar proteomic profile to p.R182W and p.R206C with enhanced association with DYNC2H1 and WDR60 with a significant reduction in association with DYNLRB1. Its profile is very similar to p.R182W in terms of cilia number and length, RPGRIP1L localization. A key difference is seen with IFT-B localization where IFT88 remains visible throughout the cilium. The SMO profile is very similar to that seen with WT GFP–WDR34.

p.T354M shows more dramatic changes in terms of the association with other dynein-2 subunits, notably DYNC2H1, WDR60 and DYNLRB2. Notably it is also the mutation most strongly predicted to be deleterious to function (Table S2). It restores cilia length to normal, contrasting with shorter cilia observed in patient fibroblasts (Huber et al., 2013). RPGRIP1L is restored to the base of the cilium, but IFT88 persists along the entire length. The SAG-induced localization of SMO to cilia is reduced. This suggests a general reduction but not complete loss of dynein-2 activity.

Both p.A341 and p.T354 are close in proximity located on the fourth β-propeller of WDR34 (Fig. S4). WDR34-p.A341V interaction with DYNC2H1–LIC3 is increased 2-fold, whereas WDR34-p.T354M interaction with the other subcomplexes are drastically decreased. Hence, the IFT88 defect associated with the two mutations might be independent of dynein-2 assembly but depend on WDR34 folding. Both residues are located on opposite ends of consecutive β-strands.

p.P390L associates more strongly with the DYNC2H1 and WDR60 modules with some reduction in association with DYNLRB1 and is also predicted to be deleterious to function (Table S2). It restores cilia length and is highly effective in restoring the localization of RPGRIP1L to the base of the cilia. We see some very small defects in IFT88 localization, but the SAG response mimics that of WT GFP–WDR34.

p.G393S shows the most dramatic changes in terms of association with other dynein-2 subunits with a reduction seen for all, and it is also the mutation that is most strongly predicted to be deleterious to function (Table S2). It cannot restore normal cilia length but is effective in terms of RPGRIP1L and IFT88 localization. The major defects seen with G393S are in Hh signalling with a near-complete failure to respond to SAG.

Similar to p.G393S, p.S410I fails to associate as effectively with other dynein-2 subunits and cannot restore cilia length. It restores RPGRIP1L to the base of the cilium, and IFT88 localization. It mirrors p.G393S with respect to absence of a robust SAG response. Together, p.G393S and p.S410I link a failure of robust dynein-2 assembly to the ability of cells to respond to SAG; broadly, this matches the predictions of loss of function (Table S2). WDR34 p.G393S was identified in patients with anencephaly (Yin et al., 2020). Both here and in their study, the p.G393S disease-causing variant impairs Hh signalling.

p.K436R shows enhanced association with DYNC2H1 and to a lesser extent with WDR60. Cilia number and length, IFT88 and RPGRIP1L localization are restored to wild-type phenotypes. This mutant has high levels of SMO in cilia at basal levels but retains a robust response to SAG stimulation. In this regard, it is highly similar to p.R182W in terms of impacts. Although p.K436R is more stable than WT WDR34 in a degradation assay (Fig. 1C), p.R182W is not, suggesting that phenotypes do not correlate with protein stability.

p.R447Q shows reduced association with most dynein-2 subunits. It rescues cilia number even above re-expression of WT GFP–WDR34 and retains the ability to restore cilia length, but not the localization of RPGRIP1L, which remains extended into the TZ. This contrasts with the shorter cilia observed in patient fibroblasts (Huber et al., 2013). However, IFT88 localization is restored, indicating effective retrograde IFT. Notably, the basal levels of SMO within cilia remain high but relocalization remains responsive to SAG stimulation.

p.R447W, unlike p.R447Q, only shows a loss of association with DYNLRB1. Both changes are strongly predicted to be deleterious to function (Table S2), yet the changes in interactions with other dynein-2 subunits do not mirror those of other mutations that are also strongly predicted to affect function (e.g. p.T354M and p.G393S). The percentage of ciliated cells is restored even above re-expression of WT GFP–WDR34 and cilia length is restored as is the localization of RPGRIP1L and IFT88. The restoration of RPGRIP1L to the ciliary base contrasts with p.R447Q. Loss of DYNLRB1 binding might underpin the severity of the p.R447W mutation (Table S1) but the differences we observe between p.R447W and p.R447Q might alternatively be due to their different expression levels or stability where p.R447W appears more susceptible to degradation even in the presence of MG132.

Overall, there are some common themes arising from this work.

We see considerable variation between the binding of dynein light chains, DYNLL1, DYNLL2 and DYNLRB1, to WDR34 mutant proteins. Previous proteomic data showed that in the absence of WDR34, the other intermediate chain, WDR60, interacts less well with DYNLRB1 (Vuolo et al., 2018). Our data also showed that many WDR34 mutations, all located within a WD repeat domain, led to weak incorporation of the DYNLRB1 into the complex. Notably, the DYNLRB light chains lie in closest proximity to the WD repeat domains in the structure (Toropova et al., 2019). Mutations that did not affect the association of DYNLRB1 include p.A22V, which is located upstream of the DYNLRB1-binding site, p.R206C which is located between two WD repeat domains, and p.K436R in WD6. It might be that other mutations within WD repeat domains commonly affect folding or stability of the β-propeller, transmitting to the region where DYNLRB1 binds. p.K436 is close in proximity to DYNC2H1 (Fig. S2), so although the residue is in a WD repeat domain, p.K436R might instead exert its mutant effect by affecting interaction with DYNC2H1.

Using deletion mutations, Tsurumi et al. (2019) showed that residues 80–93 of WDR34 are required for its interaction with DYNLL2. Our data show that the interaction is also weakened by the p.A341V mutation, again showing that mutations downstream of the region, and further along in the WD domains are also able to affect interaction with proteins at the N-terminal end. This further supports a key role for WDR34 in the incorporation light chains to the dynein-2 complex (Hamada et al., 2018; Toropova et al., 2019; Tsurumi et al., 2019; Vuolo et al., 2020).

In general, IFT140 localization remained consistent at the ciliary base in cells expressing GFP only or WDR34 missense mutations. IFT140 was enriched along the axoneme in cells expressing the p.Q158* truncation, similar to other models expressing short WDR34 N-terminal truncations (Tsurumi et al., 2019; Vuolo et al., 2018). IFT-A localization is affected in WDR60 KO cells (Vuolo et al., 2018) and DYNC2H1 mutant cells (Wu et al., 2017) suggesting that WDR34 might have no direct role in defining the steady-state localization of IFT-A. Localization of the IFT-B component IFT88 is abnormal in cells depleted of dynein-2 heavy chain, LIC3, WDR34, WDR60 or TCTEX1D2, or those retaining expression of short WDR34 fragments (Hamada et al., 2018; Kessler et al., 2015; Merrill et al., 2009; Taylor et al., 2015; Tsurumi et al., 2019; Vuolo et al., 2018). Expression of p.A341V and p.T354M mutations in WDR34 KO cells both failed to restore IFT88 to its normal distribution, suggesting that these residues have a more direct impact on retrograde IFT.

Localization of the TZ protein RPGRIP1L is defective in WDR34 KO cells. Nearly all WDR34 mutations examined here restored this localization except for the C-terminal p.R447Q mutation. Surprisingly, p.R447Q, but not p.R447W leads to significant RPGRIP1L mis-localization, possibly due to differences in charge or steric changes impacting folding. WDR60 is also required for RPGRIP1L localization (Vuolo et al., 2018). Since WDR34- p.R447 does not lie in the vicinity of the interface with WDR60 (Toropova et al., 2019), it may be that the intermediate chains control the TZ by different mechanisms.

Most mutants fail to fully restore basal levels of SMO (Fig. 8) to WDR34 KO cells but most partially restore function in this assay. These data require careful interpretation because of the low levels of SMO detected under basal conditions and the fact that WT GFP–WDR34 partially restores the basal localization of SMO but not to an extent that is statistically significant using ANOVA (that said, this is significant in a single t-test between only these samples). Although statistical testing does not define full restoration of the normal basal level of SMO in cilia, it is evident from the images and quantification (Fig. 7B, dotted line) that there is some partial restoration of function on expression of the GFP–WDR34 mutants. Hh signalling is impaired in cells depleted of dynein-2 heavy chain, WDR60, LIC3 and those retaining expression of short WDR34 fragments (Hamada et al., 2018; Taylor et al., 2015; Tsurumi et al., 2019; Vuolo et al., 2018; Wu et al., 2017). SAG-stimulated levels of SMO are impaired in some WDR34 KO cells. p.G393S and p.S410I mutations fail to restore SAG-induced translocation into cilia; p.T354M shows a partial response but this is not statistically detectable as significant. The basal level of SAG seen with the p.Q158* mutant is very high and not further increased by SAG. All other WDR34 mutants are able to restore the response to SAG.

Two of the mutations investigated here, p.T354M and p.R447W, were previously analysed in fibroblasts from patients with these disease-causing variants. Contrary to our results, both were found to result in shorter cilia (Huber et al., 2013). One reason for the inconsistency is that we are overexpressing the mutant proteins at much higher levels than normal, to be able to quantify large sample numbers, which might restore some function.

As we were preparing this work for publication, Antony and colleagues (Antony et al., 2022 preprint) released a preprint that included analysis of two of the WDR34 mutants, p.R182W and p.G393S, included here using base editing of IMCD3 cells. This more precise approach to engineering mutants clearly has advantages in terms of engineering mutations into the endogenous locus and therefore avoiding limitations of overexpression. However, it comes with huge challenges regarding efficiency, complicated by the near-triploid state of IMCD3 cells. Their work defines more subtle, and sometimes contrasting phenotypes to ours. For both mutants, no significant changes to the ability of cells to form cilia or in cilia length were found (Antony et al., 2022 preprint). Both mutants showed an increase in IFT88 at the ciliary tip, consistent with a defect in retrograde IFT. The data show a moderate increase in expression of GLI3 in both mutants as well as impaired Gli processing on SAG addition. These data are consistent with Shh signalling defects. In contrast, we find that overexpression of these mutants in a WDR34 KO background shows more clear defects with G393S showing the most significant changes in binding to other components of the dynein-2 complex and in Shh signalling. In contrast, our data show that p.R182W shows enhanced binding to dynein-2 subunits and can restore defects in ciliogenesis and IFT but not basal SMO accumulation in cilia. These differences are perhaps unsurprising given the very different model systems and in our case, we conclude that overexpression of p.R182W and p.G393S leads to more exaggerated outcomes than one would see with locus base editing. Indeed, overexpression of p.G393S could even be leading it to act as a dominant-negative mutation. That said, expression levels of WT and mutant forms of WDR34 are broadly comparable in our experiments, and we see robust functional rescue of many of the defects seen following KO of WDR34. Both approaches clearly have their merits. Base editing likely better reflects the clinical situation and gives a useful platform to screen for ways to alleviate phenotypes; our overexpression system provides some more insight into the biology of the system potentially leading to ways to achieve such interventions. Examples could include stabilizing assembly of the dynein-2 complex, controlling expression of mutant forms of WDR34, or targeting IFT or Hh signalling more selectively.

Patients with Jeune and short-rib polydactyly syndrome experience multi-organ abnormalities (Huber et al., 2013; Schmidts et al., 2013; You et al., 2017). Here, we have only studied a limited engineered cell system using epithelial cells. Cilia function is both cell type and tissue specific (Wheway et al., 2018). That said, dynein-2 is a core ciliary machine, so many functions would be highly conserved. Of the mutations investigated in this study, polydactyly was reported in individuals with p.R182W, p.R206C and p.R447W mutations (Huber et al., 2013; Schmidts et al., 2013; You et al., 2017). Although, in this study, none of the mutations, other than p.Q158*, led to an increased baseline of SMO in cilia, other elements of Hh signalling were not assayed including GLI trafficking and processing, or target gene induction.

All WDR34 mutations resulted in at least one cilia defect. Our results support the established roles of WDR34 in cilia extension, IFT-B maintenance, TZ integrity and SMO signalling. Notably, they also suggest that WDR34 does not have a single role within the dynein-2 complex and likely has multiple complex functions in assembly of IFT trains, activation of dynein-2 and retrograde IFT. Some patients carry compound disease-causing variants in other dynein-2 or IFT proteins, which likely intensify cilia defects and the clinical phenotype. This is the case with p.A22V, p.P390L and p.R447Q. This highlights the need and challenge of understanding dynein-2 assembly, not only in the context of the holoenzyme, but also in terms of co-assembly with IFT-A, IFT-B and kinesin-2.

To make predictions of the functional impact of WDR34 mutations using the Uniprot Q96EX3 accession, we used PolyPhen-2 v2.2.3r406 (Adzhubei et al., 2010). Sequences used were from UniProtKB/UniRef100 Release 2011_12 (14-Dec-2011); structures from PDB/DSSP Snapshot 25-May-2021 (178229 structures); genes from UCSC MultiZ46Way GRCh37/hg19 (08-Oct-2009). We also ran predictions using PROVEAN (Choi et al., 2012), which uses the NCBI non-redundant database (September 2012), BLAST v2.2.24+, and CD-HIT v4.5.4 using the default cut-off of −2.5. Sorting Intolerant From Tolerant (SIFT; Ng and Henikoff, 2003) and Grantham (Grantham, 1974) scores are also included.

A plasmid encoding GFP–WDR34 was used as a template for site-directed mutagenesis performed by DC Biosciences Ltd (now Aruru Molecular Ltd, Dundee, UK). All were confirmed by Sanger sequencing. The Lenti-XTM HTX Packaging System (Clontech, Saint-Germain-en-Laye, France) was used with the appropriate plasmid (generated by DC Bioscience) to generate lentiviral particles in HEK293T cells. The resultant viral supernatant was used to transduce RPE1 cells (see below) with a p.Val24Alafs*74 mutation in WDR34 (Tsurumi et al., 2019), which were then cultured in complete medium (see below) with the addition of 5 µg/ml puromycin. Cells were sorted by fluorescence-activated cell sorting using the University of Bristol Flow Cytometry Facility. Pools were selected that expressed the lowest detectable levels of GFP as verified by fluorescence microscopy. We analysed pools to avoid any differences arising from growing our individual clones.

hTERT-RPE1 and HEK293T cells were obtained from ATCC. WDR34 KO (WDR34 KO-1-5), derived by genome engineering from RPE1 cells are described in Tsurumi et al., (2019). HEK293T cells were cultured in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% FBS, whereas RPE1 cells were cultured in DMEM/F12 supplemented with 10% FBS (denoted complete medium). For ciliogenesis, cells were washed in PBS followed by serum starvation, by incubation in medium without FBS, for 24 h. All cells were grown at 37°C and 5% CO₂.

Five of the WDR34 KO cell lines expressing GFP proteins were contaminated with mycoplasma. They were treated with plasmocin (Invivogen) for 2 weeks as per the manufacturer's guidelines. All cells were from then on grown in the absence of plasmocin and verified independently to be free from mycoplasma contamination (MWG Eurofins).

To induce Hh signalling, cells were seeded, and the next day treated with 200 nM SAG, together with serum starvation for 24 h before harvesting of cells.

Antibodies used were against the following proteins: acetylated tubulin (Sigma-Aldrich, Cat# T6793, RRID:AB_477585, used 1:1000 for immunofluorescence, methanol fixation); ARL13B (Proteintech, Cat# 17711–1-AP, RRID:AB_2060867, 1:1000 for immunofluorescence, methanol fixation); ARL13B (Proteintech, Cat# 66739-1-Ig, RRID:AB_2882088, 1:1000 for immunofluorescence, methanol or 4% PFA fixation); GAPDH (Proteintech, Cat# 60004-1-Ig, RRID:AB_2107436, 1:5000 for immunoblotting); GFP (BioLegend, Cat# 902601, RRID:AB_2565021, 1:5000 for immunoblotting, methanol fixation); IFT88 (Proteintech, Cat# 13967-1-AP, RRID:AB_2121979, 1:300 for immunofluorescence, 4% PFA fixation); IFT140 (Proteintech, 17460-1-AP, RRID:AB_2295648, 1:100 for immunofluorescence); RPGRIP1L (Proteintech, Cat# 55160–1-AP, RRID:AB_10860269, 1:200 for immunofluorescence, Methanol fixation); and SMO (Santa Cruz Biotechnology, Cat# sc-166685, RRID:AB_2239686, 1:100 for immunofluorescence, 4% PFA fixation).

Uncoated #1.5 glass coverslips (Thermo Fisher Scientific) were used for the culturing of cells for immunofluorescence microscopy. Depending on the antibody to be used, cells were fixed with −20°C methanol for 5 min, or with 4% PFA for 10 mins before extraction using 0.1% Triton X-100 in PBS. Cells were blocked in 3% bovine serum albumin (BSA) in PBS for 30 min, followed by incubation in primary antibody solution in 3% BSA in PBS for 1 h, then in secondary antibody and DAPI (0.5 μg/ml) solution in 3% BSA in PBS for 1 h, all with PBS washes in between. Coverslips were mounted on microscope slides with Mowiol 4-88 mounting medium.

Cells were imaged using an Olympus IX-71 widefield microscope with a 63× objective, and excitation and emission filter sets (Semrock, Rochester, NY) controlled by Volocity software (version 4.3, Perkin-Elmer, Seer Green, UK). Images were processed using FIJI (Schindelin et al., 2012).

For mutant stability assays, confluent cells were treated with or without 100 µg/ml cycloheximide (Merck) and with or without 10 µM MG132 (Merck) for 16 h prior to lysis in RIPA buffer. Cells were lysed in RIPA buffer [50 mM Tris-HCl pH 7.5, 150 mM NaCl, 1% Triton X-100, 1% sodium deoxycholate, 0.1% SDS, 1 mM EDTA and protease inhibitors (539137, Millipore)]. LDS buffer was added to clarified lysates. Protein samples were separated on 4–12% Bis-Tris gels and transferred onto nitrocellulose membranes by wet transfer. Membranes were blocked with 5% milk in PBST for 1 h, incubated in blocking solution with primary antibodies overnight, then incubated with HRP-conjugated secondary antibodies for 1 hour. Blots were developed by chemiluminescence.

Cells ready for harvest were first incubated with 1 mM DSP (Thermo Fisher Scientific #22585) for 30 min on ice to crosslink proteins. The reaction was quenched by the addition of 500 mM Tris-HCl pH 7.5 for 15 min. Cells were then washed with PBS and scraped off the plate in lysis buffer (10 mM Tris-HCl pH 7.4, 150 mM NaCl, 0.5 mM EDTA, 0.5% IGEPAL and protease inhibitors) followed by incubation on ice for 30 min. Clarified lysates were retrieved by centrifugation (25,000 g for 15 min) and incubated for 3 h on a rotating wheel at 4°C with 20 µl GFP-Trap beads (ChromoTek). Beads were washed three times in washing buffer (10 mM Tris-HCl pH 7.4, 150 mM NaCl and 0.5 mM EDTA). Beads for proteomics analysis were resuspended in minimal volume of wash buffer.

Proteomic analysis including Nano LC mass spectrometry and primary data analysis was undertaken as described in Vuolo et al. (2018). Briefly, RPE1 cells stably expressing WDR34 proteins were washed with PBS and incubated with crosslinker solution 1 mM DSP, for 30 min on ice. The reaction was quenched by adding 500 mM Tris-HCl pH 7.5 for 15 min. Immunoprecipitation of lysates of hTERT-RPE1 cells stably expressing HA-tagged WDR60 or WDR34 was performed using anti-HA agarose beads (Sigma-Aldrich). Lysis buffer containing 50 mM Tris-HCl, pH 7.4, 1 mM EDTA, 150 mM NaCl, 1% Igepal and protease inhibitors (539137, Millipore) was used for HA immunoprecipitates and a buffer of 10 mM Tris-HCl pH 7.4, 50 mM NaCl, 0.5 mM EDTA, protease inhibitors and 0.5% Igepal was used for GFP immunoprecipitates. Subsequently, cells were incubated on a rotor at 4°C for 30 min and then lysates were centrifuged at 13,000 g at 4°C for 10 min. Cell lysates were added to the equilibrated HA or GFP beads and incubated on a rotor at 4°C. Next, the HA beads were washed in washing buffer containing 50 mM Tris-HCl pH 7.4, 150 mM NaCl, 0.5 mM EDTA, 0.3% Triton X-100 and 0.1% SDS, and GFP beads were washed in a buffer of 10 mM Tris-HCl pH 7.4, 50 mM NaCl and 0.5 mM EDTA. Subsequent proteomic analysis by nano-LC MSMS using an Orbitrap Fusion Tribrid mass spectrometer (Thermo Fisher Scientific) was described previously (Vuolo et al., 2018). Data relating to GFP–WDR34 interactions are from three independent experiments. These data have been deposited to the ProteomeXchange Consortium via the PRIDE (Perez-Riverol et al., 2019) partner repository with the dataset identifier PXD032758. Table S3 gives raw and normalized abundances as well as the normalized abundance ratio of dynein-2 components, between GFP-WDR34-Gln158* and GFP-WDR34-FL cells, in two independent experiments. These data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository with the dataset identifier PXD025686.

We thank members of the Stephens, Roberts and Nakayama labs for continued discussion and help with this work; the University of Bristol Flow Cytometry and Wolfson Bioimaging Facilities for support and advice; Marc van der Kamp for help and advice on structural modelling.

The peer review history is available online at https://journals.biologists.com/jcs/lookup/doi/10.1242/jcs.260073.reviewer-comments.pdf.