Dynein-2 intermediate chains play crucial but distinct roles in primary cilia formation and function

The dynein-2 microtubule motor is the retrograde motor for intraflagellar transport. Mutations in dynein-2 components cause skeletal ciliopathies, notably Jeune syndrome. Dynein-2 contains a heterodimer of two non-identical intermediate chains, WDR34 and WDR60. Here, we use knockout cell lines to demonstrate that each intermediate chain has a distinct role in cilium function. Using quantitative proteomics, we show that WDR34 KO cells can assemble a dynein-2 motor complex that binds IFT proteins yet fails to extend an axoneme, indicating complex function is stalled. In contrast, WDR60 KO cells do extend axonemes but show reduced assembly of dynein-2 and binding to IFT proteins. Both proteins are required to maintain a functional transition zone and for efficient bidirectional intraflagellar transport. Our results indicate that the subunit asymmetry within the dynein-2 complex is matched with a functional asymmetry between the dynein-2 intermediate chains. Furthermore, this work reveals that loss of function of dynein-2 leads to defects in transition zone architecture, as well as intraflagellar transport.

have also been reported. The role of dynein-2 heavy chain has been 79 extensively studied in Chlamydomonas, C. elegans, and mice. In all cases, loss of dynein heavy chain 80 results, in short, stumpy cilia that accumulate IFT particles at the tip, consistent with a role of 81 dynein-2 in retrograde ciliary transport (Hou and Witman, 2015). Recently, more interest has been 82 focused on the role of the subunits associated with DHC2. Two studies in Chlamydomonas and in 83 human patient-derived fibroblasts revealed that LIC3 (D1bLIC in Chlamydomonas) plays an important 84 role for ciliogenesis and stability of the entire dynein-2 complex (Li et al., 2015;Taylor et al., 2015). 85 Similarly, loss of Tctex2b (TCTEX1D2) destabilizes dynein-2 and reduces IFT in Chlamydomonas 86 (Schmidts et al., 2015). 87 Previous work from our lab and others has shown that loss of function of dynein-2 intermediate 88 chains, WDR34 and WDR60, is associated with defects in cilia. Knockdown of WDR60 or WDR34 in 89 hTERT-RPE1 cells results in a reduction of ciliated cells, with an increase or decrease of the cilia 90 length, likely depending on depletion efficiency (Asante et al., 2014). Mutations in WDR34 have also 91 been shown to result in short cilia with a bulbous ciliary tip in patients fibroblast cells affected by 92 SRP (Huber et al., 2013). Consistent with the results obtained in patient cells, loss of WDR34 in mice 93 also results in short and stumpy cilia with an abnormal accumulation of ciliary proteins and defects 94 in Shh signaling (Wu et al., 2017). Similarly, mutations in WDR60 patient fibroblasts are associated with a reduction in cilia number, although the percentage of ciliated cells was variable in different 96 affected individuals (McInerney-Leo et al., 2013). These findings are all consistent with roles for 97 WDR34 and WDR60 in IFT. 98 In this study, we sought to better understand the role of dynein-2 in human cells using engineered 99 knockout (KO) cell lines for WDR34 and WDR60. We define a functional asymmetry within the 100 complex, where WDR34 is absolutely required for ciliogenesis, while WDR60 is not. In contrast, 101 WDR60 is essential to maintain the integrity of the ciliary transition zone and for retrograde 102 trafficking of IFT particles. Furthermore, by expressing HA-tagged WDR34 in WDR60 KO cells and HA-103 tagged WDR60 in WDR34 KO cells, we found that WDR34 is not required for the other subunits to 104 assemble, whereas loss of WDR60 leads to significant defects in dynein-2 holocomplex assembly. We 105 propose a model where dynein-2 requires WDR34 for axoneme extension but not for the assembly 106 of the other subunits of the complex, whereas WDR60 is crucial for dynein-2 stability, IFT, and ciliary 107 transition zone assembly and/or maintenance. Analysis of disease-causing patient mutations further 108 defines the role of dynein-2 in cilia formation and function. 109

WDR34 or WDR60 play different roles in cilia function 112
To understand the function of WDR34 and WDR60, we generated KO human telomerase-113 immortalized RPE1 (hTERT-RPE1) cells using CRISPR-Cas9. We derived two WDR34 KO clones (1 and 114 2) using guide RNAs (gRNAs) targeting exons 2 and 3, and one KO clone for WDR60, targeting exon 3. 115 Genomic sequencing of these clones identified insertion/deletion mutations on the targeted 116 sequences (Fig. S1). All cell clones were analyzed for protein expression by immunoblot using 117 polyclonal antibodies. Neither WDR34 nor WDR60 was detected in the respective KO cells compared 118 to the controls ( Fig. S2A and Fig. S2B) which provides evidence that downstream initiation sites are 119 not being used. To mitigate against the possibility of any off-target effects, we grew KO cells 120 alongside control CRISPR cells which had been transfected and treated the same way as the KO, but 121 genomic sequencing showed no mutation at the target site. These cells (WDR34 KO CTRL and 122 WDR60 KO CTRL) did not present any cilia defects when stained with Arl13b or IFT88 ( Fig. S2C and 123 S2D). Images in all figures show WT cells where indicated. Indistinguishable results were obtained 124 using these control cell lines. Defects in ciliogenesis in both WDR34 and WDR60 KO cells were 125 rescued by overexpressing WT proteins, confirming that the phenotypes we observed were not due 126 to off-target mutations (described below). 127 Loss of WDR34 severely impaired the ability of these cells to extend a microtubule axoneme (Fig. 1A,  128 B), although Arl13b localized within those few cilia that did form. In contrast, loss of WDR60 did not 129 significantly affect the ability of cells to extend an axoneme (Fig. 1B). Cilia were shorter in both 130 WDR60 and WDR34 KO cells (Fig. 1C). Next, we examined the assembly and structure of primary cilia 131 in WDR34 and WDR60 KO cells by transmission electron microscopy (EM). After 24 hr of serum 132 starvation, WT RPE1 cells extend a defined axoneme surrounded by a ciliary membrane (Fig. 1D). In 133 contrast, WDR34 KO cells failed to extend an axoneme ( Fig. 1E) but showed a large docked pre-134 ciliary vesicle, consistent with the small Arl13b-positive structures seen by immunofluorescence. 135 WDR60 KO cells showed apparently normal cilia ( Fig. 1F) with normal basal body structures and 136 axoneme extension. However, when an entire cilium was captured in WDR60 KO serial sections (Fig.  137 1G), we observed a bulged cilia tip containing accumulated electron dense particles (Fig. 1H). To our 138 surprise, we also observed the ciliary membrane bulged at a second point along the axoneme and 139 this region contained intraciliary vesicular structures (Fig. 1Hi). 140

Loss of WDR34 and WDR60 causes accumulation of proteins at the ciliary tip 141
The abnormal structure of cilia in the KO cells led us to analyze the steady-state localization of the 142 IFT machinery. After 24 hr serum starvation, IFT88 (part of IFT-B) was found almost exclusively at the core transition zone marker, RPGRIP1L (also known as MKS5), is no longer restricted to an area 177 adjacent to the mother centriole in WDR60 KO cells (Fig. 5A, quantified in 5Ai). Conversely, TMEM67 178 (also known as MKS3), which in WT cilia extends from the basal body through a more distal region, 179 becomes much more tightly restricted to the base of the cilium in WDR60 KO cells (Fig. 5B, 180 quantified in 5Bi). We also determined the transition zone organization in WDR34 KO cells. The few 181 cilia found in the WDR34 KO cells recapitulate the same phenotype observed in the WDR60 KO cilia 182 with an expansion of RPGRIP1L to a more distal position and a reduction of the TMEM67 domain 183 ( Fig. 5A and Fig. 5B). In contrast to TMEM67 and RPGRIP1L, no changes were observed for the 184 transition zone marker TCTN1 in both WDR34 KO and WDR60 KO cells with respect to the control 185 throughout the cilium, as it was in WDR60 KO cells. Next, we tried to mimic the compound 211 heterozygosity occurring in patient cells, generating a stable cell line that expresses both WDR60 212 [T749M] and [Q631*] mutants. When the two mutants were co-expressed in the same WDR60 KO 213 cells we saw no additive effects or dominant negative effects, but cilia appeared normal with IFT88 214 only localized to the base (Fig. S3B, S3C) as was seen with the HA-WDR60[T749M] mutant. 215 In parallel, we also analyzed the phenotype of WDR34 KO cells stably expressing WT and mutant 216 forms of WDR34. We found that expression of WT mGFP-WDR34 restored ciliogenesis and axoneme 217 extension in WDR34 KO cells and that, unexpectedly, this was also true of cells expressing WDR34 218 incorporating either [A22V] or [C148F] mutations ( Fig. 7D and 7Di). These disease-causing mutations 219 were chosen as they lie in regions without specific domain prediction. The cilia that formed in 220 WDR34 KO cells expressing either mGFP-WDR34[A22V] or mGFP-WDR34[C148F] were also positive 221 for Arl13b (Fig. 7D). Moreover, WT and both WDR34 mutants were able to rescue IFT88 localization 222 to the basal body (Fig. 7E). 223 To better understand the function of WDR60 we analyzed how the complex is assembled by 224 performing immunoprecipitation of WT and mutant WDR60 proteins. We found that 225 immunoprecipitation of HA-WDR60 expressed in WDR60 KO cells effectively pulls down the 226 chaperone NudCD3, known to interact with dynein-2 via its WD repeat domains. As expected, 227  (Fig. 8B). Next, we tested the interactions with IFT proteins, the primary cargo of the 233 ciliary motors. We found that WT WDR60 can bind to IFT140, IFT88, and IFT57; WDR60[T749M] 234 binds to all 3 IFT subunits tested but binds less well to IFT140 (Fig. 7H) (Fig. 8A). Next, we sought to determine the 247 effect of WDR60 and WDR34 loss on the localization of other dynein-2 subunits. We found that LIC3 248 (DYNC2LI1) localized in cilia of WT cells, but this localization was lost in WDR34 or WDR60 cells ( Our data provide evidence that the structural asymmetry with the dynein-2 motor is matched by 282 functional asymmetry. Perhaps most strikingly, we find that WDR34 is essential for axoneme 283 extension during early steps of ciliogenesis, whereas WDR60 is not required for ciliogenesis. Both 284 subunits are necessary for maintaining proper cilia protein composition. Depletion of WDR34 using 285 RNAi is also associated with ciliary defects (Asante et al., 2013) and patient fibroblasts have shorter 286 cilia with a bulbous tip (Huber et al., 2013). Fibroblasts from WDR34 knockout mice also have 287 stumpy cilia and defects in Shh signaling (Wu et al., 2017). It is intriguing that some cells missing 288 WDR34 can still extend a rudimentary cilium but even here, ciliary protein localization is severely 289 disrupted. Since WDR60 cells can extend an axoneme, WDR34 and WDR60 clearly have distinct but 290 overlapping functions in cells. We consider that WDR34 plays an essential role in ciliogenesis to 291 ensure delivery of a key factor required for axoneme extension. It is noteworthy that in the absence 292 of WDR34, WDR60 can still assemble effectively with the other subunits of dynein-2. While these 293 interactions are likely reduced compared to the normal situation, this shows that it is specifically 294 WDR34 that is required at this early stage of ciliogenesis. Our EM data show that it acts at a stage 295 after docking of the ciliary vesicle, immediately before axoneme extension. Paradoxically, our data 296 also show that in the absence of WDR60, the dynein-2 holocomplex cannot form effectively yet 297 axoneme extension occurs normally. This raises the possibility that WDR34 is itself required for 298 axoneme extension, possibly outside of the context of the dynein-2 complex. We cannot rule out 299 that there are dynein-2-independent functions of WDR34 and WDR60 but all data provide strong 300 evidence that they co-exist in the dynein-2 holoenzyme. 301

Assembly of the dynein-2 holocomplex 302
In WDR34 and WDR60 KO cells, LIC3 is no longer detected in primary cilia, while TCTEX1 localization, 303 a dynein light chain that is also a component of dynein-1, was unperturbed. Coupled with our 304 proteomics data, this suggests that the localization of LIC3 to cilia is a good reporter of dynein-2 305 assembly. Surprisingly we found that DHC2 levels at the ciliary base are reduced in WDR34 KO cells 306 compared to WT and WDR60 KO cells. Our data show that in WDR34 KO cells, the remaining 307 subunits can coassemble into partial dynein-2 complexes. We do not, however, know if this is a 308 functional or indeed processive motor. According to the current model based on work in C. elegans In addition to defects in retrograde IFT, incomplete assembly of dynein-2 in WDR60 KO cells could 338 cause a block in IFT train assembly and impaired entry of cargo into cilia. Indeed, work in C. elegans 339 has shown that IFT-B is required for entry of dynein-2 into cilia (Yi et al., 2017). In addition to these 340 defects, we show that KAP3, a subunit of kinesin-2, accumulates at the ciliary tip of WDR60 KO cells. that these proteins might not be effectively retained within cilia, but leak out through the diffusion barrier. An alternative, as discussed above, is that these proteins are less effectively loaded into cilia. 379 We did not find a difference in the intensity levels of overexpressed Rab8a in WDR60 KO compared 380 to WT cells suggesting that at least some proteins can enter normally. It seems likely that the defects 381 we see in both entry to, and exit from, cilia in these KO cells are caused by defects in transition zone 382 structure.  Fig. 7 were imaged using Leica SP5 confocal microscope (Leica Microsystems, Milton Keynes, 517 UK). Live images in Fig. 3 were imaged using Leica SP8. All images were acquired as 0.5 µm z-stacks. 518 All graphs show mean and standard deviation. 519

Rescue experiments 520
For 'rescue' experiments, stable WDR60 KO cell lines overexpressing wild-type and mutants HA-521 tagged WDR60 were generated. Similarly, WDR34 KO#1 cells were stably transfected with WT and 522 mutants WDR34 tagged with a GFP. Cells were serum starved for 24 h, fixed and processed for 523 immunofluorescence analysis. 524

Electron microscopy 525
Cells were serum starved 24 h and fixed in 2.5% glutaraldehyde for 20 min. Next, the cells were 526 were cut and stained with 3% uranyl acetate then lead citrate, washing 3x with water after each.

Fluorescence intensity measurement 541
Quantification of fluorescence intensity was performed using original images. Measurement of 542 intensity was performed using the average projections of acquired z-stacks of the area of the ciliary 543 marker acetylated tubulin. Fluorescence intensity along the ciliary axoneme was measured using 544 ImageJ plot profile tool. Fluorescence intensity in at the ciliary base was measured drawing same 545 diameter circles at the ciliary base. inhibitors (539137, Millipore). Subsequently, cells were incubated on a rotor at 4˚C for 30 min and 552 then lysates were centrifuged at 13,000 g at 4˚C for 10 min. Cell lysates were added to the 553 equilibrated anti-HA-Agarose beads (Sigma A2095, batch number 026M4810V) and incubated on a 554 rotor at 4˚C. Next, the beads were washed three times by centrifuging at 2000 g for 2 min at 4˚C with 555 1 ml of washing buffer (50 mM Tris-HCl, 150 mM NaCl, 0.5 mM EDTA, Triton X-100 0.3% SDS 0.1%) 556 containing protease inhibitors (539137, Millipore). Samples used for SDS-PAGE and immunoblotting 557 were resuspended in 50 µl of LDS sample buffer (Life Technologies) containing sample reducing 558 agent (Life Technologies) and boiled at 95˚C for 10 min. 559

Proteomic analysis 560
For TMT Labelling and high pH reversed-phase chromatography, the samples were digested from the 561 beads with trypsin (2.5 µg trypsin, 37°C overnight), labeled with Tandem Mass Tag (TMT) six-plex 562 reagents according to the manufacturer's protocol (Thermo Fisher Scientific, Loughborough, UK) and the labeled samples pooled. The pooled sample was then desalted using a SepPak cartridge 564 according to the manufacturer's instructions (Waters, Milford, Massachusetts, USA)). Eluate from 565 the SepPak cartridge was evaporated to dryness and resuspended in buffer A (20 mM ammonium 566 hydroxide, pH 10) prior to fractionation by high pH reversed-phase chromatography using an 567 Ultimate 3000 liquid chromatography system (Thermo Fisher Scientific). In brief, the sample was 568 loaded onto an XBridge BEH C18 Column (130 Å, 3.5 µm, 2.1 mm X 150 mm, Waters, UK) in buffer A 569 and peptides eluted with an increasing gradient of buffer B (20 mM ammonium hydroxide in 570 acetonitrile, pH 10) from 0-95% over 60 min. The resulting fractions were evaporated to dryness and 571 resuspended in 1% formic acid prior to analysis by nano-LC MSMS using an Orbitrap Fusion Tribrid 572 mass spectrometer (Thermo Fisher Scientific). 573

Nano-LC Mass Spectrometry 574
High pH RP fractions were further fractionated using an Ultimate 3000 nano-LC system in line with 575 an Orbitrap Fusion Tribrid mass spectrometer (Thermo Fisher Scientific). In brief, peptides in 1% 576 (vol/vol) formic acid were injected onto an Acclaim PepMap C18 nano-trap column (Thermo Fisher 577 Scientific). After washing with 0.5% (vol/vol) acetonitrile 0.1% (vol/vol) formic acid, peptides were 578 resolved on a 250 mm × 75 μm Acclaim PepMap C18 reverse phase analytical column (Thermo Fisher 579 Scientific) over a 150 min organic gradient, using 7 gradient segments (1-6% solvent B over 1 min., 6-580 15% B over 58 min., 15-32% B over 58 min., 32-40% B over 5 min., 40-90% B over 1 min., held at 90% 581 B for 6 min and then reduced to 1% B over 1 min.) with a flow rate of 300 nl min−1. Solvent A was 582 0.1% formic acid and Solvent B was aqueous 80% acetonitrile in 0.1% formic acid. Peptides were 583 ionized by nano-electrospray ionization at 2.0 kV using a stainless steel emitter with an internal 584 diameter of 30 μm (Thermo Fisher Scientific) and a capillary temperature of 275°C. 585 All spectra were acquired using an Orbitrap Fusion Tribrid mass spectrometer controlled by Xcalibur 586 2.0 software (Thermo Fisher Scientific) and operated in data-dependent acquisition mode using an 587 SPS-MS3 workflow. FTMS1 spectra were collected at a resolution of 120 000, with an automatic gain 588 control (AGC) target of 400 000 and a max injection time of 100 ms. Precursors were filtered with an 589 intensity range from 5000 to 1E20, according to charge state (to include charge states 2-6) and with 590 monoisotopic precursor selection. Previously interrogated precursors were excluded using a 591 dynamic window (60 s +/-10 ppm). The MS2 precursors were isolated with a quadrupole mass filter 592 set to a width of 1.2 m/z. ITMS2 spectra were collected with an AGC target of 10 000, max injection 593 time of 70 ms and CID collision energy of 35%. 594 For FTMS3 analysis, the Orbitrap was operated at 30 000 resolution with an AGC target of 50 000 595 and a max injection time of 105 ms. Precursors were fragmented by high energy collision dissociation (HCD) at a normalized collision energy of 55% to ensure maximal TMT reporter ion yield. 597 Synchronous Precursor Selection (SPS) was enabled to include up to 5 MS2 fragment ions in the 598 FTMS3 scan. 599