Roles for CEP170 in cilia function and dynein-2 assembly

ABSTRACT Primary cilia are essential eukaryotic organelles required for signalling and secretion. Dynein-2 is a microtubule-motor protein complex and is required for ciliogenesis via its role in facilitating retrograde intraflagellar transport (IFT) from the cilia tip to the cell body. Dynein-2 must be assembled and loaded onto IFT trains for entry into cilia for this process to occur, but how dynein-2 is assembled and how it is recycled back into a cilium remain poorly understood. Here, we identify centrosomal protein of 170 kDa (CEP170) as a dynein-2-interacting protein in mammalian cells. We show that loss of CEP170 perturbs intraflagellar transport and hedgehog signalling, and alters the stability of dynein-2 holoenzyme complex. Together, our data indicate a role for CEP170 in supporting cilia function and dynein-2 assembly.

Intraflagellar transport is required for the func on of primary cilia.In this work, we show that Centrosomal Protein 170 (CEP170) interacts with the IFT motor dynein-2 and loss of CEP170 causes defects in dynein-2 assembly and cilia func on.

Introduc on
Primary cilia (also called non-mo le cilia) are hair-like organelles that protrude from nearly every vertebrate cell and are important for cell signalling.Muta ons in cilia-related genes lead to a range of diseases, o en developmental and skeletal, classified as ciliopathies (Huber and Cormier-Daire, 2012;Mill et al., 2023).Sonic hedgehog (Shh) signalling -one of the key pathways in vertebrate development -requires the primary cilium for proper signal transduc on via Smoothened (Smo) (Corbit et al., 2005;Huangfu et al., 2003;Mill et al., 2023).Understanding how a cilium is built and regulated is cri cal in advancing our molecular understanding of ciliopathies.
At the core a cilium is the axoneme, a structure formed of doublet-microtubules with 9-fold radial symmetry (Hall and Hehnly, 2021).The cilium is separated from the cell body by the transi on zone (TZ), a region of protein fibres that restrict protein entry into the cilium (Garcia-Gonzalo et al., 2011;Garcia-Gonzalo and Reiter, 2017).At the base of the primary cilium is the basal body origina ng from the mother centriole.The mother centriole, along with the daughter centriole and pericentriolar material, cons tute the centrosome (Gonczy and Hatzopoulos, 2019;Tischer et al., 2021).Centrioles have a 9-triplet microtubule-based structure with 9-fold radial symmetry (Gonczy and Hatzopoulos, 2019;Tischer et al., 2021).The mother centriole is dis nct from the daughter in that it has addi onal proteinaceous structures protruding radially from the distal end, called distal appendages (DAPs) and sub-distal appendages (sDAPs) (Hall and Hehnly, 2021).DAPs are required for cilia to form as, in non-dividing cells, ciliogenesis is ini ated when pre-ciliary vesicles dock with DAPs, which eventually fuse with the plasma membrane forming the ciliary membrane (Ishikawa and Marshall, 2011;Tanos et al., 2013).DAPs form a barrier between the mother centriole and axoneme (Yang et al., 2018).There is conflic ng evidence surrounding the role of sDAPs in ciliogenesis.Accumulated data support a model where sDAPs assemble a stabilized microtubule network that acts in ciliary vesicle docking ((Hehnly et al., 2012;Sorokin, 1962) reviewed in Hall and Hehnly (2021)).
The assembly and maintenance of a cilium requires intraflagellar transport (€FT) (Kozminski et al., 1993).IFT is an evolu onary conserved process by which material is transported in and out of the cilium, and is required to overcome the barrier of the TZ and concentrate cargoes within cilia (Lechtreck, 2015).IFT requires the assembly of polymeric trains composed of the protein complexes IFT-A and IFT-B, as well as the motor protein complexes, kinesin-2 and cytoplasmic dynein-2 (herea er referred to as dynein-2) (Pigino, 2021;Webb et al., 2020).Anterograde trafficking is powered by kinesin-2, a heterotrimeric complex formed of the kinesin-family motor proteins, KIF3A and KIF3B, and kinesin-associated protein (KAP) (Webb et al., 2020).When the trains reach the p of the cilium, the trains rearrange, and subsequent retrograde transport is driven by dynein-2.The correct assembly of IFT trains is vital to ensure proper ciliary composi on and signalling.As well as receptors, signalling proteins, and cargo adaptors, kinesin-2 and dynein-2 motor proteins are themselves inac ve cargoes during the direc onal transport step that they do not drive directly (Jordan et al., 2018;Toropova et al., 2017;Toropova et al., 2019).
Previous work has defined interac ons of WDR60 and WDR34 with other dynein-2 components and IFT proteins (Hiyamizu et al., 2023b;Shak et al., 2023;Vuolo et al., 2018).In this study, we extend that work to define the sDAP protein CEP170 as a dynein-2 interactor.We show that dynein-2 interacts with CEP170 and that cells lacking CEP170 inefficiently assemble the dynein-2 holoenzyme as well as having minor ciliary defects.Our data suggests a role for CEP170 in suppor ng cilia homeostasis as well as dynein-2 assembly.

CEP170, but not CEP170B, interacts with dynein-2
Previously, we have used proteomic data generated using tagged WDR60 and WDR34 (Hiyamizu et al., 2023b;Shak et al., 2023;Vuolo et al., 2018) to iden fy dynein-2 interac ng proteins.The most consistent hit in these data sets, CEP170, was iden fied in 12 new and previously published tandem mass tag (TMT) datasets, 11 of which having a Log2 abundance ra o greater than 1 (Fig. 1A, Table S1).By comparison, the related protein, CEP170B was only found in 5 of those datasets, and only in 2 with a Log2 abundance ra o of greater than 1 (Fig. 1A, Table S1).We also previously iden fied CEP170 binding in non-TMT proteomic methods (Table S1) (Hiyamizu et al., 2023b).We validated the binding of CEP170 to WDR34 by coimmunoprecipita on (Fig. 1B).CEP170B has 33.4% iden ty with CEP170 and similar centrosomal localisa on to CEP170 (Fig. S1).However, we could not detect CEP170B in co-immunoprecipita on assays with HA-WDR34 (Fig. 1C) sugges ng that the interac on is specific to CEP170.Interes ngly, immunoprecipita on with HA-WDR34 in the background of WDR60 KO cells (Vuolo et al., 2018) led to the isola on of more CEP170 (Fig. 1B).The fact that we reliably iden fy CEP170 in experiments with both WDR34 and WDR60 in these experiments strongly suggests an interac on with the dynein-2 holocomplex.
CEP170 is a centrosomal protein, known to locate to sDAPs on mature centrioles (Guarguaglini et al., 2005).We therefore sought to inves gate whether CEP170 may have a role in cilia func on.S1.Lines represent mean.Co-immunoprecipita on of CEP170 (B) but not CEP170B (C), in WT RPE1 cells expressing HA-WDR34.When performed in WDR60 KO cells more CEP170 is pulled down.

CEP170 KO cells can s ll form cilia
To further study the role of CEP170 in ciliogenesis and cilia func on, we generated CEP170 KO RPE1 and mouse IMCD3 cells (Fig. S2) and performed serum-starva on induced cilia on assays (Fig. 2).CEP170 KO RPE1 cells were able to extend cilia at the same propor on as WT RPE1 cells (Fig. 2A and B).In one of the RPE1 CEP170 KO clones (23G7) there was a modest, but consistent, increase in cilia length (Fig. 2C).In IMCD3 cells, CEP170 KO cells were s ll able to ciliate (Fig. 2D) and the cilia were of comparable lengths to WT cells (Fig. 2E).However, we did observe that CEP170 KO IMCD3 cells ciliated at a reduced propor on compared to WT IMCD3 cells (Fig. 2F).
CEP170 is recruited to sDAPs by ninein (Graser et al., 2007).In ciliated cells, ninein was s ll recruited to the ciliary basal body (Fig. S3A), indica ng that our CEP170 KO and WDR60 KO cells s ll have sDAPs.We were able to further confirm this by electron microscopy (EM) on non-ciliated cells (Fig. S3B).
The ciliary TZ zone is an important barrier that separates the axoneme of the cilium to the rest of the cytoplasm, and disrup on to the TZ is associated with numerous ciliopathies (Garcia-Gonzalo et al., 2011;Garcia-Gonzalo and Reiter, 2017).In CEP170 KO cells the TZ proteins RPGRIP1L and TMEM67 were both localized correctly indica ng that the TZ is not grossly impacted by loss of CEP170 (Fig. S4).
Previous work has suggested a role for CEP170 in cilia disassembly (Lamla, 2009).We tested this by readdi on of FBS to serum starved cells.CEP170 KO cells remained ciliated for a much longer period than WT cells or cells lacking WDR60 (Fig. S5A and B).Live cell imaging showed that once detected, cilia excision proceeds similarly in both WT and CEP170 KO cells i.e. with considerable variability in ming but visually the same with respect to GFP-Arl13B (Fig. S5C).

IFT88 accumulates at the ciliary p in CEP170 KO cells
We next sought to address whether loss of CEP170 had any effect on IFT.IFT88 is a component of IFT-B and accumula ons in ciliary ps can be used to infer defects in IFT (Hou and Witman, 2015).Serum-starved cells were fixed and labelled to detect IFT88.In CEP170 KO RPE1 cells, we saw an increase in the propor on of cells that had IFT88 accumula ons at the ciliary p, similar to what we see in WDR60 KO cells (Fig. 3A and   B).In IMCD3 cells, CEP170 KO, but not CEP170B KO, caused an increase in the rela ve ciliary p frac on of IFT88 whereas there was no difference in the rela ve frac on at the base (Fig. 3C-E).These data indicate that loss of CEP170 could lead to mild defects in IFT leading to an accumula on of IFT88 at the ciliary p.

CEP170 KO cells do not display major defects in IFT dynamics
To further examine the IFT88-p accumula ons, we generated a CEP170 KO (Fig. S4) in IMCD3-FlpIn-IFT88-NeonGreen x3 (NG3) cells (Liew et al., 2014;Mukhopadhyay et al., 2010;Ye et al., 2018) to allow us to follow IFT in real-me using total internal reflec on fluorescence (TIRF) microscopy.Live imaging (Movies 1-3) and kymograph analysis (Fig. 4A and B) was used to measure anterograde and retrograde veloci es and events (Fig. 4C-F).We observed a slight increase in anterograde velocity for one of CEP170 KO clone (Fig. 4D) but there was no change in retrograde velocity (Fig. 4C) nor the number of events for transport in either direc on (Fig. 4E and F).

Heightened Smo response in CEP170 KO cilia
Sonic hedgehog (Shh) signalling requires func onal cilia (Breslow et al., 2018).In basal condi ons, the membrane receptor Smoothened (Smo), part of the Shh pathway, localizes to cilia but is rapidly exported by a process involving retrograde IFT.When Shh signalling is s mulated, by Shh or an agonist such as Smoothened agonist (SAG), Smo instead accumulates within the cilium (Corbit et al., 2005;Hamada et al., 2018;Vuolo et al., 2018).In absence of SAG, Smo is largely absent from cilia in WT and CEP170 KO cells (Fig. 5A and C).In the presence of SAG, we found that a greater propor on of cilia in CEP170 KO cells accumulated Smo than in WT cells (Fig. 5B and C).

Dynein-2 holocomplex assembly is disrupted in CEP170 KO cells
Muta ons in dynein-2 have been linked to Shh dysregula on (May et al., 2005).Therefore, we sought to examine the localiza on of the dynein-2 heavy chain (DHC2) directly.Notably, immunofluorescence showed a decrease in DHC2 localiza on to cilia in CEP170 KO cells compared to WT (Fig. 6A-C).We also noted a decrease in DHC2 labelling at the base of the cilium (Fig. 6D-E).No differences in expression of DHC2 or other dynein-2 subunits were observed by immunoblot (Fig. 6F), sugges ng that the reduc on in ciliary and basal body DHC2 does not relate to changes in dynein-2 overall expression.As we have done previously (Vuolo et al., 2018), we used proteomic profiling to allow us to look at dynein-2 assembly (Fig. 7).In WT RPE1 cells, using HA-tagged WDR60, we could detect DHC2, the intermediate chain WDR34, dynein-2 light chains and individual components of IFT-B as well as the IFT-A component, IFT140 (Fig 7A).This latest dataset is consistent with our earlier findings that CEP170, but not CEP170B, is reliably coimmunoprecipitated with HA-WDR60 (Fig 7A and Table S1).We repeated the analysis in CEP170 KO cells (Fig. 7B).We could detect DHC2, WDR34 and light chains as well as the chaperone NUDCD1 (known to be important in WD-repeat protein folding (Asante et al., 2014;Taipale et al., 2014;Vuolo et al., 2018)p.Other than the light chain proteins, DYNLT1 (TCTEX1) and DYNLT3 (TCTEX3) (which are also known subunits of dynein-1 (Vuolo et al., 2018;Vuolo et al., 2020)), all iden fied WDR60 binding partners had a Log2 abundance ra o below 0. These data strongly suggest a role for CEP170 in the assembly or stabiliza on of the dynein-2 holocomplex.and IFT components and CEP170 are detectable in RPE1 cells (dynein-2 subunits, green circles, IFT subunits, purple triangles, CEP170).We did not detect CEP170B in this dataset (see also Table S1).(B).HA-WDR60 interac on with dynein-2 components is decreased (Log2 rela ve abundance <0) in CEP170 KO cells.Three independent experiments were performed.Data are normalized to HA-WDR60 levels.Bars represent means.

Discussion
A func onal cilium requires an intricate assembly of many mul -protein complexes including dynein-2, IFT-A, and IFT-B.Our data show that CEP170 interacts with dynein-2 and promotes its assembly.Loss of CEP170 does not prevent cilia on or cause significant defects in cilia func on, sugges ng a possible role for CEP170 as a modulator of cilia.CEP170 was originally iden fied as a binding partner of polo-like kinase 1 (PLK1) in a yeast two-hybrid screen and shown to localise to sDAPs on mother centrioles (Guarguaglini et al., 2005).From this ini al work it was clear that CEP170 has a role in microtubule organisa on and subsequent studies have also linked it with mitosis and DNA damage responses (Qin et al., 2019;Rodriguez-Real et al., 2023;Welburn and Cheeseman, 2012;Zhang et al., 2019).CEP170 has also been shown to interact with the TZ protein RPGRIP1L (Gupta et al., 2015).CEP170 may act as a hub, connec ng the cell cycle and cilia disassembly with dynein-2 and cilia func on.

CEP170 and sDAPs in ciliogenesis and cilia func on
The mother centriole is dis nct from the daughter owing to the addi on of DAPs and sDAPs.The precise components of these structures are not fully defined, with new partners being iden fied and be er microscopy techniques con nuing to enable be er defini on of their exact loca on on centrioles.DAPs contain a core set of proteins, including ODF2 (Ma et al., 2023).sDAPs are comprised of ODF2, CEP128, centriolin, CCDC120, CCDC68, ninein, α-and γ-taxilin, and CEP170 (Ma et al., 2023).The consensus is that whilst DAPs are required for ciliogenesis, sDAPs are mostly seen to be redundant and are more important for microtubule anchoring (Hall and Hehnly, 2021;Mazo et al., 2016).Indeed, there are currently no reported ciliopathies associated with any known sDAP gene.Loss of ODF2 has been shown to affect ciliogenesis, but whether this is from their role in the sDAPs is not clear as ODF2 is also present in DAPs (Ishikawa et al., 2005;Kashihara et al., 2019).CEP170 is located at the p (distal) point of sDAPs, but addi onal non-sDAP centrosomal localisa on has been observed (Guarguaglini et al., 2005;Kashihara et al., 2019;Mazo et al., 2016).Previous studies showed that deple ng either CCDC120, CCDC68, or ninein is sufficient to remove CEP170 from sDAPs, and that this did not affect cilia on (Huang et al., 2017;Mazo et al., 2016).Our results are broadly consistent with these observa ons, cells lacking CEP170 were s ll able to generate cilia of comparable lengths to WT cells.Whilst we observed normal length cilia in IMCD3 CEP170 KO cells, we do note that they ciliated to a lower degree, sugges ng a more general role for CEP170 and sDAPs in cilia spa al control as previously observed (Mazo et al., 2016).However, we also see defects in cilia func on, not previously studied in this context.Takahara et al. (2018) reported that whilst KO of the IFT-A component, IFT122, led to severe ciliogenesis defects, KO of other IFT-A genes had minor effect on ciliogenesis but with impaired trafficking in ciliary proteins (Hirano et al., 2017;Takahara et al., 2018).In CEP170 KO cells, whilst we observed near-normal IFT veloci es (Lechtreck, 2015), we did see accumula on of IFT88 at the ciliary p, as well as impacts on SAG-dependent Smo-localisa on.
Cilia assembly and disassembly is linked to the cell cycle, only non-dividing cells are capable of primary cilium assembly (Breslow and Holland, 2019;Ishikawa and Marshall, 2011;Mirvis et al., 2018).Accordingly, during cell cycle progression the primary cilium must be disassembled.The processes that govern this are not fully defined and mul ple mechanisms have been suggested (Breslow and Holland, 2019;Mirvis et al., 2018;Phua et al., 2017;Pugacheva et al., 2007).In our study, CEP170 KO cells had delayed cilia disassembly, as shown previously using deple on of CEP170 by RNA interference (Lamla, 2009).In contrast, in WDR60 KO cells cilia are disassembled at a rate comparable to WT cells, sugges ng that the delay occurs independently to dynein-2.CEP170 may, therefore, act as a point of integra on between the cell cycle and cilia disassembly, possibly via already known interac ons between CEP170 and the microtubule depolymerising kinesins, KIF2A/B (Maliga et al., 2013;Miyamoto et al., 2015;Welburn and Cheeseman, 2012).

CEP170 and dynein-2 assembly
Along with cilia defects following loss of CEP170, the main conclusion of this work is the iden fica on of, CEP170, as an interactor of dynein-2.To date, the majority of published dynein-2 interac ons have only been reported in the context of IFT-A and IFT-B (Hiyamizu et al., 2023a;Hiyamizu et al., 2023b;Shak et al., 2023;Toropova et al., 2019;Tsurumi et al., 2019;Vuolo et al., 2018;Zhu et al., 2021) these interac ons have given us great insight into how dynein-2 interacts with IFT-A and IFT-B.Whilst we cannot say for certain which subunit of dynein-2 CEP170 binds to, CEP170 can be robustly detected using either WDR34 and WDR60 in immunoprecipita ons.Immunoprecipita on of HA-WDR60 from CEP170 KO cells compared to WT cells reveals roles for CEP170 in dynein-2 holocomplex assembly and in maintaining DHC2 localiza on at the basal body.Given the reduc on in dynein-2 holocomplex it might be expected that the cilia defects would be much more severe.However, in Chlamydomonas, deple on of endogenous DHC2 by approximately 50% does not result in any dras c defects in cilia (Reck et al., 2016).Even when DHC2 levels were more substan ally reduced, cilia were s ll able to form.Moreover, loss of WDR60 in nematodes and in RPE1 cells, does not ablate cilia forma on or func on (De-Castro et al., 2022;Vuolo et al., 2018), sugges ng that cilia require only small amounts of dynein-2 holocomplex to func on.This raises ques ons as to why cells maintain a larger pool of dynein-2 than is needed to maintain cilia and IFT.
Dynein-2 light, light-intermediate and intermediate chains (WDR34 and WDR60) are required for dynein-2 assembly (Vuolo et al., 2018), stabilise dimeriza on of the heavy chain, and support forma on of the autoinhibited state (Perrone et al., 2003;Rompolas et al., 2007;Toropova et al., 2019).In this study, we find that immunoprecipita on of HA-WDR34 from WDR60 KO cells captures more CEP170 than from WT cells.Considering these data together, it is possible that CEP170 interacts with intermediate chains during dynein-2 assembly and/or stabilises the complete holocomplex.CEP170 could also act in capturing dynein-2 as retrograde trains disassemble on exit from cilia (van den Hoek et al., 2022).In these ways, CEP170 could serve to enrich dynein-2 at sDAPs to promote IFT train assembly.The KIF3A subunit of the anterograde IFT motor kinesin-2 also localizes to sDAPs (Kodani et al., 2013), perhaps indica ng a broader role for sDAPs ac ng as a "shun ng yard" in recruitment of components for eventual IFT train assembly.

Cell culture
Human telomerase-immortalized re nal pigment epithelial cells (hTERT-RPE1, ATCC CRL-4000) were grown in DMEM-F12 (Gibco) supplemented with 10% fetal bovine serum (FBS) (Gibco) at 37°C with 5% CO2.Cells were not validated further a er purchase from ATCC.WDR60 KO RPE1 cells were generated previously (Vuolo et al., 2018).Mouse inner medullary collec ng duct (IMCD-3, ATCC CRL-2123) and IMCD3-FlpIn-IFT88-NG3 (a gi from M. Nachury, UCSF) cells were grown in DMEM/F-12(HEPES) (Gibco) supplemented with 5% FBS, 100 U/ml penicillin-streptomycin at 37°C with 5% CO2.Ciliogenesis and cilia disassembly RPE1 and IMCD3 cells were washed twice in phosphate-buffered saline (PBS) and incubated in serum-free medium for 24 hrs to induce ciliogenesis.For Smo experiments, confluent cells were placed in serum-free media and treated with Shh agonist SAG (Selleckchem (from Stratech Scien fic, Ely, UK) Catalog No.S7779) at the final concentra on of 400 nM for 24 hrs.A cilium disassembly assay was performed as described in Zhang et al. (2019).Specifically, cells were starved in serum-free medium for 48 hrs to induce cilium forma on.Serum was then added back to the medium to s mulate cilium resorp on.Cells were harvested at various me points for immunolabelling assays.To monitor cilia excision, RPE1 WT and CEP170 KO cells (clone 24H8) stably expressing GFP-Arl13B, were grown on imaging dishes and serum starved.Serum was added 30 minutes before imaging at 37°C on inverted an inverted fluoresence Leica-TIRF microscope using a 60×/1.40oil objec ve.Images were collected every 2-4 minutes un l the cilia became out of focus or photobleached.

Fluorescence intensity measurements
Quan fica on of fluorescence intensity was performed using average z-stack projec ons of original images in ImageJ (Schindelin et al., 2012).Local background normalised fluorescence intensity was measured at the cilary base or along the axoneme using the plot profile tool a er manually tracing the axoneme in the cilary marker channel.

TMT-labelling and proteomic analysis
Proteins on HA-agroase beads were digested with trypsin, TMT-labelled and analysed by mass spectometry as previously described (Vuolo et al., 2018).Three independent experiments were performed with data displayed as normalised Log2 abundance ra os.Live-cell IFT88 TIRF imaging Cells were seeded (1x10 5 ) on a 35 mm glass bo om imaging dish (MatTek) in normal media (DMEM/F12, HEPES with 5% FBS, Gibco).The day a er, cells were washed twice in PBS before being serum starved in phenol red-free media (DMEM/F12, HEPES, Gibco) for 24 hours.Cells were imaged at 37°C with CO2 on an Olympus/Abbelight SAFe360 system with two Hamamatsu Fusion sCMOS cameras, 488 nm diode laser and 100x oil immersion objec ve.The pixel size was 99.7 nm.Cell surface plane was found using automated TIRF angles in the mNeonGreen channel.Abbelight NEO so ware was used for acquisi on of movies, with frames captured every 100 ms for no longer than 1 minute per movie and no more than 1 hour per dish.

IFT88 TIRF kymograph analysis
Movies of individual cilia were imported into ImageJ (Schindelin et al., 2012).Each cilium was manually traced and kymographs (in both direc ons) were generated using the KymographClear plug-in for ImageJ (h p://www.nat.vu.nl/~erwinp/downloads.html,Prevo et al. (2015)).Individual events were manually traced, and the gradient was converted to a speed (µm⋅s -1 ) by dividing by frame length (100 ms) and mul plying by pixel size (99.7 nm).The average velocity from each cilium was plo ed.As we cannot tell the p from the base, anterograde IFT was defined as being the faster of either direc on.Where no obvious trace could be unambiguously traced, the velocity was called as 0 µm⋅s -1 .Electron microscopy Cells were grown on 35 mm dishes (Corning) before being fixed in 2.5% glutaraldehyde for 20 minutes and processed as previously described in Vuolo et al. (2018).Sec ons were imaged on an FEI (Cambridge, UK) Tecnai12 transmission electron microscope.

Figure 1 :
Figure 1: Dynein-2 interacts with CEP170 but not CEP170B.(A) Previous data sets from either GFP-WDR34, HA-WDR34 or HA-WDR60 interac on tandem mass tag (TMT) proteomics have reliably iden fied CEP170 being pulled-down with a Log2 abundance ra o above 1.Data is presented as Log2 abundance ra o, normalized to HA-GFP expression.A breakdown is shown in TableS1.Lines represent mean.Co-immunoprecipita on of CEP170 (B) but not CEP170B (C), in