Human NPHP3 (aa 1-203) (ENSG00000113971) was used as a ciliary targeting signal in both IMCD3 and RPE1 cells, and hereafter referred to as NPHP3. NPHP3-HA-Ubiquitin (ENST00000339647.6), or NPHP3-HA-Ubiquitin K0 (G75A-G76A, all Lys mutated to Ala, and C-terminal Gly mutated to Ala) fusions were incorporated into a Gateway pDONR201 vector. Using traditional Gateway cloning (Thermo Fisher Scientific) the ubiquitin sequences (kindly donated by Dr. Bert van der Reijden) were cloned into pDONR201 via a BP reaction (following the manufacturer’s instructions), while the NPHP3 and HA sequences were inserted using a PCR-based large insert cloning strategy. The NPHP3-HA-Ubiquitin fusions were further subcloned into a Gateway destination vector via an LR reaction.

The C terminal UBA domains (aa 185–409) of human RAD23B (ENST00000358015.8) (kindly provided by Dr. David P Toczyski (Mark et al., 2016), were cloned into pDONR201 serving as a polyubiquitin ligase trap. The BioID2 sequence was obtained from Addgene (MCS-BioID2-HA was a gift from Kyle Roux; Addgene plasmid #74224; http://n2t.net/addgene:74224; RRID:Addgene_74224) (Kim et al., 2016) and incorporated into a pgLAP1 vector via PCR-based large insert cloning. Subsequently, NPHP3 was inserted upstream of BioID2 using the same strategy to create a pgLAP1-NPHP3-BioID2 Gateway destination vector suitable for N-terminal tagging. This plasmid was used to perform an LR reaction with pDONR201-RAD23B (aa 185–409).

CAV1 wild type (ENST00000341049.7) and CAV1 lysine-less mutant sequences, previously described (Kirchner et al., 2013) and kindly provided by Dr. Hemmo Meyer, were cloned into pDONR201 and subsequently subcloned into a pgLAP5 vector (pgLAP5 was a gift from Peter Jackson; Addgene plasmid # 19706; http://n2t.net/addgene:19706; RRID:Addgene_19706) to create C-terminal eGFP fusions.

RAB8A and ARL13B with N-terminal mCherry fusion were generated as Gateway Entry plasmids (pENT220) using standard molecular biology techniques. Subsequently, lentiviral transfer vectors were generated by recombination into Gateway Destination vector pCDH.EF1A.GW.IRES.Blast (gift of Kay Schink) (Campeau et al., 2009) via LR reaction. Lentiviral particles were generated in HEK293Tcells using second generation lentiviral packaging vectors pMD2.G and pCMVΔ-R8.2 (kindly provided by Carlo Rivolta).

Murine inner medullary collecting duct 3 (IMCD3) Flp-In cells (a kind gift from Dr. Maxence Nachury) were stably transfected using the Thermo Fisher Scientific Flp-In™ technology to express NPHP3-BioID2 (hereafter called BioID2-control) or NPHP3-BioID2-RAD23B (aa185-409) (hereafter called BioID2-Uiquitin Binding Domain, BioID2-UBD). In summary, cells were transfected with equal amounts of plasmid DNA encoding the gene of interest and the Flp-In recombinase pOG44 using Lipofectamine 2000 (Life technologies; 11668019), followed by selection with 400 ug/ml Hygromycin (Sigma-Aldrich; H3274).

Human TERT-immortalized retinal pigment epithelial 1 (RPE1) cells were stably transfected with plasmids expressing NPHP3-HA-Ubiquitin or NPHP3-HA-Ubiquitin K0. Briefly, cells were transiently transfected with linearized plasmid DNA using electroporation (Amaxa Cell Line Nucleofactor Kit V, cat #VCA-1003), selected with G418 geneticin (Sigma-Aldrich; G8168), and single cell sorted to generate monoclonal lines.

RPE1 Flp-In cells were obtained from Ximbio (cat # 153242) and stably transfected with plasmids coding for wiltype CAV1-eGFP or mutant CAV1-eGFP as described. Stable transfectants were selected using 500 ug/ml Hygromycin.

RPE1, RPE1 Flp-In, and IMCD3 Flp-In cells were cultured in DMEM (Dulbecco’s Modified Eagle’s Medium, Sigma-Aldrich, D0819): F12 (Ham′s Nutrient Mixture F12, Sigma-Aldrich, N6658) 1:1, supplemented with 10% Fetal Calf Serum (Sigma-Aldrich, F0392) or 0.2% in the case of starvation medium, 1% Sodium Pyruvate (Sigma-Aldrich, S8636) and 1% Penicillin-Streptomycin (Sigma-Aldrich, P4333). To induce cilia formation, cells were grown in starvation medium for 48 hrs. Shortly, cells were washed in PBS, then fixed and permeabilized in either 2% paraformaldehyde (PFA) for 20 min, or ice-cold methanol for 5 min, followed by extensive washing with PBS. After blocking in 5% Bovine Serum Albumin (Sigma-Aldrich, A6003), cells were incubated with primary antibodies for 1.5 hrs at room temperature. The following primary antibodies were used: mouse anti-HA (1:1000, Sigma-Aldrich clone HA-7, H9658), chicken anti-BioID2 (1:1500, BioFront Technologies, BID2-CP-100), or Streptavidin-Alexa 488 conjugate (1:1000, Thermo Fisher Scientific, S32354), rabbit anti-CAV1 (1:500, Proteintech, 16447-1-AP), and rabbit anti-CAVIN1 (1:500 Cell Signaling Technology, clone D8C1D, 46379). In addition, rabbit anti-ARL13B antibodies (1:500, Proteintech, 17711-1-AP) were used to mark cilia, guinea pig anti-RPGRIP1L (1:500, home-made) was used as a ciliary transition zone marker, mouse anti-PCM1 (1:500, Santa Cruz, SC-398365) was used to stain centriolar satellites, and rabbit anti-EHD1 (1:500, Abcam, ab109311) was used to label the ciliary pocket. To wash off the primary antibodies, cells were extensively washed in PBS with 0.05% Tween (PBST). Subsequently, cells were incubated with secondary antibodies, Alexa Flour 488 (1:800, Invitrogen), Alexa Fluor 568 (1:800, Invitrogen), and Alexa Fluor 647 (1:800, Invitrogen) for 45 min followed by washing with PBST. Finally, cells were shortly rinsed in deionized water and samples were mounted using Vectashield with DAPI (Vector Laboratories, H-1200). Images were taken using an Axio Imager Z2 microscope with an ApoTome (Zeiss) at 63x magnification.

For live cell imaging, cells were seeded on glass-bottom dishes (35/22 mm; HBSt-3522; WillCo Wells) and serum-starved the next day. Imaging was performed after 24–48 hrs of serum-starvation. Before imaging, medium was replaced by phenol-free medium DMEM. For SiR-Tubulin experiments, cells were incubated with phenol-free DMEM containing 1 µM SiR-Tubulin-Cy5 probe for 1 h. Imaging was performed under regulated temperature (37°C) and humidity (5%) on a fully motorized Olympus IX83 inverted microscope equipped with a spinning disk (Yokogawa) and a Hamamatsu ORCA-Flash 4.0 digital camera (C13440) with 100× numerical aperture (NA) 1.4 oil objective (Olympus). The 488 nm, 561 nm and 640 nm laser lines were used for imaging eGFP, mCherry and SiR-Tubulin-Cy5, respectively. The time-lapse sequences ranged from 5 to 10 s.

RPE1 cells were processed for conventional TEM as follows: cells were grown on 35 mm glass bottom dishes (MatTec Cooperation), serum-starved for 48 h, and fixed with 2.5% glutaraldehyde (EM grade, EMS) in 0.1 M cacodylate buffer (CB, EMS) pH 7.4 for 1 h on ice. Cells were post-fixed in 1% osmium tetroxide (EMS) in CB, incubated with 1% low molecular weight tannic acid (EMS) in CB for 30 min at room temperature, and stained en bloc with 2% uranyl acetate in distilled water overnight. Specimen were dehydrated using a graded ethanol series and embedded in Durcupan resin (Sigma-Aldrich). Cells were sectioned parallel to the substratum using a diamond knife. 70–80 nm semithin sections were picked up on formvar- and carbon-coated EM slot grids and imaged on a TecnaiT12 TEM (Thermo Fisher Scientific) operated at 120 kV equipped with a 4k × 4k Eagle camera (Thermo Fisher Scientific). APEX labeling on the CAVIN1-APEX2-eGFP RPE1 cell line was performed as described previously (Ludwig 2020).

SDS-PAGE and western blotting were performed using the NuPAGE system from Thermo Fisher Scientific according to the manufacturer’s instructions. Proteins expressing the HA-tag were detected using a mouse monoclonal anti-HA antibody (1:1000, Sigma-Aldrich, clone HA-7). Recombinant proteins carrying a BioID2-tag were identified via chicken polyclonal anti-BioID2 antibodies (1:5000, BioFront Technologies, BID2-CP-100). For immunodetection of proximity labelled proteins a Streptavidin-IRD800 conjugate (1: 10000, Li-cor, 926–32230) was used and the blots were analyzed on an Odyssey DLx (Li-cor).

Proximity labelling was performed using the BioID2-UBD line described above, with a BioID2-control as negative control. Cells were plated at a confluency of 15% and cultured for 24 h in normal medium containing DMEM/F12 1:1 with 10% FCS and supplemented as described above. Subsequently, cells were stimulated for ciliogenesis for 48 h using starvation medium (0.2% FCS). Proximity labelling was induced for the last 24 h by supplementing the medium with 10 μM Biotin (Sigma-Aldrich, B4501). The cells were lysed in RIPA buffer (50 mM Tris-HCl pH 7.5, 150 mM NaCl, 1% NP-40, 0.1% SDS (Life Technologies, 15553-027), 0.5% Sodium Deoxycholate, 1 mM UltraPure EDTA pH 8.0 (Invitrogen, 11568896), followed by sonication. Cell lysis was continued for 1 h at 4°C while rotating. Lysates were cleared by centrifugation at 14,000 rpm for 20 min at 4°C, snap frozen and at −80°C stored until sample enrichment. Sample enrichment for biotinylated substrates was performed for 3 h at 4°C using StrepTactin Superflow beads (IBA, 2-1206-025). After incubation the beads were washed two times in 4°C 1xTBS (20 mM Tris, 150 mM NaCl) prior to on-bead digestion for 1 h at 27°C and 800 rpm, in Trypsin digestion buffer (2 mM Urea, 50 mM Tris-HCl (pH 7.5), 5 μg/sample trypsin (Serva, 9002-07-7). After digestion the beads were washed two times in digestion washing buffer (2 mM urea, 50 mM Tris-HCl (pH 7.5), 1 mM DTT). The on-bead digestion solution and two bead washes were pulled per sample to be snap frozen and stored at −80°C until mass spectrometry (MS) analysis.

RPE1 cells stably expressing the ubiquitin variants described above were seeded at 40% confluency and cultured in normal medium for 24 h, prior to ciliogenesis induction with starvation medium for an additional 48 h. The HA-based ubiquitin pulldown was performed under non-denaturing conditions. Cell lysis was performed as described above for the BioID2 samples, snap frozen and stored at −80°C until HA-purification. After thawing on ice, HA-purification was performed for 3 h at 4°C using anti-HA affinity beads (Thermo Fisher Scientific, 26181). After incubation the beads were washed, and an on-bead digestion was performed as described above for the BioID2 samples. Sample were snap frozen and stored at −80°C until MS analysis.

The trypsinized HA pull down and BioID2 proximity labelling samples were analysed on the Q Exactive Plus mass spectrometer as described in (Beyer et al., 2021). Next, Label-free quantification (LFQ) was performed, using Maxquant (V.1.6.1.0) (Cox and Mann 2008; Cox et al., 2014; Beyer et al., 2021). In this, protein identification was done using a human or mouse Swissprot database subsets form November 2019 (20,367 protein entries) or August 2019 (17,027 protein entries) respectively. Identified proteins were further analyzed for statistical enrichment. Identified proteins were classified in a three-Tier system. Tier 3 proteins were filtered to be present in at least 66% of the replicates for the sample of interest (excluding control samples). From these, Tier 2 proteins were classified if they were also Significance A (SignA) positive as calculated in R (version 4.0.4) (R Core Team 2021). Significant proteins were identified using a one-sided SignA outlier test with a Benjamini–Hochberg FDR correction, performed as described in Cox and Mann, SignA ≤0.05 (Cox and Mann 2011). Of these Tier 2 proteins, Tier 1 proteins were identified as being also Welch’s t-test positive. For this, LFQ ratios were used in a one-sided paired Welch’s t-test in combination with a Benjamini–Hochberg p-value correction, both from the stats package in R, with a q-value ≤0.05.

Tier 1 significant proteins per analyzed dataset were used for pathway enrichment analysis. When required, mouse gene IDs were translated to human gene IDs to allow comparison of the different dataset. Translation was performed using the homologene package from R (Mancarci 2019). GO Term enrichment was performed in three sets: biological processes (BP), molecular functions (MF), and cellular component (CC). Enriched terms were identified using the TopGo package in R, nodesize = 3, filtered for an Fisher’s exact test value ≤0.05 (Alexa and Rahnenfuhrer 2020).

The combined data from the UBD-PL and UAP approaches was used for active subnetwork identification using DOMINO, as previously described (Levi et al., 2021). Active subnetworks were scored based on STRING interaction data. Networks were displayed using Cytoscape (v.3.9.1.) (Shannon et al., 2003).

To determine percentage of ciliated cells and ciliary length, three biological replicates and three technical replicates were performed. For both analyses, ARL13B was used a cilia marker and ALPACA was used for automated image analysis of the cilium phenotype (Doornbos et al., 2021). Statistical significance was calculated using One-way ANOVA with a Kruskal–Wallis test (GraphPad Prism software). Four asterisks indicate significance p-value < 0.0001. Error bars indicate mean with SD.

To study cilia-specific ubiquitination, two different proteomics approaches were developed based on either affinity purification or proximity labeling. In both instances, we used a previously described ciliary targeting method utilizing amino acids 1-203 of the NPHP3 protein as an efficient ciliary localization signal (Mick et al., 2015).

In IMCD3 cells, the Ubiquitin-binding domain Proximity Labeling (UBD-PL) approach was developed (Figure 1A). Using the NPHP3 targeting strategy, the C-terminal ubiquitin-binding domain (UBD) of the RAD23B protein was fused to the BioID2 proximity labeling tag (Kim et al., 2016). UBDs are small modular domains which can form non-covalent bonds with ubiquitin (Dikic et al., 2009). Ubiquitin-associated domains (UBAs) are among the most common types of UBDs. The RAD23B protein contains a tandem of two UBAs at its C terminus (Chen et al., 2001), which was utilized in this study as an indirect bait for polyubiquitinated substrates. These polyubiquitinated targets will become biotinylated when they come in close proximity with the BioID2 tag, which promiscuously attaches biotin moieties to primary amines within an estimated radius of 10 nm (Kim et al., 2016). As a control, an NPHP3 (1-203)-targeted BioID2 recombinant protein was used. Immunostaining of the control and bait fusion proteins revealed accurate localization of both proteins to the ciliary compartment, as confirmed through co-localization with ARL13B (Figure 2A, upper panels). The same was true for the proximally biotinylated targets upon overnight treatment with biotin (Figure 2A, lower panels). Both recombinant proteins were detected at the expected size on a western blot (Figure 2B). Furthermore, biotin supplementation, followed by immunoprecipitation and western blotting, indicated a ladder of biotinylated substrates of various sizes confirming the presence of proximity labeled proteins in the samples (Figure 2C).

For the Ubiquitin Affinity Proteomics (UAP) approach in RPE1 cells, the NPHP3 ciliary targeting sequence was fused to either wild type (WT) ubiquitin, or a lysine-less ubiquitination-impaired K0 ubiquitin mutant which is unable to become conjugated to substrates and build polyubiquitin chains, as the C-terminal Gly residues were mutated to Ala (Figure 1B). To facilitate affinity purification and visualization in cells, the hemagglutinin (HA) tag was included at the N-terminus of ubiquitin in both recombinant proteins. Both ubiquitin variants were correctly localized to the cilium when transfected cells were co-stained with an antibody against the ciliary membrane marker ARL13B (Figure 2D). Moreover, immunoprecipitated WT ubiquitin showed the characteristic high molecular weight “smear” indicative of polyubiquitination of target substrates, whereas the ubiquitin K0 mutant did not (Figure 2E).

Altogether, we conclude that both the UBD-PL, as well as the UAP-based proteomics approaches, should yield cilia-specific polyubiquitinated proteins in transfected IMCD3 and RPE1 cell lines, respectively.

Mass spectrometry analysis following biotinylation of the cilia targeted UBDs in IMCD3 Flp-In cells resulted in the identification of a total of 1650 proteins across 6 experimental replicates per cell line. These were further subdivided into three Tiers based on stringency criteria. Tier 3 contained all the proteins which were not significantly enriched. Tier 1 (q-value ≤0.05 & Sign. A ≤ 0.05) comprised 70 highly significantly enriched proteins, whereas an additional 59 proteins were present in Tier 2 (p-value ≤0.05) when a less stringent cut-off was applied (Supplementary Table S1). RAD23B, which was used as a bait in this context, was found to be the most highly enriched protein in the dataset, thus technically validating the approach.

To test the specificity of our UBD-PL approach, we compared our dataset to a variety of ciliary screens and databases generated over the last decade (Sang et al., 2011; Ishikawa et al., 2012; Nakata et al., 2012; van Dam et al., 2013; Firat-Karalar et al., 2014; Gupta et al., 2015; Mick et al., 2015; Roosing et al., 2015; Wheway et al., 2015; Kim J. H. et al., 2016; Boldt et al., 2016; Toriyama et al., 2016; Kohli et al., 2017; Breslow et al., 2018; Pusapati et al., 2018; van Dam et al., 2019; Lv et al., 2021). We compared the proteins in Tier 1 to these published cilia-specific datasets. 76% of the significantly enriched proteins present in Tier 1 were previously identified in ciliary studies. Among the most frequently found proteins (11 out of 15 included studies) were for instance PCM1, OFD1, C2CD3, CEP131, and SQSTM1. Moreover, CYLD, which is also often present in the aforementioned ciliary screens, is known to regulate PCM1 protein levels via its function as a deubiquitinating enzyme (Eguether et al., 2014; Douanne et al., 2019) further validating our approach as specific to both ciliary and ubiquitination-related processes. Several novel candidate ciliary proteins were identified in our study (Figure 3A).

Importantly, we identified a large group of proteins involved in ESCRT-dependent clathrin-mediated endocytosis (Le Roy and Wrana 2005; Shields and Piper 2011; Haglund and Dikic 2012; Kaksonen and Roux 2018; Vietri et al., 2020; Mosesso et al., 2019; Roach et al., 2021; Shinde et al., 2022). The ESCRT-0 STAM2 (Tier 2), and ESCRT-1 TSG101, VPS28, and VPS37A (Tier 1) components, as well as additional members of the clathrin endocytic pathway including SH3RF1, SH3BP4, CLINT1, EPN2, EPN3, EPS15, EPS15L1, PICALM, ITSN1, ITSN2, and TOM1 (Tier1) and TOM1L2, PIK3C2A, AP2B1 (Tier 2) were enriched in our dataset. Clathrin-mediated endocytosis is a major pathway in the internalization and recycling of proteins involved in signaling. Not surprisingly, another significantly enriched group of proteins in our dataset is involved in signaling pathways, such as Wnt, Notch and TGF beta (NOTCH1, NOTCH2, AMER1, TAB1, TAB2, TAB3, TNRC6B, NF2, RELA, and RPS6KC1). A third group of highly enriched proteins in our UBD dataset, consists of proteins involved in centrosomal and centriolar satellite integrity: CEP131, CEP152, CEP85, C2CD3, CCDC14, HAUS5, HAUS8, NEDD1, PCM1, OFD1 (Tier 1) and CEP164, CEP350 (Tier 2) (Figure 3B).

Comprehensive GO term enrichment analysis (Ashburner et al., 2000; Carbon et al., 2021) was performed in order to gain further insight into the specific functional contribution of the proteins identified in our dataset to the regulation of cilia function via ubiquitination. This analysis was subdivided into three categories: biological processes (BP), molecular functions (MF), and cellular component (CC).

A total of 259 BP-related GO terms were significantly enriched in the Tier 1 UBD-based proximity labelling dataset, of which the top 50 most enriched terms can be found in Supplementary Figure S1. Of these, the 11 most enriched BP terms (Fisher’s exact test value ≤9.5E-5) indicate a role for ubiquitination in centrosome regulation, ciliary basal body-plasma membrane docking, vesicle-mediated transport, and signaling (Figure 4A). In the MF-related GO term category 24 terms were significantly enriched. Of these, the 8 most enriched MF terms (Fisher’s exact test value ≤5.5E-3) indicated that this dataset is most specific for K63-linked polyubiquitination. All other enriched functions were involved in protein trafficking and signaling, and included ‘phospholipid binding’, ‘signaling receptor binding’/‘receptor ligand activity’, and ‘clathrin binding’. These functions are in line with biological processes pertaining to vesicle transport via ESCRT-mediated K63 ubiquitination-linked endocytosis of signaling receptors (Figure 4B). Finally, in the last category, 46 CC GO terms were significantly overrepresented. The 15 most enriched terms (Fisher’s exact test value ≤7.5E-4) were again mostly linked to either regulation of processes occurring at the basal body/centriolar satellites or clathrin-coated vesicles (Figure 4C).

Altogether, the UBD proximity labeling approach yielded a highly specific dataset indicating the important and novel role of K63-linked polyubiquitination in regulating ciliary function in IMCD3 cells via the ESCRT-dependent clathrin-mediated endocytosis pathway.

To expand the ciliary ubiquitinome dataset, the UAP approach was developed in RPE1 cells. Similar to the UBD-PL approach, 6 replicates per cell line were subjected to mass spectrometry analysis following HA-based immunoprecipitation. This yielded a total of 1272 proteins that were present in at least 4 of the 6 WT samples. Statistical analysis revealed a total of 44 proteins that were Tier 1 enriched (q-value ≤0.05 & Sign. A ≤ 0.05), and additional 22 proteins that were enriched in Tier 2 (p-value ≤0.05 & SignA ≤0.05) (Supplementary Table S2). Comparing the significantly enriched proteins to published cilia-specific datasets revealed that 66% of the proteins identified in our UAP-based approach have previously been linked to cilia. Among these proteins was PCM1, which was also identified in our UBD screen (Figure 5A).

Since ubiquitin itself was used as a bait for this approach, by far the largest group of proteins identified with this method included proteins related to the ubiquitin cascade. This comprised E1 enzymes (UBA1, UBA6), the deubiquitinating enzyme USP5, and a number of E3 ligases (BIRC6, CUL4B, HERC4, HUWE1, HECTD1, ITCH, FBXO22, and LMO7). UBA1 and UBA6 have previously been found in the cilia proteome of IMCD3 cells (Mick et al., 2015). A variety of targets involved in the tightly interconnected groups of proteins involved in cytoskeletal organization and membrane trafficking was identified (Mick et al., 2015), such as: AFAP1, LIMA1, RAI14, ZNF185, EPB41L2, AP1M1, MYOF, RABEP2, and GJA1. Fewer proteins were identified in the categories of signaling and cell metabolism. Interestingly, three out of six the proteins identified in the latter category are involved in cholesterol metabolism (ACAT2, HMGCS1 and NCEH1). The most striking group of proteins identified in this screen, however, were several main structural components of caveolae, flask-shaped cholesterol-enriched invaginations on the cell membrane (Ariotti and Parton 2013; Parton and del Pozo 2013). These proteins were namely CAV1, CAVIN1 (PTRF) and EHD2 (Figure 5B). Moreover, two proteins with known involvement in CAV1 ubiquitination were also present in our screen, VCP (Tier 1) and ANKRD13A (Tier 2) (Hayer et al., 2010; Kirchner et al., 2013; Burana et al., 2016).

GO term enrichment analysis was performed as described in the previous section. GO BP analysis revealed a total of 180 enriched terms, of which the 50 most enriched terms are indicated in Supplementary Figure S2. Of these, the 19 most enriched terms (Fisher’s exact test value ≤1.5E-3) can be divided into three groups, “Metabolic process”, “Signaling”, and “Regulation of developmental process” (Figure 6A). GO term MF enrichment analysis yielded a total 13 significantly enriched MF terms (Fisher’s exact test value ≤0.05), most of which pertained to functions involved in the ubiquitination cascade (Figure 6B). The last enrichment analysis regarding CC GO terms resulted in a total of 15 significantly enriched terms (Fisher’s exact test value ≤0.05). The most enriched terms in this category were related to the ciliary basal body, centriolar satellites, and caveolae (Figure 6C).

In conclusion, our UAP-based approach in RPE1 cells identified potential novel key players in cilia-specific ubiquitination and more specifically, the regulation of CAV1 in the context of caveolae via ubiquitination.

In addition to the GO term enrichment analysis, DOMINO active network identification was performed. For this analysis, the combined UBD-PL and UAP data were used in order to identify the modules of the complete network that are involved in and regulated by ciliary ubiquitination (Figure 7). Despite the use of only Tier 1 proteins for the DOMINO analysis, the identified modules also contained many of the Tier 2 and Tier 3 proteins. This results in an overview of mainly enriched, but also some depleted proteins in the modules, highlighting the internal regulations within each module. Most significant are the ESCRT, transport, and endocytosis modules. The latter also includes the caveolae proteins CAV1, CAVIN1, and EHD2. A list of the GO terms enriched in each of the modules can be found in Supplementary Table S3.

Several previous studies have documented the presence of caveolae or CAV1 at the cilium-centrosome axis in different contexts, and have demonstrated important roles for CAV1 in regulating ciliary length and signaling (Schou et al., 2017; Rangel et al., 2019). To better understand the role of caveolae and ubiquitination of CAV1 in ciliary function, localization experiments were carried out in RPE1 cells. Although not present in all instances, in co-localization experiments using antibodies against CAV1 (Supplementary Figure S3A) or CAVIN1 (Supplementary Figure S3B) together with the ciliary membrane marker ARL13B, the two caveolae proteins were detected in the proximal ciliary region suggestive of TZ and CiPo localization (Supplementary Figure S3), in agreement with previous work (Schou et al., 2017; Pedersen et al., 2016). To confirm this observation, co-staining was performed using an antibody against EHD1 as a marker for the CiPo compartment (Lu et al., 2015). In RPE1 Flp-In cells stably expressing CAV1 fused C-terminally to the eGFP fluorescent tag, CAV1 co-localized with EHD1 in the CiPo, when both proteins were present in this compartment (Figure 8B). Native expression of the recombinant CAV1 protein was confirmed via a co-staining using an antibody against endogenous CAV1 and ARL13B was used to visualize the ciliary compartment (Figure 8A).

Next, we sought to better characterize the localization of CAV1 by means of live imaging microscopy. Using lentiviral transduction, mCherry-RAB8A or mCherry-ARL13B were stably introduced into the RPE1 CAV1-eGFP Flp-In cell line to mark the cilia. Tracking the localization of CAV1 over time (5–20 min) revealed a dynamic distribution of the protein at the CiPo (30% of total cilia, n = 75) and the base of cilia (Figures 8C, D; Supplementary Video S1). To confirm that this effect is not a consequence of mCherry-RAB8A expression, we performed identical experiments where mCherry-ARL13B was expressed as a cilia marker instead. CAV1-eGFP exhibited the dynamic pocket localization observed in RAB8A-expressing cells (Figure 8E, Supplementary Video S2). Similar observations were made when cilia were labeled with the microtubule dye SiR-Tubulin (Supplementary Figure S4). When comparing the SiR-Tubulin images to those of mCherry-RAB8A/ARL13B, it became evident that the basal CAV1-eGFP localization that we observed directly distal to ARL13B or RAB8A is not the basal body (labelled by the SiR-Tubulin dye), but very likely corresponds to the TZ (Supplementary Figure S4) (Schou et al., 2017).

Finally, the presence of caveolae at the CiPo was interrogated be means of transmission electron microscopy (TEM). Clusters of caveolae and clathrin coated vesicles (CCV) were readily detected close to the CiPo and the base of the primary cilium (Figure 8F) (Ariotti and Parton 2013; Parton and del Pozo 2013; Pedersen et al., 2016). To confirm that the rosette-like structures found in conventional TEM micrographs are indeed caveolae, we specifically stained caveolae using an RPE1 cell line stably expressing CAVIN1-APEX2-eGFP (Supplementary Figure S3C). Confocal microscopy analysis showed that CAVIN1 localized to the CiPo compartment (Supplementary Figure S3D), as expected. APEX2 staining followed by TEM revealed clusters of caveolae closely associated with the CiPo and the ciliary base (Supplementary Figures S3E–I).

The presence of VCP (Tier 1), a known ciliogenesis regulator (Raman et al., 2015), and ANKRD13A (Tier 2) in our ciliary UAP screen pointed to the specific contribution of CAV1 ubiquitination in the regulation of its ciliary localization and function (Kirchner et al., 2013; Burana et al., 2016). To study this, a ubiquitination-impaired CAV1 Flp-In cell line model was established. A CAV1 lysine-less mutant, which has been previously described (Kirchner et al., 2013), was fused C-terminally to eGFP and stably expressed in RPE1 Flp-In cells. The ALPACA (Accumulation and Length Phenotype Automated Cilia Analysis) (Doornbos et al., 2021) tool was used to compare ciliation levels, as well as cilia length, in RPE1 cells stably expressing either the wild type or mutant CAV1 fusion protein. To verify that the stably transfected wild type CAV1 does not itself affect these variables, as an additional control, the parental untransfected RPE1 Flp-In cell line was used. ARL13B was used as a ciliary marker. While the percentage of ciliated cells was not affected across the three cell lines (Figure 9A), cilia length was significantly longer (p < 0.0001) in the CAV1 mutant cells compared to both the CAV1 wild type, as well as the parental RPE1 Flp-In cell line (Figure 9B). Since previous work showed that CAV1 depletion similar causes ciliary lengthening (Schou et al., 2017; Rangel et al., 2019) these results suggests that the ubiquitination-impaired CAV1 mutant functions as a dominant negative.

We next wondered whether changes in the localization of the ubiquitination-impaired CAV1 mutant protein were the cause for this ciliary lengthening phenotype. As described in the previous section, mCherry-RAB8A was stably expressed in addition to the mutant CAV1-eGFP in RPE1 Flp-In cells, to serve as a marker for cilia (Figure 9C). Live cell imaging demonstrated that mutant CAV1 was still able to localize at the CiPo, however the protein accumulated in lumps associated with the proximal region of the cilium. These were rarely observed in the case of the wild type CAV1 protein (Figure 9D; Supplementary Video S3). This accumulation could indicate inefficient turnover of the mutant CAV1 protein at the CiPo, and a role for timely CAV1 ubiquitination in the regulation of ciliary length.