The BBSome assembly is spatially controlled by BBS1 and BBS4 in human cells

Bardet-Biedl Syndrome (BBS) is a pleiotropic ciliopathy caused by dysfunction of primary cilia. Most BBS patients carry mutations in one of eight genes encoding for subunits of a protein complex, BBSome, which mediates the trafficking of ciliary cargoes. Although, the structure of the BBSome has been resolved recently, the mechanism of assembly of this complicated complex in living cells is poorly understood. We generated a large library of human retinal epithelial cell lines deficient in particular BBSome subunit and expressing another subunit tagged with a fluorescent protein. We performed a comprehensive analysis of these cell lines using biochemical and microscopy approaches. Our data revealed that the BBSome formation is a sequential process including a step of the pre-BBSome assembly at pericentriolar satellites nucleated by BBS4, followed by the translocation of the BBSome into the ciliary base mediated by BBS1.


Introduction
Bardet-Biedl Syndrome (BBS) is a multi-organ genetic disorder caused by the dysfunction of the primary cilia, microtubule-based sensory organelles. BBS is primarily characterized by retinopathy, polydactyly, genital and renal anomalies, obesity, and cognitive impairment [1].
Mutations in any of the BBSome subunits can cause the BBS, suggesting that every subunit is essential for the complete BBSome function [2]. The structure of the BBSome was a longstanding enigma until recently, because the indirect approaches such as the yeast two-hybrid system [16], co-precipitation [17], co-expression of the individual subunits followed by lowresolution cryo-electron microscopy (EM) [18], and structural analysis of individual BBSome subunits [19,20] provided only a very limited insight into the overall BBSome structure.
Moreover, some authors proposed that BBS2, 7, and 9 form the BBSome core enabling the subsequent recruitment of other subunits [21], whereas others showed that BBS2 and BBS7 are dispensable for the assembly of the remaining six subunits [18,22].
Three independent studies resolved the BBSome structure using cryo-EM with molecular modeling using the BBSome isolated from bovine retina [23] or using high-resolution cryo-EM analysis of bovine BBSome [24] or of human BBSome (lacking BBS2 and BBS7) expressed in insect cells [22]. These studies showed a high level of interconnectivity among the BBSome subunits, raising the question of how these BBSome proteins assemble in the cells.
Quantitative fluorescence microscopy techniques can reveal the formation and dynamics of functional protein complexes in living cells [25,26]. In this study, we applied multiple microscopy and biochemical techniques to a large library of genetically engineered human RPE1 cells to describe spatially resolved steps of the BBSome formation in cells. pSpCas9(BB)-2A-GFP (PX458) vector kindly provided by Feng Zhang (Addgene plasmid #48138) [28].
sgRNA sequence with PAM motif (3′ end) for respective BBS genes are listed below:

Co-immunoprecipitation and western blotting
RPE1 cells were grown in 15 cm cell culture dishes and lysed in the lysis buffer (20mM HEPES at pH 7.5, 150mM NaCl, 2mM EDTA at pH 8, 0.5% Triton X-100) supplemented with protease inhibitor cocktail (complete, Roche, #05056489001). Lysates were cleared by centrifugation at 15,000 × g for 15 minutes at 4°C. Protein concentration was adjusted using the Pierce™ BCA Protein Assay Kit (Thermo Scientific). For co-immunoprecipitation assay, protein lysates were immunoprecipitated with anti-GFP antibody conjugated to protein A-coupled polyacrylamide beads (#53142 Thermo Scientific) for 2 h at 4°C. Beads were washed three times in the lysis buffer and co-precipitated proteins were denatured in 1× Laemmli buffer. For protein expression analysis, protein lysates were immediately denatured in reducing 4× Laemmli buffer. Denatured protein samples were analyzed by Western blotting following standard protocols. Membranes were probed with primary antibodies overnight at 4°C and secondary antibodies for 1 h at room temperature and developed using chemiluminescence immunoblot imaging system Azure c300 (Azure Biosystems, Inc.).

Flow cytometry
RPE1 cells expressing the YFP-tagged subunits were cultivated in 24-well dishes, trypsinized and centrifuged at 1000 × g for 1 min. The pellets were washed in PBS and re-suspended in FACS buffer (2mM EDTA, 2% FBS and 0.1% sodium azide). Measurements were taken on Aurora™ (Cytek Biosciences). Geometric means of fluorescence intensity (GMFI) of YFP positive cells were obtained and used to examine the protein expression. Flow cytometry data was analyzed using FlowJo (BD).

Immunofluorescence
RPE1 cells were cultured on 12 mm coverslips and serum starved for 24 h. Cells were fixed (4% formaldehyde) and permeabilized (0.2% Triton X-100) for 10 minutes. Blocking was done using 5% goat serum (Sigma, G6767-100ml) in PBS for 15 minutes and incubated with primary antibody (1% goat serum/PBS) and secondary antibody (PBS) for 1 h and 45 minutes, respectively in a wet chamber. The cells were washed after each step in PBS 3×. At last, the cells were washed in dH2O, air-dried and mounted using ProLong™ Gold antifade reagent with DAPI (Thermo Fisher Scientific).

Fluorescence microscopy
Image acquisition for cilia length rescue assays was performed on the Delta Vision Core microscope using the oil immersion objective (Plan-Apochromat 60× NA 1.42) and filters for DAPI (435/48), FITC (523/36) and TRITC (576/89). Z-stacks were acquired at 1024 × 1024 pixel format and Z-steps of 0.2 microns. The cilia length was measured using the Fiji ImageJ.

Expansion microscopy and basal body length quantification
Protocol is based on [29]. RPE1 cells were cultured on 12 mm coverslips and serum starved for 24 h.
Coverslips with cells were fixed with 4% formaldehyde/4% acrylamide in PBS overnight and then washed 2× with PBS. The gelation was performed by incubating coverslips face down with 45 μl of monomer solution (19% (wt/wt) sodium acrylate, 10% (wt/wt) acrylamide, 0.1% (wt/wt) N,N´methylenbisacrylamide in PBS supplemented with 0.5% TEMED and 0.5% APS, prepared as described in [29] in a pre-cooled humid chamber. After 1 min on ice, chamber was incubated at 37°C in the dark for For basal body length, only those perfectly or nearly perfectly oriented longitudinally to image focal plane were selected for measurements. The line scan and plot profile tools in Fiji ImageJ were used to determine the width at the proximal end of basal body marked with Ac-tub antibody. Length of basal body was measured as a distance between proximal end (Ac-tub signal) and distal end stained by NPHP1 antibody. At least four independent experiments were measured in each condition (WT, n=56; BBS1 KO, n=56; in total). For quantification of basal body length, in every experiment, the average value of widths was compared to ideal basal body width in full expansion (4.2×) to define the difference of ideal to real expansion factor (note, the range of expansions was between 3.35× and 4.1×). Then, as the length is subject of variability, all lengths were standardized by recounting the difference to full 4.2 expansion factor in particular experiment, and second, the average length was counted.
Fluorescence recovery after photobleaching, data processing, and analysis.
Cells were seeded in glass-bottom 8-well chambers (CellVis, USA) and serum starved for 24-48 h to induce ciliogenesis. Prior to fluorescence recovery after photobleaching (FRAP) experiments, cells were washed and supplemented with HBSS media (+Ca and +Mg) with 20mM HEPES. FRAP measurements were performed using a Leica TCS SP8 confocal microscope equipped with an oil immersion objective HC Plan-Apochromat 63× NA 1.4 oil, CS2 at room temperature. Data acquisition was performed in 512×512 pixel format with pinhole 2.62 Airy, at a speed of 1000 Hz in bidirectional mode and 8-bits resolution. Photobleaching (0.3 s) was performed with a circular spot 1 µm in diameter (centrosome, pericentriolar satellites (PS)) or with a rectangle spot 1 µm in length (ciliary base and tip) at 100% intensity using a 20 mW 488 nm solid-state laser. Fluorescence recovery was monitored at low laser intensity (3-10%) at 0.26 s intervals until reaching the plateau of recovery, in total for 60 to 80 s after photobleaching. A total of 25-30 separate FRAP measurements were performed for each sample. All FRAP curves were normalized to fluorescence loss during acquisition following the subtraction of background fluorescence. Curve fitting was performed in GraphPad Prism software using the one-phase association fit. All individual curves were fitted at once to obtain the mean and 90% or 95% confidence intervals of the desired parameters, halftime T 1/2 and the mobile fraction F m (see Supplemental Tables 2   and 3). Since the initial recovery after bleaching at the centrosome/ basal body and PS was contaminated by the diffusion towards these compartments, for simplicity and to estimate only the halftime of the protein fraction bound at these compartments, we restricted fitting of our data to times t > 2 s as we did in our previous study [26].
Fluorescence correlation spectroscopy, data processing, and analysis. The instrument was aligned and the detection volume was calibrated using 20 nM Abberior STAR 488 carboxylic acid (Abberior GmbH, Germany) calibration dye. TTTR offline data processing and fitting was performed in a custom made "TTTR Data Analysis" software. The analysis pipeline starts with splitting each recording into 10 equivalent parts, calculation of autocorrelation curve for each part, removal of outlier curves that contained signal from aggregates, estimation of standard deviation for each time point of the correlation curve and weighted non-linear least-square fitting by a standard one or two component anomalous 3D diffusion in a Gaussian shape detection volume model [30] with triplet state fraction [31].  Table 1). The fits enabled us to normalize the amplitude of correlation curves to 1 for lag time 10 μs, which allowed us to visually examine the differences between the measurements obtained from WT and KO cell lines and thus evaluate the presence of any potential second component. Only in case of YFP-BBS4 in WT cells we observed such dynamic behavior and therefore for this data we performed fitting also with the 2-component anomalous diffusion model (see

Generation of a cell line library for studying BBSome formation
To study the BBSome formation in cells, we initially established stable RPE1 cell lines expressing BBSome subunits fused with super yellow fluorescent protein SYFP2 (YFP) [32]. BBS1, BBS4, BBS8, and BBS18 were tagged at the N-termini, whereas the BBS7 and BBS9 were tagged at C-termini, because the N-terminal tag interfered with their function (data not shown). All YFP-tagged BBSome subunits localized to the primary cilia of RPE1 cells and showed weak diffuse signals throughout the cytoplasm (Fig. 1A).
In the next step, we generated a library of RPE1 cell lines deficient in BBS1, BBS2, BBS4, BBS7, TTC8/BBS8, BBS9, or BBIP1/BBS18 using the CRISPR-Cas9 technology (Fig. S1A-B). Deletion of any of the BBSome subunits prevented the ciliary localization of YFP-tagged BBS1, BBS4, BBS5, BBS7, BBS8, BBS9, and BBS18 (Fig. S2A). Accordingly, endogenous BBS9 localized to primary cilia both in WT and reconstituted KO cell lines, but not in the KO cells, where it was rather diffuse throughout the cytoplasm (Fig. S3A). These data suggested that only the intact BBSome, but not BBSome intermediates or subunits alone could enter cilia.
Despite substantial overexpression of the YFP-tagged subunits over their endogenous counterparts (Fig.   S3B), the fluorescence signal was still rather weak and mostly comparable in the respective WT and KO cell lines (Fig. S2A, S3C).
Cells deficient in BBS1, BBS4, BBS5, BBS7, BBS9, or BBS18 formed significantly shorter cilia in comparison to WT cells (Fig. 1B), corresponding to a previous observation in BBS4 deficient cells [33]. This phenotype was rescued by expressing YFP-tagged variants of the missing genes (Fig. 1B), documenting both that the cilia shortening was indeed caused by the gene deficiencies and that the YFPtag does not interfere with the function of the BBSome subunits. Intriguingly, deletion of BBS8 resulted in prolonged cilia, which was reverted by the expression of YFP-BBS8 (Fig. 1C).

BBSome subunits interact in the cytoplasm
Deficiency of one BBSome subunit reduced the cellular level of other subunits (Fig. S1A-B), indicating that the BBSome and/or BBSome intermediates are more stable than free individual subunits. In particular, deficiency in any of the subunits forming the proposed core of the BBSome (i.e., BBS2, BBS7, or BBS9) [21] substantially reduced the cellular levels of other subunits, whereas the absence of BBS1, BBS4, BBS8, or BBS18 had less dramatic effects on the stability of other subunits. The interdependence of the individual BBSome subunits suggests that they are present predominantly in the form of the BBSome or BBSome intermediates in WT cells.
We addressed whether ciliogenesis (induced by serum starvation) is coupled with de novo formation of the BBSome or whether the BBSome is pre-formed in non-ciliated cells (non-starved). BBS2, 5, 7, 8, and 9 co-immunoprecipitated with YFP-BBS4 in non-starved cells, showing that the BBSome assembles in non-ciliated cells (Fig. 2A). The serum starvation increased the amount of co-precipitated BBSome subunits ~2-fold, indicating that the BBSome formation is augmented upon ciliogenesis in RPE1 cells ( Fig. 2A-B).
The presence of the BBSome in non-ciliated cells suggests that the BBSome or BBSome intermediates are present in the cytoplasm. Using FCS, we estimated the diffusion speed of YFP-tagged subunits in WT and KO cell lines (Supplemental Table 1). Because the large proteins and complexes diffuse slower than small complexes, these measurements reveal the information about the relative size of the respective complexes. We observed that the diffusion speed of YFP-BBS4 was significantly faster in BBS9 KO cells than in WT cells, providing evidence that a fraction of BBS4 resides in a BBS9-dependent complex in the cytoplasm (Fig. 2C, F). We performed a two-component fitting of the WT data and found that the diffusion speed of this complex is about 10 times slower than the free YFP-BBS4 (Fig. 2C). Significant, but less pronounced effects were observed in BBS1 KO and BBS7 KO (Fig. 2D-F). These results show that cytoplasmic BBS4 exists in a form of a complex with BBS9 and likely other BBSome subunits.
YFP-tagged BBSome subunits, other than BBS4, showed significant differences in their mobility between WT and KO cells only in some rare cases (Fig. S4A-D). It is possible that their over-expression masked the effect of BBSome disruption by favoring the monomeric forms ( Fig. S3B-C). In case of YFP-BBS1, we could observe a mild loss in the slower fraction in BBS7 or BBS9 KO cells (Fig. S4B). Overall, these data imply the existence of heterogenic complexes of BBSome subunits in the cytoplasm.
The pre-BBSome assembles at the pericentriolar satellites BBS4 interacts with PCM-1 [34], which leads to its enrichment at the pericentriolar satellites (PS) both in ciliated and non-ciliated cells (Fig. S5A). BBS9 localizes to PS as well, albeit to a much lesser extent than BBS4 (Fig. 3A-B) indicating that a fraction of BBS9 and potentially other BBSome subunits reside at the PS in starved WT cells. Surprisingly, in BBS1 KO cells, both the BBS4 and BBS9 were highly enriched at the PS (Fig. 3A-B). In addition, we observed that all other BBSome subunits (i.e., YFP-tagged BBS5, BBS7, BBS8, BBS9, and BBS18 and endogenous BBS2 and BBS9) localize to PS in BBS1 KO cells (Fig.   3C, S5B). The enrichment of the BBSome subunits at the PS was dependent on BBS4, because it was not observed in BBS4-deficient cells and BBS1/BBS4 DKO cells (Fig. 3D). To analyze the localization of BBS9 with higher resolution, we employed the expansion microscopy [29]. In line with the previous observations, BBS9 localized to the cilium in WT cells, localized to PS in BBS1 KO cells, and was absent from the ciliary/centrosomal proximity in BBS4 KO cells (Fig. 3E).

FRAP analysis revealed comparable recovery half-times for all YFP-tagged BBSome subunits in BBS1
KO cells at the PS (Fig. 3F-G, Supplemental Table 2), possibly because they are predominantly engaged in a single complex. YFP-BBS4 showed the highest immobile fraction, which is consistent with its direct binding to the PS via PCM-1 (Fig. 3F). Most likely, BBS4 recruits other subunits to the PS leading to the formation of a pre-BBSome complex. The accumulation of the BBSome intermediates at the PS in BBS1 KO cells results in the mobilization of BBS4 (Fig. 3H-I, Supplemental Table 2), indicating that the pre-BBSome binding to PS is less stable than binding of BBS4 alone. BBS1 is crucial for the completion of the full BBSome, which prevents the BBSome subunits, with the exception of BBS4, from the arrest at the PS.

BBS1 facilitates BBSome translocation from the basal body to the cilium
As all BBSome subunits, BBS1 localizes to the cilia in WT cells (Fig. 1A). However, BBS1 localizes to the centrosome in non-ciliated cells and to the centrosome/basal body in some cells with short nascent cilia (Fig. 4A). Moreover, BBS1 localizes to the centrosome/basal body in BBS4 KO cells, whereas other BBSome subunits show diffuse localization in the cytoplasm of these cells (Fig. 4B). Deficiency of BBS4 increased the turnover of BBS1 at the centrosome/basal body (Fig. 4C-D), suggesting that the interaction of BBS1 with the pre-BBSome stabilizes BBS1 at the centrosome/basal body, possibly directing the whole complex towards the cilium.
We performed the analysis of the dynamic behavior of the BBSome subunits at the ciliary tip and the basal body and the transition zone using FRAP (Fig. 5A-C, Supplemental Table 3). If the subunits are incorporated in the whole BBSome quantitatively, their recovery rate should be comparable. However, BBS1 behaved differently from the other subunits, because it was very mobile at the base of the cilium (Fig. 5C, S5D). Indeed, BBS1 was the only subunit which was more dynamic in the ciliary base than in the ciliary tip (Fig. 5D, S5C-D). These data indicate that BBS1 exists in two forms at the ciliary base, as a monomeric protein with fast turnover between the ciliary base and the cytoplasm and as a part of the BBSome, whereas other BBSome subunits are recruited to the ciliary base only in the form of the complete BBSome complex.
Interestingly, the absence of BBS1 not only stalls BBSome at the PS, but alters the structure of the ciliary base resulting in slightly prolonged basal body and transition zone (Fig. 5E-F), suggesting a role of the BBS1 in the organization of the ciliary base.
Based on our data, we propose a model of the BBSome formation in cells with the main roles for BBS4 and BBS1 in the spatial regulation of the full complex assembly (Fig. 5G). BBS4 resides on PS and via its association with BBS9, it recruits other BBSome subunits to form the pre-BBSome complex. BBS1 localizes to the basal body and guides the pre-BBSome to the ciliary base and subsequently to the cilium.

Discussion
Because the BBSome consists of 8 subunits, it has been proposed that it assembles in a step-wise manner [21]. However, in vitro experiments provided contradictory data concerning the sequence of the individual steps [18,21,22]. Although the resolution of the structure of the BBSome complex represented a break-through in the field [22][23][24], these studies could not reveal how the BBSome assembles in living cells, including the spatial regulation of individual steps. We generated a library of 64 RPE1-derived cell It has been shown that some BBSome subunits stabilize each other in cells [21,35]. Our systematic analysis demonstrated that actually all BBSome subunits are interdependent. Deficiency in any of the subunits (with the exception of BBS5 which was not tested) leads to the substantial decrease of the cellular levels of other subunits, indicating that only a minor fraction of the BBSome subunit molecules are monomeric. Deficiency of BBS2, BBS7, or BBS9 had the most pronounced effects on the abundance of other subunits, suggesting that these three structurally related proteins form the core of the BBSome in vivo [21]. We showed that BBSome is readily assembled in non-ciliated cells and ciliogenesis only modestly augments the BBSome formation. Our FCS approach detected some direct or indirect cytoplasmic interactions of BBSome subunits, most notably BBS4-BBS9. Altogether our data show that BBSome and/or its intermediates are present in the cytoplasm and that the cilia is not required for the BBSome formation.
The PS organize proteins involved in the centrosome maintenance and ciliogenesis [36,37]. It has been shown previously that BBS4 localizes to the PS, where it interacts with resident proteins PCM-1, Cep290, and AZI1 [38][39][40]. Our data indicated that BBS4 recruits other BBSome subunits to the PS, where a pre-BBSome complex is formed. Absence of PCM-1 disrupts the PS and reduces ciliogenesis [3], but does not prevent the ciliary localization of the BBS4 [40]. However, other PS proteins, Cep72 and Cep290, mediate ciliary localization of BBS4 and BBS8, suggesting that they orchestrate the pre-BBSome formation [40]. Our data show that the pre-BBSome is released from the PS and targeted to the basal body via BBS1, the only subunit that shows an intrinsic affinity for the centrosome/basal body.
Depletion of AZI1 leads to enhanced localization of BBS4 into the cilium and restores the ciliary localization of BBS9 in the absence of BBS5, partially in the absence of BBS2 and BBS8, but not in the absence of BBS1 [38]. Although these experiments used RNA interference which usually does not lead to the complete loss of the target protein, the data are in line with the fact that BBS1 is essential for the ciliary BBSome targeting. AZI1 might negatively regulate the release of the pre-BBSome from the PS.
Interestingly, deficiency of AZI1 in zebrafish leads to the typical BBS symptoms [38], suggesting that the role of AZI1 is important for the BBSome function. LZTFL1 is a BBSome-interacting partner showing a similar loss-of-function phenotype to AZI1, including the ciliary localization of the incomplete BBSome [41] and the development of BBS symptoms, in this case in humans and mice [42,43]. It is tempting to speculate that AZI1 and/or LZTFL1 mediate a quality control mechanism blocking the release of incomplete pre-BBSome from the PS. As BBS5 is a relatively peripheral subunit of the BBSome [22][23][24] that is sub-stoichiometrically incorporated into the complex in vitro [22], a control mechanism checking the incorporation of BBS5 into the pre-BBSome in cells seems to be plausible.
Earlier studies suggested that the sequestration of the BBS4 at the PS might deplete the pool of BBS4 accessible for the BBSome formation [38,40]. However, the experimental data presented in these studies are consistent with our model of the pre-BBSome assembly at the PS and its subsequent release.
Moreover, the putative quality control function of LZTFL1 and AZI1 might explain why the depletion of these proteins causes a seemingly paradoxical coincidence of enhanced ciliary localization of BBSome subunits as well as the induction of BBS-like phenotypes. Last, but not least, it would be difficult to reconcile the model of BBS4 sequestration at the PS with our observation that all BBSome subunits are strongly enriched at the PS in the absence of BBS1.
Our model proposes that BBS1 is incorporated into the BBSome as the last subunit, because of its affinity to the centrosome/basal body. However, we cannot exclude that the last step of the BBSome formation is not the incorporation of BBS1 itself, but another BBS1-dependent event. It could be a conformational change in the BBSome, caused by a BBS1-interacting GTPase, BBS3/ARL6 [9,19,24]. In any case, BBS3 seems to be involved in the translocation of the (pre-)BBSome from the PS to the ciliary base, as the BBSome-BBS3-GTP does not co-purify with PCM-1 [3,9] and BBS3 was proposed to mediate the interaction of the BBSome with ciliary membranes [9,24]. BBS1 has been shown to interact with RABIN8, a guanine-exchange factor for RAB8, which is involved in vesicular trafficking and might regulate BBSome targeting to the ciliary base [3]. Overall, it seems likely that BBS1 facilitates an energyconsuming conformational change of the complex that eventually confines the BBSome in the cilia.
Chaperonins BBS6, BBS10, and BBS12 assist the formation of the BBSome [21,44] and the loss of any of them leads to the BBS [45][46][47]. Because the deficiency in BBS1 or BBS3 generally presents with a less severe BBS disease than the deficiency in BBS2, BBS7, or BBS9 core proteins or in the BBS chaperonins [2], it is probable that the chaperonin complex assists any of the initial steps of the BBSome formation as suggested previously [21] rather than the BBS1-dependent translocation into the ciliary base. However, this should be experimentally addressed in further studies.
Overall, our study reveals that the BBSome formation is a sequential process with spatially compartmentalized steps. The formation of the pre-BBSome is nucleated by BBS4 at the PS, whereas BBS1 drives the translocation of the BBSome to the ciliary base. It shall be addressed in follow-up studies whether some of the multiple reported causative mutations [2] interfere with the process of the BBSome assembly.    Table 1). The means of n>20 measurements are shown.     (G) Schematic model of the spatially resolved formation of the BBSome in human cells. BBS4 is a resident protein at PS, while BBS1 resides at the basal body. BBS4 associates with BBS9 in the cytoplasm, which directs the BBSome subunits to the PS to assemble the pre-BBSome. Pre-BBSome is stabilized by BBS1 at the basal body resulting in the full BBSome competent to enter the cilia.