Targeting of β-Arrestin2 to the Centrosome and Primary Cilium: Role in Cell Proliferation Control

Background The primary cilium is a sensory organelle generated from the centrosome in quiescent cells and found at the surface of most cell types, from where it controls important physiological processes. Specific sets of membrane proteins involved in sensing the extracellular milieu are concentrated within cilia, including G protein coupled receptors (GPCRs). Most GPCRs are regulated by β-arrestins, βarr1 and βarr2, which control both their signalling and endocytosis, suggesting that βarrs may also function at primary cilium. Methodology/Principal Findings In cycling cells, βarr2 was observed at the centrosome, at the proximal region of the centrioles, in a microtubule independent manner. However, βarr2 did not appear to be involved in classical centrosome-associated functions. In quiescent cells, both in vitro and in vivo, βarr2 was found at the basal body and axoneme of primary cilia. Interestingly, βarr2 was found to interact and colocalize with 14-3-3 proteins and Kif3A, two proteins known to be involved in ciliogenesis and intraciliary transport. In addition, as suggested for other centrosome or cilia-associated proteins, βarrs appear to control cell cycle progression. Indeed, cells lacking βarr2 were unable to properly respond to serum starvation and formed less primary cilia in these conditions. Conclusions/Significance Our results show that βarr2 is localized to the centrosome in cycling cells and to the primary cilium in quiescent cells, a feature shared with other proteins known to be involved in ciliogenesis or primary cilium function. Within cilia, βarr2 may participate in the signaling of cilia-associated GPCRs and, therefore, in the sensory functions of this cell “antenna”.


Introduction
An increasing number of reports have highlighted the function of the primary cilium (PC) in the control of several physiological processes. The PC is a hair-like cellular extension found at the surface of most vertebrate cells. This sophisticated microtubulebased organelle has been shown to sense multiple mechanical and chemical stimuli from the environment and to elicit specific cellular responses, which play crucial roles in embryonic development and homeostatic processes in adulthood. The PC has also recently been implicated in the regulation of cell cycle progression and, as a consequence, a lack of PC was associated with increased proliferation [1][2][3].
PC formation (ciliogenesis) takes place in quiescent or differentiated cells. PCs are assembled from the mother centriole of the unique centrosome present in these cells, which therefore corresponds to the basal body of PC. The mother centriole is docked at the membrane through its distal appendages and gives rise to the microtubule-based 9+0 axoneme, which forms the skeleton of this ''antenna'' like extension of the plasma membrane.
Whereas the basal body shares many properties with classical centrosomes, made of two centrioles and of a pericentriolar matrix, the axoneme represents a unique domain, characterized by the exclusion of many proteins and the enrichment of specific soluble, cytoplasmic, as well as membrane-associated components [1][2][3]. This sorting is achieved through a complex process mediated by highly conserved machineries, involved in both the selection of ciliary proteins, which likely contain specific motifs and the transport along the axonemal microtubule doublets. The components of these machineries can either be specifically devoted to ciliary-protein transport and/or ciliogenesis, as IFT (intraflagellar transport) proteins [4], or participate in other cellular processes, as reported for the aPKC-par3-par6 polarity cassette [5,6], the 14-3-3 adaptor protein [5] and Kif3A, a kinesin required for the anterograde transport towards the tip of the PC [4]. Polycystins, proteins involved in mechano-sensation of tubular renal cells [7] and growth factor receptors [3] figure among the proteins that are highly enriched in ciliary membranes.
Receptors belonging to the G-protein coupled receptor (GPCR) family are involved in the sensing of many different kinds of molecules including odorants, ions, amines, proteins or light, and thus regulate a large array of physiological processes. Some GPCRs accumulate at the PC, such as the somatostatin type 3 receptor, which is localized at PCs in neurons [8], or smoothened (smo), the GPCR-like transmembrane protein controlling the Hedgehog pathway [9], for which translocation to the PC is essential for signalling activity [10,11]. Most GPCRs are regulated by non visual arrestins, arrestin2 and arrestin3, also known as b-arrestin1 (barr1) and b-arrestin2 (barr2), which uncouple activated receptors from Gproteins, promote their endocytosis through clathrin-coated pits and mediate receptor-dependent activation of MAP kinases [12,13].
barrs regulate numerous key physiological and developmental processes as shown by the fact that the lack of both isoforms results in early embryonic lethality [14]. They are highly conserved among higher eukaryotes, although only vertebrates express the two isoforms, which show a high sequence homology and are encoded by two separate genes. barr isoforms share most of their partners and functions, however, several isoform-specific roles have also been described. In particular, only barr2 displays an active nucleocytoplasmic shuttling, which redistributes nuclear binding partners to the cytoplasm, whereas regulation of histone acetylation at certain promoters was only reported for barr1 [15,16]. Interestingly, barr2, not barr1, was found in the cilia of olfactory neurons [17,18], suggesting that the former might regulate odorant receptors within these structures, which are very similar to PCs. Here, we report that barr2 is specifically localized to the centrosome of cycling cells. Since most PC-associated proteins are also present at the centrosome in cycling cells, we investigated if barr2 could also be localized to the PC.

Results
barr2 localizes to the centrosome, in the proximal region of the centrioles, independently of microtubules When expressed as a GFP-fusion, barr2 showed a diffuse cytoplasmic distribution except one or two bright spots close to the nucleus. This localization was suggestive of the centrosome and, confirming this hypothesis, the spots were also decorated with pericentrin ( Figures 1A and S1A), a centrosomal marker. The centrosomal targeting of barr2-GFP was specific, as barr1-GFP or GFP alone were not enriched at sites of pericentrin staining ( Figure 1A and S1A).
The accumulation of barr2-GFP at the centrosome was quantified, based on GFP-associated fluorescence. Normalized fluorescence at the centrosome for both GFP alone and barr1-GFP was close to 1 (1.15 and 1.20, respectively), whereas that of barr2-GFP (2.45) clearly showed accumulation at this organelle ( Figure 1B). This new specific localization of barr2 was independent of the tag fused to the protein, as barr2 fused with Cherry (barr2-Ch), a different fluorescent protein [19], was also enriched at the centrosome ( Figure S1B).
Endogenous barr2 targeting to the centrosome was subsequently investigated using barr2-specific antibodies and mouse embryonic fibroblasts (MEFs), derived from wild-type (WT), barr2 (2KO), barr1 (1KO) and barr1/2 double knock-out (1/2KO) embryos [14] as controls ( Figures S2 and S3). Both antibodies specifically stained bright spots close to the nucleus, which colocalized with pericentrin or c-tubulin (Figures 2, S3 and S4). In addition, localization of barr2 at the centrosome was found in b-Arrestin2 at Primary Cilium both interphasic (Figures 2A and S4) and mitotic cells ( Figure 2B). No colocalization was detected with a-tubulin in mitotic cells ( Figure 2B), indicating that barr2 is not associated with the mitotic spindle. Altogether, these results show that barr2 is associated with the centrosome throughout the entire cell cycle.
The centrosome is composed of two centrioles which are involved in distinct functions [20]. Careful analysis of the staining patterns revealed that the distribution of barr2 within the centrosome was restricted to structures close to, but distinct from c-tubulincontaining areas (Figure 2A, insets, and Figure S4B). To more precisely characterize barr2 localization at the centrosome, we used centrin as a marker of the distal part of centrioles [20]. Both barr2 antibodies and barr2-Ch stained two spots juxtaposed to each centrin-decorated structures ( Figure 2C), indicating that barr2 is targeted to the proximal region of centrioles. This specific localization was further confirmed with 3D reconstruction of deconvoluted images, in which juxtaposition of barr2 and centrin was clearly visible ( Figure 2D). Combined, these results establish that barr2 is targeted to the proximal region of the centrioles ( Figure 2E) and that this localization is not modified during the cell cycle.
Targeting of proteins to the centrosome can be achieved through microtubule dependent transport or independently of microtubules by a dynamic exchange with cytoplasm [21]. As shown in Figure 3A, barr2 was still present at the centrosome in cells treated with nocodazole or taxol, drugs which destabilize or stabilize microtubules, respectively. Normalized fluorescence of barr2 at the centrosome was indeed similar in drug-treated cells compared to control ( Figure 3B), suggesting that targeting of barr2 to the centrosome is independent of microtubules. This hypothesis was further tested following the dynamic of barr2 at the centrosome by fluorescence recovery after photobleaching ( Figure 3C). Live cells were treated with nocodazole for one hour at 37uC before and during dynamic analysis, a condition which did result in inhibition of microtubule-based transport ( Figure S5). In the presence of nocodazole, centrosome-associated barr2-GFP fluorescence was recovered after photo-bleaching ( Figure 3C), with similar kinetics as in control cells (data not shown). These results suggest that the localization of barr2 at the centrosome likely results from a dynamic exchange between a centrosomal and a cytoplasmic pool.
Since the basic functions of the centrosome are the nucleation and anchoring of microtubules [21], we investigated if barr2 could affect these processes. Neither process was perturbed by absence or overexpression of barr2 ( Figure S6). In addition, we could not detect any increase of multinucleated cells in barr-deficient MEFs (data not shown), suggesting that these proteins did not show any role in cytokinesis, another key function of the centrosome [21]. These data therefore suggest that barrs in general are not involved in classical functions of the centrosome.
barr2 is localized to the primary cilium Since most proteins found at the PC in quiescent cells are found at the centrosome of cycling cells, we investigated whether localization of barr2 at the centrosome might reflect some role at the PC.
PC formation can be induced in vitro in both fibroblasts and RPE1 cells, a widely used model to study ciliogenesis, by growing cells to confluence and this process can be enhanced upon serum starvation (see methods). PC can then be identified using antiacetylated-tubulin (AT) antibodies, which stain the stabilized array of microtubules forming the axoneme. The basal body, which corresponds to the unique centrosome of these cells, can be identified using centrosomal markers. After induction of PC assembly in RPE1 cells (Figure 4), endogenous barr2 was found in the axoneme, as indicated by its colocalization with AT ( Figure 4A) and corroborated by 3Dreconstruction of deconvoluated images ( Figure 4B). The specific targeting of barr2 to the axoneme was confirmed using transfected barr2-Ch which, at low expression levels, did colocalize with AT ( Figure 4C). In addition, both endogenous and transfected barr2 were also present in two spots at the base of the axoneme (Figure 4, arrows), which colocalized with pericentrin ( Figure 4D), thus corresponding to the basal body. Finally, barr2-Ch colocalized with the active form of smoothened (smo*) at the level of the axoneme but not at the basal body ( Figure 4E), indicating a possible function in the regulation of cilia-dependent signalling pathways (see discussion).
To establish barr2 targeting to PC in vivo, distribution of barr2 was analyzed in mouse kidney sections ( Figure 4F), where PCs are located at the luminal side (Lu) of tubular epithelial cells [22]. The staining pattern of the barr2 antibody was similar to that found in cultured cells, showing a colocalization of barr2 and AT at the apical membrane of epithelial cells ( Figure 4F). Together, these

Lack of barr2 results in ciliogenesis defects and uncontrolled proliferation
Examples in the literature have established that depletion and/ or overexpression of cilia-associated proteins may result in ciliogenesis defects [23][24][25][26]. We took advantage of the fact that WT MEFs can form PCs to investigate the potential role of barrs in the control of PC formation.
The ability of the 2KO MEFs to form PC upon serum starvation was compared to that of WT MEFs. As shown in Figure 5A and 5B, 76.5% of WT MEFs displayed an assembled PC, indicated by a unique AT-positive 2 mm long rod-like structure, while only 51.5% of the 2KO MEFs exhibited a PC. Similar results were obtained when antibodies against polygluta- . barr2 is found in the axoneme and basal body of primary cilia. (A) Confluent RPE1 cells were serum-starved for 24 hours, then fixed and stained for acetylated-tubulin (AT), which is highly enriched in primary cilia (axoneme) and for endogenous barr2 using rARR antibody. (B) Z-stacks images of representative cells were deconvoluated as in Figure 2 and a 3D reconstruction of a representative cilium is shown. (C and D) RPE1 cells transfected with plasmids encoding for the barr2-Cherry fusion (barr2-Ch), were serum-starved for 24 hours after transfection, then fixed and stained for AT (C) or for the basal body marker pericentrin (D). In coloured images, barr2 staining is in red, centrosome or cilia markers in green and nuclei stained with DAPI are in blue. (E) RPE1 cells transfected with plasmids encoding for Flag-tagged active form of smoothened (smo*) and the barr2-Ch fusion were serum-starved for 24h after transfection, then fixed and stained for AT and smo*, using a rabbit polyclonal anti-Flag antibody. In coloured image, barr2 staining is in red, AT in blue and smo* in green. Insets show higher magnifications of representative areas. Arrows stress basal bodies. Scale bars represent 5 mm. (F) Tissue sections from adult mouse kidney were stained for AT and for barr2 using the rARR antibody. In coloured image, barr2 staining is in red, AT in green and nuclei stained with DAPI in blue. The lumen of a representative tubule is indicated (Lu). Insets show higher magnifications of a representative ciliated tubular epithelial cell. Arrows stress AT positive structures. doi:10.1371/journal.pone.0003728.g004 mylated-tubulin were used to visualize PCs ( Figure S7), indicating that the observations based on AT stainings were not just due to side effects of microtubule acetylation.
Because barrs are redundant for most functions, we investigated the effect of depleting either barr1 alone or both barr1 and barr2 on the number of PCs. MEFs lacking barr1 were not affected in their ability to form PCs, with a similar proportion of ciliated cells being measured, compared to wild-type cells ( Figure 5A and 5B). This result is consistent with the fact that RPE1 cells do not express detectable amounts of barr1 ( Figure S3B) and do form PC ( Figures 4, and 5). In contrast, 1/2KO MEFs were greatly impaired in their ability to form PC, with only ,18% of ciliated cells ( Figure 5A and 5B), indicating that ciliogenesis is severely affected in cells completely devoid of barrs. As expected from these observations, depletion of endogenous barr2 in RPE1 cells with two different small interfering RNA (siRNA), resulted in markedly reduced ciliogenesis in low serum conditions, compared to a non relevant siRNA (luciferase, Figure 5C and 5D).
A close link likely exists between PC assembly and control of cell cycle progression. It is assumed that only cells exiting the cell cycle and entering into G 0 phase can form a cilium. On the other hand, ciliogenesis defects result in cell cycle progression and uncontrolled proliferation [2,27]. Since our data indicate that cells lacking barr2 are affected in their ability to form PCs, we investigated whether these cells would also exhibit defects in proliferation.
We first found that the presence of a PC was correlated with exit from the cell cycle and entry into G 0 phase. Wild-type MEFs ( Figure 6A) were grown in low serum for two days after confluence to induce PC formation. Cells were then stained for both AT and Ki-67, a nuclear protein expressed in cells cycling from G 1 to M Figure 5. barr2 deficiency results in ciliogenesis defects. (A) WT, 1KO, 2KO and 1/2KO MEFs were grown in low serum (0,5%) for 48h, then fixed and stained for AT (green). Nuclei were stained with DAPI (blue). (B) The percentage of cells with a normal primary cilium was quantified. Values are the means (+/2 SD) of at least 300 cells from three independent experiments (*: p,0,01; **: p,0,001). (C and D) RPE1 cells were treated with control luciferase siRNA (Luc), si-barr2#1 or si-barr2#2 to deplete endogenous barr2. (C) Expression of barr2 was analyzed by western-blot with the rARR barr2 antibody. Expression of the c subunit of the AP-1 clathrin adaptor complex was tested as a control. (D) Cells from the same experiment were also seeded on coverslips and the percentage of cells with primary cilia was determined following AT staining. Values are the means (+/2 SD) of ,300 cells from a representative experiment done in triplicate. doi:10.1371/journal.pone.0003728.g005 [28]. Although in high serum conditions most cells (.80%) showed nuclear Ki-67 staining ( Figure 6B), in low serum conditions ( Figure 6A and 6B), ciliated cells (arrows) were not positive for Ki-67 (blue nuclei) whereas adjacent non-ciliated cell expressed this proliferation marker (pink (blue and red) nucleus). These results confirmed that ciliated cells were in G 0 phase.
We subsequently confirmed that ciliogenesis defects observed in barr2-depleted cells are correlated with cell cycle dysregulation. In WT MEFs, serum starvation induced a decrease of the proportion of cells positive for Ki-67, from 80% in cells grown in high serum, to 50% and 25% for cells grown in low serum for 24 and 48 hours, respectively ( Figure 6B). Interestingly, the percentage of Ki-67 positive cells was inversely correlated with the number of ciliated cells (compare with Figure 5B). When 2KO and 1/2KO MEFs were grown in the same conditions, the number of cells positive for Ki-67 moderately decreased upon serum starvation but remained constant from 24 to 48 hours with 60 to 70% of the cells remaining positive for Ki-67 at 48h ( Figure 6B). Since the percentage of ciliated cells was decreased in barr2 deficient cells ( Figure 5B), it appears that the defect in PC formation is correlated with an absence of response to low serum conditions and impaired exiting from cell cycle to enter in the G 0 phase. Similarly, an increased proportion of Ki-67 positive cells was also found in RPE1 cells depleted for barr2 and grown in low serum conditions ( Figure 6C).
In cystic kidney disease, a pathological condition associated with impaired formation and/or function of PCs, cyst formation is due to both loss of planar polarity and increased mitosis in tubular cells [22]. Moreover, recent studies showed that Polaris, a protein responsible for a mouse model of polycystic kidney disease, controls cell cycle progression [26]. When MEFs were grown in high serum conditions, 2KO and 1KO cells showed increased proliferation compared to wild type cells, a phenotype more pronounced in 1/2KO cells ( Figure 6D, p,0.01 at 72h). Interestingly, cells lacking both barr isoforms kept growing even in low serum conditions ( Figure 6E, p = 0.001 at 72h). Therefore, the marked defect in ciliogenesis observed in 1/2 KO cells is likely to result from the inability of these cells to respond to signals that inhibit cell proliferation.
barr2 interacts with 14-3-3 proteins and kinesin Kif3A 14-3-3 proteins (comprising b, c, e, g, f, t and s isoforms) are molecular adaptors, which often interact with consensus phosphorylated serine/threonine motifs of many proteins, thereby controlling a wide array of processes including signalling, cell cycle and apoptosis [29]. Interestingly, the 14-3-3f isoform was found to interact with the aPKC-Par3-Par6 polarity cassette [30], whereas depletion of 14-3-3g, which was found in a molecular complex with Par3 and the kinesin Kif3A, resulted in ciliogenesis defects [5]. Unpublished yeast two-hybrid data revealed that barrs interact with 14-3-3 proteins, in agreement with a recent proteomic study [31]. The implication of 14-3-3g in ciliogenesis and its connection with intraciliary transport through Kif3A, prompted us to characterize these interactions with barr2. Endogenous 14-3-3 proteins were precipitated by a GST-barr2 fusion ( Figure 7A) and the interaction between barr2 and the 14-3-3f isoform was confirmed by co-immunoprecipitation experiments showing that endogenous 14-3-3f interacts with Flag-tagged barr2 ( Figure 7B). The barr2 C-terminus contains a motif (RPQSAP), similar to phosphorylated 14-3-3 consensus binding sites ( Figure 7B). However, mutation of S361 within this motif did not affect the interaction of barr2 with endogenous 14-3-3f, which co-immunoprecipitated as efficiently with both the S361A and S361D mutants of barr2 ( Figure 7B), indicating that the interaction of barr2 with 14-3-3 might be constitutive. This hypothesis is consistent with the observations that 14-3-3 proteins interact with recombinant GST-barr2 and that the 14-3-3/barrs interaction was not affected by GPCR activation (data not shown and ref [31]). A phosphorylation-independent interaction of barr2 with 14-3-3 proteins would not be unique, since it has already been reported for other partners of 14-3-3 [29].
Interaction of barr2 with 14-3-3f was extended to the other isoforms and we found that flag-tagged barr2, could coimmunoprecipitate with almost all 14-3-3 proteins (data not shown), including 14-3-3g ( Figure 7C), the isoform which has been implicated in ciliogenesis [5]. Finally, the possible colocalization of 14-3-3 proteins with barr2 was analyzed at the centrosome and PCs. In contrast to what was observed in kidney cells [5], 14-3-3 proteins were not detected on the axoneme of PCs in RPE1 or MEFs. In these cells, 14-3-3 proteins were only found at the centrosome or basal body where they colocalized with c-tubulin ( Figures S8 and data not shown) and with barr2 ( Figure 7D and data not shown).
Endogenous Kif3A was also precipitated by GST-barr2 ( Figure 8A), an interaction confirmed by colocalization studies. Indeed, as reported in vivo [32], Kif3A was found in the cytoplasm and at the tip of the axoneme where it was colocalized with barr2 ( Figure 8B). Finally, because Kif3A was reported to co-immunoprecipitate with 14-3-3g [5,30], we investigated whether barr2 could be present in the same molecular complex. Supporting this hypothesis myc-14-3-3g co-immunoprecipitated with both Flagtagged barr2 and endogenous Kif3A ( Figure 8C).

Discussion
Our data show that barr2 shares many of the hallmarks of proteins found at the primary cilium or involved in ciliogenesis: it is targeted to the centrosome in cycling cells and to the basal body and axoneme of PC in quiescent cells; its depletion results in accelerated and uncontrolled cell growth resulting in impaired ciliogenesis.
Similar to many PC proteins, such as, Polaris/IFT88, IFT20 or IFT57 [26,33], barr2 was found at the centrosome in cycling cells and more precisely at the proximal part of the centrioles (Figures 1  and 2). Although we could not find evidence for a role of barr2 in the basic functions of the centrosome, barr2 shares with other centrosome-associated proteins a role in cell cycle regulation. The centrosome participates in several different cell cycle regulatory events, such as G1/S transition, cytokinesis, and monitoring of DNA damage, functions which involve the recruitment of specific sets of proteins [34]. Recent studies showed that depletion of structural centrosomal proteins, such as PCM-1 or pericentrin results in a p53-dependent G1/S arrest [35,36], suggesting that the centrosome itself is involved in cell cycle control. Consistent with these observations, we found that barrs-deficient cells do not respond properly to serum starvation, as shown by their persistent growth in low serum and their failure to enter in G 0 phase, whereas they proliferate faster in high serum conditions ( Figure 6). The strong additive effect of the simultaneous depletion of both barrs likely reflects the fact that each isoform may have specific points of impact.
A role of barr1 was reported in G1/S transition downstream of IGF receptor [37], and via a receptor-independent enhancement of p27 transcription, which, in turn, inhibits G1/S transition [38]. Consistent with our observations on 1KO MEFs (Figures 5 and 6), depletion of barr1 in the latter study was shown to increase cell proliferation. barr2 was also reported to control cell growth in response to nerve growth factor in PC12 cells [39]. However, the mechanism by which barr2 controls cell cycle appears to be different. barr2 interacts with mdm2, the E3 ubiquitin ligase controlling the stability of p53, a transcriptional factor which plays a major role in cell cycle regulation [40]. barr2 was specifically reported to actively exclude mdm2 from the nucleus [41], to stabilize p53 by this mechanism, leading to either induction of apoptosis [41] or cell cycle arrest at G2/M transition [41,42]. Thus, the lack of barr2 would favour destabilization of p53 and then promote cell cycle progression and increased proliferation, as observed here in cells depleted for barr2 ( Figures 5 and 6). Altogether, these data support the idea that barrs have distinct but converging roles in cell cycle regulation and control of cell proliferation.
Our results strikingly paralleled those reported on Polaris, a protein found at the centrosome in cycling cells, which is involved in intra-ciliary transport and required for ciliogenesis. Overexpression of Polaris prevented G1/S transition and induced apoptosis, whereas its depletion promoted cell-cycle progression and increased cell growth [26]. A role in cell cycle control is shared by other IFT proteins such as IFT27 [43] but not all [33]. Altogether, these notions highlight the functional connections between centrosome-associated proteins, IFT proteins and p53 and highlight the centrosome as a meeting point for both proliferative and anti-proliferative controllers [34].
Another similarity between barr2 and Polaris is that the increased proliferation observed in barrs-depleted cells is correlated with reduced ciliogenesis in response to serum starvation, as observed in barr-deficient MEF and siRNA-treated RPE1 cells ( Figures 5 and 6 and ref [26]). In a recent study, which also reported the localization of barrs at PC, barrs-deficient MEF cells did not show ciliogenesis defects [44]. The discrepancy with our findings is likely due to differences in cell culture conditions. Indeed, the authors of the previous study found that only 20% wild-type MEF cells formed PC, consistent with what was previously described for MEF grown in high serum conditions [45]. In the present study, we analyzed ciliogenesis in response to serum starvation, a widely used experimental condition to induce PC assembly. In these conditions we repeatedly observed that ,70% of wild-type MEF cells were ciliated ( Figure 5), as described elsewhere [46]. Altogether, these data suggest that the effect of barrs expression on ciliogenesis is only observed upon serum starvation.
In addition, from our data, it appears that the effect on ciliogenesis is likely linked to uncontrolled proliferation rather than direct effect on the ciliogenesis itself. We observed a striking correlation between the inability to enter in G 0 in response to serum starvation, increased proliferation and reduced ciliogenesis. Therefore, if barr-deficient cells are unable to enter in G 0 even in response to serum starvation they would not be able to build a cilium. In addition, we observed an increased proportion of cycling cells among ciliated single barr1 and barr2 KO cells ( Figure S9). Interestingly, an aberrant outgrowth of PC in cycling cells was recently described upon depletion of Cep97 and CP110 proteins [47]. These data suggest that depletion of barrs may also result in a disconnection between the presence of a PC and cell cycle arrest.
Although the barr-dependent control of PC formation reflects the role of these proteins on cell proliferation, a direct role in ciliogenesis or in the transport of PC proteins cannot be excluded. The constitutive interaction of barr2 with 14-3-3 and Kif3A (Figures 7 and 8), which are both involved in ciliogenesis, can support this hypothesis. In addition, barr2 interaction with Kif3A and 14-3-3 proteins might control its transport within PCs. Indeed, in photoreceptor cells, visual arrestin (varr) regulates the signalling activity of rhodopsin, the light-sensing GPCR. In response to light varr is transported from the inner segment (cell body) to the outer segment (rhodopsin containing compartment), through the connecting cilium, which is a modified PC [48]. In the absence of Kif3A, varr is unable to reach the outer segment in response to light [49], suggesting that Kif3A is responsible for its transport through the connecting cilium. Finally, in the absence of light, varr is also localized in the connecting cilium and at the basal body [50][51][52][53], paralleling our observations showing a constitutive targeting of barr2 to PC. The possible role of Kif3A in the targeting of barr2 to PC could not be tested in our cellular models, since depletion of Kif3A resulted in major ciliogenesis defects ( Figure S10) as previously reported [25,32,54].
While this article was in preparation, a recent study reported that barrs mediate the interaction of smo with Kif3A leading to the targeting of active smo to PC [55]. Whether this observation could account for a more general function of barrs in the targeting of GPCRs to PCs is a key question to be addressed. Our preliminary results indicate that the Somatostatin type 3 receptor (SST3R), a GPCR described in neuronal PC [8], was efficiently targeted to the axoneme in barr-deficient cells ( Figure S11). These data suggest that barrs are not implicated in the targeting of all GPCRs to PC. Interestingly, contrasting with what was reported for smo, activation of SST3R is not required for PC targeting, which rather appears constitutive ( Figure S11 and [56]) and dependent on Bardet Bield Syndrome proteins [56].
Another open issue is the exact function of barr2 within the cilium in the context of GPCRs physiology. Our data indicate that barr2 is constitutively localized to PC: it was found within the axoneme in serum-starved cells and this localization was not modified by the expression of constitutively active form of smo ( Figure 4 and data not shown). One of the key functions of barr2 at the plasma membrane is to mediate internalization of agonistactivated GPCRs through clathrin-coated pits. However, despite the presence of clathrin-coated pits at the base of primary cilia (unpublished observations), there is no evidence in the literature for internalization of proteins found in the membrane of PC. Finally, barr2 might participate in the desensitization of activated GPCRs localized at PC, as shown for odorant receptors in olfactory neurons [17,57,58] and reminiscent of the function of varr in the outer segment of visual cells. Interestingly, activated rhodopsin of the outer segment is not internalized in response to light [48], suggesting that in PCs, receptors might be sequestered from classical downregulation pathways involving clathrin-mediated endocytosis.
Alexa Fluor conjugated secondary antibodies were from Molecular Probes (Invitrogen). Cy3-labeled donkey anti-goat antibody and horseradish peroxidase-conjugated donkey antirabbit or anti-mouse IgG were from Jackson ImmunoResearch.

Microtubules and induction of primary cilia
HeLa cells were synchronised by a treatment with nocodazole (1 mM) overnight and further release by transfer into basic culture media. Microtubule re-growth experiments were performed on MEFs and transfected HeLa cells. Briefly, cells grown on coverslips were treated with nocodazole (10 mM) for 45 minutes at 4uC or with Taxol (10 mM) for 45 minutes at 37uC to depolymerize or stabilize microtubules respectively. Treated cells were then either immediately fixed after a rapid wash in cold PBS, or washed twice in warmed PBS, then incubated for 5 or 10 minutes in pre-warmed (37uC) serum free DMEM, and finally fixed. Nocodazole and Taxol were from Sigma.
To induce ciliogenesis, MEFs or RPE1 cells were grown to confluence on coverslips treated (for MEFs) or not with polylysine (Sigma) in the presence of serum and then grown in low serum containing media (0,5% FBS) for 24 or 48 hours.

Transfections
Transfections were done following the recommended procedure of the Genejuice (Novagen) or of the FuGENE HD (Roche) transfection reagents. Basic transfection conditions were used for HeLa and HeLa-Centrin-GFP. MEFs or RPE1 cells were grown on coverslips up to 70%, then transfected and immediately grown in low serum conditions.
For siRNA experiments, RPE1 cells were treated with previously described control siRNA (Luciferase, Luc) or which target barr2 (si-barr2#1) or both barr1 and barr2 (si-barr2#2) oligos [63,64] using a protocol described elsewhere [65]. For the targeting of Kif3A, a smart pool from Dharmacon (ON-target plus SMART pool L-004964-00-0005) was transfected following the same protocol. Briefly, siRNA duplexes were transfected using Oligofectamine (Invitrogen) according to the manufacturer's instructions. Sub-confluent RPE1 cells (70%) were transfected the first day with 200 pmol of siRNA, then splitted the day after and transfected again with 200pmol of siRNA. Transfected cells were grown in low serum (0.5%) the third day and then processed for immunofluorescence or biochemistry on the fourth day.

FRAP analysis
Dynamic of barr2 at the centrosome was analyzed by FRAP (fluorescence recovery after photobleaching). Hela cells expressing barr2-GFP were treated or not with nocodazole (1h, 37uC, 10 mM) and analyzed using a laser scanning confocal microscope (TCS SP2 AOBS, Leica) after excitation with a 488-nm laser line from an argon laser as previously described [66]. Briefly, a region containing the centrosome was exposed to two consecutive pulses of 1 second with 100% of the laser intensity, and recovery of fluorescence was analyzed for 90 seconds taking an image every 1.3 seconds. Images were then analyzed using Metamorph to quantify normalized fluorescence at the centrosome (see Methods). The final images were generated using NIH image (http://rsb. info.nih.gov/nih-image/) or scion image (http://www.scioncorp. com) and Photoshop (Adobe Systems Inc.).

Immunofluorescence and immunohistochemistry
Cells grown on coverslips were washed twice in PBS and fixed in methanol (methanol/acetone: 1/1) at 220uC for 5 minutes or in 4% paraformaldehyde (PFA) for 30 minutes at 4uC followed by a 10 minutes incubation in PBS-NH 4 Cl (50 mM). Cells were incubated with primary antibodies in permeabilization buffer (PBS with 1 mg/mL bovine serum albumin (PBS-BSA) and 0.1% triton-X-100 (Sigma)) for 45 minutes at room temperature. After two washes with PBS-BSA, cells were incubated for 30 minutes at room temperature in PBS-BSA containing secondary antibodies. After one wash with PBS-BSA and two washes in PBS, cells were laid down on microscope slides in a PBS-glycerol mix (50/50) using the SlowFade Light Antifade Kit with DAPI from Molecular Probes (Invitrogen).
Kidneys from 4 weeks old-mice were harvested and embedded in OCT and snap-frozen in isopentane/liquid nitrogen for cryostat sections. Immunofluorescence labelling was performed on 6-mmthick sections fixed in acetone for 10 minutes, and incubated over night at 4uC with anti-barr2 or anti-AT antibodies diluted in incubation buffer (PBS-BSA; 0.1% triton containing 10% donkey serum). A mounting media containing DAPI (VECTASHIELD, Vector Laboratories, Burlingame, CA) was used to label the nuclei.

Image analysis using Metamorph
To calculate normalized fluorescence of GFP or Cherry fusions at the centrosome, transfected HeLa cells were fixed and stained for pericentrin and the pericentrin staining was then used to define a region corresponding to the centrosome. Briefly, the option ''Auto Threshold for light objects'' allowed us to transform stainings in objects which were then circled by selecting the option ''Create regions around objects'' in Metamorph. The resulting regions were then transferred to the GFP or Cherry corresponding images, and the fluorescence intensity corresponding to GFP/ Cherry in the centrosome (CE) was measured. To normalize these values to the local background, a region of the same size was selected outside the cell to measure the noise of the camera (CN) and another one in the cytoplasm (CY) of the same cell allowing us to normalize centrosome-associated signal to the expression level of the GFP/Cherry fusions in each cell. Normalized fluorescence at the centrosome was then calculated as follows: NF = (CE2CN)/ (CY2CN). To measure the expression level of Ki-67 in nuclei of MEFs, DAPI staining was used to define a region corresponding to the nucleus of each cell as indicated above for centrosomes. Ki-67associated fluorescence was then measured within these regions. To discriminate between Ki-67 negative and positive cells, the average fluorescence intensity of Ki-67 was measured in ciliated cells and nuclei were considered as positive if their Ki-67associated fluorescence was above this average value.

Deconvolution
Epifluorescence images were obtained with an epifluorescence microscope (Zeiss) using a 1006 objective (plan-apo) coupled to a ''piezzo'' enabling acquisition of images every 200 nm in the Z plane. Deconvolution of z-stacks was achieved with metamorph and 3D reconstruction of deconvoluated images with the Imaris software (Bitplane, Scientific solutions, Zurich, Switzerland and Minneapolis, USA). Movies or single images can be extracted from Imaris and then used to obtain the final views used in the figure.

Proliferation Test
The proliferation test was performed using UptiBlue reagent (Interchim, Montluçon, France) according to the manufacturer's instructions. Briefly, MEFs cells were seeded at 1000 cells/well (100 mL) in 96-well microtiterplates. Each assay was performed in triplicate. After 0, 24, 48 or 72 hours, 10 mL of UptiBlue working solution was added to each well and fluorescence was read at 590 nm on Typhoonß 9400 scanner (GE Healthcare, Piscataway, NJ, USA; with settings: excitation laser at 532 nm, filter 580BP30, PMT 350 V). For each cell line, proliferation rate was determined as a ratio of the fluorescence intensity emitted at l = 590 nm for time t less associated background above the fluorescence intensity emitted for time t = 0 less associated background.

Immunoprecipitation and immunoblotting
For Western Blot experiments, cells were lysed by incubation in lysis buffer (0.02M Tris HCl pH 7.5, 1% NP40, 0,1 M NH 4 SO 4 , 10% Glycerol, 10 mM protease inhibitor cocktail (Sigma)) for 30 minutes at 4uC. After centrifugation, cleared lysates were separated by polyacrylamide gel electrophoresis (SDS-PAGE) and transferred onto polyvinylidene fluoride transfer membranes (PVDF, GE Healthcare) using the NuPage electrophoresis system (Invitrogen). Immunoblotting was performed using the indicated primary antibodies and revealed using the ECL + Detection Kit (GE Healthcare).
GST-barr2 fusion protein and GST were expressed in BL21(DE3)pLysS (Invitrogen) and purified on a GSTrap FF column (GE Healthcare) according to the manufacturer's instructions. GST fusions were eluted with 10 mM glutathione were desalted on a HiTrap desalting column (GE Healthcare) in PBS and analyzed by 10% SDS-PAGE and Coomassie blue staining. For in vitro binding assays, 25 mg of GST fusion proteins were immobilized on 20 ml glutathione-Sepharose beads for 1h at 4uC in PBS. Beads were washed twice in 1 ml PBS and twice in binding buffer (50mM Tris HCl pH7.5, 150 mM NaCl, 1% NP40, 0.5% sodium deoxycholate supplemented with protease inhibitors (Sigma)). Cell lysates (500 mg of total proteins) were then incubated for 12 h at 4uC with GST loaded beads, then washed twice with buffer 1, twice with buffer 2, and twice with buffer 3. Complexes were separated on 10% SDS-PAGE, transferred proteins were revealed by Ponceau Red staining of the membrane and precipitated proteins were analyzed by Western blot. Figure S1 Targeting of GFP and Cherry barr2 fusion at the centrosome in RPE1 cells. RPE1 (retinal pigment epithelial) cells were transiently transfected with plasmids encoding for barr2-GFP fusion or GFP alone (A), or barr2-Cherry fusion or Cherry alone (B), then fixed and stained for the centrosomal marker pericentrin. Insets show higher magnifications of representative areas. Scale bar represents 5mm. (C) Pericentrin staining was used to determine the centrosome-associated fluorescence intensity for GFP or Cherry (see Methods) which was then normalized to the cytoplasmic signal in the same cells. Values are the means (+/2 SD) of at least 15 cells from three independent experiments ( **: p,0.001). Found at: doi:10.1371/journal.pone.0003728.s001 (8.56 MB TIF) Figure S2 Characterization of anti-barr2 antibodies. Description of the immunogenic peptides used to generate anti-barr2 polyclonal antibodies: The rARR rabbit polyclonal antibody is sold as an antibody against both barr2 and barr1 but a single amino-acid difference in the immunogenic peptide makes it more specific for barr2. The gBARR2 goat polyclonal was raised against a peptide specific of human barr2 and not conserved in barr1. However, a single amino acid difference between human and rodent barr2 is likely to explain its poor reactivity against murine endogenous barr2 observed in both western blot and immunofluorescence (data not shown). HeLa cells were transiently transfected with plasmids encoding for barr2-GFP, then fixed and stained for the anti-barr2 antibodies, including the rabbit polyclonal rARR anti-arrestin (A) and the goat polyclonal gBARR2 anti barrestin2 (B). In coloured images, barr2-GFP staining is in green, endogenous barr2 in red and nuclei stained with DAPI are in blue. Insets show higher magnifications of representative areas corresponding to the centrosome containing region of cells expressing (1) or not (2) GFP-barr2 in the same field. GFP-barr2 expressing cells showed increase staining with the anti-barr2 antibodies showing that they did work for immunofluorescence. In non-transfected cells, the antibodies showed a diffuse staining in the cytoplasm and illuminated two bright spots, suggesting that both antibodies are able to detect both overexpressed and endogenous barr2. Scale bars represent 5mm. Found at: doi:10.1371/journal.pone.0003728.s002 (9.82 MB TIF) Figure S3 The rARR antibody is specific for barr2 and stains the centrosome. It has to be stressed here that, independently of the commercial source, we observed a variability between batches of commercial anti-barr2 antibodies; while almost all batches did detect overexpressed barr2, some were unable to detect endogenous barr2 in neither western-blot or immunofluorescence experiments. The efficiency of each batch was then tested by western-blot using WT and barrs-KO MEFs as described below. (A and B) barr2 expression was assessed in mouse embryonic fibroblasts (MEFs) derived from wild type (WT), barr2 deficient (2KO), both barr1 and barr2 deficient (1/2KO) mice, RPE1 (retinal pigment epithelial cells) or HeLa cells by western blotting (WB) using the rabbit polyclonal antibody against barr2 (rARR, (A)) or a monoclonal antibody against barr1 (mb1, (B)). An antibody against a-adaptin subunit of the clathrin adaptor complex AP2 was used as a loading control. (C) WT or 1/2KO MEFs were fixed and stained for the centrosomal marker c-tubulin (c-tub) and endogenous barr2 (rARR). Insets show higher magnifications of representative areas. Scale bar represents 5mm. (D) Centrosome-associated fluorescence intensity corresponding to rARR staining in 2KO and 1/2KO MEFs was normalized to that found for WT MEFs. Values are the means (+/2 SD) of at least 20 cells from three independent experiments (**: p,0.001). The data show that the signals observed with the rARR antibody in both western-blot and immunofluorescence experiments do depend on the expression of barr2. Found at: doi:10.1371/journal.pone.0003728.s003 (9.11 MB TIF) Figure S4 Colocalization of endogenous barr2 with centrosomal markers. HeLa cells were fixed and stained for both the centrosome, using either mouse monoclonal antibody against ctubulin (A and B, green) or rabbit polyclonal antibody against pericentrin (C, green), and barr2, using either rARR (A, B, red) or gBARR2 (C, red) polyclonal antibodies. Insets show higher magnifications of representative areas. In coloured images, barr2 staining is in red, centrosome markers in green and nuclei stained with DAPI are in blue. Scale-bars represent 5mm. Found at: doi:10.1371/journal.pone.0003728.s004 (9.23 MB TIF) Figure S5 Targeting of barr2 to the centrosome does not depend on microtubules. To confirm that microtubles were effectively affected in live cells treated with nocodazole in Figure 3, control or nocodazole-treated cells were fixed and stained using antibodies against a-tubulin (a-tub) and Giantin, a Golgi marker. As expected, treatment of the cells with nocodazole resulted in disruption of microtubules and dispersion of the Golgi stacks in cell periphery. Found at: doi:10.1371/journal.pone.0003728.s005 (7.08 MB TIF) Figure S6 barr2 is neither involved in nucleation nor in anchoring of microtubules to the centrosome. (A) WT or 1/ 2KO MEFs untreated or treated with nocodazole to depolymerize microtubules were washed, then directly fixed or incubated in DMEM (37uC) for 5 or 10 minutes. Cells were stained for microtubules (a-tubulin, red). Nuclei appear in blue (DAPI). Insets show higher magnifications of microtubule-forming asters around centrosomes. (B) HeLa cells expressing barr2-GFP tagged fusion were treated with nocodazole then washed in PBS and incubated in pre-warmed DMEM for 5 or 10 minutes. Cells were then fixed and stained for centrosomes (pericentrin) and microtubules (atubulin). Insets show higher magnifications of microtubule-forming asters around the centrosome in cells expressing (1) or not barr2-GFP (2). Scale bars represent 5mm. In coloured images, 14-3-3 staining is in red, centrosome and cilia markers in green and nuclei stained with DAPI are in blue. (C) RPE1 cells grown in high serum conditions were transiently transfected with plasmids encoding for a GFP-14-3-3g fusion, fixed and stained for pericentrin (Peric.). In coloured image, 14-3-3 staining is in green, centrosomal markers in red and nuclei stained with DAPI are in blue. (D) RPE1 cells transiently transfected with plasmids encoding for a GFP-14-3-3g fusion were grown for 24 hours in low serum, then fixed and stained for pericentrin (Peric.) and acetylated tubulin (AT). In coloured image, 14-3-3 staining is in green, pericentrin in blue and AT in red. Insets show higher magnifications of representative areas containing the centrosome or the PC. Scale bars represent 5mm.