Direct role of Bardet–Biedl syndrome proteins in transcriptional regulation

Primary cilia are antennae-like extensions that emanate from the cellular membrane and are typically present in most phyla and, in vertebrates, in most cell types. Recent data have linked primary cilia to diverse sensory processes, including chemo- and mechano-sensation as well as the transduction and/or interpretation of different paracrine signaling cascades (Cardenas-Rodriguez and Badano, 2009; Christensen et al., 2008; Gerdes et al., 2009; Goetz and Anderson, 2010; Wallingford and Mitchell, 2011). Consistent with their broad incidence and diversity of function, ciliary defects can result in a range of clinical manifestations that are shared, to a variable extent, among clinically distinct human genetic disorders, known collectively as ciliopathies (Badano et al., 2006b; Fliegauf et al., 2007). One such ciliopathy is Bardet–Biedl syndrome (BBS; OMIM 209900), a pleiotropic disorder characterized by retinal degeneration, obesity, learning difficulties, polydactyly and gonadal and renal malformations. BBS is a genetically heterogeneous disorder for which 16 genes have been identified to date: BBS1–BBS12, MKS1, CEP290, FRITZ/C2ORF86 and SDCCAG8 [Kim et al., 2010; Leitch et al., 2008; Otto et al., 2010; Stoetzel et al., 2007 (and references within)].

All BBS proteins tested to date localize to cilia, basal bodies and centrosomes (Ansley et al., 2003; Badano et al., 2006a; Fan et al., 2004; Kim et al., 2004; Kim et al., 2005; Li et al., 2004; Marion et al., 2009; May-Simera et al., 2009). In the context of ciliary biology, the BBS proteins appear to have both structural and functional roles. Studies in Caenorhabditis elegans have demonstrated that bbs7 and bbs8 are necessary for intraflagellar transport (IFT), a mechanism that enables and regulates the trafficking of proteins along the ciliary axoneme (Blacque et al., 2004). Moreover, in cultured cells, several of the BBS proteins can form a complex, the BBSome, that plays a role during ciliogenesis (Jin et al., 2010; Nachury et al., 2007). More recently, it has been shown that the BBS complex is able to recognize sorting signals in a number of ciliary proteins and plays a role transporting this cargo into the ciliary compartment (Jin et al., 2010). In addition to their participation in the formation and maintenance of cilia, several BBS proteins have also been shown to modulate paracrine signals. In zebrafish embryos, loss of BBS proteins leads to Shh-dependent migration phenotypes; similar genetic manipulations in zebrafish, mouse and cultured mammalian cells also cause Wnt signaling defects by altering the balance between the different outcomes of the pathway (Gerdes et al., 2007). Depletion of the BBS proteins leads to defective planar cell polarity (PCP) signaling and the concomitant upregulation of canonical signaling, possibly through the stabilization of β-catenin, the main effector of the pathway (Gerdes et al., 2007; Ross et al., 2005).

To gain further insight into the biological role of this group of proteins, we have initiated a detailed characterization of the sequence and binding partners of the BBS proteins, with primary emphasis on BBS1, BBS2 and BBS7, which together account for 35–40% of the genetic load in the disorder. This is in contrast to the second-most frequently mutated group of the three type II chaperonins (BBS6, BBS10 and BBS12), whose primary sequence exhibits similarity over loosely defined domains of unknown function (Badano et al., 2003). Here we show that not only BBS1, BBS2 and BBS7, but also the plurality of bona fide BBS proteins are predicted to possess nuclear export signals with concomitant detection of BBS proteins in the nucleus of mammalian cells. This subcellular distribution is probably important to the pathomechanism of BBS because mutations found in BBS patients can alter this nuclear localization pattern. Consistent with a nuclear role for these proteins, we also show that BBS7 physically interacts with the polycomb group (PcG) member Ring Finger Protein 2 (RNF2) and controls its protein level, probably by mediating the rate of its degradation by the proteasome. Moreover, we show that other BBS proteins also participate in the process. Depletion of BBS7 leads to increased RNF2 protein levels and results in the transcriptional misregulation of a number of RNF2 target genes, both in cultured cells as well as in vivo in Danio rerio (zebrafish). Finally, our data indicate that the role of the BBS proteins in gene regulation might not be restricted to RNF2 but probably extends to other PcG members. These studies point to a hitherto unknown facet of BBS protein activity and lead to the surprising observation that this group of proteins could have a direct role in transcriptional regulation that might contribute to the pathogenesis of BBS in humans.

We have shown recently that whereas overexpression of wild-type BBS7 in HeLa and NIH3T3 cells results in a general cytoplasmic staining and a general exclusion from the nuclear compartment, expression of two BBS7 mutants (H323R and T211I) reported in BBS patients results in ubiquitous cellular staining, including the nucleus (Badano et al., 2003; Zaghloul et al., 2010). Given that the BBS proteins characterized to date localize to the centrosome, basal body and the cilium, this observation raised the possibility that at least BBS7 might also localize to the nucleus. As a first test of this hypothesis, we performed an in silico analysis of the amino acid sequence of the 16 BBS proteins to identify possible nuclear localization signals (NLS) or nuclear export signals (NES) using prediction software for both NLS (NLStradamus; http://www.moseslab.csb.utoronto.ca/NLStradamus/) and NES (NetNES 1.1; http://www.cbs.dtu.dk/services/NetNES/) motifs (la Cour et al., 2004; Nguyen Ba et al., 2009). Interestingly, although 15 out of the 16 BBS proteins, CEP290 being the exception, seem to lack an NLS (data not shown), the analysis for NES predicts the presence of a NES motif in each of the BBS proteins tested, with the exception of BBS8, suggesting that these proteins might localize to the nucleus at least transiently (supplementary material Figs S1 and S2).

Given our previous results with the BBS7 mutants, we determined the localization pattern of endogenous BBS7 in mammalian cells by performing immunocytochemistry in NIH3T3 cells. Co-staining of BBS7 with γ- and acetylated tubulin showed that BBS7 not only localizes to centrosomes and basal bodies in a pattern that was indistinguishable from that of other BBS proteins reported to date, but it also colocalized with the ciliary axoneme (Fig. 1). Interestingly, not all ciliated cells showed positive staining for BBS7 in the cilium (Fig. 1, compare B with C), suggesting that the exact localization of BBS7 in the context of this organelle might be regulated.

We reasoned that, if BBS7 is also found in the nuclear compartment, it might be localized there transiently or at low levels. Consistent with that notion, when we used higher concentrations of our primary anti-BBS7 antibody in immunocytochemistry assays we observed nuclear staining (Fig. 2A, upper panels). BBS7 is predicted to have a NES motif centered on amino acid residues leucine 625 and isoleucine 634 (supplementary material Fig. S2, upper panel) and therefore the fact that it appears to be present in the nucleus at low levels might be the result of its active export from this cellular compartment. We reasoned that if this is the case, inhibiting nuclear export might result in increased BBS7 nuclear signal. We therefore treated cells with leptomycin B (LMB) and N-ethylmaleimide (NEM), chemicals that block nuclear export mediated by CRM1 through NES motifs, and assessed the localization of BBS7 (Fornerod et al., 1997; Holaska and Paschal, 1998; Kudo et al., 1999). Treatment with LMB resulted in increased BBS7 nuclear signal (Fig. 2A, lower panels). To better visualize the effect of LMB, we processed the microscopy images using pseudocolor to highlight the relative density of the BBS7 signal. This analysis showed that cells treated with LMB presented more BBS7 signal (increased green coloring) in the nucleus that correlated with a reduction in the cytoplasmic signal (supplementary material Fig. S3). Similarly, treatment with NEM resulted in nuclear BBS7-positive staining, even when using normal concentrations of the primary antibody (Fig. 2B). To confirm these observations, we tested the subcellular localization of BBS7 by cell fractionation, in which we isolated the cytoplasmic and nuclear fractions from NIH3T3 cells and assessed the presence of BBS7 by western blot. Consistent with our immunofluorescence data, although most BBS7 appeared in the cytoplasmic fraction, the protein was also observed in the nucleus (Fig. 2C).

We next tested whether the predicted NES in BBS7 is functional. We performed site-directed mutagenesis, exchanging an isoleucine residue with alanine at position 634 (I634A) in our Myc-tagged mammalian BBS7 expression construct (Myc–BBS7 NES), a change that was predicted to disrupt the NES completely (supplementary material Fig. S2, lower panel). We transiently transfected NIH3T3 cells with both the Myc–BBS7 and the Myc–BBS7 NES expression constructs and performed immunocytochemistry using the monoclonal anti-Myc antibody. As reported previously, overexpression of wild-type BBS7 does not recapitulate the localization pattern of the endogenous protein because high levels of signal are observed along the entire cytoplasm (Zaghloul et al., 2010). However, the nuclear signal in Myc–BBS7-expressing cells was low in the majority of cases (Fig. 2D). By contrast, Myc–BBS7 NES was readily observed both in the cytoplasm and the nuclear compartment (Fig. 2D). To quantify this effect, we performed transfections in triplicate of both wild-type and mutant constructs and assessed this nuclear localization pattern by counting at least 70 cells per experiment. Our data indicates that at least 50% of cells transfected with the NES mutant construct showed positive nuclear staining compared with 8% of wild-type Myc–BBS7-transfected cells (Fig. 2C; χ²=95.9, P<0.001). We quantified the effect reported previously for the BBS7 missense changes H323R and T211I (Zaghloul et al., 2010) and found an intermediate phenotype, with 20–25% of cells presenting nuclear staining (Fig. 2D). Importantly, the nuclear accumulation of the NES mutant compared with wild-type BBS7 could also be observed by using cell fractionation assays (Fig. 2E).

Collectively, our data indicate that BBS7 localizes to centrosomes, basal bodies and the ciliary axoneme, at least in some cases, and can also be found in the nucleus from where it appears to be actively exported. Moreover, a NES motif located between residues 625 and 634 is likely to be functional and might mediate this process, at least in part, given that neither the H323R nor the T211I mutations are predicted to disrupt the motif but do affect the rate of BBS7 nuclear export.

In parallel to characterizing the cellular distribution of BBS proteins, we have been working towards dissecting the protein complex in which these proteins participate. Focusing on BBS7, we cloned its full-length open reading frame (ORF) into the pSOS construct (pSOS-BBS7) and used this fusion protein as bait in a cytoplasmic yeast two-hybrid screen using a fetal brain cDNA library as prey (Cytotrap assay). From the initial 337 putative positives obtained in the screen, we selected 13 clones that fulfilled our validation criteria: the ability to grow at 37°C only in galactose (Fig. 3). Sequencing of the pMyr (library) plasmids revealed that 11 of these clones corresponded to RNF2 (also known as RING1B, BAP1, DING, BAP-1, HIPI3 and RING2), a PcG member with a role in chromatin remodeling and gene regulation. The PcG proteins were described originally in Drosophila as transcriptional repressors of homeotic genes during development (Lewis, 1978). RNF2 is a nuclear protein, member of the polycomb repressive complex PRC1, that is thought to act as an E3 ubiquitin ligase that ubiquitylates lysine 119 of histone H2A (Schwartz and Pirrotta, 2007; Wang et al., 2004).

We first confirmed the BBS7–RNF2 interaction by testing for bait–prey dependency in yeast, showing that only cells containing both pSOS-BBS7 and pMyr-RNF2 were able to grow at the restrictive temperature of 37°C (Fig. 3A). To test whether the BBS7–RNF2 interaction can occur in mammalian cells, we performed co-immunoprecipitation assays (CoIP). We coexpressed epitope-tagged BBS7 (Myc–BBS7) and RNF2 (HA–RNF2) in cells and performed CoIPs. We were able to detect a single band of the expected size for RNF2 only in the samples where HA–RNF2 was coexpressed with Myc–BBS7 but not with the empty vector (Myc–EV; Fig. 3B, left panel). In addition, we immunoprecipitated BBS7 using our polyclonal anti-BBS7 antibody and tested for the presence of RNF2 with a monoclonal antibody. We detected the RNF2 band only in the BBS7 immunoprecipitate but not when we used a purified rabbit IgG to perform the immunoprecipitation (Fig. 3B, right panel). This result supports the notion that BBS7 might exert at least part of its biological role in the nucleus. As mentioned earlier, up to 16 proteins have been causally linked to BBS. We therefore performed CoIPs using other Myc-tagged BBS proteins (BBS1, BBS2, BBS4, BBS5, BBS6, BBS8 and BBS10) and both the empty vector and a non-related Myc-tagged protein as controls (Myc–EV and Myc–Control). HA–RNF2 was co-immunoprecipitated when coexpressed with all the BBS proteins tested but not with the controls (Fig. 3C). Thus, our data from both yeast and mammalian cells indicate that the nuclear protein RNF2 can interact with BBS7 and a number of other BBS proteins.

To understand the biological relevance of the BBS–RNF2 interaction, we analyzed the effect of depleting the protein level of several BBS proteins on basic parameters addressing the functionality of RNF2, such as its localization and expression both at the RNA and protein level. We first assessed the efficiency of our short hairpin (shRNA) constructs (pSUPER) by comparing BBS1, BBS2, BBS4 and BBS7 mRNA levels between cells transfected with pSUPER EV (control empty vector) and the corresponding targeting constructs (Fig. 4A; Fig. 5A; supplementary material Fig. S4). Upon evaluating endogenous RNF2 localization by immunocytochemistry in cells transfected with pSUPER BBS7, we did not observe any changes in the pattern of subcellular distribution of RNF2, which retained its nuclear localization (data not shown). Thus, BBS7 does not appear to be required for RNF2 to enter the nucleus.

Next, we looked at whether the BBS proteins might affect the protein levels of RNF2 by comparing the total amount of RNF2 in cells depleted not only of BBS7 but also of the other two members of this BBS sub-group of proteins, BBS1 and BBS2, as well as BBS4. We transfected HeLa cells (in triplicate) with either pSUPER EV or pSUPER BBS1, BBS2, BBS4 and BBS7, and analyzed RNF2 by western blot. We observed a marked increase in the protein levels of RNF2 upon depletion of each of the four BBS proteins (Fig. 4A,B and data not shown). Moreover, although the effects of depleting BBS1, BBS4 and BBS7 were comparable (approximately 50% increase in RNF2 levels with respect to control cells), inhibition of BBS2 resulted in significantly higher amounts of RNF2 (>200% increase; see Fig. 4B). Although this result might reflect differences in the efficiency of the corresponding shRNAs (supplementary material Fig. S4), it is interesting to note that BBS2 is readily observable in the nuclear compartment by immunocytochemistry (supplementary material Fig. S5).

We then assessed whether the changes in RNF2 levels were due to altered gene expression by performing both semi-quantitative reverse transcription PCR (RT-PCR) and real time PCR. We observed that in cells depleted for BBS7, the amounts of RNF2 mRNA were decreased (Fig. 5A,B), suggesting that RNF2 might regulate its own transcription through a negative feedback loop. We have shown recently that depletion of BBS4 results in defective proteasome-mediated protein clearance, leading to the accumulation of β-catenin in BBS4 knockdown cells (Gerdes et al., 2007). Therefore, we tested whether depletion of BBS7 resulted in a similar proteasomal phenotype using the HEK293 ZsGreen proteasome sensor cell line in which the green fluorescent protein (GFP) is targeted constitutively for degradation (see Materials and Methods). Cells transfected with our pSUPER BBS7 construct showed an increase in GFP signal compared with pSUPER EV control cells (Fig. 4C). A quantification of total luminescence by flow cytometry showed a threefold increase in GFP signal in pSUPER BBS7 cells (Fig. 4D), indicating that the efficiency of proteasome-mediated protein degradation is compromised in BBS7 knockdown cells. Cumulatively, our results show that BBS1, BBS2, BBS4 and BBS7 have the capacity to downregulate RNF2 at the protein level, probably by affecting its rate of degradation.

Our data have raised the possibility that, in the absence of BBS proteins, RNF2 levels might be abnormally increased, possibly affecting the regulation of downstream targets of this PcG protein. To test this possibility, we selected a panel of 11 known RNF2 target genes (Boyer et al., 2006) and measured their mRNA levels both in control and in BBS7-depleted cells. We transfected HeLa cells with either the pSUPER EV or pSUPER BBS7 constructs, extracted total RNA 72 hours post-transfection and performed both semi-quantitative RT-PCR (Fig. 5A) and real time PCR (Fig. 5B). We observed significant changes in the levels of expression of all the genes tested. Consistent with the known role of RNF2 as a transcriptional repressor, the mRNA levels of BCL11A, CDX2, DLX2, HAND1, HOXB8, LEF1, POU3F2 and SIX1 were significantly lower in cells in which BBS7 was downregulated. By contrast, TBX15, FOXL2 and FOXQ1 were upregulated in BBS7-depleted cells. It is important to note that RNF2 has also been shown to be associated with transcriptionally active genomic regions, although the underlying mechanism is not well understood (Bracken et al., 2006). To determine whether at least part of this effect is mediated by basal body- or cilia-related functions of the BBS proteins, we assessed the expression of a subset of RNF2 target genes in cells where ciliary function was perturbed. Although the effect was milder than the effects of knockdown of BBS7, depletion of the IFT anterograde molecular motor KIF3A phenocopies pSUPER BBS7, whereas depletion of the retrograde IFT particle IFT139 resulted in the opposite effect (supplementary material Fig. S6), suggesting that specific ciliary components might play a role in the regulation of RNF2 target genes.

To assess whether BBS-dependent misregulation of RNF2 target genes might occur in vivo, we turned to a zebrafish model. We used translation-blocking morpholinos (MOs) at doses determined previously to affect bbs7 or bbs4 function (Zaghloul et al., 2010). Embryos injected with each MO presented the characteristic gastrulation defects reported previously, including poor somitic definition, shortened body axes and kinked notochords (Gerdes et al., 2007; Leitch et al., 2008; Ross et al., 2005; Stoetzel et al., 2006; Stoetzel et al., 2007; Zaghloul et al., 2010). We then performed whole-mount in situ hybridization to assess the expression of bcl11a, six1b, pou3f2, hoxb8a and lef-1 at 30 hours post-fertilization (hpf; developmental stage Prim-15) in two independent injections and assessed 25–30 morphant embryos per probe. bcl11a, pou3f2 and lef-1 showed a reduction in expression in morphant embryos (Fig. 6A,C,E). Uninjected controls and embryos injected with a standard control MO expressed bcl11a prominently in the telencephalon and olfactory placode and mildly in the mesencephalon and hindbrain, whereas the bbs7 and bbs4 morphants exhibited significant reduction of expression in all structures, consistent with our cell-based predictions (Fig. 6A; supplementary material Fig. S7). Similarly, both pou3f2 expression (Fig. 6C; supplementary material Fig. S7), normally observed in all structures of the brain and spinal cord, and lef1 expression (Fig. 6E), which is typically expressed in the hypothalamus, diencephalon, tectum, midbrain hind–brain boundary, pectoral fin and branchial arches, were reduced in morphant embryos. Levels of expression for six1b and hoxb8a did not appear to differ significantly between morphants and control embryos (Fig. 6B,D). However, the pattern of six1b expression was altered; whereas control embryos displayed expression in the branchial arches and otic vesicles, additional expression in the diencephalon of bbs7 morphants and in the pectoral fin bud and hindbrain of bbs4 morphants could be observed (Fig. 6B). hoxb8a expression patterns remained unchanged in morphants as compared with controls (Fig. 6D). Importantly, we did not observe gross morphological defects that could be causing these changes in expression patterns (Fig. 6D; supplementary material Fig. S8). To confirm that these changes in gene expression are specific effects of depleting the bbs proteins, we used the standard control MO as an additional control and rescued the phenotype by co-injecting human BBS4 and BBS7 mRNA together with the corresponding MO. Importantly, whereas injecting the standard control MO neither affected the expression of bcl11a nor pou3f2, co-injection of BBS4 and BBS7 mRNAs with their corresponding MOs fully rescued the expression of both genes (supplementary material Fig. S7).

To determine whether the reduction in expression of bcl11a, pou3f2 and lef-1 in the bbs MOs was dependent on Rnf2, we investigated possible genetic interactions by using double injections of bbs7 and rnf2 or of bbs4 and rnf2. The expression of lef1 in the double morphants of bbs7 and rnf2 or bbs4 and rnf2 was partially rescued in all regions compared with bbs7 and bbs4 single morphants. Importantly, the expression of both bcl11a and pouf32 was not only rescued but was also expanded compared with controls. Expression of bcl11a, for example, was increased in the midbrain and expanded to the hindbrain, probably as a consequence of reducing rnf2 to lower levels than in controls (Fig. 6A). In summary, our data indicate that rnf2 target genes are misregulated in vivo in the absence of the bbs proteins, corroborating and assigning physiological relevance to our in vitro data.

Finally, we evaluated whether the role of the BBS proteins in gene regulation is restricted to their interaction with RNF2 or whether it extends beyond this particular PcG protein. We therefore performed a microarray-based analysis to visualize, in a global fashion, the changes in gene expression associated with the loss of BBS7 function. We transfected HeLa cells (in triplicate) with either pSUPER EV or pSUPER BBS7, extracted RNA and hybridized our samples onto a 4×44K Human Genome Oligo Microarray interrogating §41,000 unique human genes and transcripts arrayed in four replicas. This allowed us to compare our three biological samples as well as a technical dye swap control in a single experiment. After stringent quality control steps, we retained 18,872 probes corresponding to 13,420 unique features.

Our initial analysis showed a significant number of affected genes (for a complete list of genes with their corresponding P-values and log2 fold change, see supplementary material Table S1). We ordered putative altered genes by their adjusted P-values and set an arbitrary cut-off at P<1×10⁻⁵ and a log2 fold change of ±2 to define a set of differentially expressed (DE) transcripts for subsequent analyses (supplementary material Fig. S9 and Table S1). Because the decision of a cut-off value is arbitrary, we did not try to estimate the number or identity of genes affected by BBS7 function. Rather, we focused on confirming and possibly expanding on our previous results and determined whether genes known to be targeted by RNF2, or the other PcG proteins (SUZ12, EED and PHC1) for which lists of putative target genes have been reported (Boyer et al., 2006), are enriched in our DE dataset. From 7,275 analyzable genes that are not present in any of the PcG lists (NO group) (Table 1), 503 (6.9%) are present in our DE list of genes. However, 17% of targets of RNF2, 22% of targets of SUZ12, 25% of targets of EED, 20% of targets of PHC1 and 21% of genes with a PcG-dependent histone H3 trimethylation modification (H3K27) are DE (Table 1). Thus, our data show a significant enrichment of targets of RNF2 and all other PcG proteins tested in our dataset of genes affected by BBS7 depletion (Pearsons Chi-squared test, χ²=457, df=6, P<2.2×10⁻¹⁶). Importantly, an equivalent analysis using a PcG-independent group of developmental genes reportedly targeted by Cyclin D1 (Ctrl-neg group) (Bienvenu et al., 2010), showed that they were not significantly different to the NO group (Table 1).

Next, we performed an unbiased analysis of our data to avoid the need of establishing arbitrary cut-offs. We hypothesized that if a group of genes is preferentially altered in the absence of BBS7, these should present lower P-values than those genes that are a priori not linked to BBS7. We first tested whether genes included in the ciliary proteome (Gherman et al., 2006) (www.ciliaproteome.org) are preferentially affected in BBS7 knockdown cells and did not find a significant difference in the mean and distribution of their P-values compared with the genes in the array that are not present in the database (Fig. 7A, NO group). We then performed a similar analysis with the target genes of RNF2, SUZ12, EED and PHC1 and the H3K27 list, and compared their P-value distributions with those of the NO and Ctrl-neg groups. Importantly, our analysis revealed that the groups of genes associated previously with the PcG proteins presented significantly lower P-values than our two controls (Kruskal-Wallis Chi-squared test, χ²=461.1358, df=6, P<2.2×10⁻¹⁶; see Fig. 7B). Collectively, our data show that BBS7 probably participates in the regulation of target genes of RNF2 and other PcG proteins such as SUZ12, EED and PHC1. Finally, we performed an ontology analysis using the list of altered genes with P-values <1×10⁻⁵ in an effort to delineate common biological processes or pathways that appear to be altered in the absence of BBS7. This analysis revealed an over-representation of genes involved in maintaining cellular homeostasis, cell organization, cellular interaction with the extracellular environment and, in general, genes playing a role in developmental processes, a function that is consistent with the commonly associated roles of PcG target genes (Fig. 7C).

Our deeper knowledge of ciliary biology has been instrumental in beginning to understand the phenotypic consequences of ciliary dysfunction. However, and despite these advances, little is known regarding the exact biological role of a number of ciliary proteins, both in the context of ciliary-dependent and ciliary-independent functions. Here, we increase the complexity associated with the BBS proteins by showing that these proteins are probably found in the nuclear compartment, albeit at reduced levels, where they are involved in the regulation of gene expression. Our data indicate that several BBS proteins are able to modulate RNF2 protein levels. Importantly, it has recently been shown that cilia are involved in the regulation of the mTOR pathway, a key regulator of protein synthesis in response to different extracellular stimuli (Boehlke et al., 2010; DiBella et al., 2009). As such, although we cannot discard the possibility that cilia-dependent perturbation of mTOR signaling might affect the protein levels of RNF2, our proteasome data indicate that the effect of the BBS proteins on RNF2 is probably achieved by affecting its turnover. Given that all the BBS proteins tested in this study are able to interact with RNF2, at least in overexpressing conditions, it is possible that this regulatory activity over RNF2 occurs in the context of a complex, either the BBSome or a different BBS complex. Supporting this possibility, whereas depletion of either BBS7, BBS1, BBS2 or BBS4 resulted in a marked increase in RNF2, overexpression of a single BBS protein was not sufficient to reduce the levels of the PcG protein (data not shown).

The physiological relevance of the nuclear localization of the BBS proteins is supported by several observations. First, the vast majority of BBS proteins present NES motifs, and in the case of BBS7 this signal appears to be functional. Second, whereas depleting BBS1, BBS2, BBS4 and BBS7 results in increased RNF2 levels, inhibition of BBS2, which appears to be abundantly present in the nuclear compartment, provoked the most prominent effect on RNF2 abundance. Third, two BBS7 mutations found in patients, H323R and T211I (Badano et al., 2003), result in increased nuclear localization of the protein. Those mutants retain the ability to bind RNF2 (data not shown) and behave as dominant negatives in a zebrafish-based in vivo rescue assay (Zaghloul et al., 2010). Therefore, our data might prove relevant to the understanding of this in vivo effect. However, we note that these mutations are not predicted to affect the NES motif tested in this work (located around residues 625–634). One possibility is that these mutations affect protein–protein interactions required for efficient nuclear export. Proteins that are subjected to NES-mediated nuclear export physically interact not only with the karyopherin receptor CRM1 (also known as exportin 1 or XPO1) but also might need adaptor proteins mediating the interaction and a number of other moieties, such as RanGTP (Fornerod et al., 1997; la Cour et al., 2004; Sorokin et al., 2007). In addition, the activity of a NES signal might be regulated by binding of additional proteins and post-translational modifications (la Cour et al., 2004), aspects that could also be affected by these mutations. Finally, there are examples of proteins, such as actin from S. cerevisiae, that contain more than one functional NES signal (Wada et al., 1998). Interestingly, one of the scores provided by the NetNES server presents a peak centered on residue 218 of BBS7 that is diminished by the T211I mutation (supplementary material Fig. S10).

In this scenario, elucidating what regulates the activity or subcellular localization of proteins such as BBS7 will be crucial to completely understand their function. In addition, and given that BBS7 and RNF2 can interact in the cytoplasm at least in yeast, understanding whether the interaction of BBS7 with RNF2 is restricted to the nuclear compartment or can also occur outside of it is also important. Our data suggest that the localization of BBS7 appears to be variable in terms of its presence in the ciliary axoneme and possibly in terms of its ability to enter the nucleus. The localization of the BBS proteins might be altered in response to specific cellular or ciliation stages or upon the activation of signaling cascades that operate through the cilium. Moreover, our data indicate that the presence of cilia might mediate the role of BBS7 in the regulation of RNF2, given that knockdown of the anterograde motor KIF3A results in a mild misregulation of RNF2 target genes that resembles the effect of depleting BBS7. Consistent with this notion, depletion of other ciliary components such as IFT139, a protein that has been causally linked to the pathogenesis of several ciliopathies (Davis et al., 2011), also participates in the regulation of different RNF2 targets although the exact mechanism remains to be elucidated.

Importantly, the ability of ciliary proteins to enter the nucleus is not restricted to the BBS proteins. Inversin, mutated in nephronophthisis (NPHP2); polycystin 1, altered in autosomal dominant polycystic kidney disease (ADPKD); and NEK1, a centrosomal protein mutated in mouse models of cystic kidney disease, have also been reported in the nucleus (Hilton et al., 2009; Nürnberger et al., 2002; Yoder et al., 2002). Likewise, OFD1, the protein mutated in the ciliopathy orofaciodigital syndrome type I (OMIM 311200), has also been shown to have this capacity and has been linked to gene regulation by its association with a histone acetyltransferase (Giorgio et al., 2007). Interestingly, recent data highlights striking similarities between the process of nuclear import and that of targeting to the ciliary compartment. It has been shown that the molecular motor KIF17 possesses a ciliary localization signal that is similar to classic NLS motifs. Furthermore, Dishinger and colleagues have shown that importin-β2 and RanGTP, key proteins in nuclear translocation, are present in the cilium and are required for KIF17 ciliary targeting (Dishinger et al., 2010). In this context, we speculate that mechanisms similar to nuclear export could also function to actively regulate the composition of the ciliary compartment.

Lastly, our data both provide an extensive list of genes whose expression is potentially affected in the absence of the BBS proteins and also show that not all the genes identified as differentially expressed in our cell-based assays will be aberrantly expressed in vivo. Nevertheless, the fact that the BBS proteins affect the expression of key developmental regulators, such as targets of the PcG, might provide an important clue to the cellular basis of the different clinical manifestations that characterize BBS and other ciliopathies. For example, we show that depletion of BBS7 results in downregulation of SIX1 in our cell-based assays and probably results in ectopic expression in zebrafish bbs morphants at the Prim-15 stage. Murine Six1 is implicated in the formation of specialized sensory structures, placodes, as well as other developmental aspects of the sensory system, findings that might be relevant to the thermosensory defects recently reported in BBS animals and patients (Christophorou et al., 2009; Tan et al., 2007). Likewise, SIX1 has been also implicated in processes such as epithelial–mesenchymal transitions (EMT) (Micalizzi et al., 2009), potentially providing clues to the increased fibrosis observed in the kidney phenotype of BBS and other ciliopathies.

In summary, moieties such as the BBS proteins might provide a link between the cilium and the interior of the cell, relaying different types of signals and, finally, exerting their role through the regulation of gene transcription. Further studies will be needed to understand the nature of the signal(s) that modulate the activity of the BBS proteins, their capacity to enter the nucleus, their association with RNF2 and possibly other PcG proteins, and the extent of their involvement in gene regulation. Understanding the type of genes that are ultimately affected by the function of proteins such as the BBSs will provide significant insight into the pathogenesis and cellular basis of the ciliopathies.

We performed the Cytotrap yeast two-hybrid screen following the manufacturer's instructions (Stratagene). We amplified the complete ORF of BBS7 from HeLa cDNA by PCR using the Pfu high fidelity DNA polymerase (Stratagene) and using primers that were tagged with SalI and NotI restriction sites. We cloned the ORF of BBS7 into the pSOS vector (bait) and used a fetal brain library cloned in the pMyr vector (prey; Stratagene) for the screen.

The complete ORF of BBS1, BBS2, BBS4, BBS5, BBS6, BBS7, BBS8, BBS10, RNF2 and an unrelated control gene (Myc–Control in Fig. 1) were cloned into pCMV-Myc and HA mammalian expression plasmids (Clontech). To generate the I634A mutant, we designed primers introducing the change and performed site-directed mutagenesis using the QuickChange Site-Directed Mutagenesis Kit following the manufacturer's instructions (Stratagene). Both the H323R and T211I mutants were previously generated (Zaghloul et al., 2010).

HeLa, Hek293 and NIH3T3 cells were maintained in Dulbecco's modified Eagle medium (at PAA or Sigma) supplemented with 10% fetal bovine serum (FBS; Invitrogen) at 37°C in 5% CO₂. Transfections were performed using the Calcium Phosphate Transfection Kit (Invitrogen). For nuclear export inhibition, cells were incubated with LMB at a concentration of 50 ng/ml.

For co-immunoprecipitations (CoIP) we grew HEK293 cells in 10-cm dishes and co-transfected them with the appropriate pCMV-Myc construct and the pCMV-HA RNF2 plasmid when needed. We harvested cells 48 hours after transfection (or at desired confluency for endogenous CoIPs) by incubating them for 15 minutes at 4°C in CoIP buffer (150 mM NaCl, 50 mM TRIS-HCl pH 7.5 and 1% Nonidet P-40) supplemented with a protease inhibitor cocktail (Sigma) and 1 mM sodium orthovanadate. Cell lysates were further incubated at 4°C for 15 minutes with vortexing and then centrifuged for 15 minutes at 13,000 g at 4°C. For CoIP, we incubated cell lysates overnight and under rotation at 4°C with either 8 μg of monoclonal anti-Myc antibody or our rabbit polyclonal anti-BBS7 antibody immobilized onto a mix of protein A and protein G Sepharose beads (Invitrogen). We washed the beads using CoIp supplemented buffer, resuspended the samples in loading buffer, boiled them for 5 minutes and analyzed them by western blot using a monoclonal anti-HA (Sigma) or a mouse monoclonal anti-RNF2 antibody (Abnova).

For immunocytochemistry, either HeLa or NIH3T3 cells were grown on glass coverslips. At 24 hours post-transfection when appropriate or at desired confluency, cells were fixed in −20°C methanol for 10 minutes, blocked with 5.5% FBS for 1 hour and incubated with the appropriate primary antibody at room temperature for 2 hours. We then used secondary antibodies coupled to either Tetramethylrhodamine or Alexa Fluor 488 (Invitrogen) for visualization by both fluorescent and confocal microscopy. We used 4,6-diamidino-2-phenylindole (DAPI) to stain DNA.

We used rabbit polyclonal antibodies raised against specific synthetic peptides for BBS7 and BBS2 for immunocytochemistry assays. We used monoclonal antibodies to Myc, HA, γ-tubulin and acetylated tubulin (Sigma), a rabbit polyclonal antibody to GAPDH (Abcam) as western blot control, a mouse monoclonal antibody against RNF2 (H00006045-M01, Abnova) and a donkey polyclonal antibody against BBS7 (Santa Cruz Biotechnology) for western blot.

For the expression of short interfering RNAs (siRNAs) we selected our 19-nucleotide target sequences in the genes of interest using the siDESIGN center (Dharmacon). We cloned the hairpins into the pSUPER vector by designing oligonucleotides containing our 19-nucleotide target sequences according to the manufacturer's instructions (OligoEngine). Positive clones were confirmed by restriction enzyme digestion and direct sequencing. We transfected cell cultures with either pSUPER EV (control empty vector), pSUPER BBS7, BBS1, BBS2, BBS4 or pSUPER RNF2 and harvested cells after 72 hours for both protein and RNA extraction. The IFT139 targeting sequence was cloned into the pCAGMir30 vector (Addgene) and the hairpins for KIF3A and IFT88 were expressed using the pSHAG Magic construct (Open Biosystems).

We obtained cytoplasmic and nuclear fractions using the Qproteome Cell Compartment Kit (Qiagen). For the gene inhibition experiments, we harvested cells 72 hours after transfection with the pSUPER constructs, prepared cell lysates as described for CoIP, quantified protein using the BCA method (Sigma), loaded equal amounts on SDS-PAGE and analyzed the samples by western blot. The quantification of western blot bands was performed using the ImageJ software (NIH).

To test both the efficiency of knockdown of our constructs as well as the effect of depleting BBS7 on different target genes (primers are available from the authors upon request), we harvested cells 72 hours after transfection and extracted total RNA with Trizol (Invitrogen) following the provider's instructions, prepared cDNA using the SuperScript First-Strand Synthesis System for RT-PCR (Invitrogen) and used this cDNA as template in either limited-cycle PCR or real time PCR. For real time PCR we followed SYBR green incorporation using the Quantimix Easy SYG KIT (BioTools) in a Rotor-Gene 6000 (Corbett Research). All real time PCR reactions were performed in biological and PCR duplicates. The threshold cycle (CT) value for each gene in the panel was normalized to GAPDH, calculating the ΔCt for each gene in all samples (replicates of pSUPER EV and pSUPER BBS7). The comparative CT method (ΔΔCt method) was used to determine the relative quantity of the target genes, and the fold change in expression was calculated as 2^-ΔΔCt.

Morpholinos (MOs) against zebrafish bbs4 (1.5 ng), bbs7 (6 ng), rnf2 (0.5 ng) and a standard control (5 ng) were obtained from Gene Tools. The standard control targets a human β-globin intronic mutation and does not have a target in zebrafish. Wild-type zebrafish embryos at a one- to two-cell stage were injected with 1 nl of each solution. Zebrafish were collected at the Prim15 stage. Whole-mount in situ hybridizations were carried out using antisense probes for bcl11a, six1b, pou3f2, hoxb8a and lef-1 made from clones (Openbiosystems) and following a published protocol (Thisse and Thisse, 2008). Further processing included clearing in methyl salicylate, flat mounting when needed, and photographing with a Nikon SMZ-745T Zoom Stereo Photo Microscope at a 5× magnification using a 5.0-megapixel DS-Fi1 color digital camera head. Rescues were performed by co-injecting the MOs with the corresponding human wild-type mRNA that was transcribed in vitro using the SP6 Message Machine Kit (Ambion).

Total RNA from pSUPER EV control and pSUPER BBS7 cells was isolated by Trizol (Invitrogen). The concentration and integrity of RNA was determined using a NanoDrop-1000 Spectrophotometer (NanoDrop Technologies) and a 2100 Bioanalyzer (Agilent) with an RNA 6000 Nano LabChip Kit, using the Eukaryote Total RNA Nano assay according to the manufacturer's instructions. RNAs with an RNA integrity number (RIN) greater than 8 were used. Microarray analysis was performed using a 4×44K Human Genome Oligo Microarray (G4112F, Agilent), in a two-color design. The Low RNA Input Linear Amplification Kit (Agilent) was used to generate fluorescent complementary RNA (cRNA) for the microarray hybridizations. Briefly, we amplified and labeled 1 μg of total RNA using the Cy5-CTP or Cy3-CTP dyes. Equal amounts of labeled samples (pSUPER EV and pSUPER BBS7) were hybridized to the arrays at 65°C for 17 hours in a rotating oven (Agilent). Arrays were washed with the wash buffers 1 and 2 and the stabilization and drying solutions from Agilent. Slides were scanned on an Agilent DNA microarray scanner, and microarray data were extracted with Agilent's Feature Extraction software (v9.5). Three biological and one technical dye swap replicates were performed and a total of four arrays were analyzed.

Data analysis was accomplished in ‘R’ (R Foundation for Statistical Computing, Vienna, Austria; http://www.R-project.org), mainly through packages in the Bioconductor suite (Gentleman et al., 2004). Differential expression was assayed using the limma software package (Smyth, 2005) and the ontology analysis conducted with GOstats (Falcon and Gentleman, 2007). Probes were flagged for filtering considering saturation, signal above background and uniformity. Probes that had any of the replicates flagged were eliminated. Absolute values of correlations between the M-values exceeded 0.83 and displayed expected bivariate distributions (data not shown). Genes were considered differentially expressed when the multiple testing adjusted P-value was <1×10⁻⁵ and the absolute value of log2 of fold change greater than 2 (supplementary material Fig. S5).

We thank Florencia Irigoín and Pablo Aguilar for their critical comments on the manuscript and Norann Zaghloul for her comments and experimental advice.