Bardet–Biedl Syndrome ciliopathy is linked to altered hematopoiesis and dysregulated self‐tolerance

Bardet–Biedl Syndrome (BBS) is a pleiotropic genetic disease caused by the dysfunction of primary cilia. The immune system of patients with ciliopathies has not been investigated. However, there are multiple indications that the impairment of the processes typically associated with cilia may have influence on the hematopoietic compartment and immunity. In this study, we analyze clinical data of BBS patients and corresponding mouse models carrying mutations in Bbs4 or Bbs18. We find that BBS patients have a higher prevalence of certain autoimmune diseases. Both BBS patients and animal models have altered red blood cell and platelet compartments, as well as elevated white blood cell levels. Some of the hematopoietic system alterations are associated with BBS‐induced obesity. Moreover, we observe that the development and homeostasis of B cells in mice is regulated by the transport complex BBSome, whose dysfunction is a common cause of BBS. The BBSome limits canonical WNT signaling and increases CXCL12 levels in bone marrow stromal cells. Taken together, our study reveals a connection between a ciliopathy and dysregulated immune and hematopoietic systems.

Bardet-Biedl Syndrome is a pleiotropic disease whose primary diagnostic features are rod-cone dystrophy, polydactyly, obesity, learning difficulties, hypogonadism, and renal anomalies (Beales et al, 1999). The immune system of patients with ciliopathies including BBS has not been studied in detail. An exception in this respect is a case report of three BBS patients suffering from autoimmune diseases in a cohort of 15 studied BBS patients (Halac & Herzog, 2012). The possible connection between ciliopathies and the immune system has not been addressed most likely because immune cells do not form primary cilia (Wheatley et al, 1996;Plotnikova et al, 2009). However, there are several lines of evidence suggesting that the BBS affects the function of the immune system.
First, the immunological synapse formed between T cells and antigen-presenting cells exhibits a striking analogy to the primary cilium (Finetti et al, 2011;de la Roche et al, 2016). The formation of these structures involves the reorganization of cortical actin and the centrosome polarization. In the same vein, some components of the IFT machinery have been shown to participate in the organization of the immunological synapse to promote T-cell activation (Yuan et al, 2014;Vivar et al, 2016). In particular, it has been shown that vesicles containing key T-cell signaling molecules, the TCR/CD3 complex and LAT, are transported toward the immunological synapse by IFT proteins (Finetti et al, 2009;Finetti et al, 2014). The BBSome may regulate T-cell activation by transporting signaling proteins into and/or out of the immunological synapse. Depending on the character of these putative cargoes and the direction of the transport, the BBSome could be a positive or a negative regulator of T-cell signaling.
Second, the BBSome is required for Sonic hedgehog (SHH) signaling (Zhang et al, 2011;Zhang et al, 2012b;Goetz et al, 2017). The SHH signaling pathway regulates multiple processes in the organism including T-cell development and differentiation (Shah et al, 2004;El Abdaloussi et al, 2006;Crompton et al, 2007;Rowbotham et al, 2008;Drakopoulou et al, 2010;Furmanski et al, 2013). Key components of the SHH signaling pathway, SMO, IHH, GLI1, and PTCH2, are upregulated in effector cytotoxic T cells and transported toward the immunological synapse in vesicles (de la Roche et al, 2013). Moreover, Smo KO/KO T cells show reduced cytotoxicity associated with defects in the polarized release of cytotoxic granules (de la Roche et al, 2013).
Third, the BBSome regulates trafficking of the leptin receptor (Guo et al, 2016). Leptin is a signaling molecule which acts as a proinflammatory cytokine (Procaccini et al, 2012;La Cava, 2017). In particular, leptin signaling inhibits the proliferation of regulatory T cells (De Rosa et al, 2007) and promotes the proliferation of effector T cells and their polarization toward Th1 helper cells (Martin-Romero et al, 2000). Moreover, T cells deficient in the leptin receptor show impaired differentiation into Th17 helper T cells in mice (Reis et al, 2015), indicating a T-cell intrinsic role of leptin signaling.
Fourth, the BBSome tunes WNT signaling by promoting the noncanonical pathway and suppressing the canonical WNT signaling (Ross et al, 2005;Gerdes et al, 2007). Since WNT signaling is an important regulator of hematopoiesis (Staal et al, 2016;Richter et al, 2017), the BBSome deficiency might lead to defective formation of blood cells including leukocytes.
Fifth, one of the major symptoms of BBS is obesity, which is believed to undermine the immune tolerance (Johnson et al, 2012). Obesity induces production of pro-inflammatory cytokines, such as TNF (Park et al, 2005) and IL-6 (Moro et al, 2010). These may predispose individuals to the development of autoimmune diseases (Gremese et al, 2014;Granata et al, 2017;Owczarczyk-Saczonek & Placek, 2017;Singh et al, 2017;Wang & He, 2018). Thus, the BBSome may have an extrinsic role in the immune system via inducing obesity.
In this study, we addressed the intrinsic and extrinsic roles of the BBSome in the immune system by investigating BBS patients and mouse models of the BBSome deficiency. We found that BBS patients show elevated prevalence of particular autoimmune diseases. Moreover, our data revealed dysregulated homeostasis of blood cells both in BBS mouse models and in BBS patients.

Autoimmune diseases are more prevalent in BBS patients
In this work, we studied the potential role of the BBSome in the immune system. Initially, we analyzed two cohorts of BBS patients from the CRIBBS NIH registry and from the Guy's Hospital of Guy's and St Thomas' NHS Foundation Trust, London, or Great Ormond Street Hospital, London. We found out that certain autoimmune and inflammatory diseases, such as type I diabetes, Hashimoto's thyroiditis, rheumatoid arthritis, and inflammatory bowel diseases, are more prevalent in BBS patients than in the overall population ( Table 1). The co-occurence of multiple autoimmune diseases was relatively small (Appendix Tables S1 and S2). The prevalence of autoimmunity does not seem to be restricted to any causative gene (Appendix Tables S3 and S4). Most frequent causative genes in our cohort were BBS1 and BBS10. Patients with mutations in BBS10 showed slightly higher prevalence of autoimmunity than patients with mutations in BBS1, which is in line with typically higher disease severity in the former group (Niederlova et al, 2019). However, the sample size was too small to draw any conclusions. Overall, these findings suggested that the BBSome has an intrinsic or extrinsic role in the immune system, particularly in the immune tolerance.

BBSome subunit genes are expressed in mouse lymphoid tissues
In the next step, we employed mouse models to study the putative role of the BBSome in the immune and hematopoietic systems in controlled experiments. First, we tested if the BBSome subunits are expressed in murine lymphoid tissues. We detected the expression of all eight BBSome subunits in the spleen, lymph nodes, and isolated T cells on the mRNA level ( Fig 1A). The expression levels of Bbs2, Bbs4, Bbs9, and Bbs18 in the lymphoid tissues were comparable to the brain and the kidney, two organs where the BBSome plays a major role (ODea et al, 1996;Davis et al, 2007;Keppler-Noreuil et al, 2011;Putoux et al, 2012). The other four subunits (Bbs1, Bbs5, Bbs7, and Bbs8) were expressed in the lymphoid tissues at 10-to 50-fold lower levels than in the brain and the kidney. Moreover, we detected BBS4 protein in isolated T and B cells ( Fig 1B). Altogether, all the BBSome subunits are variably expressed in lymphocytes and lymphocyte-rich tissue, despite of the fact that hematopoietic cells are commonly considered as non-ciliated cells. This suggested that the BBSome as a whole or some individual BBSome subunits might have an intrinsic role in lymphocytes.
Mouse models for studying the role of the BBSome in the immune system Our next step was to obtain a mouse model of BBS. We decided to use the Bbs4-deficient mouse for the following reasons: (i) BBS4 is an essential part of the BBSome (Klink et al, 2017), (ii) Bbs4 KO/KO mouse has been shown to have a relatively severe phenotype in comparison to other BBSome-deficient mice (Rahmouni et al, 2008;Guo et al, 2011), (iii) Bbs4 had a relatively high expression in lymphoid tissues (Fig 1A and B). In the following experiments, we used mice with an interrupted Bbs4 gene with a gene trap (GT allele) cassette, mice with deleted Bbs4 exon 6 (KO allele), and mice with Bbs4 exon 6 flanked with LoxP sites (FL allele) enabling the conditional deletion of this exon (cKO; Fig 1C).
BBS4 protein was not detectable in the testes, thymi and brain from the Bbs4 GT/GT mice (Fig 1D and E). As expected, the Bbs4 deficiency lead to the absence of sperm flagella in testes of 30-days-old Bbs4 GT/GT males ( Fig 1F). However, we did not observe two previously reported features of BBS mouse models in Bbs4 GT/GT mice. Although mating of Bbs4 +/GT heterozygotes resulted to 17% Bbs4 GT/ GT pups at weaning, only 2% Bbs4 KO/KO pups were produced in mating of Bbs4 +/KO heterozygotes ( Fig 1G). This suggests that Bbs4 KO/KO , but not Bbs4 GT/GT , mice suffer from frequent pre-weaning lethality. Moreover, Bbs4 KO/KO mice, but not Bbs4 GT/GT , developed obesity (Fig 1H-J). As expected, adult Bbs4 KO/KO mice, suffering from obesity, had elevated levels of leptin in blood plasma in comparison to non-obese Bbs4 GT/GT mice, and pre-obese young Bbs4 KO/KO mice ( Fig 1K). Because leptin has been proposed to act as a pro-inflammatory molecule, it might have an impact on the immune system of obese BBS mice.
To explain the phenotypic difference between Bbs4 GT/GT and Bbs4 KO/KO mice, we analyzed the respective Bbs4 transcripts via RT-PCR. Although transcript of Bbs4 KO allele lacked the exon 6 as expected, we detected a transcript containing all exons 5-11 in Bbs4 GT/GT mice ( Fig EV1A). Unexpectedly, the RNA polymerase II is able to read through two transcription termination sites in the gene trap cassette in Bbs4 GT/GT tissues to generate Bbs4 mRNA at~10% of the wild-type (WT) levels ( Fig EV1B). However, the PCR product of Bbs4 GT mRNA was longer than the respective amplicon of the WT allele (Fig EV1A and B). Sequencing revealed that Bbs4 GT allele is mis-spliced and includes a short insert originating from the gene trap cassette, which induces a frame shift in the open reading frame ( Fig EV1C). Expression of this Bbs4 GT cDNA in HEK293T cells resulted in a very tiny production of truncated BBS4 (Fig EV1D), possibly via an alternative transcription/translation start. It is plausible that low levels of this truncated BBS4 are present in Bbs4 GT/GT tissues as well. Moreover, we could detect a small amount (~1% of WT levels) of properly spliced Bbs4 mRNA in Bbs4 GT/GT brain ( Fig EV1E). Overall, these data indicate that Bbs4 KO is a null allele, whereas Bbs4 GT is a hypomorphic allele.

Altered B-cell compartment in Bbs4 deficient mice
To analyze the role of the BBSome in the formation of adaptive immune cells, we analyzed the development and homeostasis of T and B cells in Bbs4-deficient mice. We did not observe any major alterations in the T-cell compartment in Bbs4 KO/KO and Bbs4 GT/GT mice (Fig EV2A-F). The only significant differences were decreased The table shows the fold difference in the prevalence (prevalence in BBS patients/general prevalence) and the odds ratio of autoimmune diseases in the CRIBBS cohort of 255 BBS patients (upper part) and in the cohort of 198 BBS patients from Great Ormond Street Hospital and Guy's Hospital in London (lower part). Normal prevalence of autoimmune diseases was adopted from (Hayter & Cook, 2012;Martel-Pelletier et al, 2016;Taylor et al, 2018). Statistical significance was calculated using exact binomial test. a Total count is not equal to sum of all the patients with autoimmune diseases because of the co-occurrence of more than one disease in 4 patients. Co-occurring diseases are listed in Appendix Table S1. b Total count is not equal to sum of all the patients with autoimmune diseases because of the co-occurrence of more than one disease in 5 patients. Co-occurring diseases are listed in Appendix Table S2.
percentage of T cells among the splenocytes ( Fig EV2C) and decreased percentage of CD44 + cells among splenic CD8 + T cells of the Bbs4 KO/KO mice ( Fig EV2F).
We found an alteration of the B-cell development and/or homeostasis in Bbs4-deficient mice. The total numbers of bone marrow B220 + B-cell lineage cells were not significantly altered in Bbs4 KO/KO or Bbs4 GT/GT mice ( Fig EV3A). However, the ratio of B220 high and B220 low cells in the bone marrow was shifted toward less mature B220 low cells in Bbs4 KO/KO mice and, to a lesser extent, in Bbs4 GT/GT mice (Fig 2A). Both Bbs4 KO/KO and Bbs4 GT/GT mice had higher percentage of IgD À IgM À B-cell precursors than controls (Fig 2B). A deeper analysis of this population showed that Bbs4-deficiency Leptin, ng/ml results in the accumulation of pre-B cells, which is the developmental stage when the pre-B-cell receptor selection occurs ( Fig 2C). We observed slightly increased numbers of B cells in the spleen of Bbs4 KO/KO , but not Bbs4 GT/GT mice (Fig EV3B and C). In the spleen and lymph nodes, Bbs4 KO/KO and Bbs4 GT/GT mice showed larger late mature B-cell population (IgD + IgM À ) than WT mice (Figs 2D and E,and EV3D and E). In addition, Bbs4 KO/KO showed~2-fold lower percentage of splenic marginal zone (MZ) B cells (Fig 2F).
To address whether the phenotype of Bbs4 KO/KO is indeed caused by the BBSome dysfunction, we generated a second mouse model of BBS based on the genetic disruption of another BBSome subunit, BBS18 (alias BBIP1; Fig EV3F). Unfortunately, we observed strong pre-weaning lethality resulting in only two adult Bbs18 KO/KO animals overall ( Fig EV3G). The comparison of these Bbs18 KO/KO animals to their WT and Bbs18 +/KO littermates showed very similar phenotype to Bbs4 KO/KO mice, i.e., increased B-cell precursors in the bone marrow ( Fig EV3H) and decreased MZ B cells in the spleen ( Fig EV3I). These data indicated that the alterations of the B-cell lineage are not unique to the deficiency in BBS4, but are likely a common feature of the BBSome disruption.
Overall, these data show that BBSome deficiency results in an abnormal development of B cells in the bone marrow, a slightly increased percentage of late mature splenic B cells and a reduction of splenic MZ B cells in the periphery. The developmental alteration was observed also in non-obese Bbs4 GT/GT mice, indicating that this phenotype is largely obesity-independent.

Bbs4 deficiency does not intrinsically influence antigen-specific T-cell and B-cell responses
As we observed an alteration of B-cell homeostasis in Bbs4-deficient mice, we decided to investigate how it affects the response of the adaptive immune system. First, we activated monoclonal B cells specific for 4-hydroxy-3-nitrophenyl acetyl (NP) from B-cell receptor transgenic B1-8 mice (Sonoda et al, 1997) and Bbs4 GT/GT B1-8 mice using NP-labeled cells. In this setup, we monitored the upregulation of the activation marker CD69 (Fig 3A and B). Moreover, we stimulated B cells from Bbs4 KO/KO , Bbs4 +/KO , and Bbs4 +/+ mice with plate bound anti-IgM antibody ( Fig 3C). In these assays, we did not observe any role of Bbs4 deficiency in the antigenic B-cell response.
Based on the homology between the cilium and the immunological synapse (Finetti et al, 2009;Finetti et al, 2011;Stephen et al, 2018), we hypothesized that the BBSome might act as a positive or negative regulator of T-cell responses. For this reason, we generated Bbs4 FL/FL Cd4-Cre mouse line where Bbs4 deficiency was restricted to T cells ( Fig EV4A). These mice did not show any obvious phenotype in the T-cell compartment (Fig EV4B-E). To study the role of the BBSome in T-cell antigenic responses, we crossed Bbs4 FL/FL Cd4-Cre mice to TCR transgenic OT-I Rag2 KO/KO mice. This mouse generates monoclonal T cells specific for K b -OVA, a model antigen originating from chicken ovalbumin. We did not observe any role of BBS4 in the conjugation of OT-I T cells with antigen-presenting cells loaded with OVA or its lower affinity variants ( Fig 3D). Moreover, WT and Bbs4-deficient OT-I T cells showed the same ability to induce autoimmune diabetes upon a transfer into RIP.OVA mice expressing ovalbumin under the rat insulin promoter and subsequent priming by Listeria monocytogenes expressing ovalbumin (King et al, 2012;Palmer et al, 2016) (Fig 3E and F). This assay examines T cells for their ability to get primed by the OVA-antigen, expand, infiltrate the pancreas, and kill the b-cells. As the onset of diabetes caused by Bbs4-deficient OT-I T cells was not different from the control, we concluded that BBS4 does not play an important intrinsic role in any of indicated steps of the T-cell-mediated immune response.

The role of BBS4 in B-cell homeostasis is not intrinsic
To investigate if the observed B-cell developmental changes in Bbs4deficient mice are B-cell intrinsic, we generated Bbs4 FL/FL Vav-iCre mice with a specific deletion of Bbs4 in the hematopoietic and ◀ Figure 1. Basic characterization of Bbs4-deficient mouse models.
A Relative expression of the BBSome subunits in the indicated organs and T cells measured by qPCR. C T values of the BBS genes were normalized to the geometric mean of the C T values of reference genes Gapdh, Tubb2a, and Eef1a1. The expression levels are normalized to those of brain (= 1). Mean + SD. Three biological replicates (the result of each biological replicate is a median of three technical replicates). B Immunoblot analysis of BBS4 expression in T and B lymphocytes isolated from the lymph nodes and spleen of C57Bl/6 mouse. b-actin staining served as a loading control. The identical immunoblot is shown in Fig EV4A. A representative experiment out of three biological replicates in total is shown. C Schematic representation of mouse models of Bbs4 deficiency used in this study. Bbs4 GT, gene trap allele interrupting the Bbs4 gene. Bbs4 FL, allele with floxed Bbs4 exon 6. Bbs4 KO and cKO alleles with deleted exon 6. D Immunoblot analysis of BBS4 expression in testes and thymi lysates of Bbs4 +/+ (WT), Bbs4 GT/+ , and Bbs4 GT/GT mice. b-actin staining serves as a loading control. A representative experiment out of three biological replicates in total is shown. E Immunoblot analysis of BBS4 expression in the brain lysates of Bbs4 +/+ , Bbs4 GT/GT , and Bbs4 KO/KO mice. b-actin staining served as a loading control. A representative experiment out of two biological replicates is shown. F Hematoxylin and eosin staining of sections of seminiferous tubules from~4 weeks old Bbs4 +/+ (n = 2 mice) and Bbs4 GT/GT (n = 4) males. Scale bar, 100 µm.
Source data are available online for this figure.
ª 2021 The Authors EMBO reports 22: e50785 | 2021 endothelial cells (Georgiades et al, 2002). We did not observe any evidence of altered B-cell homeostasis in Vav-iCre mice (Fig 4A-C). These results suggest that the B-cell compartment in Bbs4 GT/GT and Bbs4 KO/KO mice is affected by factors extrinsic to the hematopoietic lineage.
The development of B cells is guided by the cytokine environment in the bone marrow niche with prominent roles of IL-7 and CXCL12 (Zehentmeier & Pereira, 2019). We observed that the expression of Cxcl12, but not Il-7, is decreased in the bone marrow of Bbs4 and Bbs18 KO mice (Fig 4E). In the next step, we generated  (Fig 4F), indicating that the BBSome regulates Cxcl12 expression in mesenchymal cells. It has been shown previously that the BBSome inhibits canonical WNT signaling in vertebrates (Gerdes et al, 2007) and that the canonical WNT signaling downregulates Cxcl12 expression in bone marrow stroma-derived ST2 cells (Tamura et al, 2011). Accordingly, canonical WNT signaling (triggered by WNT3A), but not non-canonical WNT signaling (triggered by WNT5A), suppressed Cxcl12 expression in ST2 cells and induced the expression of canonical WNT responsive gene Alpl (Fig 4G). We produced Bbs4-deficient ST2 cells using CRISPR/Cas9 (Fig EV5A,  Appendix Table S6). Two out of four Bbs4 KO/KO ST2 cell clones showed significantly reduced expression of Cxcl12 ( Fig EV5B).
Overall, our data indicate that the BBSome deficiency amplifies canonical WNT signaling in bone marrow stromal cells leading to decreased production of CXCL12 and subsequent partial defects in B-cell development.

BBS-induced obesity affects blood homeostasis
To investigate the possible factors predisposing BBS patients to the development of autoimmune diseases, we examined the blood test results of BBS patients. Intriguingly, immunity-related parameters, such as counts of total white blood cells, neutrophils, and eosinophils, were increased in BBS patients, when compared to age and gender-matched controls (BMI-random controls) (Fig 5A). To address the role of obesity in BBS patients, we used an additional set of controls with body mass indexes (BMI) matching to those of BBS patients (BMI-matched controls). We did not observe any difference when we compared the indicated leukocyte parameters between BBS patients and BMI-matched controls. We analyzed the peripheral blood of Bbs4-deficient mice to further explore the role of the BBSome and the potential involvement of obesity in blood cell homeostasis. In agreement with the analysis of the patients' blood tests, we did not observe major differences between WT and non-obese Bbs4 GT/GT mice. However, obese Bbs4 KO/KO mice showed higher total white blood cell count than WT controls (Fig 5B), which corresponded with the data from patients as well (Fig 5A). These results indicate that obesity in BBS patients and in Bbs4-deficient mice has an impact on the leukocyte homeostasis.
We found that elevated C-reactive protein (CRP) was more frequent in BBS patients than in BMI-matched and BMI-random controls (Fig 5C), which indicates that obesity is not the only factor influencing the immune system of BBS patients. Notably, the homeostasis of red blood cells was altered in BBS patients as well as in the Bbs4 KO/KO mice, although at different levels ( Fig 5D  and E). BBS patients showed low overall hemoglobin levels caused by a mild decrease in the red blood cell count and low red blood cell hemoglobin (Fig 5D), indicating a possibility of a reduced oxygen transport capacity. Interestingly, the comparison of BBS patients with BMI-matched controls showed that the alteration of the erythroid compartment was not caused by obesity. The mouse model of BBS showed a decreased number of red blood cells, which was compensated by enlarged red blood cell volume ( Fig 5E).
In addition, we observed decreased platelet counts both in BBS patients and Bbs4 KO/KO mice (Fig 5F-G). The reduction of platelets in BBS patients was not obesity-dependent. Furthermore, Bbs4 KO/KO mice showed higher mean platelet volume and lower plateletcrit than WT mice (Fig 5F-G), indicating enhanced removal of platelets in the periphery.
Altogether, our results suggest a role of the BBSome in the immune tolerance, hematopoiesis and/or blood homeostasis. Most of the effects seem to be extrinsic to the hematopoietic compartment as revealed by using tissue-specific knock-out mouse model and comparison of patients to BMI-matched controls. However, altered CRP levels, red blood cell, and platelet homeostasis in BBS patients are obesity-independent.
Source data are available online for this figure.   (Sindhu et al, 2015). Obesity is also considered as a risk factor for autoimmune disorders (Versini et al, 2014). Indeed, we have shown elevated prevalence of autoimmune disorders (namely T1DM, inflammatory bowel diseases, rheumatoid arthritis, and hypothyroidism) in two independent BBS cohorts. Obesity potentially contributes to the increased prevalence of autoimmune diseases in BBS patients, as obesity is associated with high prevalence of hypothyroidism, Hashimoto's thyroiditis, and rheumatoid arthritis (Song et al, 2019;Ohno et al, 2020). We also observed alteration of the white blood cell count in BBS patients and in a BBS mouse model. The comparison of BBS patients and non-BBS obese   Percentage of B-cell precursors (B220 + , IgM À IgD À ) in the bone marrow. Representative experiment out of four in total is shown. Bbs4 FL/FL (n = 6 mice), Vav-iCre Bbs4 FL/FL (n = 9). (C) Percentage of CD19 + B cells in spleens and lymph nodes of Vav-iCre Bbs4 FL/FL (n = 12 mice) and Bbs4 FL/FL (n = 9) mice, six independent experiments. (D) Percentage of late mature B cells (CD19 + , IgM À , IgD + ) in spleens of Vav-iCre Bbs4 FL/FL (n = 12 mice), and Bbs4 FL/FL mice (n = 9). Representative experiment out of six in total is shown. Statistical significance was calculated using two-tailed Mann-Whitney test, P > 0.05 for all tests. Median is shown. E Expression of Cxcl12 and Il-7 in the bone marrow from Bbs4 +/+ Bbs18 +/+ (WT) mice (black circles, n = 7 mice), Bbs4 +/KO (Het, green circle, n = 3), Bbs18 +/KO (Het, blue circle, n = 3), Bbs4 KO/KO (green square, n = 5), and Bbs18 KO/KO (blue square, n = 2) was analyzed by RT-qPCR in five independent experiments. The expression was normalized to Gapdh. The statistical significance of the difference between the BBSome-deficient mice (Bbs4 KO/KO and Bbs18 KO/KO ) and controls were calculated using two-tailed Mann-Whitney test. Median is shown. F MEF cell lines were derived from a single Bbs4 +/+ (WT) or a single Bbs4 KO/KO embryos. The expression of Cxcl12 in these cell lines was analyzed by RT-qPCR. The expression was normalized to Gapdh and to Bbs4 +/+ MEFs (= 1) for each experiment. Three biological replicates. One sample t-test was used for the statistical analysis. G Expression of Cxcl12 and canonical WNT responsive gene Alpl in untreated (NT) ST2 cells, or ST2 cells treated with WNT3A or WNT5A was analyzed by RT-qPCR. The expression was normalized to Gapdh and untreated ST2 sample (= 1) for each experiment. Four biological replicates were performed. One sample t-test was used for the statistical analysis.
Source data are available online for this figure.    controls, as well as the comparison of obese and non-obese BBS mouse model, shows that this effect is caused by obesity.
One of the major players in obesity-associated inflammation is leptin, an adipocyte-derived hormone which acts as a pro-inflammatory cytokine (Lord, 2006;Paz et al, 2012). The BBSome is proposed to transport the leptin receptor to the plasma membrane, which makes the BBSome-deficient cells unresponsive to leptin (Guo et al, 2016). It has been shown that leptin signaling in the central nervous system regulates immune responses and prevents lethal overreaction of the immune system to sepsis in mice (Tschop et al, 2010). Thus, it is possible that defective leptin signaling in the nervous system directly contributes to high prevalence of autoimmunity in BBS patients.
Leptin also promotes T-cell activation and proliferation (Martin-Romero et al, 2000;Milner & Beck, 2012), suggesting a T-cell intrinsic phenotype of the BBSome deficiency. Moreover, T cells repurpose a number of ciliary transport proteins for immunological synapse formation (Finetti et al, 2009;Vivar et al, 2016;Stephen et al, 2018). Last but not least, the BBSome is functionally connected to IFT20 and SHH signaling components (Wei et al, 2012;Williams et al, 2014;Nakayama & Katoh, 2018;Niederlova et al, 2019), which were shown to regulate early T-cell development (El Abdaloussi et al, 2006;Michel et al, 2013;Yuan et al, 2014;Vivar et al, 2016). For all these reasons, we tested whether the BBSome plays a positive or negative intrinsic role in T-cell biology. However, the development and function of T cells were largely unaffected in Bbs4 KO/KO mice and in Cd4-Cre Bbs4 FL/FL mice.
Bardet-Biedl Syndrome patients exhibited additional defects in blood cell homeostasis, which cannot be solely attributed to obesity. First, BBS patients had elevated blood CRP levels, which might be a sign of ongoing inflammation. Second, BBS patients and Bbs4-null mice exhibit low platelet counts. Furthermore, platelets from Bbs4 KO/KO mice had larger volume than WT controls. This indicates that a greater fraction of platelets in Bbs4 KO/KO mice are immature. As these mice also have low platelet counts, their bone marrow is likely compensating for platelet loss on the periphery (Stasi, 2012). This state can be a result of immune-mediated processes such as immune thrombocytopenia (Schmoeller et al, 2017;Norrasethada et al, 2019). Unfortunately, data concerning mean platelet volume are not available for BBS patients, but we can presume that the peripheral destruction of platelets is a cause of thrombocytopenia in BBS patients as well as in mice.
Our analysis of BBS mouse models revealed the role of the BBSome in B-cell development. Bbs4 KO/KO , Bbs18 KO/KO , and nonobese hypomorphic Bbs4 GT/GT mice showed an increased frequency of B-cell precursors in the bone marrow, as well as low numbers of MZ B cells. These effects were not caused by BBSome deficiency in hematopoietic cells. Instead, BBSome-deficient bone marrow stromal cells produce low levels of CXCL12, most likely as a consequence of enhanced canonical WNT signaling (Gerdes et al, 2007;Tamura et al, 2011). Our explanation that the BBSome regulates Bcell development by increasing CXCL12 availability is supported by a previous study (Nie et al, 2004). This study showed that B-cellspecific deletion of CXCR4, the receptor for CXCL12, manifests with a phenotype resembling our BBS mouse models, i.e., increased frequency in immature IgM À IgD À B cells in the bone marrow and low numbers of MZ B cells in the spleen. Accordingly, deletion of WBP1L, a negative regulator of CXCR4 signaling, shows the opposite phenotype, i.e., decreased frequency of pre-B cells in the bone marrow and increased frequency of MZ B cells in the spleen (Borna et al, 2020). The importance of ciliary signaling in non-hematopoietic bone marrow cells for hematopoiesis should be addressed in further studies.
To our knowledge, this is the first study revealing the dysregulation of the immune and hematopoietic systems in patients suffering from a ciliopathy. Obesity, a common feature of ciliopathies such as BBS and Alstr€ om syndrome (Vaisse et al, 2017), leads to elevated concentration of white blood cells in BBS patients. However, BBS patients also exhibit elevated blood CRP levels and low platelet counts in an obesity-independent manner. A detailed analysis of BBS mouse models revealed low CXCL12 production by BBSome-deficient stromal cells resulting in defects in B-cell development. Given the importance of the CXCL12-CXCR4 axis in immunity and hematopoiesis, there might be additional effects of BBSome deficiency beyond our findings. Altogether, we identified BBS as a risk factor for autoimmune diseases.
Source data are available online for this figure.

Mice
All mice were 5-25 weeks old and had C57Bl/6J background if not indicated otherwise. Mice were bred in specific-pathogen-free facility (IMG) (Flachs et al, 2014). Animal protocols were approved by the Czech Academy of Sciences, in accordance with the laws of the Czech Republic. Males and females were used for the experiments. If possible, age-and sex-matched pairs of animals were used in the experimental groups. If possible, littermates were equally divided into the experimental groups. No randomization was performed since the experimental groups were based solely on the genotype of the mice. In case of the RIP.OVA mice used for the autoimmune diabetes model, mice were assigned to experimental groups randomly (defined by their ID numbers) prior to the visual contact between the experimenter and the mice. The experiments were not blinded since no subjective scoring method was used. For animal studies, minimal sample size was estimated using resource equation approach. As the number of mutant mice was limited due to pre-weaning lethality, all the available mutant mice (and their WT littermates) were used for experiments. For the diabetes experiments, the number of mice was estimated based on our previous experience with this method. B1-8 (Sonoda et al, 1997), RIP.OVA (Kurts et al, 1998), OT-I Rag2 KO/KO (Palmer et al, 2016), Vav-iCre (Shimshek et al, 2002;de Boer et al, 2003), Cd4-Cre (Lee et al, 2001) strains were described previously. Mouse Bbs4 tm1a(EUCOMM)Hmgu allele with a gene trap cassette (abbreviated as Bbs4 GT ) was obtained from KOMP (UC Davis, CA, USA) as frozen sperm and used for in vitro fertilization. Bbs4 tm1a(EUCOMM)Hmgu mice were crossed with a Flp-deleter mouse strain, Gt(ROSA)26Sor tm2(CAG-flpo,-EYFP)Ics (MGI: 5285396, Philippe Soriano), in order to obtain exon 6 flanked by LoxP sites for conditional or germ line Bbs4 ablation (Bbs4 tm1c(EUCOMM)Hmgu ). In order to generate germ line knock-out mice, the exon 6 was removed by a general Cre-deleter strain, Gt(ROSA)26Sor tm1(ACTB-cre,-EGFP)Ics (MGI: 5285392, Philippe Soriano), resulting in Bbs4 tm1d(EUCOMM)Hmgu (abbreviated as Bbs4 KO ) as described previously (Skarnes et al, 2011). Bbs4 +/+ and Bbs4 GT/GT or Bbs4 KO/KO littermates were generated by intercrossing heterozygous animals.
Mice with a null-mutation in Bbs18 (alias Bbip1) were generated in a C57BL/6N background using a CRISPR genome-editing system. For this purpose, Cas9 protein and gene-specific single guide (sg) RNAs (Integrated DNA Technologies, Coralville, IA, USA) were used for a zygote electroporation using a protocol described previously (Chen et al, 2016). sgRNA sequences with the PAM motif in bold (3 0 end) were as follows: sgRNA target 1: CTCTTCCCTGAAAATCGGTGAGG sgRNA target 2: GGAATAACCAACTGGTCTTTAGG The correct genome editing was confirmed by PCR amplification in the founder mouse with the primers listed in Appendix Table S5. Generation of Bbs18 KO mice was part of the International Mouse Phenotyping Consortium project (mouse ID 12685, Bbip1 em1(IMPC)Ccpcz ). The mutant allele was backcrossed for two generations to C57BL/6N mice and for one generation to C57BL/6J mice.
We did not confirm the efficacy of the Bbs4 deletion in Bbs4 FL/FL Vav-iCre strain. However, this Vav-iCre mouse line was proven as an efficient gene deleter in the hematopoietic bone marrow cells and splenocytes in our facility previously (Kardosova et al, 2018).

Cell culture
T2-Kb is a H-2Kb transgenic subline of the TAP1c2-deficient line (Alexander et al, 1989). The bone marrow-derived stroma cell line, ST2, was kindly provided by Dr. Jana Balounova (IMG). The HEK293T cells were kindly provided by Dr. Tomas Brdicka (IMG). All cell lines were tested for Mycoplasma contamination by PCR on a regular basis. MEF cells were derived from E13.5 mouse embryos. After the removal of the placenta, yolk sac, head, liver, and hear, the tissue was trypsinized and homogenized. The adherent MEFs were genotyped and cultured. All cell lines were cultivated in DMEM supplemented with 10% fetal calf serum (FCS; Gibco), 100 U/ml penicillin (BB Pharma), 100 lg/ml streptomycin (Sigma-Aldrich), and 40 lg/ml gentamicin (Sandoz).

Cloning, transfection, and generation of BBS4-deficient cell lines
Bbs4 WT and GT ORFs were amplified from cDNA obtained from the kidney of respective mice. FLAG-tag was fused to the Bbs4 WT and GT C-terminally and cloned into pXJ41 vector (kindly provided by Dr. Tomas Brdicka, IMG) using EcoRI/XhoI restriction sites.
HEK293T cells at~50% confluency were transfected with the Bbs4 WT or Bbs4 GT encoding pXJ41 plasmid using the following protocol. Thirty microgram of DNA plasmid was mixed with 75 lg polyethylenimine in 0.5 ml of 0.5% serum antibiotic-free media for 10 min at room temperature. The mixture was then added onto cells in 3 ml of 0.5% serum and antibiotic-free media. The media was replaced by DMEM with 10% serum and antibiotics after 3 h. Twenty-four hours after transfection, the expression of BBS4 was analyzed by a-FLAG antibody via immunoblotting.

RT-qPCR and end-point PCR
Total RNA was extracted from tissues (kidney, brain, lymph nodes, spleen, bone marrow) and T cells from indicated mice, ST2 cells or mouse embryonic fibroblasts using Trizol LS (Invitrogen, # 10296010) and transcribed using RevertAid reverse transcriptase (Thermo Fisher, # EP0442) with oligo(dT) 18 primers according to the manufacturer's instructions. cDNA was used for end-point PCR or quantitative PCR using indicated primers (primer sequences are listed in the Appendix Table S5). Samples for quantitative PCR were measured by LightCycler 480 (Roche) in technical triplicates. Obtained C T values were normalized to indicated reference genes.

Enrichment of T and B lymphocytes
T and B lymphocytes for immunoblotting were enriched by positive selection using the Dynabeads Biotin Binder kit (Invitrogen, #11047), and biotinylated a-TCRb and a-CD19 antibodies, respectively. B cells for B-cell activation assay were enriched by negative selection using biotinylated a-CD4 and a-CD8 antibodies with the automated cell separator AutoMACS Pro system (Miltenyi Biotec).

Immunoblotting
For analysis of BBS4 expression, freshly isolated murine organs (testicles, thymi, brain), enriched lymphocytes, ST2 cells, or HEK293T cells were lysed and denatured in Laemmli buffer supplemented with 20mM dithiothreitol. The resulting lysates were separated on a polyacrylamide gel and transferred to nitrocellulose membrane using standard immunoblotting protocols. Membranes were probed with indicated antibodies followed by secondary antirabbit or anti-mouse HRP-conjugated antibodies. Staining for bactin served as a loading control. The signal was visualized using chemiluminescence immunoblot imaging system Azure c300 (Azure Biosystems, Inc.).

Histological analysis
Testes isolated from 30-day-old male mice were collected, immediately dipped into Bouin solution, and fixed for 24 h at 4°C. Paraffinembedded tissue blocks were cut with a microtome (Leica RM2255), and the sections were stained with hematoxylin/eosin using standard technique. The images were taken using microscope system Axioplan 2 imaging (Zeiss) using 10×/0.50 NA objective.

Weighting of mice
Bodyweight of Bbs4 +/+ , Bbs4 GT/GT , Bbs4 KO/KO mice was recorded weekly starting at 5 weeks of age. All the mice were kept in sexmatched cages together with their littermates (≤ 6 per cage) and fed a standard chow diet ad libitum.

ELISA
Blood from Bbs4 KO/KO , Bbs4 GT/GT , and their age/sex-matched controls was collected by submandibular bleeding (Golde et al, 2005) into EDTA-coated tubes and centrifuged for 15 min at 1,000 g at 4°C in order to separate plasma. Obtained plasma samples were assayed immediately or stored at À80°C for later use. Leptin concentration was measured by mouse leptin ELISA Kit (Cloud-Clone Corp., SEA084Mu) according to the manufacturer's instructions.

Flow cytometry
Live cells were stained with the above-listed antibodies on ice. LIVE/DEAD near-IR dye or Hoechst 33258 were used for discrimination of live and dead cells. Flow cytometry was carried out using an LSRII (BD Bioscience). Data were analyzed using FlowJo software (TreeStar).

T-cell conjugation assay
T-cell conjugation assay was performed as previously shown (Palmer et al, 2016). Briefly, OT-I T cells from Bbs4 FL/FL Cd4-Cre À or Bbs4 FL/FL Cd4-Cre + littermates were stained with CFSE cell tracker dye, and splenocytes isolated from C57Bl/6 mice were stained with DDAO cell tracker dye. Splenocytes were loaded with OVA peptide or with indicated altered peptide ligands for 3 h in RPMI/10% FCS, mixed with OT-I T cells at 2:1 ratio, and centrifuged (1,000 g, 1 min). After 20 min of co-culture at 37°C/CO 2 incubator, cells were fixed by adding formaldehyde (2% final, 35 min). Cells were centrifuged (1,000 g, 2 min), resuspended in PBS/0.5% gelatin, and analyzed by flow cytometry. The result of each of the four biological replicates was an average of technical duplicates.

Model of autoimmune diabetes
The model of autoimmune diabetes has been described previously (Drobek et al, 2018). Briefly, OT-I cells from Bbs4 FL/FL Cd4-Cre À or Bbs4 FL/FL Cd4-Cre + sex-matched littermates were adoptively transferred into a host RIP.OVA mice intravenously. On the following day, the host mice were immunized with 5,000 CFU of OVA expressing Listeria monocytogenes (Lm). Lm strain expressing OVA has been described previously (Chan et al, 2013). Level of glucose in the urine of RIP.OVA mice was monitored on a daily basis using test strips (GLUKOPHAN, Erba Lachema). The animal was considered to suffer from diabetes when the concentration of glucose in the urine reached ≥ 1,000 mg/dl for two consecutive days. On day 7 post-infection, blood glucose was measured using contour blood glucose meter (Bayer).

Analysis of the clinical data of BBS patients
Fully anonymized medical records of 255 BBS patients were obtained from the Clinical Registry Investigating BBS (CRIBBS) by the NIH through the National Center for Advancing Translational Sciences and the Office of Rare Diseases Research (https://grdr. hms.harvard.edu/transmart). Data were extracted from the following categories of the CRIBBS: Clinical data/ Physical findings/ Bone and joint information, Endocrine, Thyroid disorders, Digestive. Although the CRIBBS registry does not contain information about the specific type of arthritis, we assume that the arthritis reported in can be classified as rheumatoid arthritis due to the following facts: (i) the mean age of onset of arthritis in the CRIBBS cohort is 28.7 years, (ii) the mean age on onset of rheumatoid arthritis in general population is 20-29 years (Hayter & Cook, 2012), (iii) the typical onset of osteoarthritis is after 50 years (Martel-Pelletier et al, 2016).
Medical records of BBS patients attending the BBS multidisciplinary clinic at Guy's Hospital of Guy's and St Thomas' NHS Foundation Trust, London, or Great Ormond Street Hospital, London, were studied in detail with focus on presence of any immune-related phenotype. In addition to the manual control, the records were also automatically searched for the occurrence of the following terms: autoimm-, immun-, thyro-, inflam-, diabet-, T1DM, ulcerative, crohn, IBD, rheuma-, arthri-, joints.
Data about the prevalence of autoimmune diseases in the two BBS cohorts were compared to normal prevalence of autoimmune diseases reported in the Autoimmune Registry or relevant publications in case of location-specific prevalence of hypothyroidism and inflammatory bowel diseases (Stone et al, 2003;Hayter & Cook, 2012;Taylor et al, 2018). Statistical significance of the difference in the prevalence between the BBS patients and overall population was tested using two-tailed exact binomial test in RStudio (function binom.test, R version: 3.6.2, RStudio version: 1.2.1335). The respective R-script is deposited in Zenodo (https://doi.org/10.5281/zenod o.3733230).
Results of blood tests of BBS patients (total white blood cell count, leukocyte populations, hemoglobin, platelet counts, mean corpuscular volume, red blood cell count, hematocrit, red cell distribution width and mean corpuscular hemoglobin), their age ranges, genes with causative mutation and body mass indices (BMI) were retrospectively ascertained from medical records stored at the BBS multidisciplinary clinic at Guy's Hospital of Guy's and St Thomas' NHS Foundation Trust, London, or Great Ormond Street Hospital, London. Blood tests were performed during regular medical examination of the patients.
Two distinct sets of controls for the analyzed set of BBS patients were selected from the 14,750 participants of the UK Biobank project (ID: 40103) (Sudlow et al, 2015). First, we selected 10 controls for each patient matching by age range (categories 41-50, 51-60, 60+ years) and sex. These controls had random BMI and thus were used as BMI-random controls. Second, we selected 10 controls for each patient matching by age range (categories 41-50, 51-60, 60+ years), sex, and BMI. These were used as BMI-matched controls. For 34 of the 42 patients, we found controls with BMI difference ≤ 0.6 kg/m 2 . For the eight patients with extreme BMI values, that precise matching was not possible, so that we selected the best-matching controls available for these cases. As the UK Biobank only includes participants older than 40 years, only BBS patients from this age group were included to this analysis.
All BBS patients gave informed consent or assent. The protocol for this study was approved by the Great Ormond Street Hospital Research Ethics Committee (Project Molecular Genetics of Human Birth Defects-mapping and gene identification, reference #08/ H0713/82) and by the ethical committee of the IMG.

Statistical analyses
The statistical significance of the observed differences between experimental groups was calculated using frequentist statistics. Appropriate tests were applied for particular experiments, and P-values were calculated using R or Prism 5.04 (GraphPad). The names of the tests are indicated in the Figure Legends  Two-sided tests were used. We adjusted P-values for multiple comparisons, when applicable. Whenever possible, we used nonparametric tests. The only exceptions were RT-qPCR experiments with very low number of replicates, where non-parametric tests were not applicable. It these cases we used one sample t-test, assuming normal distribution of the data (which could not be statistically tested because of low number of replicates). We indicate the exact P-value with the exception of some tests where the statistical software did not calculate the exact P-value if it was very small (P < 0.0001).

Data availability
No primary large datasets have been generated or deposited. Rscripts used for the analysis of human patient data were deposited in Zenodo (https://doi.org/10.5281/zenodo.3733230).
Expanded View for this article is available online.