Mutations in TMEM216 perturb ciliogenesis and cause Joubert, Meckel and related syndromes

The neurological features of JSRD include hypotonia, ataxia, psychomotor delay, irregular breathing pattern and oculomotor apraxia and are variably associated with multiorgan involvement, mainly retinal dystrophy, nephronophthisis (NPH) and congenital liver fibrosis. JSRD are genetically heterogeneous, and all known genes encode proteins localized at or near the primary cilium¹. We previously mapped the JBTS2 (also known as CORS2) locus to chromosome 11p12-q13.3 in a large Sicilian family and in three consanguineous pedigrees from the Middle East^2-3. Aligning the two datasets suggested a minimal candidate interval between D11S1344 and D11S1883 (46.123-63.130 Mb)⁴ (Fig. 1a).

Overlapping with JSRD is MKS, characterized by occipital encephalocele and other posterior fossa defects, cystic dysplastic kidneys, hepatic bile duct proliferation and polydactyly, and the two conditions are known to be allelic at four loci^5-8. The MKS2 locus was initially mapped in families of North African and Middle Eastern ancestry to a chromosome 11q region telomeric to JBTS2⁹, but our subsequent identification of additional families, as well as SNP re-analysis of the initial family, indicated allelism with JBTS2 between rs1113480 and rs953894 (48.014-62.518 Mb) (Supplementary Fig. 1). Because JSRD and MKS are considered ciliopathies, of the 200 total candidate genes, we first sequenced the exons and splice sites of genes listed in the cilia proteome databases^10-11 in one affected subject from each JBTS2/MKS2 family, but no mutations were identified.

Tetraspan transmembrane proteins are characterized by four hydrophobic, putative transmembrane domains (TM1-TM4), forming two extracellular and one intracellular loop, which regulate signaling and trafficking properties of their partner proteins in multiple cellular contexts¹². While little is known about their function, they can act with Wnt receptors¹³, and their ability to form complexes with a wide variety of membrane and cytosolic proteins¹⁴ suggests that they may participate in the formation of membrane domains that regulate signaling and sorting processes. Transmembrane proteins also represented attractive candidates, due to similarities to MKS3/TMEM67 encoding Meckelin, which is mutated both in JSRD and MKS^5,15. Therefore we additionally sequenced the eight genes encoding transmembrane proteins, eventually identifying homozygous deleterious mutations in TMEM216 in six of the 12 JSRD/MKS families compatible with linkage to the locus (Table 1). Interestingly, residue p.R73 was mutated both in a Sicilian family with JSRD (COR000, p.R73L) and in a Turkish family, in which MKS and JSRD coexisted in the same sibship (COR114/F37, p.R73H). The p.G77A mutation in two Palestinian families (F56, F58) resulted from a substitution (c.230G>C) that affects the first base of exon 5, leading to the use of an alternative splice site in intron 4, the inclusion of an additional 46bp and resultant premature protein termination (p.T78KfsX30) (Supplementary Fig. 2). None of these mutations were identified in over 500 controls from ethnically matched cohorts.

We next screened an additional 460 JSRD and 132 MKS probands (Supplementary Note) and identified mutations in 12 and two further cases, respectively (Table 1). Twelve of 14 JSRD families shared the same homozygous p.R73L founder mutation, including two families from Sicily and ten families of Ashkenazi Jewish descent. Saturation of the region surrounding the p.R73L mutation with 17 SNP/microsatellite markers indicated that these families shared the same ancestral haplotype, spanning 472 Kb around the mutation (Supplementary Fig. 3), that could be dated back at least 20 generations. The carrier frequency in the Ashkenazi population was determined to be about 1:100, as we identified two heterozygous healthy unrelated carrier individuals among a screened cohort of 212 Ashkenazi individuals, making carrier detection possible at least in this population. A similar carrier frequency of 1:92 has also been determined for the p.R73L mutation in a distinct study on eight Ashkenazi JBTS2 families and 2766 unaffected controls¹⁶. Microsatellite analysis also detected shared haplotypes in the two Palestinian (F56, F58) and in the two Tunisian families (F2, F5), homozygous for the same mutations (Supplementary Fig. 1).

Overall, 20 JSRD patients and 11 MKS fetuses carried TMEM216 mutations (Fig. 1b,​ Table 1). All of the nonsynonymous changes occurred in evolutionarily conserved residues (Fig. 1c-d), and led to unstable protein when transfected into heterologous cells (Fig. 1e,​ Supplementary Fig. 4). Although truncating mutations were identified in both the middle and end of the protein, p.R73 transversions predominated (Fig. 1c), with the p.R73L clearly a founder mutation. Among JSRD, the phenotype was characterized by frequent occurrence of NPH (9/20) and polydactyly (9/20), while retinal dystrophy and congenital hepatic fibrosis were never observed. In keeping with this, sequence analysis of 96 patients with Bardet-Biedl syndrome identified no homozygous mutations, since retinopathy is a key feature of this condition. In two JSRD patients (MTI161 and MTI467), polydactyly was associated with either tongue tumors or multiple oral frenula, corresponding to the Oro-Facio-Digital type VI (or Varadi-Papp) syndrome¹⁶ (OMIM%277170), indicating that TMEM216 mutations are the first known identified cause. In the 11 MKS fetuses with TMEM216 mutations, distinctive clinical features were skeletal dysplasia, including intrauterine growth retardation or bowing of the long bones in six fetuses, cleft palate in four, and anencephaly in two (Table 1,​ Supplementary Fig. 5).

TMEM216 is a poorly annotated gene, with RefSeq predicting a protein of just 86 aa, suggesting potential alternative splicing. To characterize this mRNA we performed Northern analysis with a commercial human fetal blot, and found a single major mature isoform at about 1.4 Kb (Supplementary Fig. 6a), agreeing with the predicted 1.3 Kb of the longest representative cDNAs. To interrogate splicing we designed primers complementary to the furthest 5′ and 3′ regions of the known cDNA, and sequenced 48 cloned PCR products from a 20 week gestation human fetal brain library. We identified four major splice isoforms, the longest and most prevalent predicting a protein of 148 aa (Supplementary Fig. 6b), which we consider to be the full-length mRNA. There is also extensive alternative splicing, encoding very short proteins (Supplementary Fig. 6b), the functions of which were not evaluated further. Importantly, we did not find any mutations in any of the putative UTRs.

To elucidate roles for TMEM216 in human development, we first examined its expression in human embryonic tissues. In situ hybridization analysis in human embryos confirmed expression in the central nervous system, limb bud, kidney and cartilage (Supplementary Fig. 6c-h), which is similar to the broad and relatively low-level expression pattern of other JSRD/MKS genes. We next raised an anti-TMEM216 polyclonal affinity-purified antibody against aa 81-90, demonstrated specificity (Supplementary Fig. 7a-b), and immunostained two different ciliated cell lines (inner medullary collecting duct [IMCD3] and retinal pigment epithelium [hRPE]). We observed localization with the base of the primary cilium or adjacent basal body in the majority of cells, as marked by either acetylated or glutamylated tubulin staining (Fig. 2a-d). TMEM216 antibody also reacted strongly to the base of cilia in organs like kidney containing ciliated cells (Fig. 2c,​ Supplementary Fig. 7c), but failed to react with these structures in hTERT-immortalized fetal TMEM216 p.R85X homozygous mutant fibroblasts (Supplementary Fig. 7d). Epitope-tagged TMEM216 showed similar localization to the base of cilia and other microtubule structures (i.e. mitotic spindle in cells undergoing late telophase, Supplementary Fig. 8).

In TMEM216 p.R85X mutant fibroblasts, we noticed a failure in ciliogenesis following 48 hr serum starvation (Fig. 3a) compared with controls. Western analysis of whole cell lysates from control fibroblasts identified a band at 19 kD (Fig. 3b), matching the predicted 148 aa full length protein, whereas this band was attenuated or lost in TMEM216 p.R85X fibroblasts or in IMCD3 cells in which Tmem216 was knockded down.

To determine the basis of the ciliogenesis defect, we performed transient transfection of monolayers of IMCD3 cells with two separate Tmem216 siRNA duplexes. Tmem216 knockdown prevented ciliogenesis in polarized cells, and blocked correct docking of centrosomes at the apical cell surface (Fig. 3c), as seen previously for Meckelin and MKS1¹⁷. This ciliogenesis defect was quantified by comparing the percentage of cells with cilia (defined as > 1 μm length) vs. those without cilia (< 1μm length), and by analyzing the percentage of cells with centrosomes located apical to the nucleus. In cells in which Tmem216 was knocked down, we observed a striking defect in both of these measurements compared with two separate control transfections (Fig. 3d-e, chi-squared test, p<0.001, for 350 cells from each condition), suggesting its requirement in centrosome docking.

The similarities in cellular phenotypes of Mks3 and Tmem216 knockdown, and subcellular localizations of Meckelin and TMEM216, prompted us to ask if the two proteins could interact. Firstly, GFP-tagged TMEM216 was immunoprecipitated with antibodies to either N- or C-terminal portions of Meckelin (Fig. 4a) and, secondly, the reciprocal IP experiment used α-GFP antibody to pull down Meckelin (Fig. 4b). Both assays detected a complex between TMEM216 and Meckelin.

Many aspects of actin-dependent polarized cell behavior, including morphogenetic cell movements¹⁸ and ciliogenesis¹⁹, are mediated by the planar cell polarity (PCP) pathway of non-canonical Wnt signaling²⁰. We therefore first examined RhoA, since the Rho family of small GTPases are key mediators of this pathway^20-21. Consistent with previous results following MKS3 loss²², we found that RhoA signaling was hyperactived in both TMEM216 p.R85X fibroblasts or following Tmem216 knockdown (Fig. 4c-d), despite normal total amounts of RhoA in these cells. Centrosome docking at the apical cell surface is prevented by the interruption of actin remodeling²³, and is dependent on both RhoA activation and regulation by the core PCP protein Dishevelled (Dvl)²⁴. We confirmed that RhoA is localized to the basal body in confluent IMCD3 cells but, following Tmem216 knockdown for 24 hr, RhoA was mislocalized to peripheral regions of the basal body and to basolateral cell-cell contacts (Fig. 4e), consistent with translocation of ectopically-activated RhoA to the cytosol²⁵. Tmem216 knockdown also showed evidence of a mislocalization of γ-tubulin at the centrosome/basal body for this timepoint, which suggests a defect in γ-tubulin nucleation, one of the earliest steps in ciliogenesis²⁶. The established role of RhoA in modulating the actin cytoskeleton in the PCP pathway then led us to evaluate MKS2 patient fibroblast lines for alterations. We found a co-localization of actin stress fibers and the actin cross-linker filamin-A in the cytoplasm of these mutant cells, which was absent in control (Fig. 4f).

We next looked at Dvl signaling in cells, since cilia negatively regulate Dvl activation²⁷, and Dvl mediates Rho activation at the apical surface of ciliated epithelial cells²⁴. We found that loss of TMEM216 increased phosphorylation of Dvl1 (Fig. 5a​ left and right panel), implying that TMEM216 modulates hyper-responsiveness of signaling pathways mediated by Dvl and RhoA. We found that Rho inhibition also increased the Dvl1 phosphorylation in ciliated cells, supporting the existence of feedback mechanism between Rho and Dvl (Fig. 5a, right panel). Unexpectedly, the constitutive Dvl1 phosphorylation associated with TMEM216 loss was blocked by Rho inhibition (Fig. 5a, right panel), suggesting that this loss in ciliated cells can modify the feedback mechanism. Although this possibility warrants further investigation, our data nevertheless suggest a working model in which Dvl1, RhoA and TMEM216 may serve as part of a complex in the pericentrosomal compartment to mediate cellular polarization and centrosomal apical docking. Previous studies have shown that Dvl and Rho contribute to a core framework for regulating the apical docking of centrosomes²⁴, and we also see evidence of a common complex containing TMEM216, Dvl1 and RhoA in TMEM216-transfected cells (Fig. 5b). Since we saw no difference in the localization of Dvl1 following Tmem216 knockdown (Supplementary Fig. 9), these data predict that the hyperactivation of Rho in the absence of TMEM216 might be responsible for the centrosome docking defect at the apical cellular surface. As expected, we found that the impaired centrosome docking caused by Tmem216 knockdown was rescued in a dose-dependent fashion using Rho inhibitor (Fig. 5c).

Meckelin is proposed to regulate centrosomal docking through the RhoA signaling pathway²², and bears similarity to the Frizzled family of transmembrane Wnt receptors¹⁵. Direct evidence of a role for Meckelin in PCP signaling stems from zebrafish embryo morphant phenotypes following morpholino knock-down of mks3²⁸. These included defects in gastrulation movement (a shortened body axis, broad notochords and misshapen somites), which are typical of defects in non-canonical (PCP) Wnt signaling, and have been observed in numerous ciliary and basal body morphants^28-29. We observed identical ciliary phenotypes in tmem216 morphants, which were largely rescued by RNA encoding human TMEM216 (Fig. 5d), and fully rescued by RNA encoding non-targetable zebrafish tmem216 (not shown). We therefore directly compared the tmem216 and mks3 morphant phenotypes in zebrafish, and noted similar defects in both live embryos and in embryros in which pronephric mesoderm, anterior neural structures, adaxial mesodermal cells, and somites were labeled with a krox20, pax2, and myoD riboprobe cocktail (Fig. 5e-f). Quantification demonstrated alteration of convergence to the midline and extension along the AP axis consistent with a PCP defect, although the AP extension defect was more pronounced in the mks3 compared to the tmem216 morphant.

Recent work has implicated the tetraspanin TSPAN12 in the regulation of Norrin signaling by the Wnt receptor Frizzled-4 and coreceptor LRP5¹³. We therefore speculate that TMEM216, a novel tetraspan protein, forms a non-canonical Wnt receptor-coreceptor complex with Meckelin. Our data support a role for both proteins in mediating PCP signaling through the RhoA pathway to cause actin cytoskeleton rearrangements, although whether Rho functions upstream or downstream of Dvl1 remains to be determined. In apical regions of the cell, such actin reorganization would be an essential step before the centrosome/basal body could dock correctly and initiate ciliogenesis. The identification of mutations in TMEM216 as a cause of JSRD and MKS therefore further emphasizes the interrelationship between cell polarity, cellular morphogenesis and signal transduction pathways.