An additional 25 consanguineous families with JS were analyzed, and three (MTI-134, MTI-610, and MTI-627, of Emirate, Turkish and Egyptian descent) showed at least one peak that overlapped with JBTS1. Two Italian families (COR-10 and COR-21) also linked to JBTS1 ⁴. Haplotype analysis suggested that MTI-007 and MTI-134 were identical-by-descent, as were COR-10 and COR-21 (data not shown). Together these analyses defined a candidate interval of 3.5 MB in distal 9q34.2-tel, containing approximately 86 annotated candidate genes (Supplemental Table 3), 57 of which were screened for sequence changes, whereas the remaining 29 either lacked an open reading frame or were excluded due to lack of developmental expression. This extensive screening identified the INPP5E gene as mutated in each JBTS1-linked family (Fig. 1a-b).

The INPP5E mutations were notable for two reasons. First, all identified mutations were amino acid transversions that clustered within the enzymatically-active phosphatase domain (Fig. 1b). Second, each altered the charge of highly basic evolutionarily conserved amino acid residues (Fig. 1c), together suggesting that these mutations might alter enzymatic activity. None of these mutations were encountered among 188 control chromosomes from healthy ethnically matched individuals, and all segregated with the disease in each family. We modeled these mutations using the published crystal structure of synaptojanin⁵, the only inositol 5-phosphatase for which structural data is available. We found that each of the altered residues (with one exception, R563) was predicted to project its charged tail toward the presumed binding pocket of the PtdIns substrate (Supplemental Fig. 2), suggesting that these mutations might alter substrate specificity.

To test the effects of INPP5E mutations on PtdIns phosphatase activity, we immunoprecipated tagged wildtype and enzymatically null (D477N ⁶), together with each mutation separately from mammalian cells (resulting in similar protein levels, Supplemental Fig. 3) and then analyzed phosphatase activity against PtdIns(3,4,5)P3 and PtdIns(4,5)P2, the presumed cellular INPP5E substrates (Fig. 2a). We found that each of the JBTS1 missense mutations severely disrupted phosphatase activity towards PtdIns(3,4,5)P3 (Fig. 2b, p < 0.05 activity for each mutant, N = 3 independent experiments). Of note, since the R512W/R515W variants were observed on a single haplotype, we reasoned that probably only one was driving the defect in enzymatic activity, and under separate analysis, the R515W mutation was shown to be the major contributor to the defective enzymatic activity (data not shown). Comparison with phosphatase activity using PtdIns(4,5)P2 as substrate showed similar although not as severely compromised results as compared with wildtype (Fig. 2c), suggesting that PtdIns(3,4,5)P3 may be the relevant substrate in JBTS1. Because overexpression of INPP5E was previously shown to block Akt phosphorylation in response to PDGF stimulation in cultured cells (presumably by depleting levels of PtdIns(3,4,5)P3 and PtdIns(4,5)P2) ⁷, we next tested the effect of overexpression of each of the patient mutations in this published assay. Not only did patient mutations fail to block Akt signaling, but elevated basal levels of pAkt were also apparent in unstimulated cells (Supplemental Fig. 4). We conclude that the JBTS1-associated missense mutations impair phosphatase activity towards putative PtdIns substrates, which can alter downstream signaling events.

Next, we tested whether the absence of endogenous catalytic INPP5E activity results in an altered PtdIns profiles in cells. Although difficult to measure absolute levels PtdIns moieties in cells, the model predicted an increased ratio of PtdIns(4,5)P2/PtdIns(4)P due to a block in this enzymatic step (Fig. 2a), and thus skin fibroblasts from MTI-610 patient 1 and 2 (brother and sister) were assessed for PtdIns(4,5)P2/PtdIns(4)P ratio. In control unstimulated growth-arrested fibroblasts we found a PtdIns(4,5)P2/PtdIns(4)P ratio of about 0.63, whereas in both patient fibroblast samples the ratio was increased by about 20% (p = 0.05, N = 3 independent experiments, Fig. 2c). We found no differences in expression levels of INPP5E in this or other patient fibroblast samples with INPP5E mutations (Supplemental Fig. 5). We conclude that catalytic activity of INPP5E is required for maintenance of homeostatic ratios of PtdIns in cells.

In order to investigate a possible connection with ciliopathies, we examined INPP5E localization in cells with well-described cilia. The RPE-hTERT line (immortalized retinal pigment epithelia) display a primary cilia in >80% of growth-arrested cells ^8,9. Coimmunostaining with pericentrin, a marker of the pericentriolar matrix, showed INPP5E immediately adjacent to the base of the cilium (Fig. 3a). This was confirmed by INPP5E coimmunostaining with acetylated tubulin, a marker of the ciliary axoneme, as well as with ARL13B, a cilia marker ^10-12, which showed tight-colocalization of INPP5E to the ciliary axoneme, with only minimal INPP5E localized to non-ciliary locations (Fig. 3b). Transfection of wild-type GFP-tagged INPP5E into IMCD-3 cells under ciliary-permissive conditions demonstrated ciliary localization (Supplemental Fig. 6), and we found no defect in ciliary localization of any of the JBTS1-associated patient missense mutations when expressed in these cells (data not shown). By contrast, cells at mitotic stages (after cilium withdrawal) showed cytoplasmic localization (Supplemental Fig. 7). The data suggests that INPP5E localizes predominantly to the primary cilium during interphase in the cell lines tested.

We stained for INPP5E in organs affected in JS from the GFP-Centrin2 transgenic mouse (marks the base of the cilium) ¹³, and identified cilia localization in each. In renal collecting tubules, cilia were evident projecting from the epithelium into the lumen. In this mouse, many cells were marked by luminal localization of the pair of centrioles, and from this site, the majority of cells had an INPP5E-positive cilium projecting into the lumen (Fig. 3c). In the developing cerebellum, cilia have been identified on a population of neuroblasts in the cerebellar internal granule layer ¹⁴. We identified these ciliated cells based upon GFP signal at the centrioles, and identified INPP5E-positive cilia projecting into the parenchyma in the vast majority of these cells (Fig. 3d), and similar ciliary localization was observed in the hepatic bile ducts (not shown). We also observed that INPP5E localized to the region adjacent to the basal body and connecting cilium in retinal photoreceptor cells (Fig. 3e-f). Therefore, INPP5E is predominantly localized adjacent to or within the cilium in each of the major organs affected in JS.

To assess the potential function of INPP5E at the primary cilium, we utilized primary fibroblasts from patient MTI-610-V-1. These cells displayed essentially normal growth characteristics, as well as percentage of ciliated cells compared with controls (81% vs. 76%, p > 0.05, Fig. 4a), suggesting that INPP5E enzymatic activity is not required for ciliogenesis or cilia maintenance, at least not in fibroblasts. However, the cilium is not a fixed structure, and is labile to the application of serum-derived factors such as PDGF ¹⁵. Since serum is a potent activator of the PtdIns pathway in cultured cells, we predicted that the altered PtdIns profile might render cilia in mutant cells more labile to the application of serum. We applied serum to control and MTI-610-V-1 cells and scored the percentage of cells retaining a cilium, based upon acetylated-tubulin and ARL13B staining at specific time intervals. By 4 hr after serum application, wild-type cells had an 8% cilia withdrawal rate whereas INPP5E mutant cells had a 33% cilia withdrawal rate, (73% vs. 43%, p < 0.01, N = 3 independent experiments). At the 8 hr and 24 hr timepoints, the percentage of cells with visible cilia had equalized to about 40% in the two genotypes, suggesting that INPP5E controls timing of cilia disassembly during early response to serum.

To determine if the reduced number of ciliated mutant cells is due to an enhanced rate of cilia withdrawal in the absence of functional INPP5E, we performed scanning electron microscopy. Prior to the addition of serum, we found that the length of the primary cilium was approximately the same in control vs. MTI-610-V-1 serum-starved fibroblasts, measuring ~3 um in length (Fig. 4b-e). At the 4 hr timepoint (where we saw the maximal difference in cilia number), however, cilium length of the MTI-610 fibroblasts was about half of control. These data were quantified in x-y sections from cells with remaining visible cilia as evidenced by immunostaining for ARL13B prior to, and then at 4 hr and 8 hr after serum addition. We found a significant decrease in average cilia length at 4 hr after serum stimulation in mutant vs. controls (2.73 vs. 1.28 um, wildtype vs. MTI-610, p < 0.01, n > 30 cells for each condition), whereas no difference was apparent either before serum or after 8 hrs (Supplemental Fig. 8). Thus ciliary disassembly is more rapid in the absence of full INPP5E activity compared with wildtype.

Following ascertainment of fibroblast samples from additional patients with INPP5E mutations, we noted some alterations in cell proliferation rates in culture, and thus tested quiescent cells for ability to reenter the cell cycle. While MTI-610 fibroblasts had no apparent alteration in cell cycle reentry (based upon Ki67 staining or BrdU incorporation), MTI-134 and MTI-627 exhibited a reduction in the percentage of mitotically active cells following serum addition (Supplemental Fig. 9). As MTI-134 and MTI-627 had more severe reduction in enzymatic activity than MTI-610 in the in vitro assay (0% and 8.5% vs.15%, presented in Fig. 2), the data suggests that residual but not full function of INPP5E may be required for this serum-based cell-cycle reentry. MTI-610 fibroblasts exhibited a normal percentage of mitotically active cells, however, a greater percentage of these mitotically synchronized cells reentered the cell cycle initially (Ki67 staining between 18-24hrs) and progressed through the cell cycle more rapidly (reduction of BduU incorporation from 24hrs to 30hrs) than control fibroblasts, suggesting that INPP5E enzymatic activity may also influence the rate of cell cycle progression. Not all serum-based responses are abnormal in these patient samples, however. For instance, there were no alterations in low-dose serum-dependent scratch-wound assay healing (Supplemental Fig. 10), which we designed to be migration-dependent but largely proliferation-independent ¹⁶. We conclude that INPP5E may play specific serum-dependent cellular roles, mediating both ciliary stability as well as cell cycle dynamics.

Our data demonstrates that hypomorphic mutations in INPP5E lead to JS, providing the first evidence of a link between PtdIns signaling and the ciliopathies. It will be important to determine if the developmental defects observed in JS are the result of defects in cilia maintenance or due to a more broadly defined disruption in PtdIns signaling. Since PDGFαα receptor has been localized to primary cilia ¹⁷, and since PDGF is one of the key serum-derived factors that mediates cilia withdrawal concurrent with cell cycle reentry ¹⁸ it is possible that PDGF and other factors that utilize cilia as signaling centers have downstream pathways that are regulated through specific localization and function of INPP5E.

Previous studies have hinted at possible connections between PtdIns signaling and the cilium. Tubby mutant mice have cilia-like phenotypes and the protein encoded from the mutant gene binds specific PtdIns including PtdIns3,4,5(P3) ¹⁹. Furthermore, Bardet-Biedl syndrome shares similar ciliopathy features to JS, and the BBS5 protein can bind PtdIns ⁹. However, our findings are the first, to our knowledge, to directly implicate defective enzymatic conversion within the PtdIns class. The fact that INPP5E displays enrichment within the ciliary axoneme makes it tempting to speculate that PtdIns signaling may not just function to regulate cilia stability, but also that this signaling may occur within the cilium. An alternative model is that INPP5E might be sequestered within the cilium, but exert its activity within the cytoplasm, similar to the models being proposed for Sonic hedgehog signaling ²⁰. The identification of a key regulatory enzyme in this process highlights an important area of research with clinical relevance.

1