Abstract
A functional primary cilium is a crucial cell appendage which is essential for normal, regulated signalling, and loss of the primary cilium is implicated in a suite of severe developmental conditions known as ciliopathies. The mechanisms of signal regulation by the cilium remain unclear. Previous studies have suggested links between the primary cilium/basal body, the ubiquitin proteasome system (UPS) and Wnt signalling. Here we provide further mechanistic insights into these processes in vivo by crossing the Mks1−/− knockout mouse with the UbG76V-GFP reporter line. We demonstrate in vivo that MKS1 is essential for normal proteasomal processing of ubiquitinated proteins, but that this only manifests as accumulation of ubiquitinated proteins when the proteasome is inhibited. We show that an increase in proteasomal enzymatic activity and Wnt signalling de-regulation in the absence of MKS1. Yeast 2-hybrid demonstrate that the UPS component UBE2E1, an E2 ubiquitin-conjugating enzyme which polyubiquitinates β-catenin, interacts with MKS1. Levels of UBE2E1 and MKS1 are co-dependent, and loss of UBE2E1 recapitulates the ciliary and Wnt signalling phenotypes observed during loss of MKS1, suggesting a functional association between the two proteins. We suggest that MKS1 regulates UBE2E1 and other UPS components at the base of the cilium, which leads to proteasomal and canonical Wnt signalling dysregulation. These findings provide further mechanistic detail of the interaction between the basal body and the UPS in regulating signal transduction through β-catenin, and confirm that the UPS plays a central role in the molecular pathogenesis of ciliopathies.
Introduction
Primary cilia are microtubule-based organelles that sense and transduce extracellular signals on many mammalian cell types during the G1/G0 phase of the cell cycle. The cilium is known to play essential roles throughout development in mechanosensation 1,2, in signal transduction by the Sonic Hedgehog (Shh), Wnt and PDGFRα signalling pathways 3–5 and in the establishment of left-right asymmetry 6. Primary cilia have a complex ultrastructure with compartmentalization of molecular components that combine in functional modules, and mutations in proteins that are structural or functional components of the primary cilium cause a suite of human inherited developmental conditions known as ciliopathies 7. Examples of ciliopathies include Meckel-Gruber syndrome (MKS) and Joubert syndrome (JBTS). Many proteins that are mutated in ciliopathies are thought to localize to the transition zone, a compartment of the proximal region of the cilium, including the MKS1 protein 8,9. Mutations in the MKS1 gene account for about 15% of all cases of Meckel-Gruber syndrome (MKS), a lethal neurodevelopmental condition that is the most severe ciliopathy 10.
The MKS1 ciliopathy protein 11, contains a B9/C2 domain with homologies to the C2 (calcium/lipid-binding) domain of the synaptotagmin-like and phospholipase families. MKS1 interacts with TMEM67/meckelin, the receptor encoded by the TMEM67 gene 12, and two other B9/C2-domain containing proteins, B9D1 and B9D2 (26638075). The B9/C2-domain proteins, including MKS1, are predicted to bind lipids in the ciliary membrane, and to all co-localise to the ciliary transition zone 13 forming components of a functional module (known as the “MKS-JBTS module”) at the transition zone. These include transmembrane proteins (TMEMs), namely the Tectonic proteins (TCTN1-3), TMEM17, TMEM67/meckelin, TMEM231 and TMEM237, as well as other C2-domain proteins (jouberin/AHI1, RPGRIP1L and CC2D2A) 14–16. Transition zone proteins are thought to form a diffusion barrier at the base of the cilium that restricts entrance and exit of both membrane and soluble proteins 17. The compartmentalization of the cilium is essential for the regulated translocation of signalling intermediates, most notably during Shh signalling 18, and mutations of transition zone components invariably cause Shh signalling defects during development 19. For example, mouse embryos from the Mks1Krc knock-out mutant line have severe Shh signalling and left-right patterning defects during early embryonic development 19. However, we have previously described the Mks1tm1a(EUCOMM)Wtsi knock-out mouse line, for which mutant embryos have additional de-regulated canonical Wnt/β-catenin signalling and increased proliferation defects in the cerebellar vermis and kidney 20.
Other studies have shown that the ciliary apparatus restricts the activity of canonical Wnt/β-catenin signalling 4,21,22, although the mechanistic detail by which signal transduction is regulated remains unclear. One regulatory pathway involves the ciliary transition zone protein jouberin/AHI1, which shuttles β-catenin between the cytosol and nucleus in order to regulate Wnt signalling 23. However, ubiquitin-dependent proteasomal degradation by the ubiquitin-proteasome system (UPS) is the best-characterized as a mechanism for regulating canonical Wnt signalling 24. In the absence of a Wnt signal, cytoplasmic β-catenin is phosphorylated in a complex of proteins (often known as the destruction complex) that include axin, adenomatous polyposis coli (APC) protein and glycogen synthase kinase 3 (GSK-3)25–27. Subsequent ubiquitination of β-catenin leads to its degradation by the proteasome, meaning that in the absence of Wnt signalling the steady state levels of cytoplasmic β-catenin are low. Part of this regulation appears to be mediated by a functional association of the ciliary apparatus with the UPS 28, and UPS components have been shown to interact with ciliopathy proteins, such as USP9X with lebercilin 29. RPGRIP1L (a ciliary transition zone protein mutated in a range of ciliopathies including MKS and JBTS) has been reported to interact with the proteasome proteins, PSMD3 and PSMD5 30. Furthermore, discrete localization of ubiquitin has been observed at the ciliary base suggesting that UPS processing can be constrained and regulated by the cilium 30. However, the mechanistic basis to substantiate the association between the UPS and ciliary apparatus remains unclear and, in particular, it is unknown if the pathomechanism of Wnt signalling defects in ciliopathies depends on defective regulation of β-catenin localization and processing by ciliary proteins.
Here, we describe the interaction and functional association of the MKS1 ciliopathy protein with components of the ubiquitination pathway, specifically the E2 ubiquitin conjugated enzyme UBE2E1 (also known as UbcH6) and the E3 ubiquitin ligating enzyme RNF34. In addition to ciliogenesis defects, MKS1 mutation or loss also causes deregulation of both proteasome activity and canonical Wnt/β-catenin signalling. These cellular phenotypes are reiterated after loss of UBE2E1. MKS1 and UBE2E1 colocalize during conditions of cilia resorption, and levels of MKS1 and UBE2E1 are co-dependent. We show that in the absence of MKS1, levels of ubiquitinated UBE2E1 and other polyubiquitinated proteins, including β-catenin, are increased. Furthermore, polyubiquitination of MKS1 is dependent on both UBE2E1 and RNF34, and lysine63-linked polyubiquitination of MKS1 is dependent on UBE2E1. This suggests that regulation of intracellular signalling, specifically canonical Wnt/β-catenin signalling, can be regulated and constrained at the primary cilium by MKS1. It provides a mechanistic explanation for Wnt signalling defects in ciliopathies and highlights new potential targets in the UPS for therapeutic intervention.
Results
MKS1 mutation causes deregulation of both proteasome activity and canonical Wnt/β-catenin signalling
Loss of ciliary basal body proteins perturbs both UPS function and Wnt signalling 28, and we have previously reported de-regulation of canonical Wnt signalling in Mks1−/− mutant mice 20. To investigate the mechanistic basis for regulation of canonical Wnt/β-catenin signalling and UPS processing of β-catenin by a ciliary protein, we first characterized these processes in cells and tissues lacking functional MKS1. We derived immortalized dermal fibroblasts from a human MKS patient, carrying compound heterozygous MKS1 mutations [c.472C>T]+[IVS15-7_35del29] causing the predicted nonsense and splice-site null mutations [p.R158*]+[p.P470fs*562] 10 (Supplementary Figure 1a) leading to loss of MKS1 protein (Supplementary figure 1b-c). MKS1-mutated fibroblasts had decreased cilia incidence and length (Supplementary figure 1d), and de-regulated canonical Wnt/β-catenin signalling (Figure 1a). MKS1-mutated fibroblasts had moderately increased levels of total β-catenin and the Wnt downstream target cyclin D1 (Figure 1a). SUPER-TOPFlash reporter assays confirmed that the increased levels of β-catenin in MKS1-mutated fibroblasts were reflected by de-regulation of canonical Wnt signalling in response to Wnt3a (a canonical Wnt ligand; Figure 1b). Treatment with the non-specific proteasome inhibitor MG-132 also increased levels of phosphorylated β-catenin (Figure 1a). Since β-catenin is phosphorylated to mark it for processing by the 26S proteasome, we also tested if proteasome enzymatic activity was affected in MKS1-mutated fibroblasts. We observed de-regulated proteasome activity, which was inhibited by treatment with lactacystin that targets the 20S catalytic core of the proteasome, as well as moderate increased levels of the proteasome subunit α7 (Figure 1c).
To substantiate an in vivo association between de-regulated canonical Wnt signalling and proteasome activity in the ciliopathy disease state, we crossed the Mks1tm1a(EUCOMM)Wtsi knock-out mouse line 20 with the UbG78V-GFP transgenic reporter line. UbG78V-GFP constitutively degrades transgenic ubiquitinated-GFP, leading to an absence of GFP signal if proteasome processing is unimpaired 31. Confirming our observations with human MKS1-mutated fibroblasts, Mks1−/− × UbG76V-GFP mouse embryonic fibroblasts (MEFs) also had de-regulated proteasome enzymatic activity that was inhibited by treatment with lactacystin (Figure 1e) compared to Mks1+/+ × UbG676V-GFP wild-type littermate MEFs. Furthermore, Mks1−/− × UbG676V-GFP mutant embryos at embryonic day E12.5 had increased levels of GFP, detected by both epifluorescence confocal microscopy and western blotting, in the neocortex (Figure 1d), heart and liver (Supplementary Figure 2a-b) compared to wild-type littermate controls after treatment with MG262, proteasome inhibitor. Autofluorescence was virtually undetectable in these tissues (data not shown). The accumulation of GFP in mutant embryonic tissues indicated that constitutively expressed GFP-tagged ubiquitin was not properly degraded, suggesting that these tissues harboured proteasomal defects. These defects accompanied an absence of ciliogenesis and increased levels of active β-catenin in the neuroepithelium of Mks1−/− × UbG76V-GFP mutant ventricular zone (Supplementary Figure 2b).
MKS1 interacts with the E2 ubiquitin-conjugation enzyme UBE2E1, with colocalization during cilia resorption
To understand why mutation or loss of MKS1 causes deregulation of both proteasome activity and canonical Wnt/β-catenin signalling, we sought to identify interacting proteins of MKS1. We performed a yeast two-hybrid screen using amino acids 144-470 of MKS1 that contain the B9/C2 domain as bait (Figure 2a). We identified the E2 ubiquitin conjugation enzyme UBE2E1 (also known as UbcH6) 32 as an interactant of MKS1 (Figure 2b), and confirmed this interaction by a “one-to-one” yeast two-hybrid assay (Figure 2c). We also identified the E3 ubiquitin ligated enzyme RNF34 and confirmed its interaction and colocalization with MKS1 (Figure 2b, Supplementary Figure 3a-b). In support of a possible role of MKS1 in the targeting of ubiquitinated signalling proteins, UBE2E1 has already been shown to function as an E2 with the E3 JADE1 during the ubiquitination of β-catenin 33. We therefore further substantiated the interaction of UBE2E1 with MKS1. We purified GST-tagged UBE2E1, and confirmed expression by SDS-PAGE analysis (Figure 2d). We used the GST-UBE2E1 to confirm the interaction between MKS1 and UBE2E1 by a GST pulldown assay (Figure 2d). The interaction between endogenous MKS1 and UBE2E1 was confirmed by co-immunoprecipitations (co-IPs) using anti-MKS1 (Figure 2e). This was further confirmed when an interaction between endogenous UBE2E1 and exogenously expressed cmyc tagged MKS1 was detected by co-IP with an anti-UBE2E1 antibody (Figure 2f).
UBE2E1 and MKS1 co-localized at the basal body in a subset of confluent, ciliated hTERT-immortalized retinal pigment epithelium RPE1 and ARPE19 cells during G0 of the cell cycle following serum starvation for 48hr 34 (Figure 3a-d). Serum starvation followed by re-addition of serum for 3hr caused rapid cilia resorption 34, with further significant colocalization of UBE2E1 and MKS1 at the basal body (Figure 3b and 3d). This suggests that the interaction between MKS1 and UBE2E1 is particularly important during the process of cilia resorption. Furthermore, unlike in normal wild-type cells, UBE2E1 co-localised with γ-tubulin at the basal body in MKS1-mutated fibroblasts (data not shown). This suggests that in normal cell conditions MKS1 excludes UBE2E1 from basal body localisation.
Mutation or loss of UBE2E1 causes ciliogenesis defects, deregulated proteasome activity and deregulated canonical Wnt/β-catenin signalling
Since correct UPS function appears to be required for ciliogenesis 30,40,41, we next asked if loss or mutation of UBE2E1 had an effect on ciliogenesis. UBE2E1 is an enzyme that transfers ubiquitin to a substrate, with or without the presence of an E3, in a reaction that is dependent on an active enzymatic domain. To assess if enzymatic activity of UBE2E1 is necessary for correct ciliogenesis, we mutated the active site cysteine residue 131 to serine 35 to make a dominant negative (DN), catalytically inactive form of UBE2E1. This cysteine residue is essential for E2 function as this forms a thioester bond with the carboxyl terminus of ubiquitin. Over-expression of the DN C131S mutated form of UBE2E1 caused significant loss and shortening of cilia in mouse inner medullary collecting duct (mIMCD3) cells (Figure 4a), suggesting that catalytically active UBE2E1 is required for normal ciliogenesis. Over-expression of wild-type (WT) UBE2E1 had a moderate dominant negative effect on cilia length only (Figure 4a). To model the effect of UBE2E1 loss on ciliogenesis, we first used both pooled and individual siRNA duplexes targeting Ube2e1 in mIMCD3 cells. This affected ciliogenesis in mIMCD3 cells, reducing cilia incidence and length, but achieved only moderate knockdown of UBE2E1 protein levels (Supplementary Figure 4).
To ensure more robust, long-term knockdown of UBE2E1, we derived stably-transfected mIMCD3 cell-lines with three different Ube2e1 shRNA constructs. Each Ube2e1 shRNA construct reduced UBE2E1 protein levels compared to cells expressing scrambled shRNA, and significantly reduced both numbers of ciliated cells and mean cilium length (Figure 4b). Since UBE2E1 and MKS1 both interact and co-localize, we next determined if UBE2E1 loss reiterates the cellular phenotypes caused by MKS1 loss or mutation. Indeed, we observed increased proteasome enzymatic activity compared to scrambled shRNA (shScr) negative control cells (Figure 4c). Furthermore, in agreement with MKS1-mutated fibroblasts and Mks1−/− MEFs, shUbe2e1 knock-down cells had de-regulated canonical Wnt/β-catenin signalling in response to Wnt3a (Figure 4d). These data highlight the important role of UBE2E1 in mediating correct protein ubiquitination, proteasome function and Wnt signalling in the context of cilia. The striking similarities in ciliary phenotypes suggest a close functional association between MKS1 and UBE2E1, and led us to hypothesize that they are placed in the same signalling pathway.
Co-dependent regulation of MKS1 and UBE2E1 protein levels
To further investigate the functional association between MKS1 and UBE2E1, we tested if de-regulated Wnt signalling could be rescued by exogenous over-expression experiments. In a control experiment, expression of cmyc tagged MKS1 partially rescued normal canonical Wnt signalling responses to Wnt3a in MKS1-mutated fibroblasts (Figure 5a). However, expression of FLAG-tagged UBE2E1 lead to almost complete rescue of normal Wnt signalling responses (Figure 5a). Conversely, expression of MKS1 in shUbe2e1 knock-down cells also rescued normal canonical Wnt signalling (Figure 5b), suggesting co-dependency between MKS1 and UBE2E1. We confirmed this following transient siRNA knockdown of MKS1 (Supplementary Figure 5a-b) that caused significantly increased levels of UBE2E1 in cells (Figure 5c). In the reciprocal experiment, MKS1 protein levels were significantly increased in shUbe2e1 knock-down cells (Figure 5c). To further support co-dependency, we over-expressed both MKS1 and UBE2E1. At higher levels of UBE2E1, we observed a moderate decrease in MKS1 levels (Figure 5d). Conversely, expression of high levels of MKS1 caused a significant decrease in UBE2E1 protein levels (Figure 5d). These results show a striking co-dependency in protein levels between MKS1 and UBE2E1, suggesting inhibitory roles for each of these proteins on the protein level of the other.
MKS1 is polyubiquitinated and its polyubiquitination depends on UBE2E1
UBE2E1 is an E2 ubiquitin conjugating enzyme, and we next tested the obvious hypothesis that it participates in polyubiquitination and targeting degradation of MKS1. We therefore investigated if MKS1 is indeed tagged with ubiquitin chains and if absence of UBE2E1 affects ubiquitination of MKS1. We determined MKS1 levels and its polyubiquitination status at different ciliogenesis conditions, namely: absent cilia (proliferating cells grown in normal medium supplemented with serum); cilia present (quiescent cells grown in serum-deprived medium); and cilia resorption (cells grown in serum-deprived medium, but then grown in normal media for the final two hours). Across three biological replicates, shUbe2e1 knock-down cells consistently had significantly increased levels of MKS1 as well as polyubiquitinated MKS1 (Figure 6a; p<0.05 paired t test between shScr and shUbe2e1). The highest levels of MKS1 were observed in cells undergoing cilia resorption, when MKS1 and UBE2E1 co-localisation is the strongest. Furthermore, expression of exogenous UBE2E1 led to moderate decreases in MKS1 levels for both shScr and shUbe2e1 knock-down cells, suggesting that UBE2E1 inhibits both MKS1 levels and polyubiquitination of this protein.
To substantiate the central role of UBE2E1 in regulating MKS1 levels, we confirmed that shUbe2e1 cells had increased levels of polyubiquitinated cmyc tagged MKS1 using TUBE (Tandem Ubiquitin Entity) assays. Total polyubiquitinated proteins from cell extracts were pulled-down using TUBEs bound to agarose beads, resolved by SDS-PAGE and western blotted with anti-c myc. TUBE assays confirmed that MKS1 was polyubiquitinated (Figure 6b, upper panel). Treatment of pull-downs with either a pan-specific deubiquitinase (DUB) and a DUB specific for lysine63-linked polyubiquitination confirmed that MKS1 polyubiquitination occurred through both lysine63 and other ubiquitin lysine sidechains such as lysine48 (Figure 6b, lower panel). Importantly, this result suggests that ubiquitination of MKS1 has dual functions in targeting this protein for degradation, as well as other regulatory functions through lysine63. Although UBE2E1 could be an E2 in an MKS1 degradation pathway, our data showed that loss of UBE2E1 caused an increase in the levels of polyubiquitinated MKS1 consistent with an inhibitory function for UBE2E1 in ubiquitinating MKS1. To test this alternative hypothesis, we therefore performed in vitro ubiquitination assays in which purified MKS1 was used as a substrate of the reaction, purified UBE2E1 was the E2 and RNF34 was a possible E3. We observed mono- and bi-ubiquitination of MKS1 by UBE2E1 in the presence of RNF34 (Figure 6c), confirming the role of UBE2E1 in regulation of MKS1 but not targeting it for degradation.
MKS1 and UBE2E1 interact to regulate β-catenin ubiquitination
Monoubiquitination by UBE2E1 has been previously described 36 and UBE2E1 has also been shown to be an E2 ubiquitin conjugating enzyme that polyubiquitinates β-catenin 33. These studies suggest that UBE2E1 has dual functions as an E2 in regulating protein function (for example, through monoubiquitination of MKS1) or targeting them for degradation (for example, polyubiquitination of β-catenin). We therefore next asked if the co-dependency of MKS1 and UBE2E1 could regulate levels of β-catenin. We first confirmed that MKS1 and β-catenin interact (Figure 6d and Table 1) and that shUbe2e1 knock-down cells have increased levels of β-catenin (Figure 6d), consistent with the de-regulated canonical Wnt/β-catenin signalling that we observed in these cells (Figure 4d). We reasoned that UBE2E1 could mediate normal levels of polyubiquitination of β-catenin, followed by subsequent targeted degradation, maintaining normal regulated levels of canonical Wnt signalling. In the event of high levels of the E2, caused by absent regulator – MKS1, β-catenin is over-polyubiquitated and its levels increase leading to dysregulation of canonical Wnt signalling (Figure 7). Consistent with the co-dependency of MKS1 and UBE2E1, TUBE pull-down assays confirmed that levels of polyubiquitinated β-catenin increased following MKS1 knockdown (Figure 6e). We also observed increased levels of phosphorylated β-catenin following loss of MKS1 (Figure 1a) and specific localization of phosphorylated β-catenin at the base of the cilium (Figure 6f). Levels of phosphorylated β-catenin were also increased at the base of the cilium in MKS1 knockdown cells, where that form of β-catenin is processed by UBE2E1 for polyubiquitination, confirming our previous findings. These observations suggest that correct β-catenin levels are tightly regulated at the primary cilium by a ciliary-specific E2 and substrate-adaptor, UBE2E1 and MKS1.
Discussion
A number of studies suggest that the primary cilium or basal body constrains canonical Wnt/β-catenin signalling activity 4,21,22, and de-regulated signalling is one of the hallmarks of the ciliopathy disease state. Canonical Wnt/β-catenin signalling is aberrantly up-regulated in several ciliopathy animal models, and, in particular, in postnatal cystic kidneys 37,38. We have shown previously that homozygous Mks1−/− mouse embryos also had de-regulated canonical Wnt signalling, reduced numbers of primary cilia and increased proliferation in the cerebellar vermis and kidney 20. Importantly, transgenic mice over-expressing an activated mutant form of β-catenin develop polycystic kidney disease early in development 39, suggesting that de-regulated Wnt/β-catenin signalling is linked to increased cellular proliferation and kidney cyst formation. The mechanistic detail of Wnt signalling de-regulation in ciliopathies remains unclear and controversial. A key question remains whether this ciliary signalling defect is a secondary consequence of cilia loss, or if it is directly and causally related to the loss of function of specific cilium proteins. Several studies support the latter hypothesis, including one study that suggests that jouberin, a component of the transition zone/basal body, may modulate Wnt/β-catenin signalling by facilitating nuclear translocation of β-catenin in response to Wnt stimulation 38. Regulation of Wnt signalling appears to be also mediated by a functional association of the basal body with the UPS 28, through which signalling pathway components such as β-catenin are degraded 24. Early studies showed that the basal body and the proteasome could colocalise 40,41 and more recently, normal, regulated Wnt signalling has been shown to be dependent on the interaction of the basal body protein BBS4 with RPN10, a component of the proteasome 28. Rpgrip1l−/− knock-out mice had decreased proteasome activity and Psmd2, a component of the proteasome, was identified as an interacting partner of Rpgrip1l 30. Furthermore, mutations in a number of genes encoding UPS proteins have been shown to be causative for ciliopathies 42, supporting a specific role for MKS1 in UPS-mediated proteostasis and signalling regulation.
As expected from previous studies, we demonstrated that loss of MKS1 caused aberrant accumulation of β-catenin (Figure 1a). However, proteasome activity was also de-regulated and aberrantly increased (Figure 1c-d). A possible explanation for these observations is that increased levels of β-catenin in the absence of MKS1 are normally compensated by ubiquitination and proteasomal destruction. However, under conditions of proteasome inhibition, or during early embryonic development when canonical Wnt/β-catenin signalling is activated 42, the increased levels of β-catenin are up-regulated or de-regulated to such an extent that the proteasome is unable to compensate for this increase. This hypothesis provided the conceptual basis for our further studies to understand the possible pathomechanisms of de-regulated Wnt/β-catenin signalling in the ciliopathy disease state.
The increase in proteasomal activity may be a non-specific response to cellular stress in the absence of MKS1, but our discovery and validation of direct interactions of MKS1 with two proteins (UBE23E1 and RNF34) in the ubiquitination cascade suggest that loss of MKS1 causes a more specific defect. We confirmed biochemical and functional interactions of MKS1 with both UBE2E1 and RNF34 through a series of yeast two-hybrid “one-to-one” assays (Figure 2c), co-immunoprecipitations (Figure 2e-f, Suppl. Figure 3), co-localisations (Figure 3, Suppl. Figure 3) and in vitro ubiquitination assays (Figure 6c). Furthermore, we demonstrated a functional interaction between UBE2E1, MKS1 and β-catenin (Figure 6d and e) and we showed that processing of phosphorylated β-catenin happens at the base of the cilium (Figure 6f).
Our data therefore suggests that MKS1 interacts with UPS components and β-catenin to regulate levels of β-catenin through normal degradation during Wnt signalling. We suggest that MKS1 could mediate the degradation of β-catenin by stabilising the localisation of UBE2E1 at the ciliary apparatus, and perhaps ensuring the correct processing of ubiquitinated β-catenin through close proximity to the proteasome at the ciliary base. This suggestion is supported by the biochemical interaction of Rpgrip1l with proteasome proteins and the discrete localization of ubiquitin at the ciliary base 30. Furthermore, we also show that catalytically active UBE2E1 can regulate ciliogenesis (Figure 6c), which implies that correct UPS function is required for ciliogenesis, a conclusion supported by previous studies 41,42. Since MKS1 contains a predicted lipid-binding B9/C2 domain, MKS1 may act as a membrane anchor to ensure the spatial organization and co-ordinated regulation of both the β-catenin destruction complex 21 and UPS components at the ciliary apparatus. Loss of MKS1 would lead to the disruption of both the structure and function of the transition zone, preventing regulated ciliary signalling and β-catenin degradation.
In summary, our results suggest that the MKS1-UBE2E1 complex plays a key role in the degradation of β-catenin, which in turn facilitates correct cell function and signalling (Figure 7). This offers new insights into the role of the primary cilium in the regulation of Wnt signalling. In our model, β-catenin translocates to the ciliary apparatus where UBE2E1 (and the cognate E3, either JADE-1 or RNF34) ubiquitinate it. We propose that MKS1 has a functional and organizational role at the basal body in ensuring that correctly ubiquitinated β-catenin is in the proximity of the proteasome for degradation. In the absence of MKS1, the defect in proteasomal targeting is partly compensated by increased proteasomal activity, but if proteasomal activity is inhibited, ubiquitinated proteins (including β-catenin) accumulate in the cell. Our data also confirm and extend the existing evidence that the ciliary apparatus and proteasome components biochemically interact and are functionally associated 28,43 during the regulation of signalling pathways 24,44,45,41,42,41,39,40This may have broader effects on other developmental signalling pathways, and UPS defects may be a fundamental pathogenic cause of the ciliopathy disease state. For example, Suppressor-of-fused (Sufu) is an essential negative regulator of the Sonic Hedgehog (Shh) pathway that is degraded by the UPS in a specific ciliary-dependent mechanism 46. The de-regulation of ciliary-mediated Shh signalling by UPS defects is a clear prediction, and the investigation of UPS components in these novel signalling roles provides an interesting and exciting new field of future research in ciliary biology.
Materials and methods
Informed consent for use of patients in research
Informed consent was obtained from all participating families or patients, with studies approved by the Leeds (East) Research Ethics Committee (REC ref. no. 08/H1306/85) on 4th July 2008.
Animals
The animal studies described in this paper were carried out under the guidance issued by the Medical Research Council in Responsibility in the Use of Animals for Medical Research (July 1993) in accordance with UK Home Office regulations under the Project Licence no. PPL40/3349. B6;129P2-Mks1 tm1a(EUCOMM)Wtsi heterozygous knock-out mice were derived from a line generated by the Wellcome Trust Sanger Institute and made available from MRC Harwell through the European Mutant Mouse Archive http://www.emmanet.org/ (strain number EM:05429). The UbG76V-GFP line (25) B6.Cg-Tg(ACTB-Ub*G76V/GFP)1Dant/J (strain number 008111) was obtained from the Jackson Laboratory, Maine, USA. Genotyping was done by multiplex PCR on DNA extracted from tail tips or the yolk sac of E11.5-E15.5 embryos, or ear biopsies of adult mice. Primer sequences are available on request. Proteasome inhibition treatment of Mks1 × UbG76V-GFP mice using MG262 was carried out as previously described 31.
Preparation of tissue sections
Mouse embryos or tissue for IF staining were lightly fixed in 0.4% paraformaldehyde, soaked in 30% sucrose/PBS, frozen in OCT embedding medium and cut into 5μm sections on a cryostat. Fresh-frozen sections were left unfixed and processed for immunofluorescent staining by standard techniques.
Cells
Mouse inner medullary collecting duct (mIMCD3), human retinal pigment epithelium cells immortalized with human telomerase reverse transcriptase (hTERT-RPE1) and immortalized adult retinal pigment epithelium (ARPE19) cells were grown in Dulbecco’s minimum essential medium (DMEM)/Ham’s F12 supplemented with 10% foetal calf serum at 37°C/5% CO2. Human embryonic kidney (HEK293) cells were cultured in DMEM with 10% foetal calf serum at 37°C/5% CO2. The derivation and culture of mouse embryonic fibroblasts (MEFs) has been described previously 47. MEFs were grown in DMEM/Ham’s F12 supplemented with 10% foetal calf serum and 1% penicillin streptomycin at 37°C/5% CO2. Fibroblasts from a normal undiseased control, a patient (MKS-562) with a compound heterozygous MKS1 mutation, and a female patient with a homozygous ASPM mutation, were immortalised following transduction with an amphotropic retrovirus encoding the hTERT catalytic subunit of human telomerase, and maintained in Fibroblast Growth Medium (Genlantis Inc. San Diego, CA) supplemented with 0.2 mg/ml geneticin. Patient MKS-562, a compound heterozygote for the MKS1 mutations [c.472C>T]+[IVS15-7_35del29] causing the predicted nonsense and splice-site mutations [p.R158*]+[p.P470fs*562], has been described previously 10. Proteasome inhibition treatment was carried out using 10μM final concentration of the inhibitor dissolved in DMSO for 16 hours (unless otherwise stated). DMSO was used as the vehicle-only negative control.
Cloning, plasmid constructs and transfections
Human MKS1 was cloned into the pCMV-cmyc vector as described previously 48. The pGEX5X-1-UBE2E1 and pCMV-UBE2E1-FLAG-cmyc constructs have been described previously 32. The c.341T>A, p.C131S active site dominant negative (DN) missense mutation was introduced into pCMV-UBE2E1-FLAG-cmyc using the QuickChange mutagenesis kit (Stratagene Inc.) and verified by DNA sequencing. For transfection with plasmids, cells at 80% confluency were transfected using Lipofectamine 2000 (Invitrogen Inc.) according to the manufacturer’s instructions and as described previously 48. Cells transfected with plasmids expressing Ube2e1 shRNA (Origene) were selected for using 0.5μg/ml puromycin for 5 passages. Transfection with Dharmacon ON-TARGET PLUS siRNAs was carried out using Lipofectamine RNAiMAX according to the manufacturer’s instructions and as described previously 48. To assess co-dependency of protein levels, 1μg of cmyc-MKS1 was co-transfected with 1, 2.5 and 5μg of FLAG-cmyc-UBE2E1. To investigate if an increased amount of MKS1 would have an effect on UBE2E1 levels, 3μg of cmyc-MKS1 were co-transfected with 1μg FLAG-cmyc-UBE2E1. After 24hr incubation with transfection complexes, cells were treated with 100μg/ml cycloheximide for 4hr. Ubiquitination of cmyc-MKS1 in mIMCD3 cells was assessed after treatment with proteasome inhibitor (MG132 at 10μM) for 3hr.
Antibodies
The following primary antibodies were used: mouse anti-cmyc clone 9E10, mouse anti-acetylated-α-tubulin clone 6-11B-1, mouse anti-HA (Sigma-Aldrich Co. Ltd.), rabbit anti-GFP (“Living Colors A.v. Peptide Antibody”) and mouse anti-UBE2E1 (BD Biosciences Inc.); rabbit-anti-γ-tubulin and mouse anti-β-actin clone AC-15 (Abcam Ltd.); mouse anti-cyclin D1 clone A-12 (Santa Cruz Biotechnology Inc.); rabbit anti-phospho-β-catenin and rabbit anti-β-catenin (Cell Signalling Technology Inc.); and mouse anti-mono- and polyubiquitinylated conjugates clone FK2 and rabbit anti-20S proteasome α7 subunit (Enzo Life Sciences Inc.). Rabbit anti-MKS1 has been described previously 49,50. Secondary antibodies were AlexaFluor488-, and AlexaFluor568-conjugated goat anti-mouse IgG and goat anti-rabbit IgG (Molecular Probes Inc.) and HRP-conjugated goat anti-mouse immunoglobulins and goat anti-rabbit immunoglobulins (Dako Inc.).
Immunofluorescence and confocal microscopy
Cells were seeded at 1.5 × 105 cells/well on glass coverslips in six-well plates, 24 hours before transfection and 48-96 hours before fixation. Cells were fixed in ice-cold methanol (5 minutes at 4°C) or 2% paraformaldehyde (20 minutes at room temperature). Permeabilization, blocking methods and immunofluorescence staining were essentially as described previously 51. Confocal images were obtained using a Nikon Eclipse TE2000-E system, controlled and processed by EZ-C1 3.50 (Nikon Inc.) software. Images were assembled using Adobe Photoshop CS3 and Adobe Illustrator CS2.
Yeast 2-hybrid screening
The B9/C2 domain of human MKS1 (amino acids 144-470; Figure 4a) was cloned into the Gal4 vector pB27 and screened against a human fetal brain RP1 prey cDNA library. Yeast-2-hybrid screens were performed by Hybrigenics SA as described previously 48. Confirmatory “1-to-1” pairwise assays for selected interactants were performed with the MatchMaker Two-Hybrid System 3 (Clontech Inc.)
GST fusion protein purification
GST-UBE2E1 fusion protein was prepared essentially as described previously 32, with protein expression induced at 20°C using 0.2mM IPTG for 4 hr.
Proteasome activity assays
Crude proteasomal fractions were prepared from cells 52 and incubated with the 20S fluorophore substrate Suc-LLVY-AMC (Enzo Life Sciences Inc.). Fluorescence of each proteasomal preparation was measured on a Mithras LB940 (Berthold Technologies Inc.) fluorimeter and adjusted against a calibration factor calculated from a standard curve to give activity measurements in pmol AMC release/μg cell lysate/hour. Treatment of cells with 10 μm of the proteasome inhibitors MG-132 or c-lactacystin-β-lactone were positive controls for the assay. Results reported are from at least five independent biological replicates.
Canonical Wnt activity (SUPER-TOPFlash) luciferase assays
For luciferase assays of canonical Wnt activity, we grew cells in 12-well plates and co-transfected with 0.5 μg SUPER-TOPFlash firefly luciferase construct53 (or FOPFlash, as a negative control); 0.5 μg of expression constructs (pCMV-cmyc-MKS1, or empty pCMV-cmyc vector); and 0.05 μg of pRL-TK (Promega Corp; Renilla luciferase construct used as an internal control reporter). We obtained Wnt3a- or Wnt5a-conditioned media from stably-transfected L cells with Wnt3a or Wnt5a expression vectors (ATCC). Control media was from untransfected L cells. Activities from firefly and Renilla luciferases were assayed with the Dual-Luciferase Reporter Assay system (Promega Corp.) on a Mithras LB940 (Berthold Technologies Inc.) fluorimeter. Minimal responses were noted with co-expression of the FOP Flash negative control reporter construct. Raw readings were normalized with Renilla luciferase values. Results reported are from at least four independent biological replicates.
Ubiquitination assays
In vivo ubiquitination assays were carried out essentially as previously described 33. HEK293 cells were transfected with 2 μg of wild-type or dominant negative UBE2E1 and 2 μg HA-ubiquitin plasmids using Lipofectamine 2000. 24 hours later, cells were treated with 10μM MG132, and IP was performed 24 hr later with anti-β-catenin antibodies. The IPs were analyzed by immunoblotting with anti-HA antibody. The same experiment was repeated in cells transfected with wild-type (WT) or dominant negative (DN) UBE2E1, His-tagged β-catenin, HA-tagged-ubiquitin, and either scrambled or MKS1 siRNA. To assess in vitro ubiquitination, we used a ubiquitination kit (Enzo) according to the manufacturer’s protocol, supplemented with MKS1-GST (Proteintech) and RNF34-HIS (Novusbio) fusion proteins in a total volume of 30μl. Samples were incubated for 1.5hr at 37°C followed by SDS-PAGE and Western blotting.
TUBE assays
Agarose-bound TUBE were used as recommended by the manufacturer (LifeSensors, Malvern, PA, USA). mIMCD3 cells were transiently transfected with cmyc-MKS1 and treated with proteasome inhibitor (MG132 at 10μM) 2hr before harvesting. Lysis buffer was based on RIPA supplemented with 50mM Tris-HCl pH7.5, 0.15M NaCl, 1mM EDTA, 1% NP40, 10% glycerol, DUB inhibitors: 50μM PR619 and 5mM 1,10-phenanthroline, protease inhibitors. BRISC was used as K63 deubiquitinating enzyme. Samples were run on SDS-PAGE followed by Western blotting.
Co-immunoprecipitation
Whole cell extracts (WCE) were prepared and co-IP performed essentially as described previously 54. Co-IPs used either 5 μg affinity-purified mouse monoclonals (MAbs), or 5-10 μg purified IgG fractions from rabbit polyclonal antisera, coupled to protein G- and/or protein A-sepharose beads (GE Healthcare UK Ltd.). Proteins were eluted from beads with 0.2M glycine HCl pH2.5. Samples were neutralised by addition of 0.1 volume 1M Tris HCl ph8.5.
Western blotting
Soluble protein was analysed by SDS-PAGE using 4-12% Bis-Tris acrylamide gradient gels and western blotting was performed according to standard protocols using either rabbit polyclonal antisera (final dilutions of ×200-1000) or MAbs (x1000-5000). Appropriate HRP-conjugated secondary antibodies (Dako Inc.) were used (final dilutions of ×10000-25000) for detection by the enhanced chemiluminescence “Femto West” western blotting detection system (Pierce Inc.). Chemiluminescence was detected using a BioRad ChemiDoc MP Imaging System and Image Lab software. Volumetric analysis of immunoblot bands was performed using Image Lab software (Bio Rad).
Statistical analyses
Normal distribution of data (for SUPER-TOPFlash assays, proteasome activity assays, cilia length measurements) was confirmed using the Kolmogorov-Smirnov test (GraphPad Software). Pairwise comparisons were analysed with Student’s two-tailed t-test using InStat (GraphPad Software). Results reported are from at least three independent biological replicates.
Conflict of interest statement
The authors declare no commercial affiliations or conflicts of interest.
Supplementary Figure legends
Supplementary Figure 1: Characterisation of MKS1-mutated human patient fibroblasts
(a) RT-PCR amplicons of exons 4-6 and 15-17 from cDNA of healthy control and MKS1 patient fibroblasts, compound heterozygote for the MKS1 mutations [c.472C>T]+[IVS15-7_35del29] causing the predicted nonsense and splice-site mutations [p.R158*]+[p.P470fs*562]. Additional smaller PCR products in MKS1 patient corresponds to skipping of exon 5 and exon 16, confirmed by Sanger sequencing, due to the frameshift mutation affecting splicing. (b) Immunoblot showing loss of MKS1 protein in MKS1-mutated patient fibroblasts compared to healthy controls; loading control is β-actin. (c) IF microscopy images of wild-type control and MKS1-mutated fibroblasts showing loss of cilia and disorganization of cytoskeleton in patient cells. Bar graphs quantify reductions in incidence and length of cilia in patient cells. Statistical significance of pairwise comparisons with control for three independent biological replicates are shown (*** p < 0.001, **** p < 0.0001; paired two-tailed Student t-test; error bars indicate s.e.m.) (d) IF microscopy images showing loss of MKS1 ciliary localization in MKS1-mutated patient fibroblasts compared to wild-type control fibroblasts (indicated by arrowheads).
Supplementary Figure 2: In vivo loss of MKS1 causes deregulated ubiquitin-proteasome processing.
(a) Proteasome defects (GFP; green) in Mks1−/− × UbG76V-GFP E12.5 embryonic heart and liver, as indicated. Scale bars = 50 μm. (b) Left panel: accumulation of active β-catenin (red) and GFP-tagged ubiquitin (green)in the neuroepithelium of E12.5 Mks1−/−xUbG76VGFP embryonic neocortex. Middle panel: loss of primary cilia (stained for acetylated α-tubulin, red) on the neuorepithelial cells lining the fourth ventricle in mutant animals. Right panel: loss of basal body localisation of UBE2E1 (red; indicated by arrowheads) in mutant animals, accompanied by accumulation GFP-tagged ubiquitin. The asterisk indicates a periventricular heterotopia. Scale bar = 10μm. White frames indicate magnified regions displayed in insets (showing only the red and green channels for clarity).
Supplementary Figure 3: The E3 ubiquitin ligase RNF34 interacts with MKS1 and co-localizes at the basal body.
(a) Co-immunoprecipitation of 30 and 40 kDa isoforms of endogenous RNF34 by rabbit polyclonal anti-MKS1, but not by an irrelevant antibody or the pre-immune. (b) Co-localization of RNF34 (red) with MKS1 (green) at the basal body (arrowheads) in mIMCD3 cells. Scale bar = 10μm.
Supplementary Figure 4: Validation of Ube2e1 siRNA knockdown in mIMCD3 cells.
Left panels: mouse Ube2e1 pooled siRNA duplexes (shUbe2e1) prevent ciliogenesis compared to scrambled (shSscr) control. Primary cilia and basal bodies are visualized by staining for acetylated α-tubulin (red) and γ-tubulin (green). Scale bar = 10 μm. Right panel: immunoblot to confirm reduction of Ube2e1 protein levels following knockdown in mIMCD3 cells using individual siRNA duplexes (#1 to #3) and pooled duplexes, with knockdown efficiency indicated normalized to β-actin loading control levels.
Supplementary Figure 5: Validation of Mks1 siRNA knockdown in mIMCD3 cells.
(a) The specificity of the anti-MKS1 antibody and siRNA were confirmed by decreased basal body staining for MKS1 (green) and decreased cilia incidence (acetylated α-tubulin; red) in mIMCD3 cells following Mks1 siRNA knockdown. Scale bars = 10 μm. (b) Knockdown by siRNA was confirmed by anti-MKS1 immunoblotting, with the specific band indicated for MKS1 at 65 kDa.
Acknowledgements
This paper is dedicated to the memory of P. Robinson, our valued collaborator, colleague and friend. We are very grateful to D. Evans, J. Bilton, C. McCartney and M. Reay for technical support. We thank E. Pitt for hTERT immortalization of MKS1 patient fibroblasts. We thank A. Monk, K. Passam and T. Simpson of Nikon UK Ltd. for technical support and advice on confocal microscopy. We are grateful to R. T. Moon (University of Washington) for the SUPER-TOPFlash and FOPFlash constructs, and S. Kang (Korea University) for the UbcH6 constructs. The anti-MKS1 antibody was a gift from N. Katsanis, Duke University Medical Center and the APC antibody was a gift from I. Näthke, University of Dundee. We acknowledge funding from the UK Medical Research Council (Doctoral Training Award, GW. The research also received funding from the European Community’s Seventh Framework Programme FP7/2009 under grant agreement no: 241955 SYSCILIA. KS was funded by Wellcome Trust Institutional Strategic Support Fund.
Footnotes
↵* joint last authors