Primary cilia start forming within the G1 phase of the cell cycle and continue to grow as cells exit the cell cycle (G0). They start resorbing when cells re‐enter the cell cycle (S phase) and are practically invisible in mitosis. The mechanisms by which cilium biogenesis and disassembly are coupled to the cell cycle are complex and not well understood. We previously identified the centrosomal phosphoprotein NDE1 as a negative regulator of ciliary length and showed that its levels inversely correlate with ciliogenesis. Here, we identify the tumor suppressor FBW7 (also known as FBXW7, CDC4, AGO, or SEL‐10) as the E3 ligase that mediates the destruction of NDE1 upon entry into G1. CDK5, a kinase active in G1/G0, primes NDE1 for FBW7‐mediated recognition. Cells depleted of FBW7 or CDK5 show enhanced levels of NDE1 and a reduction in ciliary length, which is corrected in cells depleted of both FBW7 or CDK5 and NDE1. These data show that cell cycle‐dependent mechanisms can control ciliary length through a CDK5‐FBW7‐NDE1 pathway.
Biogenesis and disassembly of primary cilia are coupled to the cell cycle by incompletely understood mechanisms. This study identifies such a link, by which a quiescence‐specific kinase controls the levels of a negative regulator of ciliary length.
Primary cilia are formed in quiescent cells.
NDE1, a negative regulator of ciliary length, is downregulated in quiescent cells.
CDK5, a kinase active in quiescent cells, phosphorylates NDE1 within an FBW7‐specific phosphodegron.
The FBW7 E3 ligase recognizes and targets phosphorylated NDE1 for degradation via the ubiquitin–proteasome system.
Destruction of NDE1 in G1/G0 allows cilia to reach their proper length.
The primary cilium is an antenna‐like organelle present in virtually every cell type of the human body. It is indispensable for human health, as structural defects in primary cilia result in pathological conditions as diverse as kidney cysts and cancer to retinal degeneration and brain malformations (Hildebrandt et al, 2011). It functions as a signaling center for receptor tyrosine kinases and G protein‐coupled receptors, as well as the Hedgehog, Notch, and Wnt pathways (Goetz & Anderson, 2010; Christensen et al, 2012). A unique feature of the primary cilium is that its biogenesis and resorption are intimately associated with the cell cycle, as primary cilia form during quiescence and resorb before mitosis, providing an exquisite means to integrate multiple signaling pathways with the cell cycle (Quarmby & Parker, 2005; Pan & Snell, 2007; Plotnikova et al, 2008; Seeley & Nachury, 2010; Kim & Tsiokas, 2011; Sung & Li, 2011). The cellular program that triggers the initiation of cilia formation is not completely understood, but depends on the combinatorial actions of numerous positive and negative regulators of ciliogenesis (Kobayashi & Dynlacht, 2011; Kim & Dynlacht, 2013; Sung & Leroux, 2013). Activity and/or levels of positive regulators of ciliogenesis reach and maintain high levels, whereas inhibitors of ciliogenesis are removed from pre‐ciliary structures or destructed at the onset of ciliogenesis (Ishikawa & Marshall, 2011; Kobayashi & Dynlacht, 2011). Ciliogenesis proceeds in an orchestrated fashion initiated by the attachment of Golgi‐derived ciliary vesicles to the distal ends of the mother centriole inside the cell, followed by centriole docking to the plasma membrane. The docked mother centriole elongates and the axonemal extension follows. Finally, cilium assembly and maintenance are achieved by intraflagellar transport (IFT) mediated by the concerted action of anterograde kinesins (IFT‐B complex) and retrograde dynein (IFT‐A complex) (Ishikawa & Marshall, 2011; Kobayashi & Dynlacht, 2011; Nigg & Stearns, 2011; Sung & Leroux, 2013). Three negative regulators of ciliogenesis have been shown to be destructed or removed from ciliary structures at the onset of ciliogenesis. First, CP110, a mother centriole capping protein (Spektor et al, 2007; Tsang & Dynlacht, 2013), which blocks centriole elongation, is removed following the action of the Tau tubulin kinase 2 (TTBK2) (Goetz et al, 2012). Second, trichoplein, a protein preventing axonemal extension, is recognized by the Cullin 3‐RING E3 ligase, KCTD17, and is degraded through the ubiquitin–proteasome system (UPS) (Kasahara et al, 2014). Cells depleted of CP110 or trichoplein show increased ciliary length. Third, the Oral‐facial‐digital syndrome 1 (OFD1) gene product is removed from centriolar satellites through selective autophagy induced by nutrient deprivation (Tang et al, 2013). These mechanisms seem to impact different stages of ciliogenesis in a complementary fashion, allowing for proper cilium biogenesis and function. However, exactly how the removal, destruction, and/or mislocalization of negative regulators of ciliogenesis are integrated within the cell cycle remains poorly understood.
We have identified the centrosomal phosphoprotein, NDE1, as a suppressor of the later stages of cilium formation affecting ciliary length by modulating the IFT process (Kim et al, 2011). We also showed that its level of expression is markedly reduced when cells enter G1 and peaks when cells are in mitosis. This oscillatory pattern of NDE1 during the cell cycle, which has been reported not only in cells in culture, but also in the human brain (Bakircioglu et al, 2011), suggests that high levels of NDE1 may not be compatible with cellular functions in quiescent cells, which could be consistent with its negative effect on ciliary length. Here, we show that CDK5, a kinase active in G1/G0 (Hellmich et al, 1992; Tsai et al, 1994; Nikolic et al, 1996; Su & Tsai, 2011; Shah & Lahiri, 2014), phosphorylates NDE1. Subsequently, CDK5‐primed NDE1 is recognized by the FBW7 E3 ubiquitin ligase (Davis et al, 2014) and degraded through the UPS. Thus, a low level of NDE1 is maintained during G1 and G0 through the actions of CDK5 and FBW7, allowing cilia to form and function properly. Downregulation CDK5 or FBW7 leads to markedly decreased ciliary length that is dependent on increased levels of NDE1.
FBW7 physically interacts with NDE1
The mechanisms responsible for the oscillatory expression of NDE1 during the cell cycle are entirely unknown. Treatment of serum‐starved RPE1‐hTERT cells with MG132 or lactacystin, known inhibitors of the UPS, minimized the overall decrease in NDE1 induced by serum starvation (Fig 1A). A similar effect was seen in the centro‐somal pool of NDE1 (Fig 1B). These results indicated that the downregulation of NDE1 upon serum starvation is mediated through the UPS. Examination of the primary structure of NDE1 for potential degradation motifs revealed the presence of KEN and D‐box motifs, both recognized by the anaphase promoting complex/cyclosome (APC/C), multiple RxL motifs recognized by cyclin F, a β‐TrCP1‐specific phosphodegron, and an FBW7‐specific phosphodegron (Fig 1C). A high‐affinity FBW7‐specific recognition motif consists of the core Threonine(T)Proline(P)XXSerine(S) motif. Phosphorylation of T at position 1 is required for binding to FBW7, whereas the phosphorylation of S at position 4 or the presence of a negatively charged residue such as glutamic acid (E) enhances binding, but is not absolutely required (Hao et al, 2007). T is often phosphorylated by GSK3, after priming at the secondary phosphorylation site. In some instances, P‐directed S/T kinases such as CDKs may mediate T phosphorylation at position 1 (Hao et al, 2007; Welcker & Clurman, 2008). Deviations from this di‐phosphorylated motif have been reported (Mao et al, 2008).
The role of APC/C in ciliary disassembly and suppression of centriole amplification through the regulation of Nek1 and STIL, respectively, have been shown (Arquint & Nigg, 2014; Wang et al, 2014). Thus, we tested whether NDE1 could also be targeted for degradation through the APC/C. Incubation of in vitro transcribed/translated NDE1 with Xenopus egg extracts in the presence of Cdh1 or non‐degradable cyclin B failed to promote NDE1 degradation (Appendix Fig S1A). Similarly, NDE1 failed to physically interact with cyclin F or β‐TrCP1 in transfected cells, suggesting that these E3 ligases are unlikely to mediate the ubiquitin‐dependent degradation of NDE1 (Appendix Fig S1B). In contrast, endogenous NDE1 co‐immunoprecipitated with endogenous FBW7 in RPE1‐hTERT cell lysates. Interestingly, only the slower migrating species of NDE1 interacted with FBW7 (Fig 1D), which would be consistent with the possibility that only phosphorylated NDE1 interacted with FBW7. Indeed, phosphatase treatment of immunoprecipitated NDE1 resulted in the elimination of the slower migrating species (Fig 1E), suggesting that phospho‐NDE1 interacted with FBW7 under physiological conditions. This should be expected for an FBW7‐specific target (Davis et al, 2014). To test whether NDE1 could be a direct target of FBW7, we mutated the primary phosphorylation site within the predicted FBW7 phosphodegron in NDE1 as shown in Fig 1C from T191 to isoleucine (I) and tested for its interaction with wild‐type FBW7α. Conversely, we used the FBW7R465C mutant, which was identified in colorectal cancers abrogating the interaction and destruction of cyclin E (Rajagopalan et al, 2004), and tested for its interaction with wild‐type NDE1. These mutations in NDE1 or FBW7 reduced the strength of interactions with wild‐type FBW7 or NDE1, respectively (Fig 1F), suggesting that T191 in NDE1 and R465 in FBW7 are critical for the interaction of NDE1 and FBW7. We confirmed that overexpressed NDE1T191I accumulated at the centrosome/basal body of transfected RPE1‐hTERT cells resulting in abnormally short, if not invisible cilia (Appendix Fig S2). To test whether NDE1 can be directly conjugated with ubiquitin (Ub) in cells, we co‐transfected 6× His‐tagged Ub (His‐Ub) with wild‐type or NDE1T191I in HEK293T cells. Cells were lysed in a denaturing solution, and the covalent conjugation of Ub to NDE1 was tested by immunoprecipitation using nickel beads followed by immunoblotting. NDE1 was heavily ubiquitylated under these conditions (Fig 1G). The NDE1T191I mutant that showed reduced association with FBW7 (Fig 1F, lane 5) also reduced ubiquitylation compared to wild‐type NDE1 (Fig 1G, lane 3), suggesting that impaired recognition of NDE1T191I by endogenous FBW7 in HEK293T cells may account for its reduced level of ubiquitylation.
Next, we examined whether depletion of FBW7 would lead to increased protein stability of NDE1 in G1/G0. Transfection of RPE1‐hTERT cells with FBW7‐siRNA#1 or #2 resulted in efficient silencing of FBW7 and increased the expression of steady‐state NDE1 levels in serum‐starved cells (Fig 2A). Downregulation of FBW7 did not cause significant changes in levels of Aurora A kinase, a known partner and target of FBW7 (Mao et al, 2004; Otto et al, 2009) with an indispensable role in ciliary disassembly (Pugacheva et al, 2007). This is not entirely unexpected since the degree of FBW7‐mediated targeting is cell‐ and context‐specific (Mao et al, 2004; Hoeck et al, 2010; Sancho et al, 2014). Nevertheless, Aurora A kinase was downregulated in serum‐starved cells consistent with its degradation by the APC/C in G1 (Littlepage & Ruderman, 2002). Depletion of FBW7 did not affect CDK5 levels. Thus, these data showed that both FBW7‐specific siRNAs downregulated FBW7, resulting in the upregulation of NDE1. To test whether the upregulation of NDE1 occurred at the centrosome/basal body, we quantified NDE1 signal intensity at this structure in the presence or absence of serum in wild‐type and FBW7‐depleted cells. Upon serum starvation, there was a significant accumulation of NDE1 at the centrosome/basal body in cells depleted of FBW7 compared to wild‐type cells (Fig 2B). The negative regulation of NDE1 by FBW7 was also confirmed in human colon adenocarcinoma DLD‐1‐FBXW7−/− cells, where both alleles of FBXW7 (formal gene name) were inactivated by homologous recombination (Rajagopalan et al, 2004). NDE1 steady‐state levels were increased upon serum starvation in cells lacking FBW7 in contrast to a marked downregulation in wild‐type cells (Fig 2C). Expression of C‐MYC, which is a known target of FBW7, was used as a positive control (Welcker et al, 2004; Yada et al, 2004). Taken together, these results demonstrate that FBW7 can serve as an effective E3 ligase for NDE1.
FBW7 co‐localizes with NDE1 at the centrosome
While NDE1 shows strong expression in the centrosome (Feng et al, 2000; Bakircioglu et al, 2011; Kim et al, 2011), the presence of FBW7 in this structure in intact ciliated cells has not been examined. Double staining of FBW7 and γ‐tubulin or NDE1 showed co‐localization of FBW7 and γ‐tubulin or NDE1 in the centrosome of exponentially growing hRPE1‐hTERT cells (Fig 3Aa–g). In addition to the centrosome, FBW7 showed punctate cytoplasmic and nuclear staining, consistent with previous reports (Welcker et al, 2004; Bhaskaran et al, 2013). In ciliated, serum‐starved RPE1‐hTERT cells (arrested in G1), FBW7 was not present in the cilium (Fig 3Ba–c), but co‐localized with γ‐tubulin in the two centrioles (Fig 3Bd–f), ninein, a marker of the distal and proximal regions of the mother centriole and proximal region of daughter centriole (Fig 3Bg–i), or centrin‐2, a marker of the two centrioles, but not pericentriolar material (Fig 3Bj–l). FBW7 also showed co‐localization with CEP164, a marker of mature mother centriole present at the distal appendages (Appendix Fig S3). These data indicated that FBW7 was present in both centrioles in exponentially growing RPE1‐hTERT cells or in ciliated cells. The presence of FBW7 in the pericentriolar material cannot be excluded. To confirm the specificity of the mouse antibody used to detect FBW7 by immunofluorescence (Fig 3Aa–g and Ba–f) and immunoprecipitation (Fig 1D), wild‐type DLD‐1 and DLD‐1‐FBXW7−/− cells were immunostained with the same antibody. No FBW7 signal was detected in the red channel in cells lacking FBW7 (Fig 3C). NDE1 is expressed in cortical and hippocampal neurons (Pei et al, 2014). Thus, we tested whether it would co‐localize with FBW7 at the centrosome in primary post‐mitotic hippocampal neurons identified as such by Tau‐positive staining (Fig 3Da). FBW7 co‐localized with γ‐tubulin (Fig 3Db–d) and NDE1 (Fig 3De–g) in these cells. These data showed that FBW7 co‐localizes with NDE1 at the centrosome and basal body of diverse cell types.
FBW7 regulates ciliary length
Given that FBW7 controls the stability of NDE1 and NDE1 negatively regulates ciliary length (Kim et al, 2011), we hypothesized that FBW7 and NDE1 may function antagonistically in the regulation of ciliary length. To test this hypothesis, FBW7 or NDE1 was knocked down separately or together in RPE1‐hTERT cells and the effects of these depletions on ciliary length were determined at 24 h following serum starvation (Fig 4A). Efficiency of siRNAs in silencing respective targets was confirmed (Fig 4B). As we have shown earlier, RPE1‐hTERT cells depleted of NDE1 show an increased ciliary length, but not a further increase in the number of ciliated cells at 24 h following serum starvation, as at this time point, ciliation has been completed (Kim et al, 2011). Knockdown of FBW7 had a strong effect on cilium formation, as only ~20% of cells had visible cilia (Fig 4C and D). To test whether depletion of FBW7 caused a block in ciliogenesis or a marked reduction in ciliary length that would make cilia invisible for counting, we performed a time course experiment in which cells depleted of FBW7 were scored for the presence of cilia at 12, 24, or 48 h following serum starvation. If depletion of FBW7 had blocked ciliogenesis, the number of ciliated cells should not increase over time. Figure 4C shows that instead, the number of ciliated cells depleted of FBW7 increased over time (Fig 4D), suggesting that FBW7 did not block cilium formation, but it rather had a strong effect on ciliary length. To test whether this effect was partially or entirely contributed to the upregulation of NDE1, cells were co‐transfected with FBW7‐ and NDE1‐specific siRNAs. Mock‐, FBW7‐, or NDE1‐specific siRNAs were co‐transfected with GFP, and the percentage of ciliated cells was determined only from GFP‐positive cells. Knockdown of FBW7 caused a reduction in the number of ciliated cells by ~3‐fold at 24 h following serum starvation, and this effect was partially rescued in cells depleted of both FBW7 and NDE1 (Fig 4C), suggesting that NDE1 can function downstream of FBW7 in the regulation of ciliary length.
To provide independent evidence that NDE1 is negatively regulated by FBW7, we examined the effect of single and double depletions on the activation of the Hedgehog pathway, which is dependent on cilia for proper expression and processing of Glioma‐Associated Oncogene Homolog 1 (GLI) transcription factors (Goetz & Anderson, 2010). Activation of the pathway at the level of Smoothened (SMO) using Smoothened Agonist (SAG) caused a significant accumulation of GLI2 in mock‐transfected cells, as expected (Fig 4E, lanes 1 and 2). However, cells depleted of NDE1 showed enhanced steady‐state levels of GLI2 that could not be increased further by SAG (Fig 4E, lanes 3 and 4). Conversely, cells depleted of FBW7 showed the opposite effect (Fig 4E, lanes 5 and 6). The reduced expression of GLI2 was partially rescued in cells depleted of both proteins (Fig 4E, lanes 7 and 8). These data supported the hypothesis that FBW7 negatively controls NDE1 levels and this downregulation might be important for the regulation of the Hedgehog pathway in ciliated cells.
NDE1 is phosphorylated by CDK5
Substrate phosphorylation is required for FBW7‐mediated recognition (Koepp et al, 2001; Yada et al, 2004; Davis et al, 2014). Thus, we attempted to identify the kinase that would prime NDE1 for recognition by FBW7. We tested the following kinases: CDK1/cyclin A, CDK2/cyclin E, CDK5/p25, and GSK3β. CDK1/cyclin A has been shown to phosphorylate NDE1 (Alkuraya et al, 2011), whereas CDK5 has been shown to phosphorylate NDEL1 (Niethammer et al, 2000; Sasaki et al, 2005), a homolog of NDE1. In addition, CDK5 is most active in post‐mitotic cells (Su & Tsai, 2011) and its expression is inversely correlated with that of NDE1 during cell cycle exit or entry to G1 (Appendix Fig S4). CDK5 requires interaction with p35 (or CDK5R1) or its proteolytic fragment, p25, for activation (Lew et al, 1994; Tsai et al, 1994; Patrick et al, 1999). GSK3β is a priming kinase for known FBW7 substrates (Welcker & Clurman, 2008; Busino et al, 2012), and its activity can be induced by an interaction with p25 (Chow et al, 2014). It should be noted that GSK3 and CDK5 show similar overall folding (Chow et al, 2014). The effect of CDK2/cyclin E on NDE1 phosphorylation was unknown. Cell lysates were separated in SDS–PAGE supplemented with 50 μM Phos‐tag™ reagent that allows step‐wise separation of phosphorylated species. The majority of transfected NDE1 migrated as a non‐phosphorylated protein (Appendix Fig S5). Co‐transfection with CDK5/p25 increased the abundance of the di‐phosphorylated species by about 5‐fold (indicated by two asterisks in Fig 5A, lane 2). It also induced phosphorylation at a single residue (indicated by one asterisk, Fig 5A, lane 2). Co‐transfection with none of the other kinases induced di‐phosphorylation to levels seen by CDK5/p25 (Fig 5B; Appendix Fig S5). Co‐transfection of NDE1T191I with CDK5/p25 led to an almost similar level of di‐phosphorylation compared to wild‐type NDE1, but there was no detectable mono‐phosphorylation (Fig 5A, lane 4). These data suggested that phosphorylation at T191 required CDK5/p25. To test whether mono‐phosphorylation of NDE1 occurs in native conditions such as in endogenous NDE1 in RPE1‐hTERT cells, lysates from these cells were analyzed side by side with lysates from HEK293T cells transfected with NDE1 alone or NDE1 plus CDK5/p25. Indeed, endogenous NDE in RPE1‐hTERT cells migrated as a mono‐phosphorylated species (Fig 5C), indicating that this species is likely to interact with FBW7 in RPE1‐hTERT cells (Fig 1D, lanes 1 and 3).
CDK5 induces downregulation of NDE1
We reasoned that if CDK5 phosphorylates NDE1 at T191 within the FBW7 phosphodegron, co‐transfection of activated CDK5 should reduce the steady‐state levels of wild‐type NDE1, but not NDE1T191I. Co‐transfection of wild‐type NDE1 with CDK5/p25 or p35 (both activators of CDK5) led to downregulation of wild‐type NDE1 (Fig 6A, lanes 4 and 7). However, this was not the case with NDE1T191I, whose levels remained unchanged in cells co‐transfected with mutant NDE1 and CDK5/p25 (Fig 6B, lane 4). Next, we tested whether NDE1 could physically interact with p35 or p25. Despite a significant downregulation of NDE1 in whole cell lysates transfected with CDK5/p25 or CDK5/p35, NDE1 co‐immunoprecipitated with p35 or p25 (Fig 6C, lanes 4 and 7). In addition to p35/p25, CDK5 can be activated by p39 (or CDK5R2) (Su & Tsai, 2011; Shah & Lahiri, 2014). Thus, we tested whether CDK5/p39 could destabilize NDE1 and whether p39 could interact with NDE1. Interestingly, CDK5/p39 did not have an effect on NDE1 levels and p39 did not physically interact with NDE1 in transfected cells (Fig 6D). Consistent with co‐immunoprecipitation data, endogenous p35/p25 co‐localized with ninein in the centrosome in cycling RPE1‐hTERT cells (Fig 6E). Depletion of p35 resulted in the loss of p35‐specific signal in the green channel (Fig 6E) and an upregulation of NDE1 levels in serum‐starved cells (Fig 6F). These data suggest that CDK5 may be recruited to NDE1 through its activators, p25 or p35, but not p39, to prime NDE1 for FBW7‐mediated recognition.
To determine whether the downregulation of NDE1 in cells co‐transfected with CDK5/p25 was due to decreased stability of NDE1, we measured the half‐life of wild‐type NDE1 or NDE1T191I in the presence or absence of CDK5/p25. NDE1 levels remained unchanged for 6 h after the addition of emetine, a protein synthesis inhibitor, in cells transfected with a dominant negative form of CDK5 plus p25 (negative control, Fig 7C and E). NDE1T191I showed the same pattern in cells transfected with activated CDK5 (Fig 7B and E). However, wild‐type NDE1 reached half of its starting concentration within ~4.5 h after the addition of emetine in the presence of activated CDK5 (Fig 7A and E), indicating that the stability of wild‐type but not mutant NDE1 protein was reduced in the presence of activated CDK5.
CDK5‐mediated downregulation of NDE1 increases ciliary length
Next, we examined whether endogenous CDK5 could regulate overall NDE1 levels in RPE1‐hTERT cells. RPE1‐hTERT cells depleted of CDK5 showed increased expression of NDE1 without a significant change in FBW7 (Fig 8A). An accumulation of NDE1 in the centrosome/basal body in cells depleted of CDK5 compared to wild type cells was also seen (Appendix Fig S6). These data show that NDE1 is negatively regulated by CDK5 when cells are induced to enter G1 or exit the cell cycle by serum starvation. Given the negative effect of NDE1 on ciliary length, we reasoned that cells depleted of CDK5 should have abnormally short cilia, which should be rescued in cells depleted of both CDK5 and NDE1. Indeed, cells depleted of NDE1 showed increased average ciliary length from 2.3 to 6.8 μm, as expected (Kim et al, 2011). Cells depleted of CDK5 had an average ciliary length of 1.8 μm, which was increased to 4.6 μm in cells depleted of both proteins (Fig 8B and C). The percentage of ciliated cells at 24 h following serum starvation did not change in cells with single and double knockdowns (Fig 8C), suggesting that the effect of CDK5 on ciliary length was not as severe as that of FBW7, but within the same pathway as NDE1.
In this study, we identify the CDK5‐FBW7‐NDE1 module as a critical regulator of ciliary length in a cell cycle‐dependent manner. This conclusion is based on the following lines of evidence: first, NDE1 can be phosphorylated by p35‐/p25‐activated CDK5 at a site (T191) located within an FBW7‐specific phosphodegron. Second, phosphorylated NDE1 physically interacts with FBW7 and cells depleted of FBW7 show increased levels of NDE1. Third, activation of CDK5 destabilizes NDE1. Fourth, the mutation of T191 to isoleucine in NDE1 diminishes the physical interaction with FBW7 and renders the protein insensitive to CDK5‐mediated destabilization. Fifth, depletion of CDK5 or FBW7 results in the accumulation of NDE1 and a reduction in ciliary length, which is corrected in cells depleted of CDK5 and NDE1 or FBW7 and NDE1. Sixth and final, the effects of single and double depletions on ciliary length correlated with effects on the activation of the Hedgehog pathway, which is dependent on primary cilia. These data lead us to propose a model in which, upon entry into G1, levels and/or activity of CDK5 rise inducing the phosphorylation of NDE1. Phospho‐NDE1 is recognized by FBW7 and targeted for UPS‐mediated degradation, allowing cilium assembly to proceed and cilia to function properly (Fig 9). Accordingly, a low level of NDE1 is maintained throughout G1 and G0. During cell cycle re‐entry, levels of CDK5 are reduced (Zhang et al, 2012) and remaining activity is further suppressed through known interactions with cyclin D1 (Xiong et al, 1992; Modi et al, 2012) and cyclin E (Odajima et al, 2011), allowing NDE1 levels to accumulate and promote ciliary resorption.
Our studies provide evidence for an essential role of FBW7 in the regulation of ciliary length. To our knowledge, this is the first time that FBW7 has been implicated in cilium formation and function. We show that depletion of FBW7 results in a dramatic decrease in the number of ciliated cells accompanied by a robust reduction in the expression levels of core components of the Hedgehog pathway. Effects on ciliary length and Hedgehog pathway are partially rescued in cells lacking both FBW7 and NDE1, suggesting that NDE1 functions downstream of and antagonistically to FBW7 in cilium formation and function. The FBW7‐mediated recognition of NDE1 and degradation by the SCFFBW7 enzyme complex are likely to occur at the centrosome, because NDE1 and FBW7 co‐localize there and MG132 treatment increases the amount of NDE1 at the centrosome. This is consistent with previous work in which Skp1 and Cul1 were identified as core components of the centrosome (Freed et al, 1999) and proteasomal subunits co‐localize with γ‐tubulin in HeLa cells (Wigley et al, 1999). Further, Polo‐like kinase 2 co‐immunoprecipitates and phosphorylates FBW7 in centrosomal preparatio n in HeLa cells (Cizmecioglu et al, 2012).
Depletion of FBW7 reduced the expression levels of GLI2, which were partially restored in cells depleted of both FBW7 and NDE1. This effect mirrored the effect of single and double deletions on ciliary length. We propose that full‐length GLI2 is stabilized in the presence of cilia and this stabilization depends on ciliary length. The depletion of NDE1 seems to affect the Hedgehog pathway in a mechanism similar to a weak allele of Ift144, which caused ectopic activation of the pathway, and activation was not induced further by SAG (Liem et al, 2012). IFT144 is part of the IFT‐A complex mediating anterograde transport. We had shown earlier that NDE1 affected ciliary length by tethering the dynein light chain, DYNLL1 (or LC8), also an IFT‐A component, at the basal body (Kim et al, 2011). Thus, the stabilization of GLI2 in the presence of cilia could involve the IFT‐A complex.
Our results show that FBW7 affects ciliary length and perhaps ciliogenesis in general, not only by suppressing the expression of NDE1, but also through an effect on other proteins. At the present time, we do not know all stage(s) and ciliogenesis‐relevant pathways affected by FBW7. The fact that some of these effects were rescued by the depletion of NDE1 indicates that FBW7 may have affected later stages of ciliogenesis/maintenance modulating the IFT machinery through NDE1. We believe that NDE1 suppresses ciliogenesis at the stage of cilium assembly and maintenance because massive overexpression of NDE1 still allowed the formation of stumpy cilia (Kim et al, 2011), suggesting that early stages of ciliogenesis, such as centriole docking, must be intact. It is tempting to speculate on a regulatory role of cyclin E in FBW7‐mediated destruction of NDE1. Cyclin E is expressed in high levels in post‐mitotic neurons where it negatively regulates CDK5 by sequestering it away from its activators (Odajima et al, 2011). Although the accumulation of cyclin E in G0 might be limited to neurons, it is conceivable that at the onset of S phase, elevated levels of cyclin E could inhibit CDK5, terminating its priming effect on NDE1 and subsequent destruction of NDE1 by the UPS. Accumulated NDE1 could contribute to ciliary disassembly.
Our data show that activation of CDK5 via p35 or p25 can induce T191 phosphorylation and destabilization of NDE1. NDE1 was also found to associate with p35 or p25, but not with CDK5 directly. Activation of CDK5 via p39 did not have an effect on NDE1, and p39 did not associate with NDE1. These data suggest that CDK5 is recruited to NDE1 through a physical interaction with its activators. The subcellular localization of p35 and p25 are not completely overlapping with p35 being predominantly localized in the plasma membrane, whereas p25 being localized in the cytosol (Shah & Lahiri, 2014). p35 was shown to be present in the centrosome (Rosales et al, 2010b). We confirmed its localization there. A splice variant of CDK5 has been shown to be present in the centrosome (Kim et al, 2010). Furthermore, CDK5 has an indispensable role in cilia formation (Rosales et al, 2010a). Although, we could not detect endogenous CDK5 in the centrosome in RPE1‐hTERT cells, it is possible that p35 transiently recruits CDK5 to the centrosome where it can phosphorylate NDE1 and possibly other substrates. Thus, priming and recognition of NDE1 by CDK5 and FBW7, respectively, may likely occur at the centrosome.
The effect of CDK5 on ciliary length seems to be in agreement with biochemical experiments showing that CDK5 primes NDE1 for FBW7‐mediated destruction, since depletion of both proteins, CDK5 and NDE1, partially restored ciliary length. These data lead us to propose the following suggestions: first, the effect of CDK5 on ciliary length is less severe than that of FBW7. This idea is due to the fact that cells depleted of FBW7 had a reduced number of ciliated cells, most likely due to a strong effect on ciliary length that would make cilia invisible for counting, whereas that was not the case in cells depleted of CDK5 where reduction in ciliary length was less dramatic (from 2.3 to 1.8 μm). This observation suggests that CDK5 could only modify a subset of FBW7 targets with a negative role in ciliary length regulation. Other kinases could prime these putative FBW7 targets. Second, CDK5 seemed to have affected other substrates, in addition to NDE1. These other substrates may have positive or negative roles in ciliary assembly. Identification of these targets may provide new information on mechanisms controlling ciliary length.
Our study shows that CDK5 and FBW7 can function in a linear pathway to maintain low levels of NDE1 and ensure proper cilia formation and function. Because CDK5 activity is high in G1/G0, it is likely that the CDK5/FBW7 module contributes to the regulation of NDE1 in the G1/G0 phases of the cell cycle. The functionality of CDK5 and FBW7 in a G1/G0‐specific function, such as cilium formation and function through NDE1, is shown here. However, it is possible that FBW7 may regulate the abundance of NDE1 in other phases of the cell cycle, where priming could occur by another kinase. NDE1 has important functions in mitosis (Feng & Walsh, 2004; Vergnolle & Taylor, 2007) and S phase (Houlihan & Feng, 2014), and naturally occurring inactivating mutations in the NDE1 gene in humans result in microcephaly (Alkuraya et al, 2011; Bakircioglu et al, 2011), a condition associated with the reduction the neuronal progenitor pool through effects on cell cycle progression and apoptosis. Given the well‐known role of FBW7 as a tumor suppressor (Crusio et al, 2010; Wang et al, 2012; Davis et al, 2014), it is tempting to speculate that an excessive accumulation of NDE1 throughout the cell cycle could contribute to tumorigenesis induced by inactivating mutations in FBXW7. In addition to its tumor suppressor activity, FBW7 can affect differentiation, survival, and reprogramming of stem cells and progenitors in tissues that are ciliated (Hoeck et al, 2010; Sancho et al, 2014). Based on our data, it is conceivable that some of these effects could be mediated through the FBW7‐mediated suppression of important regulators of cilium formation and function, such as the Hedgehog pathway.
Here, we have identified a pathway that negatively regulates the abundance of NDE1 in quiescent cells providing mechanistic leads into the connection of the cell cycle with cilium formation, assembly, and function. These findings can spur future studies to identify regulators of cilium biogenesis, assembly, and/or maintenance. They also have important implications in clinical conditions associated not only with well‐known ciliopathies, but with cancer and microcephaly.
Materials and Methods
HEK293T, BALB/C 3T3, and RPE1‐hTERT cells were obtained from ATCC, and cell cultures were maintained in Dulbecco's Modified Eagle's Medium (DMEM) plus 10% fetal bovine serum. DLD‐1 and DLD‐1‐FBXW7−/− cells were obtained from Dr. Wei (Harvard).
Primary hippocampal neuronal cultures and antibody staining were prepared as follows: Hippocampi from newborn mice were dissected away from the cortex and digested with Papain and DNase‐I for 25 min. Hippocampi were then triturated with a glass Pasteur pipette, and cell suspensions were plated on poly‐L‐lysine‐coated coverslips. Primary hippocampal cultures were maintained for 3 weeks in Neuralbasal media supplemented with B27 and Glutamax‐I with treatment of the mitotic inhibitor 5′‐fluoro‐2′‐deoxyuridine on day 3 in culture. Cells were fixed in 4% paraformaldehyde and stained for designated primary antibodies for 1 h at 37°C and complementary secondary antibodies for 30 min at 37°C. Coverslips were mounted on slides with ProLong Gold, and nuclei were labeled with DAPI. Tau+ neurons were imaged the following day on a Leica SP2 scanning confocal microscope with an oil immersion 63× objective.
Nocodazole, emetine, lactacystin, and MG132 were purchased from Sigma. Phos‐tag reagent was purchased from Wako Chemicals. SAG was purchased from Millipore.
Mouse and human NDE1 were obtained from Open Biosystems and were subcloned into myc‐pFlag‐CMV‐2 vector (Kim et al, 2011). GFP‐CDK5, GFP‐DN‐CDK5, MYC‐p35, GFP‐p25, FLAG‐β‐TrCP1, HA‐CDK1, HA‐CDK2, HA‐GSK3β, cyclin A, cyclin E, MYC‐FBW7, and MYC‐FBW7R465C were obtained from Addgene. tGFP‐p39 was obtained from Origene. MYC‐NDE1T191I and FLAG‐NDE1T191I were generated using QuikChange mutagenesis kit according to the manufacturer's instructions (Stratagene).
siRNAs were transfected into RPE1‐hTERT cells using Lipofectamine 2000 (Invitrogen) according to the manufacturer's instructions. Plasmids were transfected into 293T cells using calcium phosphate method, as described (Kim et al, 2011).
HEK293T, BALB/C 3T3, or RPE1‐hTERT cells were lysed in 1% Triton X‐100, 150 mM NaCl, 10 mM Tris–HCl at pH 7.5, 1 mM EGTA, 1 mM EDTA, 10% sucrose and protease inhibitor cocktail (Roche Applied Science), and phosphatase inhibitors 0.2 mM Na3VO4 and 1 mM NaF at 4°C for 30 min. Antibodies against NDE1, (Proteintech), c‐Myc, Flag (Sigma‐Aldrich), CDK5 (Cell Signaling Technology), p35 (used to detect both p35 and p25) (C‐19:sc‐820, Santa Cruz Biotechnology), β‐actin (Santa Cruz Biotechnology), mouse‐FBW7 (Abnova), rabbit‐FBW7 (A301‐720A; A301‐721A) (Bethyl Laboratories), rabbit‐GLI2 (Cell Signaling), rabbit‐Aurora A Kinase (Cell signaling), and α‐tubulin (Sigma Aldrich) were used at a 1:1,000 dilution. Densitometric quantification of autoradiograms was analyzed by ImageJ software.
Lambda (λ‐) phosphatase treatment
RPE1‐hTERT cells were lysed in 1% Triton X‐100, 150 mM NaCl, 10 mM Tris–HCl at pH 7.5, 1 mM EGTA, 1 mM EDTA, 10% sucrose and protease inhibitor cocktail (Roche Applied Science), and phosphatase inhibitors 0.2 mM Na3VO4 and 1 mM NaF at 4°C for 30 min. NDE1 was immunoprecipitated with rabbit α‐NDE1, and complexes were washed three times with lysis buffer followed by a wash with phosphate‐buffered saline (PBS). A second wash was performed with λ‐phosphatase buffer containing MnCl2 (1 mM). Half of immunoprecipitated NDE1 was treated with 1.25 units of λ‐phosphatase (Biolabs, New England) for 1 h at 30°C followed by boiling for 5 min at 96°C.
RPE1‐hTERT cells were grown on glass coverslips and fixed in 1:1 methanol/acetone or 4% paraformaldehyde, permeabilized in 0.2% Triton X‐100 in PBS, blocked in 2% heat‐activated goat serum/0.2% Triton X‐100 in PBS (blocking buffer), and incubated overnight with primary antibodies diluted in blocking buffer at 4°C. Primary antibodies were used against mouse acetylated α‐tubulin, at 1:1,000 (Sigma Aldrich), rabbit γ‐tubulin 1:1,000 (Sigma‐Aldrich), mouse γ‐tubulin 1:1,000 (Abcam), rabbit Flag 1:1,000, rabbit CEP164, 1:1,000 (Novus Biologicals), mouse ninein 1:500, mouse centrin‐2 1:500 (Millipore), rabbit ARL13B at 1:500, rabbit NDE1 at 1:1,000 (Proteintech), mouse FBW7 1:1,000 (Abnova), and rabbit FBW7 (A301‐720A; A301‐721A) (Bethyl Laboratories). Cells were washed three times with PBS and incubated for 2 h at 4°C with appropriate combinations of fluorescence‐labeled secondary antibodies at 1:2,000 dilution. Secondary antibodies were donkey anti‐rabbit Alexa 488, donkey anti‐mouse Alexa 568, goat anti‐mouse Alexa 488, donkey anti‐mouse Alexa 568, or donkey anti‐rabbit Alexa 647 (Molecular Probes). Excess of secondary antibodies were removed by four washes in PBS. Coverslips were mounted with ProLong Gold antifade reagent with DAPI (Molecular Probes). Images were acquired with Leica SP2MP confocal microscope and processed with ImageJ software and Adobe Photoshop 6.0. Quantification of fluorescence intensity ratio of NDE1/γ‐tubulin (red/green) at the centrosome (Fig 2B; Appendix Fig S6) was done as described earlier (Kim et al, 2011).
Cell cycle analysis
BALB/C 3T3 and RPE1‐hTERT cells were fixed with 70% ethanol and stored at −20°C for no longer than 24 h. After fixation, cells were stained with 40 μg/ml propidium iodide (Sigma Aldrich) plus 200 μg/ml RNase A (Sigma Aldrich) in PBS and subjected to flow cytometry analysis. At least 15,000 cells were analyzed per sample using a FACSCalibur™ flow cytometer.
The human non‐specific siRNAs 5′‐UUCUCCGAACGUGUCACGU‐3′ and 5′‐ACGUGACACGUUCGGAGAA‐3′ were obtained from QIAGEN. Human NDE1‐specific smart pool siRNA (5′‐GGACCCAGCUCAAGUUUAA‐3′, 5′‐GCGCAGACCAAAGCCAUUA‐3′, 5′‐GCUGAAGCCUGUUCUUGGU‐3′, 5′‐GCAGCACUCUGAAGGCUAC‐3′), human p35 (CDK5R1)‐specific smart pool siRNA (5′‐GCAGAUAAAUGCCGACCCA‐3′, 5′‐UCACGCACCUCAACAAUGA‐3′, 5′‐GGAAGGCCACGCUGUUUGA‐3′, 5′‐UCACACAGGUCUUCUCCGA‐3′), and human CDK5‐1 specific smart pool siRNA (5′‐UAUAAGCCCUAUCCGAUGU‐3′, 5′‐CCGGGAGACUCAUGAGAUC‐3′, 5′‐GGGCUGGGAUUCUGUCAUA‐3′, 5′‐GGAUUCCCGUCCGCUGUUA‐3′) were obtained from Dharmacon (Thermo scientific) and human‐specific CDK5‐2 5′‐UAUGACAGAAUCCCAGCC‐3′, 5′‐GGGCUGGGAUUCUGUCAUA‐3′. Human FBW7‐1‐specific siRNAs 5‐GCAUUUCUCUCUCCAGAGAAGGU‐3′ and 5′‐ACCUUCUCUGGAGAGAGAAAUGC‐3′ and FBW7‐2‐specific siRNAs 5‐UCUCCAGUCUCUGCAUUCCACAC‐3′ and 5′‐GUGUGGAAUGCAGAGACUGGAGA‐3′ were initially obtained from Dharmacon (Thermo Fisher) and Integrated DNA Technology.
In vivo ubiquitylation assay was done according to Li et al (2013). Briefly, HEK293T cells were transiently co‐transfected with FLAG‐tagged NDE1 or FLAG‐tagged Nde1T191I and pcB6‐His ubiquitin (gift from R. Baer). After 24 h, cells were treated with 10 μM MG132 (Peptides international) for 5 h, and cells were lysed and briefly sonicated in buffer A (100 mM Na2PO4, 10 mM Tris–HCl, 6 M guanidine‐HCl, 10 mM imidazole pH 8.0). Cell lysates were incubated with nickel‐NTA resin (Qiagen) to precipitate His‐tagged ubiquitylated proteins for 2–3 h at room temperature (24°C). Ni‐NTA beads were washed three times with buffer A, two times with washes A/TI (1 volume buffer A and 1 volume TI buffer) (TI buffer, 25 mM Tris–Cl pH 6.8 and 20 mM imidazole), and then wash with TI buffer. Finally, Ni‐NTA beads were eluted in SDS loading buffer containing 200 mM imidazole, separated by SDS–PAGE, and detected by immunoblotting.
Data are presented as mean ± SEM. One‐way ANOVA followed by Newman–Keuls post‐test was used to determine statistical significance among multiple measurements. *P < 0.05, **P < 0.01, ***P < 0.005, nsP > 0.05.
DM performed biochemical and imaging experiments in HEK293T, RPE1‐hTERT, and DLD‐1 cells; MCM and GL performed imaging experiments and analyzed data on primary hippocampal neurons; SeoK performed biochemical experiments; SehK performed APC/C assays; ECO performed initial mutagenesis experiments; and LT supervised the project, performed biochemical experiments, analyzed data, and wrote the paper together with DM.
Conflict of interest
The authors declare that they have no conflict of interest.
We would like to thank Dr. Axel Behrens for helpful suggestions and Drs. Wei for DLD‐1 cells, R. Baer for pcB6‐His ubiquitin plasmid, S. Rankin for help with the APC/C assay using Xenopus egg extracts, and M. Ahmad for help with the preparation of hippocampal cultures. This work was supported by PBBEP3‐141439 from the Swiss National Science Foundation (DM); GM074692 from NIH (GL); DK59599 from NIH, Oklahoma Center for the Advancement of Science and Technology; and the John S. Gammill Endowed Chair in Polycystic Kidney Disease (LT).
FundingSwiss National Science FoundationPBBEP3‐141439
- © 2015 The Authors