Localization of Protein Kinase NDR2 to Peroxisomes and Its Role in Ciliogenesis*

Nuclear Dbf2-related (NDR) kinases, comprising NDR1 and NDR2, are serine/threonine kinases that play crucial roles in the control of cell proliferation, apoptosis, and morphogenesis. We recently showed that NDR2, but not NDR1, is involved in primary cilium formation; however, the mechanism underlying their functional difference in ciliogenesis is unknown. To address this issue, we examined their subcellular localization. Despite their close sequence similarity, NDR2 exhibited punctate localization in the cytoplasm, whereas NDR1 was diffusely distributed within the cell. Notably, NDR2 puncta mostly co-localized with the peroxisome marker proteins, catalase and CFP-SKL (cyan fluorescent protein carrying the C-terminal typical peroxisome-targeting signal type-1 (PTS1) sequence, Ser-Lys-Leu). NDR2 contains the PTS1-like sequence, Gly-Lys-Leu, at the C-terminal end, whereas the C-terminal end of NDR1 is Ala-Lys. An NDR2 mutant lacking the C-terminal Leu, NDR2(ΔL), exhibited almost diffuse distribution in cells. Additionally, NDR2, but neither NDR1 nor NDR2(ΔL), bound to the PTS1 receptor Pex5p. Together, these findings indicate that NDR2 localizes to the peroxisome by using the C-terminal GKL sequence. Intriguingly, topology analysis of NDR2 suggests that NDR2 is exposed to the cytosolic surface of the peroxisome. The expression of wild-type NDR2, but not NDR2(ΔL), recovered the suppressive effect of NDR2 knockdown on ciliogenesis. Furthermore, knockdown of peroxisome biogenesis factor genes (PEX1 or PEX3) partially suppressed ciliogenesis. These results suggest that the peroxisomal localization of NDR2 is implicated in its function to promote primary cilium formation.

In most cases, NDR1 and NDR2 appear to share common cellular and physiological functions (5,6,9,11). However, we recently showed that NDR2, but not NDR1, plays a critical role in the formation of primary cilia (12), which are antenna-like sensory organelles that sense and transmit a variety of chemical and mechanical signals from outside of the cell (13,14). Because primary cilia are essential for the development and homeostasis of numerous tissues, defects in cilium formation cause diverse human diseases, including polycystic kidney disease, retinal degeneration, polydactyly, and brain malformation; these are commonly known as ciliopathies (13,14). In the early stage of ciliogenesis, membrane vesicles are transported and fused to the distal end of the mother centriole to generate the "ciliary vesicle" (15). As the axoneme grows from the distal end of the mother centriole, the ciliary vesicle elongates and fuses with the plasma membrane, resulting in the extrusion of the cilium from the cell surface (15). The process of ciliary vesicle formation requires Rabin8 (a GDP-GTP exchange factor for Rab8)-mediated activation of Rab8 on the centrosome (15)(16)(17). NDR2 was * This work was supported by Ministry of Education, Culture, Science, Sports, and Technology of Japan Grants for Scientific Research 15H04347 and 15H01197 (to K. M.) and 24247038 and 26116007 (to Y. F.). The authors declare that they have no conflicts of interest with the contents of this article. □ S This article contains supplemental Movie S1. 1 To whom correspondence should be addressed. Tel.: 81-22-795-6676; Fax: 81-22-795-6678; E-mail: kmizuno@biology.tohoku.ac.jp. 2 The abbreviations used are: NDR, nuclear Dbl-related kinase; NDR1(ϩL), NDR1 with an additional leucine residue at the C terminus; NDR2(⌬L), NDR2 with a loss of a C-terminal leucine residue; CFP, cyan fluorescent protein; CFP-SKL, CFP with a Ser-Lys-Leu sequence at the C terminus; Pex, peroxin or peroxisome biogenesis factor; PNS, post-nuclear supernatant; shown to be crucial for the early step of ciliogenesis by phosphorylating Rabin8 and promoting local activation of Rab8 in the vicinity of the centrosome (12). Recent studies have also identified NDR2 (STK38L) as the causal gene for canine early retinal degeneration, the genetic disorder corresponding to the human ciliopathy, Leber congenital amaurosis (18,19), which further suggests a role for NDR2 in primary cilium formation. NDR1 and NDR2 have similar structures, sharing 87% amino acid sequence identity, and similar substrate specificities (2,20). In vitro kinase assays showed that both NDR1 and NDR2 have the potential to phosphorylate Rabin8 (9,12). However, depletion of NDR1 had no apparent effect on ciliogenesis, whereas depletion of NDR2 significantly suppressed ciliogenesis (12). Notably, the subcellular localizations of NDR1 and NDR2 differ; NDR2 exhibits vesicular localization in the cytoplasm, whereas NDR1 is distributed diffusely throughout the cytoplasm and the nucleus (12,20). Thus, the functional difference between NDR2 and NDR1 in ciliogenesis might be due to their distinct subcellular localizations. However, the questions of which vesicles or organelles NDR2 localizes to and how NDR2, but not NDR1, localizes to vesicular particles remain unsolved. Furthermore, it is also unclear whether the vesicular localization of NDR2 is correlated to its function in ciliogenesis.
In this study, we found that NDR2, but not NDR1, localizes to the peroxisome. We provide evidence that the C-terminal GKL sequence of NDR2 functions as the peroxisomal targeting signal. Furthermore, the expression of wild-type NDR2, but not its peroxisome-non-targeting mutant, rescued the inhibitory effect of NDR2 knockdown on ciliogenesis. Knockdown of the PEX gene (PEX1 or PEX3) partially suppressed ciliogenesis. Collectively, our results suggest that the peroxisomal localization of NDR2 is crucial for its function to promote primary cilium formation.

NDR2 Localizes to the Peroxisomes-Previous studies have
shown that NDR2 exhibits a punctate localization in the cytoplasm, whereas NDR1 is distributed diffusely in both the cytoplasm and the nucleus (12,20). To define the subcellular localization of NDR2, human telomerase-immortalized retinal pigment epithelial-1 (RPE1) cells were transfected with the plasmid encoding N-terminally yellow fluorescent protein (YFP)-tagged NDR2 and immunostained with specific antibodies against several organelle marker proteins. Comparison of the punctate localization of YFP-NDR2 with the localization of organelle marker proteins revealed that YFP-NDR2 co-localized well with catalase (a marker for peroxisomes) but not with early endosome antigen 1 (EEA1, a marker for early endosomes), cis-Golgi matrix protein of 130 kDa (GM130, a marker for cis-Golgi networks), microtubule-associated protein 1 light chain 3 (LC3, a marker for autophagosomes), or lysosome-associated membrane protein-1 (LAMP1, a marker for lysosomes) (Fig. 1). These results suggest that NDR2 localizes to the peroxisome.
To further examine the localization of NDR2, we compared the localization of YFP-NDR2 with that of CFP-SKL (cyan fluorescent protein carrying a typical PTS1 sequence, Ser-Lys-Leu, at its C terminus). YFP-NDR2 mostly co-localized with CFP-SKL (Fig. 2). In contrast, control YFP and YFP-NDR1 distributed diffusely in the cell (Fig. 2). Time lapse fluorescence analysis of RPE1 cells co-expressing YFP-NDR2 and CFP-SKL further demonstrated that these proteins co-localized well during the observation period (supplemental Movie S1). These results indicate that NDR2 localizes to the peroxisome.
Subcellular Fractionation Analysis-The intracellular localization of NDR2 was further examined by subcellular fractionation analysis. When the post-nuclear supernatant (PNS) fraction of YFP-NDR2-expressing HeLa cells was fractionated into the organellar and cytosolic fractions, YFP-NDR2 was detected in both fractions (Fig. 3A). Upon further fractionation of the PNS fraction by iodixanol density gradient ultracentrifugation, YFP-NDR2 was detected and co-sedimented well with a peroxisomal protein, Pex14p (Fig. 3B, lanes 8 and 9), which further supports the peroxisomal localization of NDR2. To examine the localization of endogenous NDR2, the PNS fraction of RPE1 cells was separated to the organellar and cytosolic fractions and analyzed by immunoblotting with an anti-NDR2 antibody. NDR2 was detected in both fractions, although less in the organellar fraction (Fig. 3C), which suggests that endogenous NDR2, at least in part, is localized at the organelles. Further fractionation or immunostaining analysis using an anti-NDR2 antibody failed to detect endogenous NDR2, probably because the expression level of endogenous NDR2 is below the detection limit of the antibody.
The C-terminal GKL Motif Is Required for the Peroxisomal Localization of NDR2-The amino acid sequences of NDR1 and NDR2 are highly homologous. The overall identity of human or mouse NDR1 and NDR2 is 87%, and the identity of their protein kinase domains is 91% (1,20). However, there is a noticeable difference in the C-terminal sequences of these proteins: NDR2 has the C-terminal sequence YMKAGKL, whereas NDR1 has YMKAAK (Fig. 4A). The difference in the C-terminal sequences of NDR1 and NDR2 is conserved in mammals. The C-terminal GKL sequence of NDR2 is closely related to the PTS1 consensus sequence (S/A/C)(K/R/H)(L/M) (26 -28), which prompted us to examine the possibility that NDR2 is localized to the peroxisome through the C-terminal GKL motif and that NDR1 is not localized to the peroxisome due to the absence of the C-terminal leucine residue. To explore this possibility, we constructed expression plasmids encoding a YFP-tagged NDR2 mutant with a deletion of the C-terminal leucine, YFP-NDR2(⌬L), and a YFP-tagged NDR1 mutant carrying an additional leucine at the C-terminal end, YFP-NDR1(ϩL) (Fig.  4A), and examined their peroxisomal localization by co-expressing them with CFP-SKL. The expression of each protein was confirmed by immunoblot analysis with an anti-GFP antibody (Fig. 4B). YFP-NDR2(WT) showed a punctate co-localization with CFP-SKL ( Fig. 2A), but YFP-NDR2(⌬L) distributed diffusely in the cytoplasm and the nucleus (Fig. 4C). Conversely, YFP-NDR1(WT) distributed diffusely in the cytoplasm and the nucleus ( Fig. 2A), but YFP-NDR1(ϩL) exhibited a punctate localization in the cytoplasm and co-localized with CFP-SKL (Fig. 4C). Quantitative analysis showed that the deletion of the C-terminal leucine significantly decreased the number of cells with punctate localization of NDR2, and the addition of leucine at the C terminus significantly increased the number of cells with punctate localization of NDR1 (Fig. 4D). These results indicate that the leucine residue in the C-terminal GKL motif of NDR2 is required for its localization to the peroxisome and that the distinct localization between NDR2 and NDR1 is primarily due to the difference between their C-terminal sequences.
NDR2, but Not NDR1, Binds to Pex5p-The PTS1 signal of peroxisomal proteins is recognized by a receptor protein, Pex5p, which binds to the PTS1 sequence and translocates cargo proteins to the peroxisome through its interaction with a docking complex (21)(22)(23)(24)(25). There are two splicing variants of Pex5p: a short form (Pex5pS) and a long form (Pex5pL), of which Pex5pS is exclusively involved in PTS1-containing protein import to the peroxisome (23,29). To examine whether NDR2 is localized to the peroxisome through its binding to Pex5pS, we analyzed the binding ability of NDR2 with Pex5pS using co-immunoprecipitation assays. The construct YFP-NDR2 or -NDR2(⌬L) was co-expressed with FLAG-tagged Pex5pS in 293T cells and immunoprecipitated with an anti-GFP antibody, and the precipitates were analyzed by immunoblotting with an anti-FLAG antibody. FLAG-Pex5pS was coprecipitated with NDR2(WT) but not with NDR2(⌬L) (Fig. 5), which suggests that NDR2 is localized to the peroxisome by the association with Pex5pS and that the C-terminal GKL motif is required for NDR2 binding to Pex5pS. Additionally, FLAG-Pex5pS was co-precipitated with NDR1(ϩL), but not with NDR1(WT) (Fig. 5), which suggests that the C-terminal AKL motif of NDR1(ϩL) is recognized by Pex5pS for localization to the peroxisome.
NDR2 Is Exposed to the Cytosol-To examine whether NDR2 is imported into the peroxisomal matrix or resides on the outside of the peroxisomal membrane, YFP-NDR2 was expressed in RPE1 cells, and the PNS fraction of the cells was subjected to a proteinase K sensitivity assay. YFP-NDR2 was detected in the mock-treated PNS fraction (Fig. 6A, lane 1) but was not detectable after proteinase K treatment (Fig. 6A, lane 2). Under similar conditions, a peroxisomal membrane protein PMP70 was sensitive to proteinase K treatment, whereas catalase, a peroxisomal matrix protein control, was mostly protected against the protease treatment (Fig. 6, B and C, lane 2). All of these proteins were degraded by the protease treatment in the presence of Triton X-100 (Fig. 6, A-C, lane 3). These results suggest that NDR2 is not a peroxisomal matrix protein and that at least a part of the molecule resides on the cytosolic side of the peroxisomal membrane.
The C-terminal GKL Motif Is Required for the Function of NDR2 to Promote Ciliogenesis-We showed previously that the knockdown of NDR2 significantly suppressed primary cilium formation but that knockdown of NDR1 had no overt effect on ciliogenesis, indicating that NDR2, but not NDR1, plays a critical role in ciliogenesis (12). To explore the role of the peroxisomal localization of NDR2 for its function in ciliogenesis, we examined whether the expression of siRNA-resistant mouse NDR2 (WT or ⌬L) could ameliorate the inhibitory effect of NDR2 knockdown on ciliogenesis. RPE1 cells were treated with control or NDR2-targeting small interfering RNA (siRNA) and cultured for 24 h and then serum-starved for 48 h. Then the number of ciliated cells was counted by staining acetylated (Ac)-tubulin. As reported (12), cell treatment with NDR2 siRNA reduced the levels of NDR2 protein (Fig. 7A) and significantly suppressed ciliogenesis, compared with control siRNA treatment (Fig. 7, B and C). Expression of YFP-tagged, siRNA-resistant NDR2(WT) significantly rescued primary cilium formation in NDR2-depleted cells, but expression of YFP-NDR2(⌬L) did not (Fig. 7, B and C), indicating that the C-terminal leucine residue is required for the function of NDR2 in ciliogenesis and that the peroxisomal localization of NDR2 is correlated to its function in ciliogenesis. The expression of YFP-NDR1 did not rescue ciliogenesis in NDR2-depleted cells (Fig. 7, B and C). Notably, the expression of YFP-NDR1(ϩL) slightly, but not significantly, restored ciliogenesis in NDR2depleted cells (Fig. 7, B and C). These results suggest that per-  oxisomal localization alone is not sufficient for NDR1(ϩL) to replace the function of NDR2 in ciliogenesis.
Knockdown of PEX1 or PEX3 Suppresses Ciliogenesis-As the peroxisomal localization of NDR2 is correlated to its function in ciliogenesis, we next examined whether the knockdown of PEX1 or PEX3 (genes essential for peroxisome biogenesis) affects ciliogenesis. RPE1 cells were transfected with siRNAs targeting PEX1 or PEX3, cultured for 24 h, and then cultured for 48 h under serum-starved conditions. Two independent siRNAs against PEX1 and PEX3 effectively decreased the expression of the respective transcripts in RPE1 cells (Fig. 8A). Treatment with PEX1 or PEX3 siRNAs considerably reduced the frequency of ciliated cells, as compared with control siRNA treatment (Fig. 8B). More severe suppression of ciliogenesis was observed by knockdown of Tau tubulin kinase-2 (TTBK2), an essential gene for ciliogenesis (30 -32). These data suggest that the peroxisome plays a role in primary cilium formation.
Effect of NDR2 Knockdown on Peroxisome Assembly-The peroxisomal localization of NDR2 raises the possibility that it might be involved in peroxisome formation. To examine this possibility, we analyzed the effect of NDR2 knockdown on the formation of peroxisomes. RPE1 cells were transfected with control siRNA or NDR2-targeting siRNAs, and the formation of peroxisomes was analyzed by immunostaining using an anti-catalase antibody. Treatment with NDR2-targeting siRNAs had no apparent effect on the punctate localization of catalase (Fig. 9), which suggests that NDR2 does not play an essential role in peroxisome formation.

Discussion
Despite high sequence similarity, NDR1 and NDR2 display distinct subcellular localizations; NDR2 localizes to the vesicu-  lar compartments in the cytoplasm, whereas NDR1 is diffusely distributed within cells. In this study, we found that NDR2 colocalizes with catalase and CFP-SKL, indicating that NDR2 localizes to the peroxisomes. Deletion of the single leucine residue from the C-terminal GKL sequence abrogated the abilities of NDR2 to bind to the PTS1 receptor Pex5pS and to localize to the peroxisome. These observations suggest that NDR2 localizes to the peroxisome by using the C-terminal GKL motif as the peroxisomal targeting signal. Previous mutation analyses revealed that mutant proteins harboring the C-terminal GKL motif are capable of localizing to the peroxisome, albeit at a lower efficiency than those containing the conventional SKL motif (33,34), which supports our finding that the C-terminal GKL motif of NDR2 acts as the peroxisomal targeting signal. To our knowledge, NDR2 is the first known endogenous protein that localizes to the peroxisome by utilizing the GKL sequence as the peroxisomal targeting signal in mammalian cells.
In contrast to NDR2, NDR1 contains a C-terminal AK sequence, which does not conform to the consensus sequence of PTS1. We showed that NDR1 is diffusely distributed in cells, but its mutant NDR1(ϩL), containing an additional leucine at the C-terminal end, localizes to the peroxisome. NDR1(ϩL), but not wild-type NDR1, bound to Pex5pS. These results suggest that NDR1(ϩL) localizes to the peroxisome by using the artificially constructed C-terminal AKL sequence as the peroxisomal targeting signal. Thus, the distinct localization of NDR2 and NDR1 is due to the presence or absence of the single leucine residue at the C-terminal end of each protein, respectively.
We previously showed that knockdown of NDR2, but not NDR1, suppressed primary cilium formation in RPE1 cells (12), indicating that NDR2, but not NDR1, plays a crucial role in primary cilium formation. In this study, we showed that the expression of an siRNA-resistant wild-type NDR2, but not NDR2(⌬L), significantly ameliorated the suppressive effect of NDR2 knockdown on ciliogenesis. Furthermore, knockdown of PEX1 or PEX3 partially suppressed ciliogenesis. These results suggest that the peroxisomal localization of NDR2 is critical for its function to promote ciliogenesis. Expression of NDR1 also failed to restore the NDR2 knockdown-induced suppression of ciliogenesis. Notably, expression of the peroxisome-targeting NDR1(ϩL) also did not significantly counter the NDR2 knockdown-induced suppression of ciliogenesis. This result suggests that peroxisomal localization alone is not sufficient for NDR1 to promote ciliogenesis and that a yet uncharacterized difference in the properties of NDR1 and NDR2, such as target specificity, efficiency of enzyme activity, or regulation mechanism of enzyme activity, might also contribute to their functional disparity in promoting ciliogenesis.
Rabin8 and its downstream target Rab8 play crucial roles in ciliary vesicle formation in the early steps of ciliogenesis (15)(16)(17). NDR2 was shown to promote ciliogenesis via Rabin8 phosphorylation, which leads to local activation of Rab8 at the centrosome (12). It is unclear why the peroxisomal localization of NDR2 is favorable for its function in cilium formation. Because time-lapse fluorescence analysis of serum-starved RPE1 cells FIGURE 6. Topology analysis of NDR2. RPE1 cells were transfected with the plasmid for YFP-NDR2, and the PNS fraction was prepared. The PNS fraction was mock-treated (Ϫ) or treated with proteinase K (Pro-K) (ϩ) in the absence (Ϫ) or presence (ϩ) of Triton X-100 (TX-100). The reaction mixture was ultracentrifuged, and the resulting pellets were analyzed by SDS-PAGE and immunoblotting (IB) with antibodies against GFP (A), catalase (B), and PMP70 (C). FIGURE 7. The C-terminal GKL motif is required for NDR2 to promote ciliogenesis. A, effect of NDR2-targeting siRNA on the expression of endogenous NDR2. RPE1 cells were transfected with human NDR2-targeting siRNA and cultured for 72 h, and lysates were analyzed by immunoblotting (IB) with an anti-NDR2 antibody. B, effects of expression of YFP, YFP-NDR2 (WT or ⌬L), or YFP-NDR1 (WT or ϩL) on the ciliogenesis of NDR2 knockdown cells. RPE1 cells were transfected with the plasmids coding for YFP or YFP-tagged mouse NDR2, NDR1, or their mutants and cultured for 4 h. Then cells were transfected with human NDR2-targeting siRNA, cultured for 24 h, and serumstarved for 48 h. Cells were fixed and stained with anti-Ac-tubulin (red) and anti-pericentrin (green) antibodies. DNA was stained with DAPI (blue). Insets, magnified images (ϫ 2.7) of the white boxes. Arrows, primary cilia. C, quantitative analysis of the percentage of ciliated cells. RPE1 cells were treated as in B. Data are expressed as the means Ϯ S.E. (error bars) of three independent experiments (Ͼ100 cells/experiment, one-way ANOVA followed by a Dunnett's test). *, p Ͻ 0.05; **, p Ͻ 0.01. revealed that NDR2-carrying vesicles (identified herein as peroxisomes) are frequently associated with the centrosome (12) and that Rabin8 accumulates to the centrosome upon serum starvation (12,(15)(16)(17), the peroxisomal localization of NDR2 might be helpful to mediate the effective phosphorylation of Rabin8 at the centrosome after serum starvation. We showed that NDR2 is sensitive to proteinase K treatment, suggesting that at least a part of NDR2 is exposed to the cytosol. Thus, it is likely that NDR2 on the cytosolic side of the peroxisome is accessible to Rabin8 accumulating on the centrosome.
Peroxisomes are generally known to function to regulate the metabolism of lipids and reactive oxygen species (21), but recent studies have demonstrated a novel function of peroxisomes as signaling platforms (35)(36)(37)(38)(39)(40). For example, the peroxisome functions as the platform for the signaling of reactive oxygen species (ROS)-induced activation of tuberous sclerosis complex 1/2 (TSC1/2) (which is a GTPase-activating protein for Rheb GTPase) and the subsequent inhibition of Rheb and mammalian target of rapamycin complex 1 (mTORC1) (36). The signaling components, TSC1/2, Rheb, and mTORC1, are localized on the cytosolic surface of the peroxisomal membrane. TSC2 binds to the peroxisomal matrix protein receptor Pex5p through the internal PTS1-like sequence motif, but it resides on the outside of the peroxisomal membrane. TSC2 mutants lacking the ability to bind Pex5p are ineffective for ROS-induced mTORC1 inhibition, indicating that the Pex5pmediated peroxisomal membrane localization of TSC2 is critical for this response (36). A protein kinase, ataxia telangiectasia-mutated (ATM), also binds to Pex5p and is localized to the outer surface of the peroxisome (37,38). ROS stimulate the kinase activity of peroxisomal ATM, which induces mTORC1 inhibition and ULK1 activation via TSC1/2 activation, as well as the phosphorylation and ubiquitination of Pex5p, resulting in the ROS-induced autophagy of peroxisomes (pexophagy) (38). Peroxisomes also function as the signaling platform for antiviral innate immunity responses (39,40). Mitochondrial antiviral signaling protein (MAVS) is an organelle-bound adaptor protein for RIG-I-like receptors, which detect the double stranded RNAs of RNA viruses in the cytoplasm (39,40). MAVS localizes on both peroxisomes and mitochondria. Upon viral infection, peroxisomal MAVS triggers the immediate expression of antiviral factors that provide a transient antiviral response; this response is independent of mitochondrial MAVS (39,40). Together, these results suggest that the peroxisome serves not simply as a metabolic organelle but also as a signaling platform. The localization of NDR2 at the peroxisome suggests the possible role of this organelle as the platform for NDR2-mediated reactions as well.
Similar to TSC2 and ATM, NDR2 binds to Pex5p through the PTS1-like sequence motif and appears to reside on the exterior of the peroxisomal membrane. This is in contrast to most PTS1-harboring proteins, which are imported into the matrix of the peroxisome (21)(22)(23)(24)(25). How NDR2 is localized to the outside of the peroxisomal membrane remains intriguing. Because Pex5p associated with peroxisomes partly behaves as an integral membrane protein (41,42), it is conceivable that such Pex5p recruits and tethers NDR2 to the cytosolic side of the  peroxisome. It is also possible that NDR2 is recruited to the peroxisome by Pex5p and then tethered to the membrane by any unknown NDR2-anchoring factor(s). Further studies are required to address this point.
Knockdown of NDR2 had no apparent effect on peroxisome assembly, suggesting that NDR2 has no or little contribution to the process of peroxisome biogenesis, at least under our experimental conditions. However, the possibility that NDR2 has the function to regulate peroxisome biogenesis under distinct conditions and/or peroxisomal metabolic functions by modifying peroxisomal enzymes cannot be excluded. Although databases such as PhosphoSite indicate that many Pex proteins and peroxisomal enzymes are phosphorylated, little is known about the physiological roles of phosphorylation of these proteins in the biogenesis and functions of peroxisomes. The roles of NDR2 in peroxisome biogenesis and function deserve further investigation.
The present study demonstrates the hitherto unrealized involvement of peroxisomes in primary cilium formation, which might open up a new line of investigation toward understanding the novel pathophysiological role of the peroxisome. Very recently, Pex6p was shown to localize to the connecting cilia of retinal photoreceptor cells, and homozygous mutation of the PEX6 gene was found to cause retinal degeneration, deafness, and microcephaly (43). These symptoms considerably overlap with those associated with ciliopathy and centriole dysfunction (13,14,44). Further studies on the possible linkage of peroxisome biogenesis to cilium and centriole formation will pave the way to a better understanding of the mechanisms underlying peroxisome biogenesis disorders, ciliopathies, and their relationships.

Experimental Procedures
Antibodies-The antibodies used in this study were purchased as follows: rabbit polyclonal antibodies against EEA1 , and Ac-tubulin (Sigma-Aldrich, 6-11-B). Rabbit polyclonal antibodies against rat Pex14p were as described (45). Rabbit polyclonal antibodies against NDR2 were raised against the peptide (amino acid residues 417-433) in the C-terminal region of NDR2, as described previously (12).
Cell Culture and Transfection-Human immortalized RPE1 cells were a gift from H. Nakanishi (Kumamoto University, Kumamoto, Japan) and were cultured in Dulbecco's modified Eagle's medium/F-12 (Wako Pure Chemical Industries, Osaka, Japan) supplemented with 10% fetal calf serum. Human embryonic kidney HEK293T cells were purchased from the American Type Culture Collection (ATCC (Manassas, VA), CRL-3216) and cultured in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum. Plasmids were transfected into cells using the FuGENE-HD or FuGENE-6 transfection reagent (Promega, Tokyo, Japan) following the manufacturer's instructions.
Immunofluorescence Staining-RPE1 cells were grown on coverslips in 35-mm dishes. Cells were fixed with 10% trichloroacetic acid for 15 min or 4% paraformaldehyde at 37°C for 15 min, followed by permeabilization with 0.1% Triton X-100 for 5 min. Cells were then incubated overnight at 4°C with appropriate primary antibodies that were diluted with 2% fetal calf serum. Alexa Fluor 633-labeled anti-mouse IgG or anti-rabbit IgG antibodies were used as secondary antibodies. DNA was stained with 4Ј,6-diamidino-2-phenylindole (DAPI). Fluorescent images were obtained using an inverted fluorescence microscope (Leica (Wetslar, Germany), DMI6000B) equipped with a Plan-Apochromat ϫ63 oil immersion objective lens (numeric aperture 1.3) and processed using ImageJ software (National Institutes of Health, Bethesda, MD).
Immunoblotting-Denatured proteins were resolved by SDS-PAGE and transferred to Immobilon-P membranes (Millipore, Billerica, MA). The membranes were blocked with skim milk and incubated with the appropriate antibodies. A horseradish peroxidase-conjugated anti-rabbit or mouse IgG antibody was used as the secondary antibody.
Subcellular Fractionation-RPE1 cells or HeLa cells expressing YFP-NDR2 were homogenized in homogenization buffer (20 mM HEPES-KOH, pH 7.4, 0.25 M sucrose, 1 mM dithiothreitol, and protease inhibitor mixture (Roche Diagnostics)) and centrifuged at 800 ϫ g for 10 min at 4°C to yield PNS fraction. Total organellar fraction and cytosolic fraction were prepared by centrifugation of the PNS fraction at 100,000 ϫ g for 30 min. For fractionation by ultracentrifugation on an iodixanol density gradient, the PNS fraction of HeLa cells was mixed with iodixanol (Alere Technologies AS, Oslo, Norway) to yield a final 30% iodixanol solution. The PNS solution was loaded beneath a 14 -26% iodixanol gradient and separated by ultracentrifugation in an SW41Ti rotor (Beckman Coulter) at 100,000 ϫ g for 90 min at 4°C. The gradient was collected into 12 fractions. Organelles in each fraction were pelleted by dilution with an equal volume of buffer and centrifugation at 100,000 ϫ g for 20 min at 4°C and subjected to immunoblot analysis.
Proteinase K Sensitivity Assay-The PNS fractions (140 g of protein each) of YFP-NDR2-expressing RPE1 cells were treated with 1 g/ml proteinase K for 2 min at 0°C in the presence or absence of 1% Triton X-100. After terminating the reaction with 3 mg/ml phenylmethylsulfonyl fluoride, the PNS fraction was centrifuged at 100,000 ϫ g for 30 min, and the resulting pellets were analyzed by SDS-PAGE and immunoblotting.
Reverse Transcription (RT)-PCR-RT-PCR was performed as described previously (47). Total RNA was isolated using an Isogen II kit (Nippon Gene, Tokyo, Japan) and reverse-transcribed to yield single-stranded cDNAs using a Transcriptor high fidelity cDNA synthesis kit (Roche Diagnostics). The cDNA fragments were amplified by PCR amplification using GoTaq DNA polymerase (Promega) and analyzed by agarose gel electrophoresis.
Assay for Primary Cilium Formation-An assay for primary cilium formation was carried out as described previously (12,32). RPE1 cells were plated on coverslips in 35-mm plates at 1.0 ϫ 10 5 cells/plate, transfected with 10 nM siRNAs using Lipofectamine RNAiMax (Invitrogen), cultured for 24 h, and then serum-starved for 48 h. In knockdown/rescue experiments, before siRNA transfection, cells were transfected with expression plasmids using FuGENE-HD and cultured for 4 h. Cells were fixed with 10% trichloroacetate and immunostained with anti-Ac-tubulin and anti-pericentrin antibodies. Alexa Fluor 568-labeled anti-mouse IgG and Alexa Fluor 633-labeled antirabbit IgG antibodies were used as secondary antibodies.
Statistical Analysis-Statistical data are represented as the means Ϯ S.E. of three independent experiments. All statistical analyses were carried out using Prism 6 (GraphPad Software, La Jolla, CA). The p values were calculated using one-way ANOVA followed by Dunnett's test for multiple data set comparisons. In all cases, p Ͻ 0.05 was considered statistically significant.  Fig. 2 and supplemental Movie S1. Y. F. produced several materials and provided intellectual input. K. M. conceived and designed the project and wrote the manuscript. All authors reviewed the results and approved the final version of the manuscript.