CORE Metadata, citation and similar papers at core.ac.uk

Provided by Elsevier - Publisher Connector

Centriolar Kinesin Kif24 Interacts with CP110 to Remodel Microtubules and Regulate Ciliogenesis

Tetsuo Kobayashi,1,5 William Y. Tsang,1,2,3,5,6 Ji Li,1 William Lane,4 and Brian David Dynlacht1,* 1Department of Pathology and Cancer Institute, Smilow Research Center, New York University School of Medicine, 522 1st Avenue, New York, NY 10016, USA 2Faculte´ de Me´ decine, Universite´ de Montre´ al, Montre´ al H3C 3J7, Canada 3Division of Experimental Medicine, McGill University, Montre´ al H3A 1A3, Canada 4Mass Spectrometry and Proteomics Resource Laboratory, Center for Systems Biology, Harvard University, 52 Oxford Street, Cambridge, MA 02138, USA 5These authors contributed equally to this work 6Present address: Institut de recherches cliniques de Montre´ al, 110 avenue des Pins Ouest, Montre´ al H2W 1R7, Canada *Correspondence: [email protected] DOI 10.1016/j.cell.2011.04.028

SUMMARY structures that are thought to be critical for membrane attach- ment and cilia formation. New centrioles, or procentrioles, are We have identified a protein, Kif24, that shares formed using pre-existing centrioles as templates in S phase homology with the kinesin-13 subfamily of motor and continue to elongate in G2 phase. During this time, the proteins and specifically interacts with CP110 and daughter centriole acquires appendages and matures to Cep97, centrosomal proteins that play a role in regu- become the mother centriole. Upon cell-cycle exit, the mother lating centriolar length and ciliogenesis. Kif24 prefer- centriole can convert to a basal body, which nucleates the entially localizes to mother centrioles. Loss of Kif24 assembly of a cilium. Kinesins are molecular motor proteins that are involved in from cycling cells resulted in aberrant cilia assembly intracellular transport (Vale, 2003). They bind to microtubules but did not promote growth of abnormally long and use ATP to slide along these filamentous structures, trans- centrioles, unlike CP110 and Cep97 depletion. We porting protein complexes, membrane vesicles, and organelles found that loss of Kif24 leads to the disappearance (Hirokawa and Noda, 2008). There are at least 45 mammalian of CP110 from mother centrioles, specifically in kinesin (Kif) genes, and their gene products are often grouped cycling cells able to form cilia. Kif24 is able to bind into three types according to the position of the motor domain and depolymerize microtubules in vitro. Remarkably, within the protein. N-kinesins and C-kinesins encode motor ectopically expressed Kif24 specifically remodels domains that reside in amino-terminal and carboxy-terminal centriolar microtubules without significantly altering regions, respectively. In contrast, the motor domain resides in cytoplasmic microtubules. Thus, our studies have the middle of the coding region in M-kinesins (also known as identified a centriolar kinesin that specifically re- internal, or Kin I, motor proteins). Broadly speaking, N- and C-kinesins function to move cargo toward the plus and minus models a subset of microtubules, thereby regulating ends of microtubules, respectively, whereas M-kinesins have cilia assembly. These studies also suggest mecha- been reported to possess microtubule-depolymerizing activity nistic differences between the regulation of microtu- (Desai et al., 1999; Moores et al., 2002; Ogawa et al., 2004; bule elongation associated with centrioles and cilia. Ovechkina et al., 2002; Shipley et al., 2004). M-kinesins can be further clustered phylogenetically and grouped into the kinesin- INTRODUCTION 13 family, which includes four human and mouse proteins: Kif2A, Kif2B, Kif2C/MCAK, and Kif24 (Ems-McClung and Walc- Centrosomes are the major microtubule-nucleating centers in zak, 2010). The Kif2 proteins are also conserved in Drosophila animal cells and play important roles in cell-cycle progression, and have been studied extensively. However, the function of cell division, and cilia formation (for review, see Bettencourt- Kif24 is currently unknown. Dias and Glover, 2007). They are composed of two perpendicu- Given that kinesins associate with microtubules, it is not larly oriented centrioles surrounded by a proteinaceous pericen- surprising that they are involved in centrosome and cilia function. triolar matrix from which microtubules emanate and elongate. For example, Kif3a is an N-kinesin that is required for antero- The two centrioles, termed mother and daughter, are morpho- grade transport of protein complexes from the basal body to logically distinct: the mother centriole is the older of the two the tip of the cilium, a process known as intraflagellar transport centrioles and assembles subdistal and distal appendages, (Cole et al., 1998; Kozminski et al., 1995; Ou et al., 2005).

914 Cell 145, 914–925, June 10, 2011 ª2011 Elsevier Inc. Likewise, Kif17 plays a role in ciliary targeting of an olfactory Sequence alignment revealed that many amino acids essential cyclic nucleotide-gated channel, which is critical for sensing for ATPase activity, including those within the P loop or Walker stimuli (Jenkins et al., 2006). In addition, it has been shown A motif, are found in the Kif24 motor domain (Figure 1B). We that Kif7 is a ciliary protein that functions in Sonic hedgehog performed immunofluorescence experiments to detect Flag- signaling during development (Endoh-Yamagami et al., 2009; tagged Kif24 and showed that the recombinant protein colocal- Liem et al., 2009). Kinesin-13 members in Giardia, Leishmania ized with centrin at centrosomes (Figure S1A available online). major (LmjKIN13-2), and Chlamydomonas (CrKinesin-13) GFP-Kif24 is likewise detected at centrosomes (Figure S5D). localize to axonemes and are involved in flagellar length control Centrosomal localization of Flag-Kif24 was not affected by treat- (Blaineau et al., 2007; Dawson et al., 2007; Piao et al., 2009), ment with the microtubule depolymerizing drug nocodazole, although the mechanisms by which these kinesin-13 family indicating that it is a stable, intrinsic component of centrosomes proteins dynamically regulate microtubule assembly and disas- (Figure S1A). To confirm interactions with Kif24 in cells, we trans- sembly are unknown. fected plasmids expressing Flag-Kif24 in HEK293 cells, per- We have previously identified a centrosomal protein, CP110, formed anti-Flag immunoprecipitations, and demonstrated that and shown that it is important for centrosome and ciliary function both CP110 and Cep97 were coprecipitated (Figure S1B). In (Chen et al., 2002; Spektor et al., 2007; Tsang et al., 2006, 2008, contrast, Cep290, a previously characterized CP110-interacting 2009). CP110 and an associated protein, Cep97, regulate the protein, did not bind to Flag-Kif24 (Figure S1B). Next, we raised length of centrioles and prevent aberrant formation of primary antibodies against Kif24 to detect the endogenous protein. We cilia (Kohlmaier et al., 2009; Schmidt et al., 2009; Spektor showed that these antibodies were specific for Kif24 by trans- et al., 2007). CP110 is destroyed at the basal body prior to cilia fecting siRNAs that depleted the full-length protein from crude assembly, and ectopic expression is sufficient to inhibit the lysates (Figure S1C). In agreement with our ectopic expression formation of cilia (Spektor et al., 2007). Here, we performed studies, we found that endogenous CP110 and Cep97 were parallel immunoaffinity purification of CP110 and Cep97 and immunoprecipitated with anti-Kif24 antibodies (Figure 1C). identified a candidate kinesin, Kif24. Endogenous Kif24 associ- Reciprocal immunoprecipitations with CP110 and Cep97 ates with CP110 and Cep97, but not with Cep290, and preferen- antibodies confirmed their physiological interaction with Kif24 tially localizes to the mother centriole. Kif24 is a member of the (Figure 1C). We also detected Kif24 in Cep76 immunoprecipi- kinesin-13 subfamily. Remarkably, depletion of Kif24 leads to tates, although this interaction was not observed in the reverse the formation of primary cilia in cycling human epithelial cells, direction (Figure 1C). However, endogenous Kif24 did not asso- a phenotype that is reminiscent of CP110 or Cep97 loss. ciate with Cep290 (Figure 1C). Furthermore, we found that Kif24 However, unlike depletion of CP110 and Cep97, loss of Kif24 associates with CP110 and Cep97 in multiple cell lines, including does not promote growth of extra-long centrioles in nonciliated HEK293, human diploid retinal pigment epithelial (RPE-1), and cells. Kif24 depletion also results in the disappearance of U2OS cells (Figure 1C and Figure S1D). Together, these data CP110 from mother centrioles in cells that are able to form primary indicate that Kif24 is a component of a CP110-Cep97 complex cilia. Furthermore, consistent with the known function of other described previously (Spektor et al., 2007). kinesin-13 family members, Kif24 possesses microtubule depoly- merizing activity in vitro, and ectopic expression of Kif24 specifi- Kif24 Localization and Expression in Quiescent cally remodels centriolar, but not cytoplasmic, microtubules. and Cycling Cells Thus, Kif24 is a centriolar kinesin that specifically regulates cilio- Next, we assessed Kif24 subcellular localization and protein genesis by dynamically controlling microtubule polymerization. levels during cell-cycle progression. We used immunofluores- cence microscopy to detect Kif24 localization at different stages RESULTS of the cell cycle in RPE-1 cells. In G1 phase cells, Kif24 anti- bodies preferentially stained one dot that overlapped substan- Identification of Kif24 as a Cep97- and tially with a canonical centriolar marker, centrin (Figure 1D). CP110-Interacting Protein This signal disappeared in Kif24-depleted cells, indicating that In our initial proteomic screen for CP110-binding partners, we Kif24 antibodies specifically recognize endogenous Kif24 at identified two centrosomal proteins, Cep97 and Cep76, each the centrosome (Figure S1E). Two Kif24 foci were apparent in of which functions in centriole duplication and ciliogenesis S and G2 phases, and they overlapped with the brighter centrin (Spektor et al., 2007; Tsang et al., 2009). However, this screen dots, rather than the weaker dots that are associated with pro- failed to uncover proteins with enzymatic activities that could centrioles (Figure 1D). We found that the Kif24 signal peaked explain the ability of CP110 to suppress ciliogenesis. Therefore, during S/G2 (Figure 1D). Once cells entered mitosis, the Kif24 we performed a second proteomic screen to identify CP110- and signal decreased considerably, becoming more diffuse and Cep97-interacting proteins. Immunoaffinity chromatography less concentrated at the centrioles (Figure 1D). In addition, we and tandem mass spectrometry identified one candidate, analyzed Kif24 localization during ciliogenesis. Upon serum named kinesin superfamily protein 24 (Kif24), which copurified withdrawal, RPE-1 cells develop primary cilia, which we de- with both CP110 and Cep97. Kif24 encodes an uncharacterized tected using an antibody against glutamylated tubulin, which protein of 1368 acids with a putative protein interaction (SAM) visualizes the primary cilium, as well as the daughter centriole domain and a predicted motor domain that binds microtubules and the mother centriole/basal body. We found that Kif24 and ATP (Figure 1A). Kif24 is conserved in mammals, and bovine, primarily localized to the mother centriole/basal body and was canine, rodent, and chicken homologs have been identified. either absent from or substantially reduced at daughter

Cell 145, 914–925, June 10, 2011 ª2011 Elsevier Inc. 915 A B 1 1368 P-loop L9-loop L11-loop

SAM Motor (ATP-binding) 1-64 220-477

C IP D G1 S/G2 Late G2 N Cep97 C Control CP110 Cep76 Cep290 Kif24 IN C C C C K Kif24 Cep290 Cep97 Prometaphase Metaphase CP110 Cep76 Kif24 E F Centrin Kif24/GT335/Merge Kif24/Cep170/Merge Merge

Anaphase Telophase

G1 G2 G Kif24/γ-tubulin/Merge H I T98G U2OS RPE1 G0 G1 G1/S S G2/M AS G1 G1/S G2 M G0 AS Kif24 Kif24 Kif24 α-tubulin α-tubulin ββ-Actin - 1.0 2.5 3.7 2.7 0.3 1.0 1.9 2.7 2.3 1.0 2.8 (Fold increase of G1) (Fold increase of G1) (Fold increase of G0) Cep97 siRNA Cep97 siRNA

Figure 1. Kif24 Interacts with Cep97 and CP110, Preferentially Localizes to the Mother Centriole, and Is Cell Cycle Regulated (A) Domains of Kif24. The SAM domain and the kinesin motor domain reside at amino acid residues 1–64 and 220–477, respectively. Amino acids 313–320 are predicted to be critical for ATP binding. (B) Amino acid alignment of the motor domains of Kif24 and other kinesin superfamily proteins. (C) Western blotting of endogenous Kif24, Cep290, Cep97, CP110, and Cep76 after immunoprecipitation from HEK293 cell extracts with antibodies indicated at the top of panel. IN denotes input. (D) RPE-1 cells in diverse stages of the cell cycle were processed for immunofluorescence with anti-Kif24 (red) and anti-centrin (green) antibodies. (E) Quiescent RPE-1 cells were processed for immunofluorescence with anti-Kif24 (red) and anti-glutamylated tubulin (GT335, green) antibodies. The percentage of cells in which Kif24 preferentially localizes to the basal body was 58% ± 9.5%. (F) RPE-1 cells were processed for immunofluorescence with anti-Kif24 (red) and anti-Cep170 (green) antibodies. (G) Ablation of Cep97 function with siRNA has no effect on the localization of Kif24 to centrosomes and does not cause Kif24 to localize along the length of an elongated centriole in U2OS cells. (H and I) T98G or RPE-1 cells were synchronized by serum deprivation and restimulation, and western blots showing endogenous Kif24 protein levels are shown. a-tubulin or b-actin was used as a loading control. U2OS cells were synchronized with mimosine (G1) or HU (G1/S), blocked with HU, and released for 6 hr (S/G2) or were nocodazole treated (M). Lysates from different cell-cycle stages were collected, and western blots showing endogenous Kif24 protein levels are shown. a-tubulin was used as a loading control. AS, asynchronous population. Amount of Kif24 was quantified using ImageJ, and the relative values are shown below the pictures. (D–G) Scale bars, 10 mm. Here and where indicated in subsequent figures, DNA staining with DAPI is shown (blue), but this individual channel is omitted for simplicity. (H and I) Synchronization of populations at different cell-cycle stages was determined by FACS analysis. See also Figure S1 and Figure S2. centrioles (Figure 1E). However, use of excessive amounts of ages (Figure 1F and Figure S2A). Moreover, Kif24 associated antibody resulted in the detection of Kif24 at both centrioles in with elongated centrioles induced upon Cep97 or CP110 deple- ciliated cells (data not shown), suggesting that the protein may tion (Figure 1G, right and data not shown) (Spektor et al., 2007), exist at lower levels at the daughter as well. Kif24 staining was but it did not extend throughout these structures, reminiscent of never observed along the ciliary axoneme (Figure 1E). Instead, the localization of other mother centriole proteins (Odf2, ninein, Kif24 localized to the distal portion of the mother centriole at or and Cep164) (Kohlmaier et al., 2009; Schmidt et al., 2009). near the transition zone. In support of our conclusion that Kif24 We also examined Kif24 levels in synchronized populations is primarily at the mother centriole/basal body, we found that and found that Kif24 levels were relatively low in quiescent cells Kif24 staining exhibited near-complete overlap with Cep170, as compared to growing cells, and its abundance increased as a mother centriole-specific marker residing at subdistal append- cells progressed through G1 (Figures 1H and 1I). Protein levels

916 Cell 145, 914–925, June 10, 2011 ª2011 Elsevier Inc. A B 1-219 220-477 1-477 478-923 924-1368 1-1368 IP: Flag Control 220-477 478-923 924-1368 1-1368 509-1368 Cep97 Flag CP110

γ-tubulin Flag

Merge IN: Cep97 IN: CP110

IP: Flag 1-477 478-709 1-709 710-923 478-923 Cep97 CP110 C CP110- Cep97- SAM Motor (ATP-binding) Amino Acids binding binding Centrosomal 1-1368 + + + Flag 1-219 NT NT - 220-477 - - + 1-477 - - + 478-709 - - - IN: Cep97 1-709 - - + IN: CP110 710-923 + + - 478-923 + + - 924- 1368 - - - 509-1368 + + - Cep97/CP110-binding domain Centrosome localization domain

Figure 2. Cep97- and CP110-Binding and Centrosomal Localization Domains in Kif24 Are Separable (A) The indicated fragments of Flag-tagged Kif24 were expressed in HEK293 cells and immunoprecipitated from lysates. Flag-Kif24 fusion proteins, CP110, or Cep97 were detected after western blotting the resulting immunoprecipitates. Input CP110 or Cep97 was detected in lysates from each transfection (IN). (B) U2OS cells were transiently transfected with plasmids expressing Flag-Kif24 truncation mutants and were stained with antibodies to Flag (red) and g-tubulin (green). Scale bar, 10 mm. (C) Summary of the results of in vivo binding experiments and colocalization studies using Kif24 truncation mutants. peaked during S and G2 phases and decreased as cells entered ing (Figures 2B and 2C). Several fragments encompassing the mitosis (Figures 1H and 1I), consistent with our immunofluores- amino-terminal portion of the protein localized to the centro- cence data. We found that the mobility of Kif24 shifted dramati- some, whereas other truncation mutants did not (Figures 2B cally during G2/M in multiple cell types, suggesting possible links and 2C). Interestingly, these results indicate that a region span- between a posttranslational modification, delocalization, and ning residues 220–477, which is distinct from the Cep97- and destruction of this protein (Figures 1H and 1I). We note that CP110-binding domain but coincides with the motor domain, Kif24 contains a potential destruction box (D box) that could targets Kif24 to the centrosome. promote Kif24 destabilization during mitosis and mitotic exit, although this possibility needs further exploration. Kif24 Depletion Results in Inappropriate Cilia Assembly, but Not Extra-Long Centrioles Characterization of Kif24 Functional Domains Next, we investigated the functional consequences of depleting To delineate the functional domains of Kif24 that enable its Kif24. CP110 and Cep97 can suppress the inappropriate binding to Cep97 and CP110, we generated a series of Flag- assembly of primary cilia in growing cells (Spektor et al., 2007). tagged Kif24 truncation mutants, transfected HEK293 cells, Because Kif24 interacts with both proteins, we investigated and performed anti-Flag immunoprecipitations (Figures 2A and whether it could regulate the dynamics of cilia formation using 2C). We found that several fragments of Kif24 containing resi- siRNA-mediated gene silencing. We transfected pools of four dues 710–923 are sufficient to interact with both Cep97 and small interfering RNAs (siRNAs) or two individual siRNAs and CP110. This region lies downstream of the predicted motor/ found that Kif24 levels were substantially reduced (by 75%– ATPase domain, indicating that the motor and Cep97/CP110- 90%) in RPE-1 cells (Figure 3A). We stained Kif24-depleted cells binding domains are separable. with antibodies recognizing a panel of ciliary markers, including Next, we transiently expressed Kif24 truncation mutants to glutamylated tubulin, detyrosinated tubulin, IFT88/polaris, and determine which portion(s) are required for centrosomal target- Arl13b. Remarkably, ablation of Kif24 led to a dramatic increase

Cell 145, 914–925, June 10, 2011 ª2011 Elsevier Inc. 917 A B C CP110 GT335 Merge Protein siRNA 30 siRNA GT335 RPE1 4 NS *

a 25 Detyr.y Tub. S N Kif2 Kif24 20 * Cep97 * Kif24 * CP110 15

Cep76 with primary cili 10 α-tubulin Detyr.Tub. Centrin Merge

% cells 5 NS 0 NS Kif24 CP110 siRNA siRNA siRNA Kif24

D E siRNA CP110 Centrin Merge F Protein siRNA U2OS NS 30

ntriole 25 * NS Kif24 Kif24 20 Cep97 15 CP110 Kif24 10 p>0.99 Cep76 ls with long cen 5 α-tubulin 0 CP110 % cell NS Kif24 CP110 siRNA siRNA siRNA

Figure 3. Depletion of Kif24 Induces the Formation of Primary Cilia, but Not Elongated Centrioles (A) Western blotting of Kif24, Cep97, CP110, and Cep76 in cycling RPE-1 cells treated with control (NS) or Kif24 siRNA for 48 hr. a-tubulin was used as a loading control. (B) Cycling RPE-1 cells transiently transfected with control or Kif24 siRNA were stained with antibodies against CP110 (red) and glutamylated tubulin (GT335) (green) or detyrosinated tubulin (red) and centrin (green). Scale bar, 5 mm. (C) The percentages of cycling RPE-1 cells with primary cilia were determined using either glutamylated tubulin (GT335) or detyrosinated tubulin (Detyr. tub.) as markers. The average of five independent experiments is shown. *p < 0.01 compared with NS. (D) Western blotting of Kif24, Cep97, CP110, and Cep76 in U2OS cells treated with control (NS) or Kif24 siRNA. a-tubulin was used as a loading control. (E) U2OS cells transiently transfected with control, Kif24, or CP110 siRNA were stained with antibodies to CP110 (red) and centrin (green). Scale bar,5mm. (F) The percentages of U2OS cells with elongated centrioles were determined by scoring for the presence of elongated centrin filaments. The average of three independent experiments is shown. *p < 0.01 compared with NS. (C and F) About 250 cells for each siRNA condition were scored each time. Error bars represent SD. (B and E) DNA was stained with DAPI (blue). See also Figure S2, Figure S3, and Figure S4. in primary cilia assembly in cycling RPE-1 cells, reminiscent of not significantly impact cell-cycle progression in RPE-1 cells, Cep97 and CP110 depletion (Figures 3B and 3C and Figures as determined by flow cytometry and immunofluorescent detec- S3A and S3B). We confirmed that these cilia were devoid of tion of the Ki-67 proliferation marker (Figures S4A and S4B). centrin and g-tubulin (Figure 3B and data not shown), as ex- Furthermore, Kif24 depletion had no effect on cell-cycle progres- pected, suggesting that they are bona fide cilia rather than elon- sion in nonciliated U2OS cells (Figure S4C). These data indicate gated centriolar filaments (Kohlmaier et al., 2009; Schmidt et al., that cilia assembly in Kif24-depleted cells does not occur as 2009). In contrast, cilia length was not affected by ablation a result of cell-cycle perturbations or induced cell-cycle exit. of Kif24 (Figure S3C). We next performed combined ablation of To test whether Kif24 silencing can also provoke assembly of Kif24 and CP110 in cycling RPE-1 cells, which resulted in extra-long centrioles, we performed analogous ablation experi- modest increases in cilia formation as compared to ablation of ments in U2OS cells, which do not form cilia. The extent of deple- individual proteins (Figures S3D and S3E), indicating that Kif24 tion in U2OS matched or exceeded what we observed in RPE-1 is likely to function in a common pathway with CP110 to cells (Figures 3D). Depletion of Kif24 did not affect the levels of suppress cilia assembly. Suppression of Kif24 expression did CP110 or two CP110-interacting proteins, Cep97 and Cep76

918 Cell 145, 914–925, June 10, 2011 ª2011 Elsevier Inc. B A siRNA GT335 CP110 Merge GT335 CP110 Merge 70 * 60

NS s 50 siNS 40 siKif24 30 Kif24 *

% of total cell 20 10 0 p C siRNA Centrin CP110 Merge D 90 >0.19 123,44< 80 Number of CP110 dots

NS s 70 siNS 60 50 siKif24 40 Kif24 30 % of total cell % 20 p>0.28 10 0 123,44< Number of centrin dots

E siRNA γTubulin CP110 Merge F siRNA Cep170 CP110 Merge

NS NS

Kif24 Kif24

Figure 4. Depletion of Kif24 Induces the Disappearance of CP110 from the Mother Centriole (A, C, E, and F) Cycling RPE-1 cells transiently transfected with control (NS) or Kif24 siRNAs were stained with antibodies against CP110 (red) and (A) glutamylated tubulin (GT335), (C) centrin, (E) g-tubulin, or (F) Cep170 (green). In (A), left and right panels correspond to cells without and with primary cilia, respectively. In (E), the percentages of cells with less than two g-tubulin dots after transfection with control (NS) and Kif24 siRNAs was 12.7% and 10.6%, respectively. In (F), the percentage of cells with one CP110 dot in daughter centrioles after transfection with Kif24 siRNA was 96.7%. DNA was stained with DAPI (blue). Scale bars, 5 mm. (B and D) The percentage of cycling RPE-1 cells with the indicated number of (B) CP110 dots and (D) centrin dots. Averaged data obtained from two to three independent experiments are shown. Approximately 150 cells for each siRNA transfection were scored each time. Error bars represent SD. *p < 0.01 compared with NS.

(Figure 3D). Furthermore, Kif24 depletion did not perturb centro- CP110 exercise control over both centriole length and somal localization of Cep97, CP110, g-tubulin, two subdistal ciliogenesis. appendage proteins (Cep170 and cenexin), or the distal appendage marker Cep164 (Figures S2B–S2E), suggesting Kif24 Recruits CP110 to Mother Centrioles in Cycling that these proteins are recruited to centrosomes independently Cells Able to Form Cilia of Kif24 and that Kif24 loss does not impede centriole matura- The distinct phenotypes that are associated with Kif24 ablation tion. Likewise, depletion of Cep97 or CP110 does not provoke in RPE-1 and U2OS cells prompted us to investigate possible loss of Kif24 from centrosomes (Figure 1G and data not shown). explanations for these results. We examined CP110 levels and Previous studies showed that depletion of CP110 in this cell line found that they were substantially reduced in Kif24-depleted leads to the formation of extended centrin and g-tubulin fila- RPE-1 cells (Figure 3A and Figure S3D). These data prompted ments (Figures 3E and 3F), and these filaments represent elon- us to examine whether centrosomal localization of CP110 in gated centrioles (Spektor et al., 2007; Kohlmaier et al., 2009; growing RPE-1 cells is affected by ablation of Kif24. Interest- Schmidt et al., 2009). Strikingly, suppression of Kif24 function ingly, depletion of Kif24 led to a decrease in the number of did not induce extra-long centrioles (Figures 3E and 3F and Fig- CP110 dots on centrioles in RPE-1 cells (Figures 4A and 4B). ure S2C), nor did it enhance or antagonize the centriolar pheno- In contrast, the number of centrin and g-tubulin dots was not type that is associated with loss of Cep97 or CP110 (data not affected by ablation of Kif24 (Figures 4C–4E), suggesting that shown). Thus, we conclude that Kif24 plays a specific role in centrosomes are not catastrophically disturbed by Kif24 deple- regulating the assembly of primary cilia, whereas Cep97 and tion. Furthermore, staining with Cep170, a mother centriole

Cell 145, 914–925, June 10, 2011 ª2011 Elsevier Inc. 919 A GFP GFP-Kif24 GFP-Kif24 GFP-Kif24 GFP-Kif24 GFP-Kif24C GFP-Kif2C /VD /KEC /VD,KEC

GFP

GT335

merge

B C D 25 60 ** dot * ot 20 * centrin 50 γTubulin 40 15 Ac.Tubulin

30 10 20 s with 1 GT335

% cells with 1 d % 5 10

% cell 0 0 control Kif24 Kif24/KEC expressed protein

expressed protein

Figure 5. Ectopic Expression of Kif24 Leads to Reduced Accumulation of Specific Proteins at Centrosomes (A) Cycling RPE-1 cells were transiently transfected with plasmids expressing GFP, GFP-Kif24, GFP-Kif24/VD, GFP-Kif24/KEC, GFP-Kif24/VD,KEC, GFP-Kif24C (924–1368), or GFP-Kif2C and were stained with an antibody to glutamylated tubulin (GT335) (red). For each panel, DNA was stained with DAPI (blue). Scale bar, 10 mm. (B and C) The numbers of (B) GT335 dots, (C) centrin, g-tubulin, or acetylated tubulin dots in GFP-positive cells were counted. Averaged data obtained from three to five independent experiments are shown. More than 100 cells in each transfection were scored. Error bars represent SD. We note that fewer than 2%of GFP-Kif24-expressing cells have no GT335 dots, most likely because Kif24 preferentially localizes to mother centrioles. (B) *p < 0.01 compared with control. (C) *p < 0.01, **p < 0.05 compared with control. Pictures of cells stained with centrin, g-tubulin, or acetylated tubulin antibodies are shown in Figures S5A–S5C. (D) Amino acid alignment of conserved and essential amino acids for depolymerizing activity in kinesin-13 subfamily proteins. See also Figure S5. marker, showed that CP110 specifically disappeared from the sequence homology to other kinesin-13 family members, but mother centriole (Figure 4F). Collectively, these data suggest its ability to remodel microtubules has not been explored. To that Kif24 is required to stabilize and/or recruit CP110 to the investigate the impact of enforced Kif24 expression on centro- mother centriole in cycling RPE-1 cells. On the other hand, some dynamics, we expressed GFP-Kif24 in cycling RPE-1 CP110 depletion had no effect on either Kif24 expression or cells and examined the impact on cytoplasmic as well as cen- localization to centrosomes (Figure S3D and data not shown). triolar microtubules. Interestingly, we found that the number of The fact that combined ablation of Kif24 and CP110 did not addi- glutamylated tubulin foci that are associated with centrosomes tively promote cilia assembly (Figure S3E) suggests that a critical was significantly reduced in cells expressing GFP-Kif24, but not function of Kif24 is to control cilia formation by regulating CP110 in cells expressing GFP alone or a carboxy-terminal fragment of dynamics. In striking contrast to RPE-1 cells, we found that the Kif24 (GFP-Kif24C) consisting of residues 924–1368 and lacking levels and centrosomal localization of CP110 in U2OS cells the motor and Cep97/CP110-binding domains (Figures 5A and were not affected by Kif24 depletion (Figures 3D and 3E and Fig- 5B). In contrast, GFP-Kif2C, a protein that is known to depoly- ure S2D). Taken together, these data suggest that CP110 plays merize cytoplasmic microtubules, exhibited a modest effect a key role in controlling assembly of cilia and centriole elonga- (Figures 5A and 5B). We also observed that the number of tion, whereas Kif24 regulates CP110 levels and function only in g-tubulin and acetylated tubulin dots significantly decreased cells that are able to form cilia. in cells expressing Kif24, whereas the number of centrin dots was not affected by Kif24 expression (Figure 5C and Figures Ectopic Kif24 Expression Remodels Centriolar Tubulin S5A–S5C). Given that glutamylated and acetylated tubulin are Members of the kinesin-13 family, including Kif2C, have been components of centriolar microtubules and a fraction of shown to depolymerize microtubules. Kif24 is related by g-tubulin localizes to centriolar microtubules (Hammond et al.,

920 Cell 145, 914–925, June 10, 2011 ª2011 Elsevier Inc. 2008; Moudjou et al., 1996), whereas centrin localizes to the Kif24 Possesses Microtubule-Depolymerizing Activity distal lumen of centrioles (Paoletti et al., 1996), these data In Vitro and In Vivo suggest that enhanced Kif24 expression preferentially affects The impact of Kif24 expression on centriolar tubulins, but not centriolar microtubules but does not completely abolish cytoplasmic microtubules, suggested that this protein could centriole integrity. Because a significant decrease in glutamy- play a role in remodeling at least a subset of microtubules. This lated tubulin dots was observed in cells with either low Kif24 prompted us to more directly examine whether the protein has expression (GFP-Kif24 localizes exclusively at centrosomes; the ability to depolymerize microtubules in vitro and in vivo. To Figures S5D and S5E) or high Kif24 expression (GFP-Kif24 examine the biochemical properties of Kif24, we used baculovi- protein aggregates were formed in the cytoplasm; Figures 5A rus to express Kif24. We were unable to produce the full-length and 5B), the observed impact on centrioles is not due to protein protein, and therefore we expressed a GST-tagged Kif24 aggregation into cytoplasmic foci. Furthermore, we found that fragment encompassing domains predicted to be required for Flag-tagged Kif24 also forms cytoplasmic foci (similar to those microtubule depolymerization (residues 93–547) (Ogawa et al., associated with high levels of GFP-Kif24 expression) and 2004) and purified the protein to apparent homogeneity. We induces a decrease in glutamylated tubulin dots (Figure S5F), incubated recombinant Kif24 with purified microtubules and confirming that the observed centriolar effect induced by found that the protein indeed binds to microtubules, similar to Kif24 overexpression is independent of the fused protein or other kinesins (Figure 6D). Next, we investigated whether Kif24 tag. Importantly, we showed that Kif24 expression does not could depolymerize microtubules, a feature shared by other lead to sequestration of centriolar proteins or observable kinesin-13 family members. Kif24 exhibited modest activity in defects in centrosome duplication (Figures S5A–S5C and data this in vitro microtubule depolymerization assay (Figures 6E not shown). Moreover, Kif24 expression does not alter the cell and 6F), in striking contrast to Kif2C, which exhibited robust cycle, and the number of glutamylated tubulin foci also microtubule-depolymerizing activity. These results, together decreased in quiescent RPE-1 cells after Kif24 expression (Fig- with our cell-based observations on Kif24 and Kif2C activity ure S4D and data not shown), indicating that this phenotype is (Figures 6A–6C), are in line with the fact that the bulk microtu- not caused by defects in centriolar duplication stemming from bules used in this assay were derived from cell extracts and cell-cycle arrest. are therefore highly enriched in cytoplasmic tubulins. Interest- Next, we asked whether the phenotype that is associated with ingly, we note that Kif24 exhibited similar depolymerizing activity ectopic production of Kif24 requires its microtubule-depolyme- in the presence of the nonhydrolyzable ATP analog, AMP-PNP, rizing activity. We constructed putative loss-of-function muta- as compared to ATP (Figure S6B). In contrast, Kif2C depolymer- tions (Kif24/VD, Kif24/KEC, and Kif24/VD, KEC) targeting several ized microtubules much more efficiently in the presence of ATP residues that are conserved in kinesin-13 family proteins and are (by nearly an order of magnitude under our assay conditions), required for microtubule-depolymerizing activity (Blaineau et al., suggesting that the depolymerizing activities that are associated 2007; Ogawa et al., 2004; Shipley et al., 2004)(Figure 5D). We with these two proteins could differ mechanistically. Further, the expressed these mutants in cycling RPE-1 cells and examined ability of Kif24 to disassemble microtubules in the presence of the impact on centriolar microtubules. We found that these AMP-PNP is reminiscent of the KinI motor of P. falciparum mutations effectively abolished the impact on the number of glu- (Moores et al., 2002). Taken together, these data indicate that tamylated tubulin, acetylated tubulin, and g-tubulin foci in cells, Kif24 exhibits microtubule-binding and depolymerizing activity suggesting that this phenotype requires Kif24-depolymerizing in vitro, but its intrinsic activity toward cytoplasmic microtubules activity (Figures 5A–5C and Figures S5A–S5E). Importantly, is much lower than that of Kif2C. these point mutations do not globally disrupt Kif24 structure As a further test of these results, we asked whether Kif24 could because each of these mutants interacts with Cep97 and specifically act on centriolar microtubules in vivo. We employed CP110 with affinities comparable to wild-type Kif24 and localizes two sensitive assays to test this premise. First, it has been shown to centrosomes (Figures S5D and S5G). that ablation of CP110 or Cep97 leads to excessive growth of Next, we examined whether cytoplasmic microtubules are centrioles, producing highly elongated structures several-fold affected by Kif24 expression. As a positive control, we ex- longer than normal centrioles (Kohlmaier et al., 2009; Schmidt pressed Kif2C, which has been shown to depolymerize cyto- et al., 2009; Spektor et al., 2007). Therefore, we examined plasmic microtubules in HeLa cells (Ogawa et al., 2004). whether elevated Kif24 expression suppresses the formation of Whereas Kif2C expression did not alter the number of glutamy- elongated centrioles that are induced upon Cep97 loss in lated tubulin dots in HeLa cells (Figure S6A), expression of this U2OS cells. Remarkably, visualization of several centriolar protein led to the disappearance of the cytoplasmic microtubule markers indicated that ectopic expression of Kif24 suppressed network, as expected (Figures 6A–6C). In striking contrast, quan- centriole elongation (Figures 7A and 7B and Figure S7A). titative comparisons demonstrated that expression of Kif24 did Second, we tested the ability of Kif24 to inhibit cilia assembly not appreciably affect depolymerization of cytoplasmic microtu- and found that its overexpression suppressed cilia formation in bules (Figures 6A–6C). Similar results were obtained in three quiescent NIH 3T3 and RPE-1 cells (Figures 7C and 7D, Fig- other cell lines (RPE-1, U2OS, and 3T3 cells; Figure 6A). Taken ure S7B, and data not shown). Importantly, expression of Kif24 together, these data suggest that the biochemical properties of mutants (Kif24/VD, Kif24/KEC, Kif24/VD, KEC) resulted in the two kinesins are different: Kif24 specifically remodels cen- marginal suppression in both assays, unlike the wild-type protein triolar microtubules, but not cytoplasmic microtubules, whereas (Figures 7B and 7D). We confirmed that Kif24/VD and Kif24/KEC Kif2C exhibits the opposite activity. exhibit a lower microtubule binding efficiency than wild-type

Cell 145, 914–925, June 10, 2011 ª2011 Elsevier Inc. 921 A B C 35 control Kif24 Kif2C 140 p>0.63 30 120 bules

merizied RPE1 25 U2OS αTubulin 100 Hela 20 3T3 80 * 15 60 % of control ls with depoly lasmic microtu 10 40 5 merge 20 % cel cytopl 0 0 control Kif24 Kif2C control Kif24 Kif2C expressed protein expressed protein D E F MT+ MT- 0 0.6 μM 1.2 μM 2.4 μM control Kif24 Kif2C IN S P SP IN S P SP SPSP SPSP SP GST- Kif24 Kif24 Kif24 GST-Kif2C Tubulin Tubulin Tubulin (%) 60 ) GST S) 2 * 40 S/(P+S) 11.2 3.1 (fold increase) ** 1 20 * (fold increase Tubulin S/(P+ Tubulin Kif24 P/(P+S) ( 0 0 MT + MT - Kif24- Kif24+

Figure 6. Kif24 Depolymerizes Cytoplasmic Microtubules with Low Efficiency In Vivo and In Vitro (A) RPE-1, HeLa, U2OS, and 3T3 cells were transiently transfected with plasmids expressing GFP, GFP-Kif24, or GFP-Kif2C and were stained with antibody against a-tubulin. Fluorescence intensity of microtubules in GFP-positive cells was quantified using ImageJ. Cells with depolymerized cytoplasmic microtubule (fluorescence intensity less than 50% of average of GFP-expressed cells) were shown (Ogawa et al., 2004). More than 30 cells for each expressed protein were scored. (B and C) HeLa cells were transiently transfected with plasmids expressing GFP and Flag, Flag-Kif24, or Flag-Kif2C and were stained with an antibody to a-tubulin. DNA was stained with DAPI (blue). Scale bar, 10 mm. (C) Fluorescence intensity of microtubules in GFP-positive cells was quantified using ImageJ (Lacroix et al., 2010). More than 20 cells for each expressed protein were scored. *p < 0.01 compared with control. (D) (Top) Microtubule binding assay using recombinant Kif24. 2.4 mM GST-Kif24 (93–547) was incubated in the presence (+) or absence (–) of polymerized microtubules (MT) and was separated into supernatant (S) and pellet (P) fractions after high-speed centrifugation. Coomassie blue staining of both tubulins and Kif24 is indicated. (Bottom) The percentage of Kif24 in the precipitated fraction for the experiment shown in the upper panel was calculated. (D) *p < 0.01 comparing reactions with (MT+) and without (MT-) microtubules. (E) (Top) Microtubule depolymerization assay was performed using the indicated amounts of recombinant Kif24 with 2 mM AMP-PNP. Coomassie blue staining of proteins separated into supernatant and pellet fractions as in (D). (Bottom) The fold increase of tubulin in the supernatant versus supernatant and pellet (tubulin [S]/tubulin [P+S]) was calculated for the maximal concentration of Kif24 (2.4 mM) in this experiment. Data from three independent experiments are shown. Error bars represent SD. *p < 0.01 comparing reactions without Kif24 (Kif24–) and with 2.4 mM Kif24 (Kif24+). (F) Microtubule depolymerizing assay was performed using recombinant 6.4 mM GST, 2.4 mM GST-Kif24, or 2.4 mM GST-Kif2C (187–589; these residues correspond to Kif24 [93–547] based on sequence alignment). Coomassie blue staining of proteins separated into supernatant and pellet fractions as in (D). The fold increase of tubulin in the supernatant versus supernatant and pellet (tubulin [S]/tubulin [P+S]) is shown below gel. *p < 0.01 compared with control. (C–E) Error bars represent SD. (D–F) Amount of protein was quantified using ImageJ. Data from three independent experiments are shown. See also Figure S6.

Kif24 using purified recombinant Flag-Kif24 proteins or lysates growth of extra-long centrioles and cilia formation. In addition, from GFP-Kif24-expressing cells (Figure S7C and data not we noticed that cilia length was somewhat reduced in Kif24-ex- shown), suggesting that these mutations affect microtubule pressing cells in comparison to controls (Figure S7E). Given that binding and therefore lead to lower depolymerizing activity. Kif24 expression does not affect cytoplasmic microtubules Furthermore, expression of the microtubule depolymerizing (Figures 6A–6C), these data indicate that Kif24 specifically domain identified in vitro (residues 93–547; Figures 6E and 6F depolymerizes microtubules of centriolar origin, and this and Figure S6B) inhibited cilia assembly, whereas other depolymerizing activity contributes to the regulation of cilia nonoverlapping truncation mutants had significantly less activity assembly. (Figure S7D). These data suggest that the microtubule-depoly- Finally, we carried out rescue experiments in which we merizing activity of Kif24 is indispensable for suppressing expressed wild-type Kif24 or the Kif24/KEC mutant in cycling

922 Cell 145, 914–925, June 10, 2011 ª2011 Elsevier Inc. AB C GFP GFP-Kif24 60 GFP GFP-Kif24 50 * * GFP 40 *

ong centriole GFP 30 * 20 10 centrin % cells with l 0 GT335

merge merge

siRNA + expressed protein

D E F 90 6 ol a 80 * cells with primary cilia 1) Kif24 st ta bilizes an d/ or 5 recruits CP110 to distal end of 70 cells with 1 CP110 dot mother centriole. 60 4 p>0.28 50 * 3 p 40 <0.02

ase over contro Kif24 CP110 with primary cili 30 2 ** 20 * 1 CP110 caps the distal 10 2) Kif24 remodels end of centriole. % cells 0 fold incre 0 centriolar microtubules. M CP110 D Kif24 expressed protein siRNA + expressed protein

Figure 7. Kif24 Possesses Centriolar Microtubule-Depolymerizing Activity In Vivo (A and B) U2OS cells treated with control or Cep97 siRNA were transiently transfected with plasmids expressing GFP, GFP-Kif24, GFP-Kif24/VD, GFP-Kif24/ KEC, or GFP-Kif24/VD, KEC. Cells were stained with antibody to centrin (red). (B) The percentages of U2OS cells with elongated centrioles were determined by scoring for the presence of elongated centrin filaments. *p < 0.01 compared with siNS+control. (C and D) 3T3 cells were transiently transfected with plasmids expressing GFP, GFP-Kif24, GFP-Kif24/VD, GFP-Kif24/KEC, or GFP-Kif24/VD, KEC and were brought to quiescence by serum starvation. Cells were visualized with antibody against glutamylated tubulin (GT335). (D) The percentages of 3T3 cells with cilia were determined using GT335. *p < 0.01 compared with control. (E) RPE-1 cells treated with control or Kif24 siRNA targeting the 30UTR of Kif24 mRNA were transiently transfected with plasmids expressing GFP, GFP-Kif24, or GFP-Kif24/KEC. Cells were stained with antibodies to glutamylated tubulin (GT335) or CP110. The fold increase in the percentage of cells with primary cilia or 1 CP110 dot was determined. *p < 0.01, **p < 0.05 compared with siNS+control. (A and C) DNA was stained with DAPI (blue). Scale bars, 10 mm. (B, D, and E) The average of two to three independent experiments is shown. More than 100 cells for each expressed protein were scored each time. Error bars represent SD. (F) Kif24 prevents inappropriate assembly of a primary cilium in cycling RPE-1 cells by stabilizing and/or recruiting CP110 at the distal end of mother centrioles and by specifically remodeling centriolar microtubules without affecting the cytoplasmic microtubule network. See also Figure S7.

RPE-1 cells depleted of endogenous Kif24. We found that the DISCUSSION two phenotypes associated with Kif24 depletion, namely, inap- propriate cilia formation and disappearance of CP110 from the The series of dynamic events that accompany the conversion of mother centriole, were largely rescued by Kif24 wild-type or a basal body to primary cilium, particularly the events that trigger KEC mutant expression (Figure 7E). We note that, whereas growth of the axoneme, are poorly understood. Presumably, both wild-type and mutant were equally capable of rescuing assembly of a primary mammalian cilium requires active remod- the loss of CP110 from the mother centriole, the KEC mutant eling of microtubules to generate micron-scale structures. Here, was less efficient in rescuing the inappropriate cilia formation we have identified and characterized a previously unexplored phenotype than wild-type (Figure 7E). These data suggest that, kinesin, Kif24, which could represent an important step toward in addition to recruiting and/or stabilizing CP110 to the centro- understanding these events. Our data suggest a model in which some, the depolymerizing activity of Kif24 also contributes to Kif24 suppresses inappropriate cilia assembly in cycling cells suppression of aberrant cilia formation. through two mechanisms (Figure 7F). First, it is able to stabilize

Cell 145, 914–925, June 10, 2011 ª2011 Elsevier Inc. 923 and/or recruit CP110, which may ‘‘cap’’ the distal ends of centri- were performed as described (Tsang et al., 2007). Additional information on oles. Because the presence of CP110 at mothers is incompatible cell lines is provided in the Supplemental Information. with cilia assembly, Kif24 could stabilize CP110 to prevent Antibodies and siRNAs assembly at inappropriate times. Interestingly, this activity For a complete list of antibodies and siRNAs, see Supplemental Information. appears to be restricted to certain cell types, as recruitment/ stabilization of CP110 did not occur in nonciliated U2OS cells. Plasmids, Recombinant Protein Expression, and Protein Purification Second, Kif24 counteracts microtubule polymerization by re- A detailed description of all cloning, expression, and purification of proteins modeling microtubules at the distal end of centrioles that could is provided in the Supplemental Information. otherwise lead to premature formation of cilia (Figure 7F). The Immunoprecipitation, Immunoblotting, and Immunofluorescence combination of these two mechanisms is likely to explain the Microscopy phenotypes in both ciliated and nonciliated cells. Thus, the ability Cells were lysed with ELB buffer (50 mM HEPES [pH 7], 250 mM NaCl, 5 mM of endogenous Kif24 to suppress cilia assembly, but not EDTA [pH 8], 0.1% NP-40, 1 mM DTT, 0.5 mM AEBSF, 2 mg/ml leupeptin, 2 mg abnormal centriole elongation, could reflect the cell type-specific aprotinin, 10 mM NaF, 50 mM b-glycerophosphate, and 10% glycerol) at 4C stabilization of CP110 in ciliated RPE-1 versus nonciliated U2OS for 30 min. For immunoprecipitation, 2 mg of the resulting supernatant after cells. Several questions remain, however, for future studies. For centrifugation was incubated with an appropriate antibody at 4 C for 1 hr example, we found that Kif24 persisted at basal bodies after cilia and collected using protein A or G Sepharose. The resin was washed with ELB buffer, and the bound polypeptides were analyzed by SDS-PAGE and assembly. This suggests that recruitment of additional compo- immunoblotting. 50 mg–100 mg of lysate were typically loaded in the input nents to a CP110 scaffold may be needed to facilitate depoly- (IN) lane. For mapping studies, Flag-tagged constructs were transfected into merization, and in the absence of CP110, the activity of Kif24 HEK293 cells. Cells were harvested 48–72 hr after transfection. Flag beads is somehow restricted. Alternatively, because Kif24 levels are were incubated with cell extract at 4C for 2 hr. After washing with ELB buffer, dramatically reduced in quiescent cells (Figure 1I), we suggest bound proteins were analyzed by SDS-PAGE and immunoblotting. Indirect that a critical enzyme concentration may be needed to antago- immunofluorescence was performed as described in the Supplemental Information. nize cilia assembly, and when concentrations of Kif24 fall below a certain threshold, cilia assembly is permissible. Microtubule Binding and Depolymerization Assay To our knowledge, Kif24 is the first mother centriole protein For the microtubule binding assay, recombinant Kif24 protein was incubated shown to possess microtubule-depolymerizing activity. with 2 mM AMP-PNP (Sigma) and 4 mM taxol-stabilized bovine tubulins Although we have not yet fully delineated the pathways that (90% of tubulins are polymerized into microtubules) (Cytoskeleton) in binding link the function of Kif24 with ciliogenesis, the identification of buffer (80 mM PIPES [pH 6.9], 1 mM EGTA, 2 mM MgCl2, 50 mM NaCl, 1 mM DTT and 20 mM Taxol) for 30 min at room temperature. After centrifugation this protein offers an important starting point for identifying addi- at 100,000 3 g for 15 min at 20C, supernatant and pellet fractions were tional components that facilitate assembly of a primary cilium. subjected to SDS-PAGE and stained with Coomassie blue. For the microtu- Because overexpression of Kif24 does not appear to destabilize bule depolymerization assay, recombinant Kif24 was incubated with 2 mM cytoplasmic microtubules, Kif24 could specifically act on cen- ATP or AMP-PNP and 4 mM taxol-stabilized microtubule in buffer (80 mM triolar microtubules. This scenario is in sharp contrast to CPAP PIPES [pH 6.9], 1 mM EGTA, 2 mM MgCl2, 50 mM NaCl, 1 mM DTT and (centrosomal P4.1-associated protein), which is required for 2 mM Taxol) for 60 min at room temperature and was subjected to ultracentri- centrosome duplication and possesses the ability to destabilize fugation as described above. cytoplasmic microtubules while apparently stabilizing centriolar SUPPLEMENTAL INFORMATION microtubules when ectopically expressed (Hsu et al., 2008; Kohl- maier et al., 2009; Tang et al., 2009). Consistent with the notion Supplemental Information includes Extended Experimental Procedures and that Kif24 and CPAP depolymerize different types of microtu- seven figures and can be found with this article online at doi:10.1016/j.cell. bules, the microtubule-depolymerizing domain of CPAP is 2011.04.028. distinct from the motor domain of Kif24 and other kinesin super- family proteins with regard to amino acid sequence. ACKNOWLEDGMENTS We do not yet know how Kif24 is specifically targeted to cen- We thank all members of the Dynlacht laboratory for constructive advice and triolar versus cytoplasmic microtubules. CP110 and Cep97 may encouragement. We thank M. Jelcic and I. Sanchez for assistance with baculo- constitute a part of the recruitment mechanism, but they cannot viral expression and thank J. Liu for experimental assistance. We thank J. Salis- completely explain this specificity. It is tempting to speculate bury, E. Nigg, K. Lee, C. Janke, T. Katada, and K. Kontani for providing anti- that one or more posttranslational modifications of tubulins— bodies. B.D.D. was supported, in part, by an Irma T. Hirschl Career Scientist which are detyrosinated, glutamylated, glycylated, and acety- Award and March of Dimes research grant, for which he expresses his gratitude. J.L. was supported by a DOD Prostate Cancer pre-doctoral Training Award. lated—that stabilize and distinguish this pool of microtubules could assist in precisely targeting Kif24. Future studies will be Received: July 20, 2010 needed to clarify this targeting mechanism. Revised: December 13, 2010 Accepted: April 28, 2011 Published online: May 26, 2011 EXPERIMENTAL PROCEDURES REFERENCES Cell-Cycle Synchronization and FACS Analysis T98G cells were synchronized by serum deprivation and restimulation as Bettencourt-Dias, M., and Glover, D.M. (2007). Centrosome biogenesis and described (Tsang et al., 2007). U2OS cells were synchronized as reported function: centrosomics brings new understanding. Nat. Rev. Mol. Cell Biol. previously (Tsang et al., 2007). Propidium iodide staining and FACS analysis 8, 451–463.

924 Cell 145, 914–925, June 10, 2011 ª2011 Elsevier Inc. Blaineau, C., Tessier, M., Dubessay, P., Tasse, L., Crobu, L., Page` s, M., and Moores, C.A., Yu, M., Guo, J., Beraud, C., Sakowicz, R., and Milligan, R.A. Bastien, P. (2007). A novel microtubule-depolymerizing kinesin involved in (2002). A mechanism for microtubule depolymerization by KinI kinesins. Mol. length control of a eukaryotic flagellum. Curr. Biol. 17, 778–782. Cell 9, 903–909. Chen, Z., Indjeian, V.B., McManus, M., Wang, L., and Dynlacht, B.D. (2002). Moudjou, M., Bordes, N., Paintrand, M., and Bornens, M. (1996). gamma- CP110, a cell cycle-dependent CDK substrate, regulates centrosome duplica- Tubulin in mammalian cells: the centrosomal and the cytosolic forms. J. Cell tion in human cells. Dev. Cell 3, 339–350. Sci. 109, 875–887. Cole, D.G., Diener, D.R., Himelblau, A.L., Beech, P.L., Fuster, J.C., and Ogawa, T., Nitta, R., Okada, Y., and Hirokawa, N. (2004). A common mecha- Rosenbaum, J.L. (1998). Chlamydomonas kinesin-II-dependent intraflagellar nism for microtubule destabilizers-M type kinesins stabilize curling of the transport (IFT): IFT particles contain proteins required for ciliary assembly in protofilament using the class-specific neck and loops. Cell 116, 591–602. Caenorhabditis elegans sensory neurons. J. Cell Biol. 141, 993–1008. Ou, G., Blacque, O.E., Snow, J.J., Leroux, M.R., and Scholey, J.M. (2005). 436 Dawson, S.C., Sagolla, M.S., Mancuso, J.J., Woessner, D.J., House, S.A., Functional coordination of intraflagellar transport motors. Nature , Fritz-Laylin, L., and Cande, W.Z. (2007). Kinesin-13 regulates flagellar, inter- 583–587. phase, and mitotic microtubule dynamics in Giardia intestinalis. Eukaryot. Ovechkina, Y., Wagenbach, M., and Wordeman, L. (2002). K-loop insertion Cell 6, 2354–2364. restores microtubule depolymerizing activity of a ‘‘neckless’’ MCAK mutant. 159 Desai, A., Verma, S., Mitchison, T.J., and Walczak, C.E. (1999). Kin I kinesins J. Cell Biol. , 557–562. are microtubule-destabilizing enzymes. Cell 96, 69–78. Paoletti, A., Moudjou, M., Paintrand, M., Salisbury, J.L., and Bornens, M. (1996). Most of centrin in animal cells is not centrosome-associated and Ems-McClung, S.C., and Walczak, C.E. (2010). Kinesin-13s in mitosis: Key centrosomal centrin is confined to the distal lumen of centrioles. J. Cell Sci. players in the spatial and temporal organization of spindle microtubules. 109, 3089–3102. Semin. Cell Dev. Biol. 21, 276–282. Piao, T., Luo, M., Wang, L., Guo, Y., Li, D., Li, P., Snell, W.J., and Pan, J. (2009). Endoh-Yamagami, S., Evangelista, M., Wilson, D., Wen, X., Theunissen, J.W., A microtubule depolymerizing kinesin functions during both flagellar disas- Phamluong, K., Davis, M., Scales, S.J., Solloway, M.J., de Sauvage, F.J., and sembly and flagellar assembly in Chlamydomonas. Proc. Natl. Acad. Sci. Peterson, A.S. (2009). The mammalian Cos2 homolog Kif7 plays an essential USA 106, 4713–4718. role in modulating Hh signal transduction during development. Curr. Biol. 19, 1320–1326. Schmidt, T.I., Kleylein-Sohn, J., Westendorf, J., Le Clech, M., Lavoie, S.B., Stierhof, Y.D., and Nigg, E.A. (2009). Control of centriole length by CPAP Hammond, J.W., Cai, D., and Verhey, K.J. (2008). Tubulin modifications and and CP110. Curr. Biol. 19, 1005–1011. their cellular functions. Curr. Opin. Cell Biol. 20, 71–76. Shipley, K., Hekmat-Nejad, M., Turner, J., Moores, C., Anderson, R., Milligan, Hirokawa, N., and Noda, Y. (2008). Intracellular transport and kinesin super- R., Sakowicz, R., and Fletterick, R. (2004). Structure of a kinesin microtubule 88 family proteins, KIFs: structure, function, and dynamics. Physiol. Rev. , depolymerization machine. EMBO J. 23, 1422–1432. 1089–1118. Spektor, A., Tsang, W.Y., Khoo, D., and Dynlacht, B.D. (2007). Cep97 and Hsu, W.B., Hung, L.Y., Tang, C.J., Su, C.L., Chang, Y., and Tang, T.K. (2008). CP110 suppress a cilia assembly program. Cell 130, 678–690. Functional characterization of the microtubule-binding and -destabilizing Tang, C.J., Fu, R.H., Wu, K.S., Hsu, W.B., and Tang, T.K. (2009). CPAP is a domains of CPAP and d-SAS-4. Exp. Cell Res. 314, 2591–2602. cell-cycle regulated protein that controls centriole length. Nat. Cell Biol. 11, Jenkins, P.M., Hurd, T.W., Zhang, L., McEwen, D.P., Brown, R.L., Margolis, B., 825–831. Verhey, K.J., and Martens, J.R. (2006). Ciliary targeting of olfactory CNG chan- Tsang, W.Y., Bossard, C., Khanna, H., Pera¨ nen, J., Swaroop, A., Malhotra, V., nels requires the CNGB1b subunit and the kinesin-2 motor protein, KIF17. and Dynlacht, B.D. (2008). CP110 suppresses primary cilia formation through 16 Curr. Biol. , 1211–1216. its interaction with CEP290, a protein deficient in human ciliary disease. Dev. Kohlmaier, G., Loncarek, J., Meng, X., McEwen, B.F., Mogensen, M.M., Spek- Cell 15, 187–197. tor, A., Dynlacht, B.D., Khodjakov, A., and Go¨ nczy, P. (2009). Overly long Tsang, W.Y., Spektor, A., Luciano, D.J., Indjeian, V.B., Chen, Z., Salisbury, centrioles and defective cell division upon excess of the SAS-4-related protein J.L., Sa´ nchez, I., and Dynlacht, B.D. (2006). CP110 cooperates with two 19 CPAP. Curr. Biol. , 1012–1018. calcium-binding proteins to regulate cytokinesis and genome stability. Mol. Kozminski, K.G., Beech, P.L., and Rosenbaum, J.L. (1995). The Chlamydomo- Biol. Cell 17, 3423–3434. nas kinesin-like protein FLA10 is involved in motility associated with the Tsang, W.Y., Spektor, A., Vijayakumar, S., Bista, B.R., Li, J., Sanchez, I., flagellar membrane. J. Cell Biol. 131, 1517–1527. Duensing, S., and Dynlacht, B.D. (2009). Cep76, a centrosomal protein that Lacroix, B., van Dijk, J., Gold, N.D., Guizetti, J., Aldrian-Herrada, G., specifically restrains centriole reduplication. Dev. Cell 16, 649–660. Rogowski, K., Gerlich, D.W., and Janke, C. (2010). Tubulin polyglutamylation Tsang, W.Y., Wang, L., Chen, Z., Sa´ nchez, I., and Dynlacht, B.D. (2007). stimulates spastin-mediated microtubule severing. J. Cell Biol. 189, 945–954. SCAPER, a novel cyclin A-interacting protein that regulates cell cycle progres- 178 Liem, K.F., Jr., He, M., Ocbina, P.J., and Anderson, K.V. (2009). Mouse Kif7/ sion. J. Cell Biol. , 621–633. Costal2 is a cilia-associated protein that regulates Sonic hedgehog signaling. Vale, R.D. (2003). The molecular motor toolbox for intracellular transport. Cell Proc. Natl. Acad. Sci. USA 106, 13377–13382. 112, 467–480.

Cell 145, 914–925, June 10, 2011 ª2011 Elsevier Inc. 925