Mapping Centrosome-targeting Domains of ALMS1
ALMS1 is a large (4169 amino acid) protein with several notable sequence features, including an extensive tandem repeat domain, a putative leucine-zipper motif, and an ALMS motif (A). Unlike many other centrosomal proteins, it has limited potential to form coiled-coils (A). To query if a particular motif or domain targets ALMS1 to centrosomes, we generated a panel of HA-tagged deletion constructs for immunolocalization analysis. We transiently expressed constructs in U2OS cells and compared their localizations to that of the centrosome marker γ-tubulin. Briefly, this suggested that the C-terminal quarter of ALMS1 (represented in construct ΔN-3175) is important for efficient centrosome-targeting, but also implicated an internal region (residues 2261–2602; , A and B, and data not shown). In general, specific and compact centrosomal localization of transfected ALMS1 constructs was observed at low expression levels, whereas additional diffuse intracellular staining was apparent at higher expression levels.
Notably, at moderate-to-high expression levels the construct representing residues 2261–2602 gave punctate staining throughout the cytoplasm and appeared to form aggregates, one of which typically coincided with the centrosome (B). Misfolded proteins can accumulate in the vicinity of the centrosome via the aggresome pathway (
Kopito, 2000 
). However, at low expression levels, construct 2261–2602 (and construct ΔN-3175) colocalized specifically with endogenous ALMS1 (Supplemental Figure S1), suggesting that at least a fraction of this construct assembles at the centrosome similarly to the endogenous protein. The antibody used to detect endogenous ALMS1 recognizes an epitope within the tandem repeat domain, which is not present in either construct 2261–2602 or ΔN-3175 (A). Deletion of the putative leucine-zipper motif in construct 2261–2602, which eliminated its short predicted coiled-coil, did not appear to alter its intracellular distribution (A and data not shown). We noted that immunostaining of endogenous ALMS1 was diminished in cells expressing either construct ΔN-3175 or construct 2261–2602 (A). At low-to-moderate levels of construct expression this effect appeared to be specific because centrosomal γ-tubulin immunostaining was not appreciably altered (B). These data imply that transiently expressed ALMS1 deletion constructs compete with endogenous ALMS1 for access to the centrosome.
We attempted to narrow the part of construct ΔN-3175 required for centrosome-targeting by generating three shorter, nonoverlapping constructs. However, each of these showed some evidence of centrosomal localization in a proportion of cells (, A and B, and data not shown). Specifically, constructs 3175–3549, 3550–3942, and ΔN-3941 gave detectable centrosomal staining in 58, 93, and 10% of cells, respectively (n = 40). Construct ΔN-3941 preferentially localized to the nucleus (data not shown). These data suggest that all or a large part of construct ΔN-3175 contributes to centrosome targeting, but that the extreme C-terminal portion including the ALMS motif may be less important.
In summary, these findings suggest that two regions of ALMS1 have roles in targeting it to centrosomes: a relatively small internal region (residues 2261–2602) and a larger C-terminal region (residues 3176–4169).
Human KIAA1731 Contains an ALMS Motif and Tagged C10orf90 and KIAA1731 Localize to the Centrosome
The ALMS motif (residues 4037–4169) appears to be the only element of either putative centrosome-targeting domain of ALMS1 that bears significant similarity to another human protein. Notably, PSI-BLAST analysis identified an ALMS motif-like sequence at the C-terminus of the human protein KIAA1731 (, A and B; Supplemental Table S1), in addition to that previously identified at the C-terminus of C10orf90 (
Collin et al., 2002 ). Interestingly, KIAA1731 was previously identified as a candidate centrosomal protein by proteomic analysis (
Andersen et al., 2003 ) and contains an N-terminal region with similarity to Ddc8, a protein that is highly expressed in testis and localizes to the tails of elongated spermatids and spermatozoa (
Catalano et al., 1997 ;
Jaworski et al., 2007 ;
Shi et al., 2009 ; A). The Ddc8-like region of KIAA1731 also shares limited similarity with
Drosophila Ana1, a protein implicated in centriole formation (
Goshima et al., 2007 ;
Dobbelaere et al., 2008 ;
Blachon et al., 2009 ).
In the absence of antibodies to C10orf90 and KIAA1731, we investigated their subcellular localizations using tagged constructs. Immunofluorescence analysis of transiently expressed constructs showed that both localized to the centrosome (C); tagged C10orf90 also localized to the actin cytoskeleton in a proportion of cells (Supplemental Figure S2). These data support centrosomal localization of KIAA1731 and provide the first evidence that C10orf90 also localizes to this organelle. The ALMS motif appears to be the only region of sequence similarity shared by ALMS1, C10orf90, and KIAA1731. However, based on our deletion analysis of ALMS1, it seems unlikely that this motif acts as a discrete centrosome-targeting domain. C10orf90 and KIAA1731 may have additional structural homologies to ALMS1 that we have been unable to detect at the sequence level or may possess unrelated centrosome-targeting domains. Notably, we found that a construct representing the N-terminus of KIAA1731 was able to localize to the centrosome, providing further evidence that the ALMS motif is not critical for centrosome targeting (Supplemental Figure S5A).
Evolutionary Conservation of the ALMS Motif
BLAST analysis also detected C-terminal ALMS motif-like sequences in predicted proteins from numerous metazoan phyla and from two unicellular ciliated eukaryotes (Supplemental Figure S3 and Table S1). These data suggest that the ALMS motif evolved earlier than other regions of ALMS1, similarities to which we have been able to detect only in chordates and vertebrates, and pinpoint residues likely to be critical for the structure and/or function of this motif. The sea urchin and lancelet proteins identified in this analysis may represent orthologues of human KIAA1731 and ALMS1, respectively, based on additional sequence similarity outside the ALMS motif (not shown; similarity to KIAA1731 is noted in the database entry for the sea urchin protein). The remaining predicted proteins identified do not appear to share additional regions of similarity with human ALMS motif proteins.
siRNA-mediated Depletion of KIAA1731 and C10orf90 Suggests that the Encoded Proteins Have Centrosomal Functions
Next we investigated the functions of
KIAA1731 and
C10orf90 by RNAi, targeting each with two different siRNA duplexes. In the absence of antibodies to the encoded proteins, siRNA-mediated knockdown was monitored at the mRNA level (A and A). RT-PCR analysis confirmed expression of each gene but suggested that
C10orf90 is expressed at a significantly lower level than
KIAA1731 and
ALMS1 in hTERT-RPE1 cells (our unpublished observation). Strikingly, a proportion of hTERT-RPE1 cells treated with
KIAA1731-directed siRNAs displayed markedly reduced or undetectable centrosomal immunostaining for acetylated tubulin, ALMS1, C-Nap1, pericentrin, and γ-tubulin (, B–D, and data not shown). Similar effects were observed in U2OS cells (data not shown). Loss of centriole (acetylated tubulin), centriole-associated (C-Nap1 and ALMS1; see below), and PCM (γ-tubulin and pericentrin) markers strongly suggests the absence of centrosomes. Because loss of centrioles can cause dispersal of the PCM (
Bobinnec et al., 1998 ) and presumably also of centriole-associated proteins, these observations may be explained by a defect in centriole formation or stability.
Failure of centriole biogenesis should result in “dilution” of preexisting centrioles during successive cell divisions (assuming that at least a proportion of cells continue to divide normally). We therefore examined centriole numbers at different time points after siRNA treatment, using acetylated tubulin as a centriole marker. We seeded fewer cells in this experiment in an attempt to increase the level of KIAA1731 knockdown, because in previous experiments <25% of cells displayed complete loss of centriolar acetylated tubulin staining after 96-h siRNA treatment (B). As shown in A, cells treated with a negative control siRNA had either two or more than two clearly visible centrioles at all time points. In contrast, cells treated with a KIAA1731-directed siRNA showed evidence of centriole loss at all time points (A). At 48 h after treatment, the number of cells containing a single centriole was almost double that of cells completely lacking centrioles. The latter increased with time, exceeding the number of cells containing one centriole at 96 h after siRNA treatment. The number of cells containing two or more than two centrioles decreased with time. These data are therefore consistent with a role for KIAA1731 in centriole biogenesis, although they do not exclude the possibility that it is required for centriole stability.
Consistent with loss of centrioles/centrosomes in
KIAA1731-depleted cells, we observed mitotic cells with only one detectable focus of acetylated tubulin/γ-tubulin (B). Metaphase cells with just one focus of γ-tubulin frequently displayed abnormal arrangements of chromosomes, suggesting failure to assemble a bipolar spindle (B). We also observed mitotic cells with highly asymmetric distributions of γ-tubulin between spindle poles (B). Similar asymmetry is seen in a proportion of cells depleted of CPAP, a human protein known to be required for centriole formation (
Kohlmaier et al., 2009 ).
In contrast to KIAA1731-depleted cells, C10orf90-depleted cells did not exhibit appreciably reduced staining intensity of centrosome markers (B and data not shown). To query if C10orf90 has a role in cilium formation or maintenance, similarly to ALMS1, we incubated siRNA-treated cells in serum-free medium to induce cilium formation and visualized ciliary axonemes by immunofluorescence. We observed decreased ciliation in cells treated with C10orf90-directed siRNAs, compared with cells treated with a negative control siRNA (B), suggesting that C10orf90 is required for primary cilium formation.
In summary, these findings provide evidence of centrosomal functions for the ALMS motif–containing proteins KIAA1731 and C10orf90, consistent with our data showing centrosomal targeting of tagged versions of these proteins (C) and with the previous mass spectrometry-based classification of KIAA1731 as a candidate centrosomal protein (
Andersen et al., 2003 ).
ALMS1 Localizes to the Proximal Ends of Centrioles and Basal Bodies
Immunofluorescence microscopy analysis suggested that ALMS1 localizes adjacent to centrioles (Supplemental Figure S1). To examine its distribution in more detail, we analyzed immunostained cells by 4Pi microscopy. To visualize ALMS1 in relation to the immature parental centriole and basal body of ciliated cells, we used acetylated tubulin as a marker for centrioles, basal bodies, and ciliary axonemes (
Piperno et al., 1987 ). This analysis revealed that ALMS1 specifically caps one end of both the immature parental centriole and the basal body (A). The basal body was unequivocally distinguished from the distal tip of the cilium by additionally labeling cells for γ-tubulin (analyzed by standard microscopy; data not shown). Because the ciliary axoneme is known to extend from the distal end of the basal body, these data demonstrate that ALMS1 localizes specifically to the proximal end of the basal body. This localization was confirmed by costaining cells for C-Nap1, a marker for the proximal ends of centrioles (
Fry et al., 1998 ), which revealed similar spatial distributions of ALMS1 and C-Nap1, with ALMS1 having a slightly more compact distribution (B). Colocalization of ALMS1 and C-Nap1 was also observed in nonciliated cells (analyzed by standard microscopy; Supplemental Figure S4). In summary, these data establish that ALMS1 is targeted specifically to the proximal ends of centrioles and basal bodies, where it partially colocalizes with C-Nap1. This localization pattern suggests that ALMS1 may have a role in maintaining centrosome cohesion.
Next we asked if tagged versions of the ALMS motif–containing proteins C10orf90 and KIAA1731 localize similarly to ALMS1. Using standard microscopy we found that, in contrast to ALMS1, the bulk of centrosomal Myc-KIAA1731 immunostaining colocalized with centriolar acetylated tubulin (Supplemental Figure S5A). Notably, a construct representing the N-terminus of KIAA1731, containing most of the Ddc8-like region of this protein, localized similarly to the full-length construct (Supplemental Figure S5A). Interestingly, this localization pattern resembles that of the
Drosophila protein Ana1, which localizes along the lengths of centrioles and is known to share sequence similarity with the N-terminus of KIAA1731 (
Blachon et al., 2009 ). Global sequence alignment of KIAA1731 and Ana1 indicated only 15% amino acid identity (27% similarity) over their entire lengths, but did identify an additional conserved motif in the C-terminal half of each protein (Supplemental Figure S5B and data not shown). Taken together with the N-terminal sequence similarity and our RNAi data, these findings suggest that KIAA1731 is evolutionarily related to Ana1. Tagged C10orf90 also colocalized with centriolar acetylated tubulin, but appeared slightly more diffuse (Supplemental Figure S5A). In summary, these data suggest that, in contrast to ALMS1, C10orf90 and KIAA1731 are not specifically targeted to the proximal ends of centrioles.
siRNA-mediated Depletion of ALMS1 Diminishes Centriolar Levels of C-Nap1
C-Nap1 dissociates from parental centrioles at the onset of mitosis, concomitant with centrosome separation. At the end of cell division, after disengagement of parent and progeny centrioles, C-Nap1 reassociates with the parent centriole and associates, for the first time, with the progeny centriole (
Fry et al., 1998 ;
Mayor et al., 2000 ). Centriolar association/dissociation of C-Nap1 is regulated by the balance in activities of a NIMA-related kinase (Nek2) and protein phosphatase 1 (
Fry et al., 1998 ;
Helps et al., 2000 ;
Mayor et al., 2002 ). Little is known about how C-Nap1 is tethered to the proximal ends of centrioles during interphase, although Cep135 has recently been implicated in this role (
Kim et al., 2008b ). The highly similar localizations of ALMS1 and C-Nap1 prompted us to investigate if ALMS1 is also involved in retaining C-Nap1 at centrioles. We abrogated ALMS1 expression by RNAi (A) and examined centrosomal levels of C-Nap1 by immunofluorescence. This revealed marked reductions in the intensity of C-Nap1 staining at the centrosome, compared with cells treated with a negative control siRNA (, B and E). Immunoblot analysis indicated that total cellular levels of C-Nap1 were not substantially altered by treatment of cells with
ALMS1-directed siRNAs (D). Similar effects were observed with two independent
ALMS1-directed siRNAs and in U2OS and HEK 293 cells (B; Supplemental Figure S6 and data not shown). U2OS cells do not normally form primary cilia, indicating that the effect is not cilium-dependent. In the reciprocal experiment, depletion of C-Nap1 by RNAi did not appreciably alter the level of ALMS1 immunostaining at the centrosome (C), suggesting that centrosomal assembly of ALMS1 does not depend on C-Nap1.
Centriolar immunostaining of acetylated tubulin was not appreciably lower in ALMS1-depleted cells, indicating that diminished C-Nap1 immunostaining was not due to compromised centriole formation or stability (, A and E). Also of note, this phenotype was not due to accumulation of cells in late G2 or M phase, because the majority of ALMS1-depleted cells that had low levels of centriolar C-Nap1 contained just two centrioles (as determined by acetylated tubulin staining) and showed noncondensed DNA staining. Centrosomal immunostaining of the PCM component pericentrin appeared to be slightly diminished in ALMS1-depleted cells (see below; E and data not shown).
Next we examined the effect of ALMS1 depletion on cohesion between parental centrioles, using direct depletion of C-Nap1 as a positive control. We chose to analyze cells in G1 to avoid including cells undergoing the physiological process of centrosome separation at the G2-M transition and also because siRNA-mediated depletion of C-Nap1 has been reported to induce G1-S arrest (
Mikule et al., 2007 ). We used cyclin B1 as a marker for cell cycle stage (A). Cyclin B1 is undetectable in G1, begins to accumulate in the cytoplasm in S phase, and then at the centrosome in late S/early G2 phase, redistributes to the nucleus in prophase and is degraded at the metaphase–anaphase transition (
Pines and Hunter, 1991 ;
Bailly et al., 1992 ). In certain cell lines the immature parental centriole has been reported to be relatively free to move within the cytoplasm in G1 (
Piel et al., 2000 ); however, we found that in ~95% of interphase cyclin B1–negative hTERT-RPE1 cells the two centrioles were closely associated (, A–C). Furthermore, treatment with
C-Nap1–directed siRNA led to a dramatic increase in the distance between centrioles in ~65% of these cells, indicating that these conditions were suitable for analyzing C-Nap1–dependent association of centrioles (, B and C). Consistent with reduced levels of C-Nap1 at the centrioles of ALMS1-depleted cells, we noted an increase in the proportion of these cells with centrioles separated by >2 μm (referred to here as split centrosomes; , B and C). The increase was modest compared with that observed in cells treated with
C-Nap1–directed siRNA, possibly because of higher residual levels of C-Nap1 at the centrosome in ALMS1-depleted cells (Supplemental Figure S7C).
Although these data imply that ALMS1 is required for localization of C-Nap1 to the proximal ends of centrioles, to date we have not been able to demonstrate coimmunoprecipitation of these two proteins, suggesting that ALMS1 might not be directly involved in physically linking C-Nap1 to centrioles. This led us to consider that diminished levels of C-Nap1 at centrioles could also result from disruption of PCM1-based intracellular transport, on which centrosomal recruitment of C-Nap1 has been reported to depend (
Hames et al., 2005 ). PCM1 is a major component of centriolar satellites, nonmembranous 70–100-nm granules that are transported toward the centrosome via the microtubule-based molecular motor dynein/dynactin (
Kubo et al., 1999 ;
Kubo and Tsukita, 2003 ). In support of a functional interaction between ALMS1 and PCM1, we found evidence that the distribution of PCM1 granules was altered in ALMS1-depleted cells, particularly in cells treated with a siRNA duplex previously reported to cause ciliary abnormalities (Supplemental Figure S7A), that centrosomal levels of the PCM1-dependent protein pericentrin (
Dammermann and Merdes, 2002 ) were also reduced (E), and that the distribution of a cytoplasmic pool of ALMS1 may be altered in PCM1-depleted cells (Supplemental Figure S7B). However, we also found that, under the conditions used in our experiments, siRNA-mediated depletion of PCM1 did not significantly alter the level of C-Nap1 at the centrosome (Supplemental Figure S7C). These data suggest that although ALMS1-depleted cells may have a defect in PCM1-based trafficking, this defect is unlikely to account for the observed reduction in centriolar levels of C-Nap1.
In summary, these findings indicate that depletion of ALMS1 leads to a reduction in the level of C-Nap1 at the centrosome, resulting in compromised centrosome cohesion (D), and implicate ALMS1 in PCM1-based intracellular transport.
*
Present address: Endocrine Sciences Research Group, Manchester Academic Health Science Centre, University of Manchester, Oxford Road, Manchester M13 9PL, United Kingdom.
This article has been 
