A highly conserved centrosomal kinase, AIR-1, is required for accurate cell cycle progression and segregation of developmental factors in Caenorhabditis elegans embryos

Several microtubule-based events must occur for a cell to complete mitosis and faithfully segregate its chromosomes to each daughter cell (Sorger et al., 1997). The interphase array of microtubules must be broken down, and the centrosome must be duplicated to produce a pair of centrosomes that separate and migrate to opposite sides of the nucleus. The microtubule-nucleating capacity of the centrosomes must be greatly increased and the microtubules emanating from the centrosomes must contact the chromosomes via an association with the kinetochores. The forces generated by a number of motor proteins present on the spindle microtubules, kinetochores and chromosomes must also cooperate to form a functional bipolar spindle (Barton and Goldstein, 1996; Nigg et al., 1996). These processes are tightly regulated by p34^cdc2 and a number of other protein kinases (Vandre and Borisy, 1989; Nicklas et al., 1993; Verde et al., 1990; Buendia et al., 1992; Nasmyth, 1996; Nigg et al., 1996; Blangy et al., 1995). However, many of the proteins involved in the phosphorylation cascades controlling these events have not yet been identified (Nasmyth, 1996).

An increasing body of evidence suggests that the Ipl1 and Aurora family of protein kinases plays a key role in regulating mitotic events (Chan and Botstein, 1993; Francisco et al., 1994; Glover et al., 1995; Gopalan et al., 1997; Kimura et al., 1997). Ipl1 (increase-in-ploidy) was identified in a screen for mitotic mutants that failed to undergo normal chromosome segregation in S. cerevisiae (Chan and Botstein, 1993). Temperature-sensitive (ts) ipl1 mutants suffer from severe chromosome nondisjunction that eventually results in the accumulation of unbudded, polyploid cells at the restrictive temperature (Chan and Botstein, 1993). Further cytological and immunochemical analysis of ipl1 mutant cells has revealed no obvious defects in spindle pole body duplication, separation or bipolar spindle formation (Francisco et al., 1994). Ipl1 encodes a putative serine/threonine protein kinase (Francisco et al., 1994).

Genes highly related to ipl1 have recently been identified in Drosophila, mouse and humans. Mutations at the Drosophila aurora locus result in pupal lethality and mitotic arrest that is characterized by the presence of monopolar spindles (Glover et al., 1995). Less severe mutations result in female sterility where embryos from aurora mutant mothers also show defects in spindle morphogenesis that apparently arise from a failure in centrosome separation (Glover et al., 1995). Two recently published reports describing a mammalian relative of Ipl1 and Aurora, called IAK-1 or AIK-1, have both shown that IAK-1/AIK-1 is a mitotically active protein kinase that is localized to the centrosomes and microtubules of the mitotic spindle (Gopalan et al., 1997; Kimura et al., 1997). Further experiments show that, although mammalian IAK-1 cannot substitute for Ipl1 function in yeast cells, it does appear to participate in the same pathway since its expression enhances the defects seen in an ipl1 ts mutant (Gopalan et al., 1997). Together, these results suggest that the Ipl1 and Aurora family of protein kinases is highly conserved and that its members regulate mitotic spindle function and chromosome segregation. Microtubule dynamics are central not only to the events of mitosis, but also to the developmental processes that generate cellular asymmetry and embryonic polarity (Nelson and Grindstaff, 1997; Guo and Kemphues, 1996a; Gonczy and Hyman, 1997; Schierenberg and Strome, 1992). In the soil nematode, C. elegans, microtubules and centrosomes are required for the proper orientation of cleavage axes, and are thought to be partially responsible for the accurate localization of P-granules and PIE-1 to germline blastomeres (Hyman and White, 1987; Hyman, 1989; Hird et al., 1996; Mello et al., 1996). To further define the role of the Ipl1 and Aurora family of protein kinases in mitosis and to address the possible role of this kinase family in microtubule dynamics during development, we have identified and characterized two members of this family found in C. elegans, AIR-1 and AIR-2 (Aurora/Ipl1 related). Here we show that AIR-1 is required for embryogenesis and that it is associated with mitotic centrosomes. In addition to participating in mitotic spindle formation and chromosome segregation, AIR-1 also appears to be required for microtubule-based developmental processes. We show that the proper segregation of P-granules and the association of PIE-1 with centrosomes can be disrupted in the absence of AIR-1. The localization of PIE-1 to germline blastomeres has been suggested to be dependent on its association with centrosomes (Mello et al., 1996). Although AIR-1 appears to be necessary for the association of PIE-1 with centrosomes, our results suggest that this association is not necessary for the localization of PIE-1 to specific blastomeres. Rather, PIE-1 may be properly segregated via its association with P-granules.

Total RNA was prepared from N2 gravid hermaphrodites by standard methods. First strand cDNA was prepared with a GeneAmp RT-PCR kit (Perkin-Elmer, Branchburg, NJ) as described by the manufacturer. air-1 and air-2-specific cDNAs were PCR amplified as described in the text. PCR products were gel-purified using Wizard columns (Promega, Madison, WI) and cloned into a pBluescript vector (Stratagene, La Jolla, CA). Automated DNA sequencing of each cDNA was performed using standard methods. The cDNA sequences confirmed the predictions of Genefinder, a program used by the C. elegans Sequencing Consortium, except as described in the text. Amino acid sequence alignments were performed using the NCBI Blast Program, and the ClustalW1.7 Multiple Sequence Alignment Program (Higgins et al., 1996), accessed through the Baylor College of Medicine Search Launcher, Houston, TX.

A peptide corresponding to the carboxyl 12 amino acids of the predicted AIR-1 protein was coupled to KLH by standard methods and injected into rabbits. Rabbits were boosted for a period of 6 months and final bleeds were collected. For affinity-purification, the AIR-1 antigenic peptide was coupled to BSA and immobilized using an AminoLink Kit (Pierce Biochemical, Rockford, IL) as described by the manufacturer. Using this column, AIR-1-specific antibodies were purified essentially as described by Walczak et al. (1996).

Immunocytochemistry experiments and gonad dissections were performed on N2 gravid hermaphrodites and air-1 antisense RNA-injected hermaphrodites essentially as described by Seydoux and Dunn (1997). Anti-α-tubulin antibodies were purchased from Sigma (St. Louis, MO) and the P-granule-specific antibody OIC1D4 (Strome and Wood, 1983) was obtained from the Developmental Studies Hybridoma Bank (University of Iowa, Iowa City, IA). Monoclonal PIE-1 and rabbit polyclonal P-granule-specific antibodies were generous gifts of C. Schubert (FHCRC, Seattle, WA) and S. Strome (Univ. of Indiana, Bloomington, IN), respectively. Immunofluorescence microscopy was performed with a Nikon Microphot FXA Microscope and a Zeiss Confocal Microscope. Images were processed using Adobe Photoshop (Adobe, Mountain View, CA).

Antisense RNA corresponding to the entire coding region of the air-1 cDNA was synthesized using a T7 or T3 In Vitro Transcription Kit (Ambion, Austin, TX) as described by the manufacturer. L4 and adult N2 hermaphrodites were injected by standard methods (Mello et al., 1991) and were allowed to recover for 12-24 hours before fixation for immunocytochemistry.

We have identified two new members of the Aurora/Ipl1 protein kinase family in the C. elegans genome database. The air-1 genomic sequence was found on cosmid K07C11 from chromosome V, and the air-2 genomic sequence was found on cosmid B0207 from chromosome I. An oligonucleotide primer corresponding to the SL1 trans-spliced leader RNA (found on the majority of C. elegans mRNAs) (Blumenthal, 1995), and primers specific for the region surrounding the predicted translation stop codon of the air-1 and air-2 sequences were used to RT-PCR amplify both cDNAs from total RNA isolated from gravid adult C. elegans hermaphrodites. Sequencing of both cDNAs revealed that they each lacked the first exon predicted by the C. elegans Sequencing Consortium, but otherwise reflected the predicted exon structures. An alignment of the predicted protein sequences for AIR-1, AIR-2, IAK-1, Aurora and Ipl1 showed that all five coding regions share a high degree of homology throughout the predicted kinase domain and only diverge significantly at their amino termini (Fig. 1). Further analysis of AIR-2 will be presented elsewhere (J. M. S., unpublished data).

To determine the subcellular location of the AIR-1 protein, a rabbit polyclonal antiserum was raised against a synthetic peptide corresponding to the carboxyl-terminal 12 amino acids of AIR-1. Immunocytochemistry using affinity-purified antibodies on fixed C. elegans embryos of varying ages revealed that AIR-1 is present on mitotic centrosomes (Fig. 2). Counter-staining with DAPI and anti-α-tubulin antibodies confirmed these conclusions (Fig. 2C-E). Immunostaining was not seen with preimmune sera and was entirely competed by preincubation of the AIR-1 antibody with the antigenic peptide (data not shown). Diffuse cytoplasmic staining was found during interphase but, as cells entered mitosis, AIR-1 became localized to the centrosomes and remained associated with them throughout the mitotic phase of the cell cycle. Developmentally, AIR-1 was diffuse throughout oocytes and was not detected on the centrosomes of the sperm pronucleus (data not shown). AIR-1 staining of centrosomes was initially detected in the first mitotic division following pronuclear fusion, and was present in all mitotic cells up to the limits of detection in the hundreds of cells comprising late-stage embryos (data not shown). In the adult, AIR-1 staining was also found on the centrosomes of mitotic germ cells in the distal gonad (data not shown). These expression patterns suggest that AIR-1 is likely to be both a maternally supplied and a zygotically expressed gene product.

We disrupted the function of AIR-1 during embryogenesis by injecting air-1 antisense RNA, corresponding to the entire coding sequence of the air-1 cDNA, into the syncytial gonads of adult C. elegans hermaphrodites. Injection of antisense, sense or double-stranded RNA into the gonads of wild-type hermaphrodites has been shown to result in gene-specific loss-of-function phenotypes in the embryos of injected mothers (Guo and Kemphues, 1996b; Powell-Coffman et al., 1996, Fire et al., 1998). This method of disrupting protein expression is commonly referred to as RNA-mediated interference (RNAi) (Rocheleau et al., 1997). Although the mechanism is not understood, it presumably acts by depleting maternally provided mRNA and may also prevent the expression of zygotic transcripts (Fire et al., 1998). Injection of air-1 antisense RNA completely abolished AIR-1 staining in the embryos of injected animals (Fig. 3), but did not affect the staining pattern of the highly related AIR-2 kinase (data not shown). Loss of AIR-1 protein expression resulted in embryonic lethality in 100% of the embryos laid 12 or more hours postinjection (>200 embryos/injected animal). By Normarski analysis, the embryos appeared to arrest as disorganized masses of more than 100 cells. DAPI staining revealed that these embryos were severely aneuploid, having numerous cells with very large polyploid nuclei, as well as cells that appeared to be completely anucleate (Fig. 3A,B). As controls, we injected RNAs corresponding to several unrelated cDNAs. None mimicked the effect of air-1 antisense injection nor resulted in the loss of AIR-1 protein expression (data not shown). Disruption of AIR-2 protein expression via the same method resulted in a very different lethal phenotype (J. M. S., unpublished data). For simplicity, we will refer to AIR-1-deficient embryos as air-1(RNAi) embryos in the remainder of the text.

To determine how the air-1(RNAi) terminal phenotype arose, we dissected younger embryos from the gonads of injected mothers and examined them by DAPI, anti-AIR-1 and anti-α-tubulin staining (Fig. 3C-K). This analysis revealed that the AIR-1 protein was no longer detectable at any stage of the cell cycle (Fig. 3C-K). Furthermore, affected embryos had a variety of chromatin and microtubule abnormalities. To quantify these defects, embryos up to the 2-cell stage were classified as follows: (1) meiotic: newly fertilized embryos with maternal chromatin undergoing meiotic divisions I and II, (2) pronuclear: both the maternal and paternal pronuclei had decondensed but had not yet fused in the posterior half of the embryo, (3) 1-cell mitotic: embryos undergoing the first mitotic division after pronuclear fusion, and (4) 2-cell: embryos consisting of two cells (regardless of the amount of chromatin in each cell). air-1(RNAi) embryos of each classification were then scored for chromatin and microtubule defects (as compared to wild-type controls) (Tables 1, 2; Figs 4, 5).

Meiotic stage air-1(RNAi) embryos did not differ significantly from wild-type controls. In pronuclear stage embryos, chromatin defects not often found in wild-type controls (Fig. 4A) were apparent in a significant fraction of the air-1(RNAi) embryos (Table 1). Defects included pronuclei that appeared to be polyploid or had string-like or hypercondensed chromatin and/or misplaced nuclei. However, by the first mitotic division, the percentage of air-1(RNAi) embryos displaying chromatin abnormalities increased dramatically. The various defects seen in these embryos are catalogued in Table 1 and examples of air-1(RNAi) embryos at various stages of the first cell cycle are shown in Fig. 4. Some of these embryos appeared to enter into the first mitotic division despite incomplete fusion of the maternal and paternal pronuclei (Table 1; Fig. 4E-L). Such events possibly contribute to the other chromatin organization problems that we have observed in these embryos. Wild-type metaphase and anaphase mitotic figures were not found in any of the 1-cell mitotic air-1(RNAi) embryos examined (n>84) (Table 1; Fig. 4M-T). Metaphase chromatin was not well condensed and did not congress in a tight band at the metaphase plate as in wild-type cells (Fig. 4M-P). In anaphase cells, chromatin bridges were often found between separating chromatin masses that were not well condensed (Fig. 4Q-T).

Of the 2-cell air-1(RNAi) embryos examined, 80% had clear chromatin defects as compared to wild-type controls (Table 1; Fig. 5). These included similar defects to those seen in 1-cell mitotic air-1(RNAi) embryos. Additionally, AB and P₁ sister cells often did not have equal amounts of chromatin. In the most extreme cases, all chromosomes were found in a single cell (Table 1; Fig. 5E-H). Interestingly, a significant fraction of the embryos also had AB and/or P₁ cells that contained two or more nuclei that had progressed into the next cell cycle (Table 1). These results suggest that, in addition to chromosome segregation defects, air-1(RNAi) embryos may also be defective in cell cycle progression and/or cytokinesis.

To address the effect of the loss of AIR-1 protein expression on microtubule dynamics during mitosis, the DAPI and anti-α-tubulin-stained air-1(RNAi) embryos described above were rescored for the presence and appearance of mitotic centrosomes and spindles (Table 2; Figs 4, 5). Tubulin staining of meiotic and pronuclear stage air-1(RNAi) embryos was indistinguishable from wild-type controls (Table 2; Fig. 4A-D, and data not shown). However, nearly all of the air-1(RNAi) embryos undergoing the first mitotic division had discernible defects in the formation and appearance of mitotic spindles. Tubulin staining of embryos that appeared to have a failure in pronuclear fusion revealed disorganized arrays of microtubules (Fig. 4G,H). Further defects at distinguishable stages of mitosis included the presence of interphase-like arrays of microtubules in cells with condensed chromatin (Table 2; Fig. 4I-P) and spindles that consisted of ill-defined, small centrosomes which nucleated fewer microtubules than wild-type centrosomes (Fig. 4I-P). Although spindles in air-1(RNAi) embryos were bipolar, they were not well organized and were distinctly different from similar-staged mitotic spindles found in wild-type cells (Fig. 4M-T). Similar microtubule defects to those seen in 1-cell mitotic embryos were also noted in almost all of the 2-cell air-1(RNAi) embryos examined (Table 2; Fig. 5). These results suggest that air-1(RNAi) embryos may be defective in the breakdown of interphase microtubule arrays and the subsequent nucleation of spindle microtubules from mitotic centrosomes.

A small fraction of 1-cell mitotic and 2-cell air-1(RNAi) embryos had cells with multiple centrosomes (these cells were invariably polyploid by DAPI staining; Table 2 and data not shown). Thus, as above, some air-1(RNAi) embryos appear to undergo multiple cell cycles in the absence of cytokinesis. Altogether, our data suggests that, in the absence of AIR-1, cell cycle progression is dramatically affected at the transition from interphase to mitosis and can also be defective in the coupling of the completion of mitosis with entry into the next cell cycle.

In wild-type embryos, the first cleavage axis lies parallel to the long axis of the embryo (Fig. 4I,J,M,N,Q,R). In the 2-cell embryo, AB invariably divides before its sister P₁, and the cleavage axes are perpendicular to one another (Fig. 5A,B) (Guo and Kemphues, 1996a). The spindle in P₁ rotates to lie parallel to the long axis of the embryo (Hyman and White, 1987; Hyman, 1989). Examination of 1-cell mitotic and 2-cell air-1(RNAi) embryos by DAPI staining revealed that a small fraction of affected embryos had DNA and/or spindles that were dividing in the wrong orientation (Table 1). However, the majority of divisions in each cell type did occur in the proper orientation.

Thus, although AIR-1 does appear to have a small affect on spindle orientation, it does not appear to be absolutely required for spindle rotation. An additional defect noted in 2-cell air-1(RNAi) embryos included a small fraction of embryos with synchronous divisions of AB and P₁ (Table 1).

Since air-1(RNAi) embryos showed chromosome segregation and microtubule defects as early as the first mitotic division, we wanted to determine whether cells resulting from such aberrant mitoses maintain their cellular identity. Therefore, we examined whether P-granules are segregated properly in these embryos. P-granules are small proteinaceous particles consisting of RNA and several recently identified proteins, many of which have been implicated in RNA metabolism (Seydoux and Fire, 1994; Draper et al., 1996; Gruidl et al., 1996; Guedes and Priess, 1997). P-granules segregated to the posterior end of the embryo prior to the first mitotic division are retained at the posterior cortex of the P₁ cell, and continue to be partitioned to the germ lineage in each subsequent division (Strome and Wood, 1983; Hird et al., 1996). Staining of both wild-type and air-1(RNAi) embryos with an antibody specific for P-granules (OIC1D4; Strome and Wood, 1983) revealed that these particles were often mislocalized in air-1(RNAi) embryos (Fig. 6). As expected, P-granules were localized in the posterior blastomere of each wild-type embryo (Fig. 6A). In contrast, P-granules in air-1(RNAi) embryos were often found at both poles of the embryos (Fig. 6B, arrows), or clumped around nuclei in the middle of embryos (Fig. 6B, arrowheads). The bipolar localization was observed in younger embryos (1- and 2-cell stage), whereas the clumping of P-granules with nuclei tended to occur in older embryos (see below).

Given that P-granule segregation is defective in air-1(RNAi) embryos, we examined the localization of another germline factor, the PIE-1 protein. PIE-1 is localized to the germ lineage in C. elegans embryos and is transiently associated with P-granules in these cells (Mello et al., 1996). In wild-type animals, PIE-1 is localized to the posterior in postmeiotic 1-cell embryos and is segregated specifically to P₁ at the first division. During this division, PIE-1 staining is apparent on the posterior centrosome. As the P₁ cell begins to divide, PIE-1 again becomes localized to the centrosomes. As mitosis proceeds, the PIE-1 protein present on the centrosome destined for the somatic daughter disappears, whereas that on the germline destined centrosome persists. Concurrently, PIE-1 is specifically localized to the P₂ daughter cell. The same set of events occurs in subsequent divisions as PIE-1 is specifically localized to the germ lineage blastomeres P₃ and P₄. It has been proposed that this differential association of PIE-1 with the germline destined centrosome is involved in the specific localization of PIE-1 to the P-lineage blastomeres (Mello et al., 1996).

To address how PIE-1 and P-granules behaved in the same embryo, wild-type and air-1(RNAi) embryos were double-stained with a P-granule-specific rabbit polyclonal antibody and a mouse monoclonal antibody specific for PIE-1. In wild-type embryos, both PIE-1 and P-granules were localized to a single posterior blastomere and PIE-1 staining of P-granules was seen (Fig. 7A-F). Strong staining of centrosomes was observed with the PIE-1 antibody (arrowhead, Fig. 7C). Increased PIE-1 staining of centrosomes was also apparent along with increased concentrations of P-granules at the future sites of germline blastomeres (arrowheads, Fig. 7E,F).

In air-1(RNAi) embryos, P-granules and PIE-1 were localized to a single blastomere in approximately half of the examined embryos (Table 3; Fig. 7G-I). In some cases, the staining blastomere was completely anucleate (data not shown). Thus, despite very significant defects in chromatin distribution, PIE-1 can still be localized to a single cell. Furthermore, although PIE-1 was readily apparent on mitotic centrosomes in wild-type embryos (Fig. 7C,F), it was never detected on centrosomes in air-1(RNAi) embryos (n>200 embryos) (Fig. 7G-O and data not shown). This result suggests that: (1) AIR-1 is required for PIE-1’s association with centrosomes, and (2) this association is not absolutely required for PIE-1 localization to germline blastomeres. However, our results do not exclude the possibility that the PIE-1 protein was masked or present at undetectable levels on centrosomes in air-1(RNAi) embryos.

Although both P-granules and PIE-1 can be localized to a single blastomere in the absence of AIR-1, single-cell localization was most often seen in younger embryos, and was less common in older embryos (Table 3). Defects in P-granule and PIE-1 localization that were observed in 1-cell and 2-cell air-1(RNAi) embryos included the bipolar localization of both P-granules and PIE-1 (Fig. 7J-L), or bipolar P-granules with PIE-1 localized to only one pole (data not shown). In older embryos, P-granules were often mislocalized to more than one blastomere and, in the majority of these cases, the PIE-1 protein was undetectable (Table 3). P-granules were often clumped around many nuclei in older embryos while PIE-1 staining was absent (Fig. 7M-O). These results suggest that mislocalized P-granules can persist more readily than mislocalized PIE-1 and that mislocalized PIE-1 may be preferentially degraded.

In summary, both P-granules and PIE-1 could be localized to a single blastomere in early air-1(RNAi) embryos. However, in older embryos, P-granule mislocalization became apparent and PIE-1 was not usually detectable. This contrasts with observations in wild-type embryos, where PIE-1 can be detected up to the 100-cell stage (Mello et al., 1996), and P-granules persist in the germline precursor cells (Z2 and Z3) throughout embryogenesis and are readily visible at all stages of postembryonic development (Strome and Wood, 1983). Ectopic P-granules were found in the absence of PIE-1, but PIE-1 was rarely found in the absence of P-granules. Although PIE-1’s association with P-granules was apparent in these embryos, PIE-1 staining of centrosomes was never detected in the absence of AIR-1.

Here we have described the subcellular location and the consequences of disrupting the expression of the C. elegans AIR-1 protein, a member of the Aurora and Ipl1 family of protein kinases. To date, the subcellular location of only one other member of this protein kinase family has been described (Gopalan et al., 1997; Kimura et al., 1997). The mammalian protein IAK-1 has been shown to be associated with the duplicated centrosomes and microtubules of the mitotic spindle in tissue culture cells. AIR-1 is localized to mitotic centrosomes throughout C. elegans embryogenesis and is also present on the centrosomes of dividing cells in the adult, including the mitotic germ cells of the distal gonad (data not shown). These data suggest that other members of this protein kinase family, including Ipl1 and Aurora, may also be associated with microtubules, centrosomes and/or other microtubule organizing centers.

Injection of antisense RNA corresponding to the entire air-1 cDNA into the gonads of adult hermaphrodites caused fully penetrant embryonic lethality in the resultant brood. Several studies in C. elegans have shown that RNA-mediated interference (RNAi) predictably reproduces phenotypes that are known to result from severe loss-of-function or null alleles (Guo and Kemphues, 1996b; Lin et al., 1995; Powell-Coffman et al., 1996; Rocheleau et al., 1997; Guedes and Priess, 1997). Most importantly, we and others have found that RNAi results in the loss of detectable protein expression in the affected embryos (Lin et al., 1995; Powell-Coffman et al., 1996).

The terminal phenotype of embryos produced by air-1(RNAi) was characterized as severe aneuploidy. Younger embryos displayed incomplete pronuclear fusion, numerous anaphase chromatin bridges, chromosome segregation defects and polyploid cells apparently arising from a failure to complete mitosis and/or cytokinesis. Mutations in both ipl1 and aurora also result in chromosome segregation defects and polyploidy (Chan and Botstein, 1993; Glover et al., 1995). However, unlike aurora, the defects seen in air-1(RNAi) embryos do not arise from a failure in centrosome separation (Glover et al., 1995). Bipolar spindles are found in air-1(RNAi) embryos, as they are in ipl1 mutant yeast cells (Francisco et al., 1994). Thus, AIR-1, like Ipl1, does not appear to be absolutely required for centrosome duplication, centrosome separation or centrosome migration to opposite sides of the nucleus.

Like air-1(RNAi) embryos, mutations in a gene encoding another component of mitotic centrosomes in C. elegans, zyg-9, also result in defects in pronuclear migration and fusion (Albertson 1984; Matthews et al., 1998). Zyg-9 appears to be involved in the control of microtubule length since microtubules in zyg-9 mutants are much shorter than those in wild-type cells (Albertson 1984). Thus, other proteins like AIR-1 that have been implicated in microtubule dynamics at centrosomes also affect the premitotic events of pronuclear migration and fusion.

Although bipolar spindles are found in the absence of AIR-1, they appear to have fewer and shorter microtubules than normal (the microtubules do not reach the cell cortex as they do in wild-type spindles), and are found in cells that still contain an interphase-like array of microtubules. The centrosomes nucleating these mitotic spindles are also small and ill-defined. Our results are consistent with AIR-1 having a role in the transition from interphase microtubule arrays to the dynamic microtubules that are nucleated by mitotic centrosomes. Defects in this transition are likely to be the primary cause of the mitotic spindle and chromosome segregation abnormalities that we have described in air-1(RNAi) embryos. A phosphorylation cascade dependent on p34^cdc2 function has been shown to be necessary for this transition (Verde et al., 1990), and phosphorylation of centrosomal components is required for the increased nucleation of microtubules from mitotic centrosomes (Karsenti, 1991; Nigg et al., 1996). Our results suggest that AIR-1 may be an important structural component of centrosomes and its kinase activity may function in the phosphorylation cascades that are required for the recruitment of other centrosomal proteins and for the formation of the mitotic spindle.

The defects observed in air-1(RNAi) embryos are reminiscent of the defects seen in the Drosophila mutants abnormal anaphase resolution (aar) and microtubule star (mts) (Gomes et al., 1993; Mayer-Jaekel et al., 1993; Snaith et al., 1996). aar mutants are characterized by anaphase chromatin bridges and the presence of apparently intact, ‘lagging’ chromosomes similar to those seen in air-1(RNAi) embryos (Gomes et al., 1993; Mayer-Jaekel et al., 1993). mts mutant embryos die at the time of cellularization and are characterized by a block in anaphase progression (Snaith et al., 1996). Multiple centrosomes are seen in mts cellularized embryos, showing that, as in air-1(RNAi) embryos, centrosome duplication and DNA replication can be uncoupled from the completion of mitosis (Snaith et al., 1996). A similar uncoupling has also been noted in mutations at the Drosophila gnu locus (Freeman et al., 1986).

Both the aar and mts loci encode subunits of the Drosophila protein phosphatase 2A (PP2A; Mayer-Jaekel et al., 1993; Snaith et al., 1996). Additionally, mutations in one of the Drosophila isozymes of protein phosphatase 1 (PP1) are also characterized by mitotic defects, including abnormal sister chromatid segregation, excessive chromosome condensation and polyploid cells with multiple centrosomes (Axton et al., 1990). Since mutations in these essential Drosophila phosphatase subunits phenocopy many of the defects seen in air-1(RNAi) embryos, AIR-1 is likely to be acting in parallel with mitotically active protein phosphatases in C. elegans. Interestingly, mutations at the glc7 locus, which encodes the PP1 of S. cerevisiae, can suppress the mitotic defects of a temperature-sensitive allele of ipl1 (Francisco et al., 1994). These experiments suggest that PP1 acts in opposition to the function of the Ipl1 kinase (Francisco et al., 1994).

Studies in Xenopus cell-free extracts have shown that PP1 and PP2A both play distinct roles in controlling mitotic microtubule dynamics (Tournebize et al., 1997). Possible substrates of these phosphatases include OP18 (Tournebize et al., 1997; Belmont and Mitchison, 1996) and microtubule-associated proteins (MAPs) whose binding to microtubules is negatively regulated by phosphorylation (Sorger et al., 1997). A number of motor proteins necessary for mitotic progression and chromosome segregation are also regulated by phosphorylation, including the centrosome-associated kinesin, Eg5 (Barton and Goldstein, 1996; Nigg et al., 1996; Blangy et al., 1995). The activity of such proteins may be modulated by the combined or opposing activities of protein kinases like AIR-1 and as yet unidentified protein phosphatases.

In addition to a requirement during mitosis, AIR-1 has a probable role in microtubule-based developmental processes as well. P-granules and PIE-1 are both germline factors that are thought to be partially dependent on microtubules for their localization. The initial posterior localization of P-granules in 1-cell embryos is dependent on a cytoskeletal-based streaming of P-granules and cytoplasm to the posterior, with a concomitant degradation of those granules that remain at the anterior (Strome and Wood, 1983; Hird et al., 1996). In 4-cell and older wild-type embryos, P-granules normally associate with the nuclear envelope and move with the nucleus to the future germline side of the cell. This process has been shown to be microtubule-dependent (Hird et al., 1996).

P-granules are localized to a single blastomere in over 50% of the 1-cell and 2-cell stage air-1(RNAi) embryos examined. This localization occurs despite gross abnormalities in chromatin distribution. However, segregation defects are apparent in older embryos (>8 nuclei), with the P-granules becoming clumped around nuclei that are not in the posterior of the embryo. The mislocalization that is seen in a portion of the younger embryos is usually characterized by P-granules being present at both poles of the embryo. Since P-granules are not found throughout air-1(RNAi) embryos, it appears that this early defect may not be in segregation per se, but rather may be due to defects in the proposed anterior degradation machinery. Indeed, segregation of P-granules in the 1-cell embryo has been shown to be dependent on microfilaments and is microtubule independent (Strome and Wood, 1983; Hill and Strome, 1988). However, in these experiments, embryos were treated with microtubule inhibitors at the pronuclear stage, thus they did not address whether there is an earlier requirement for microtubules in the establishment or maintenance of the proposed anterior degradation machinery (Hird et al., 1996).

Since the mechanism by which the anterior degradation machinery functions is entirely unknown, the possible role of AIR-1 and/or microtubules in this process remains to be determined.

In P₂ and P₃, P-granules associate with germline-destined nuclei in a process that is microtubule-dependent (Hird et al., 1996). Hence, the mislocalization of P-granules that we find in older air-1(RNAi) embryos could be due to a primary defect in nuclear movement. Alternatively, since we have noted cases in which P-granules are present in anucleate cells, AIR-1 may be affecting a microtubule-based P-granule segregation system that is independent of nuclear behavior. Whether the two processes are coupled through a common regulatory mechanism that affects microtubule dynamics and whether AIR-1 is involved in a such a mechanism are important questions worthy of further investigation.

In addition to P-granules, the localization of another germline factor, PIE-1, is also affected in air-1(RNAi) embryos. PIE-1 represses transcription in the germline precursor cells, thus preventing somatic differentiation in these cells (Mello et al., 1992; Seydoux et al., 1996; Seydoux and Dunn, 1997). The localization of PIE-1 to the germline blastomeres has been thought to be dependent on its association with centrosomes and its perdurance on the centrosome that is destined for the germline daughter in each division (Mello et al., 1996). Although centrosome-staining of PIE-1 is readily apparent in wild-type embryos, we have never observed PIE-1 staining of centrosomes in air-1(RNAi) embryos. This suggests that the presence of AIR-1 on centrosomes is necessary for the association of PIE-1 with centrosomes. However, PIE-1 is still localized to a single blastomere in over 50% of the 1-, 2- and 4-cell embryos that we have examined. Thus, the association of PIE-1 with centrosomes does not appear to be essential for the localization of PIE-1 to germline blastomeres.

Our results instead suggest that the localization of PIE-1 is linked to its association with P-granules. Ectopic P-granules can exist in the absence of PIE-1, but the converse is very rarely true. Similar conclusions were reached in a recent study where the localizations of PIE-1 and P-granules were examined in a number of mutants in which P-granules are known to be mis-segregated (Tenenhaus et al., 1998). Immunofluorescence analysis of mex-1 mutants also suggested that PIE-1 localization is linked to the segregation of P-granules (Guedes and Priess, 1997). However, neither of these studies addressed whether PIE-1’s centrosome-association is necessary for its localization to germline blastomeres.

AIR-1 is a highly conserved centrosomal protein whose function is required during C. elegans embryogenesis. In the absence of AIR-1 protein expression, mitotic centrosomes are ill-defined and nucleate abnormal mitotic spindles that do not properly segregate chromosomes. Furthermore, P-granules are not properly localized at stages when P-granule segregation has been shown to be microtubule-dependent and the PIE-1 protein no longer associates with AIR-1-deficient mitotic centrosomes. In addition, a small fraction of air-1(RNAi) embryos also had defects in the orientation of cleavage axes.

Centrosomes are involved in establishing the orientation of the division axes in the early cleavages of C. elegans embryos (Hyman and White, 1987; Hyman, 1989). The reorientation of the mitotic spindle that occurs in these divisions is influenced by microtubules that emanate from the centrosome to the cell cortex (Hyman and White, 1987). A number of mutations that affect the orientation and timing of these early cleavages have been identified. They are called partitioning-defective, or par mutants, and many of them are defective in the segregation of P-granules (Guo and Kemphues, 1996a; Nelson and Grindstaff, 1997; Cheng et al., 1995). Many of the par protein products have been identified and some have been shown to be asymmetrically distributed on the cell cortex (Guo and Kemphues, 1995; Boyd et al., 1996; Etemad-Moghadam et al., 1995; Levitan et al., 1994; Watts et al., 1996). PAR-1 is asymmetrically localized in the early embryo and is homologous to the mammalian MARK (MAP/microtubule affinity-regulating kinase) family of protein kinases (Guo and Kemphues, 1995; Drewes et al., 1997). The MARK protein kinases phosphorylate MAPs, and MARK overexpression results in the hyperphosphorylation of MAPs and in the disintegration of microtubule arrays in tissue culture cells (Drewes et al., 1997). Although the PAR proteins clearly influence the orientation of the early cleavage axes and may do so by affecting microtubule behavior, none of the PAR proteins have been shown to be associated with centrosomes. This is surprising given the central role of centrosomes in the orientation process (Hyman and White, 1988; Hyman, 1989). Perhaps the pleiotropic nature of the defects seen in air-1(RNAi) embryos is characteristic of other centrosomal proteins that influence microtubule dynamics, and thus may have prevented their identification in genetic screens for PAR phenotypes.

We thank G. Gopalan for initiating this work, T. Copeland for synthesizing peptides and generating AIR-1 antisera, C. Schubert and S. Strome for antibodies, B. Bowerman, S. van den Heuvel, S. Strome, G. Seydoux, J. Strathern, R. Braun, and D. Garfinkel for helpful discussions, and G. Seydoux, B. Byers, B. Bowerman and D. Garfinkel for critical reading of various versions of the manuscript. This work was supported by the National Cancer Institute, DHHS, under contract with ABL. J. M. S. is a fellow of the Leukemia Society of America.