A phospholipid kinase regulates actin organization and intercellular bridge formation during germline cytokinesis

In metazoan organisms, cytokinesis is initially driven by the purse-string-like constriction of an actomyosin-based contractile ring and its associated plasma membrane. At the end of cytokinesis, the contractile ring disassembles and the two daughter cells separate (reviewed in Satterwhite and Pollard, 1992; Fishkind and Wang, 1995). Although a large number of protein effectors of cytokinesis have been characterized, crucial mechanistic aspects of the process remain unexplained. For example, it is not clear how filamentous actin (F-actin), myosin II and other cleavage furrow proteins are assembled at the equator of the cell (for review, see Field et al., 1999), how the contractile ring maintains its association with the cell cortex during furrowing, nor how the terminal stage of cytokinesis, which requires proteins distinct from those involved in the previous steps, is regulated (Adachi et al., 1997; Madaule et al., 1998; Powers et al., 1998; Raich et al., 1998; Swan et al., 1998).

A remarkable variation of the universal process of cytokinesis occurs during gametogenesis in many organisms: instead of separating, the daughter cells develop as a syncytium, with clonally related cells connected by intercellular bridges called ring canals (Cooley, 1995). During cytoplasmic bridge formation in the Drosophila male germline, structural proteins normally associated with the contractile ring persist to line the ring canal wall. These include the actin-binding protein anillin, and the septins, homologs of yeast bud neck filament proteins (Field and Alberts, 1995; Longtine et al., 1996; Hime et al., 1996). Although F-actin and myosin II are present in constricting contractile rings, they are largely disassembled during intercellular bridge formation in the male germline and do not persist as major components of the resulting ring canals (Hime et al., 1996).

Incomplete cytokinesis in the male germline of Drosophila occurs following each of the four mitotic and two meiotic divisions, resulting in syncytial cysts of 64 spermatids connected by 63 ring canals (Lindsley and Tokuyasu, 1980). During meiosis, each primary spermatocyte undergoes two meiotic divisions to generate four early round spermatids, each with a single haploid nucleus. Immediately after meiosis, all of the mitochondria in a spermatid cell aggregate and fuse to form a single mitochondrial derivative (reviewed in Lindsley and Tokuyasu, 1980; Fuller, 1993). Early round spermatids from males defective in meiotic cytokinesis, for example klp3A or tsr, or treated with the anti-microfilament agent cytochalasin B exhibit a ‘four wheel drive’ phenotype, in which four haploid nuclei accompany an abnormally large mitochondrial derivative (Gunsalus et al., 1995; Williams et al., 1995; Giansanti et al., 1998).

Here we describe four wheel drive (fwd), a gene required for proper organization of the constricted F-actin ring, accumulation of phosphotyrosine epitopes, and formation of stable intercellular bridges during male meiosis in Drosophila. The fwd gene encodes a predicted phosphatidylinositol 4-kinase (PI4K), the enzyme that performs the first committed step in the synthesis of the important membrane phospholipid phosphatidylinositol 4,5-bisphosphate (PIP2). Our results suggest a crucial role for phosphoinositide synthesis in regulating membrane and cytoskeletal dynamics during the endgame of cytokinesis.

Flies were raised on standard cornmeal molasses agar at 25°C (Ashburner, 1990). Visible markers, balancer chromosomes and large chromosomal deletions are described in Lindsley and Zimm (1992). Df(3L)B71 is the deficiency segregant of T(Y;3)B71. The fwd¹ allele was identified as a male sterile mutation in an insertional mutagenesis screen (Cooley et al., 1988). However, detailed deletion mapping revealed that the P element in the line was not associated with the lesion in fwd¹. Four additional EMS-induced alleles (fwd^2-5) were isolated from 2377 mutagenized third chromosomes tested for failure to complement fwd¹ in the screen described by Lin et al. (1996). fwd alleles were maintained over TM6B or TM6C.

Small deficiencies in the region were generated by transposase-induced mobilization of two different P elements by the method of Zhang and Spradling (1993). l(3)10512, a PZ element located in trachealess (Isaac and Andrew, 1996), was used to generate Df(3L)7C. 0751/02, a non-lethal PlacW insertion in 61B (Deak et al., 1997), was used to generate Df(3L)17E and Df(3L)2D. The source of transposase for these screens was a TM3 balancer chromosome carrying Δ2-3 on a defective Hobo element (gift of W. Gelbart).

fwd mutant phenotypes were evaluated in males bearing allele/Df combinations. Data for fwd²/Df(3L)emc^E12 are shown in Figs 1-4. Detailed examination of fwd³/Df(3L)emc^E12 and fwd⁵/Df(3L)emc^E12 was also carried out and revealed no appreciable differences. In the experiments in Table 1 and Fig. 7B-E, Df(3L)7C was used in place of Df(3L)emc^E12. The subtle phenotypic differences between fwd² and fwd³ versus fwd⁴ and fwd⁵ in Table 1 were most likely due to a difference in background chromosomes (see Molecular Biology, below). Oregon R was used as the wild-type control for the phenotypic analysis. Viability counts in Table 1 were obtained by counting male progeny from crosses of fwd allele/TM6C or Df(3L)17E/TM6C with Df(3L)7C/TM6C flies. At least 690 flies were counted for each cross. Germline-deficient flies used for northern analysis (Fig. 7A) were obtained as progeny of oskar mutant mothers (Lehmann and Nusslein-Volhard, 1986).

Squashed testis preparations were performed as described in Regan and Fuller (1990). The data in Table 1 were obtained by counting early spermatids in ten testes. Fidelity of karyokinesis was assessed by examining the relative sizes of nuclei in early round spermatids (Gonzalez et al., 1989). To view live spermatocytes undergoing cytokinesis, flies were dissected on a coverslip under Voltalef 20S halocarbon oil and spermatocytes were visualized using phase-contrast microscopy essentially as described in Church and Lin (1985) except with an upright microscope (Zeiss Axiophot). Meiotic cells from three (wild type) or four (fwd) males were examined, and images were recorded at five-minute intervals using Kodak T-Max 100 film. For immunofluorescence staining, testis sample preparation and rhodamine phalloidin and antibody sources and dilutions were exactly as described (Hime et al., 1996). Images were captured using a Photometrics cooled CCD camera connected to a Zeiss Axiophot microscope (kindly made available by Bruce Baker). Images of separate fluorochromes from multiply-stained tissues were collected individually and combined using Adobe Photoshop.

Genomic DNA sequences flanking the breakpoints of Df(3L)7C, Df(3L)17E and Df(3L)2D were cloned by plasmid rescue of the P elements used to generate the deficiencies (Uemura et al., 1989; Preston et al., 1996). Fourteen genomic phage forming a contiguous walk across the Df(3L)7C interval (spanning approximately 80 kilobases (kb)) were isolated from the Tamkun λEMBL3 library (Tamkun et al., 1992) by hybridization with the following probes: 0.8 kb flanking the Df(3L)7C distal breakpoint; the 3.5 kb EcoRI fragment from cosmid 13C12 (European Drosophila Genome Project; EDGP); the 4.7 kb XbaI fragment from cosmid 146B11 (EDGP); and a 2.3 kb NotI-NcoI end fragment from an NcoI deletion derivative of P1 DS00539 (Berkeley Drosophila Genome Project; BDGP). Southern blots and other molecular techniques were carried out essentially as described in Sambrook et al. (1989). Radioactive probes were labeled with [α-³²P]dCTP (NEN) using the Megaprime or Rediprime kits (Amersham). Poly (A)+ RNA for northern blots (Fig. 7A) was purified from whole RNA made using the Trizol Reagent (Gibco/BRL) by batch binding to oligo(dT) cellulose (Pharmacia). BamHI fragments used as probes on Southern and northern blots are indicated (*,**, Fig. 5B). The ribosomal protein probe Rp49 was from a plasmid kindly provided by M. Schubiger and S. McNabb (O’Connell and Rosbash, 1984).

Sequencing reactions were carried out using the Big Dye kit (Amersham) on a Perkin Elmer 9600 GeneAmp PCR machine, using primers obtained from Operon (Emeryville, CA). Unincorporated nucleotides were removed using Autoseq G-50 columns (Pharmacia) and reactions were run on an ABI sequencer at the Biochemistry Sequencing facility (University of Washington). Sequencing of fwd alleles was carried out on aggregate PCR products pooled from at least two independent reactions. Genomic DNA isolated from fwd/Df males was amplified using the Pfu Turbo enzyme (Stratagene) and pairs of sequencing primers. For each allele, 7-10 overlapping PCR products spanning 6.1 kb of contiguous genomic DNA were sequenced, including all but approx. 100 nucleotides at the C terminus of the predicted protein. The genomic sequences of fwd² and fwd³ differed from fwd⁴ and fwd⁵ in a manner consistent with different respective background chromosomes (data not shown). PCR products and probe fragments were gel purified using the Qiaquick kit (Qiagen). Homology searches and sequence comparisons used the BLAST and ClustalW programs on the NCBI and BCM web sites, respectively. Accession numbers of the proteins compared in Fig. 8 were fwd (AF242375), H. sapiens (AAC51156), A. thaliana (CAB37928), C. elegans (AAA83177), D. discoideum (AAA85725), S. pombe (CAA93903) and S. cerevisiae (CAA96174).

cDNAs corresponding to ESTs in the Berkeley Drosophila Genome Project database that matched the genomic sequence were purchased from Research Genetics (GH11214, GM01278, GM08014, LD25627) and Genome Systems, Inc. (LD5780, LD06959, LD07966, LD29916). The 5′ ends of these cDNAs show evidence for alternative splicing, resulting in alternate amino-terminal predicted amino acid sequences in a non-conserved part of the protein (data not shown). The predicted protein sequence shown in Fig. 6 is from GH11214, a cDNA from the head EST library (BDGP) that contains in-frame stop codons in the 5′ untranslated region upstream of the first methionine. Partial cDNAs were also obtained by screening the BS testis cDNA library (generously provided by J. Andrew and B. Oliver) with the 1.3 kb BamHI genomic fragment (*, Fig. 5B) and with an 0.8 kb SalI fragment internal to the 3.0 kb BamHI fragment (**, Fig. 5B). The most amino-terminal amino acid from a conceptual translation of the partial testis cDNAs is aa191 of the sequence shown in Fig. 6. The rest of the testis cDNA sequence produced a predicted protein nearly identical to that shown in Fig. 6, except for several conservative amino acid substitutions, reflecting the different strain backgrounds used to make the two libraries (data not shown).

RNA in situ hybridization was carried out as described (White-Cooper et al., 1998), using as probes DIG-labeled (Boehringer-Mannheim) single-stranded sense and antisense products made with T3 (Boehringer-Mannheim) or T7 (New England Biolabs) polymerases, respectively, from LD06959. This cDNA includes 4.0 kb that is common to all of the cDNA clones (see above) and would be expected to hybridize to all known alternative transcripts. Testes were photographed on a Nikon Microphot-FX microscope with DIC optics (kindly made available by Gerold Schubiger).

Males mutant for fwd had abnormal early round spermatids, suggesting a defect in cytokinesis during male meiosis. In wild-type early round spermatids, the haploid nucleus and the mitochondrial derivative in each cell appeared approximately the same diameter in unfixed squashed preparations viewed by phase-contrast microscopy (Fig. 1A). In testes from flies bearing either EMS-induced mutant alleles of fwd or a deletion of the entire gene, early round spermatids commonly contained four (or, less often, two) normal-sized nuclei associated with an abnormally large mitochondrial derivative (Fig. 1; Table 1). The spermatids with four nuclei and a large mitochondrial derivative (Fig. 1B) most likely resulted from failure of cytokinesis at both meiotic divisions followed by the aggregation and fusion of mitochondrial material equivalent to that of four cells. Spermatids with two nuclei and an intermediate-sized mitochondrial derivative (Fig. 1B) most likely failed cytokinesis at only one of the two meiotic divisions. Spermatids with one nucleus and a mitochondrial derivative of similar size, suggesting successful meiosis, were rare in the mutants. The nuclei observed in all cells were normal in size, indicating that karyokinesis was unaffected by mutations in fwd.

Occasionally cells were observed with nucleus:mitochondrial derivative ratios other than 1:1, 2:1 or 4:1 (Table 1). Cells with three nuclei per mitochondrial derivative could have arisen from successful separation of a single spermatid during meiosis II from a cell that had failed cytokinesis during meiosis I. Cells containing 8 or 16 (or, more rarely, 6 or 12) haploid nuclei associated with extremely enlarged mitochondrial derivatives probably derived from failure of one or more of the preceding mitotic divisions in addition to failure of meiosis I and meiosis II (Table 1, c.f. fwd⁴/Df(3L)7C, fwd⁵/Df(3L)7C, Df(3L)17E/Df(3L)7C). Indeed, primary spermatocytes with two sets of prometaphase chromosomes were occasionally observed in orcein-stained squashed preparations (data not shown).

The phenotype of failure of meiotic cytokinesis in fwd mutants was highly penetrant; greater than 99% of haploid nuclei observed were in multinucleate spermatids (Table 1). All five alleles of fwd, one from a P-element screen (fwd¹) and four obtained by EMS mutagenesis (fwd^2–5), showed similar phenotypes as hemizygotes (allele/Df; Table 1). The mutants appeared to be null alleles, as a viable combination of deletions (Df(3L)17E/Df(3L)7C) that removed the entire fwd gene had a similar cytokinesis defect (Table 1).

The spermatogenesis defect in the fwd mutants appeared to be restricted to cytokinesis, in that there was no arrest of spermiogenesis in fwd males. Indeed, flagellar elongation (Fig. 1D) and nuclear shaping (not shown) appeared morphologically normal. All late-stage spermatid bundles were extensively elongated, although the spermatids within the bundles appeared thick and irregularly packed (compare Fig. 1E,F). Cross-sections of elongating spermatids examined by transmission electron microscopy revealed sets of two or four axonemes accompanying enlarged mitochondrial derivatives, consistent with the formation of multinucleate cells (data not shown). No motile sperm were observed, as fwd spermatids failed individualization. In addition, although all fwd mutant males examined exhibited severe meiotic cytokinesis defects and were sterile, females mutant for fwd were fertile and showed no obvious defects in oogenesis when ovarioles were examined by light microscopy (data not shown). fwd flies of both sexes were fully viable (males, Table 1; females, not shown).

To determine the stage of meiotic cytokinesis affected in fwd mutants, we viewed living cells from wild-type and fwd mutant testes dissected under oil (see Materials and Methods). Cells undergoing meiotic division are easily identified by the phase-dark mitochondria, which line up along the meiotic spindle (Fig. 2A, arrow). In nearly all (12/13) wild-type cells undergoing cytokinesis during either meiosis I or meiosis II, cleavage furrows ingressed, bisected the cell at the equator (Fig. 2C, arrowhead) and persisted for the remainder of the observation period (at least fifteen minutes; Fig. 2C-F). The remaining cell did not form a cleavage furrow within the time frame of the observation. In the majority of fwd mutant cells (13/15; Fig. 2G-R), cleavage furrows initiated and ingressed (Fig. 2G,O, arrowheads), as in wild type. In five of these 13 cells (38%), the cleavage furrows did not persist and the daughter cells fused back together within 15 minutes (Fig. 2I,R), suggesting a role for fwd in stable intercellular bridge formation. The remaining eight fwd cells formed stable cleavage furrows. However, five of these cells had failed a previous division, as assessed by the presence of two or more nuclei in each of the daughter cells (not shown). Of the two fwd cells that did not form a detectable cleavage furrow, one had failed multiple previous divisions, as it contained six nuclei (not shown). These data likely underestimate the frequency of cytokinesis failure in fwd, as grossly abnormal structures deriving from defects in multiple earlier divisions were not scored in the analysis. Consistent with formation and ingression of cleavage furrows observed in living fwd cells, fluorescence microscopy revealed that early stages of cytokinesis appeared normal in fwd mutant cells. F-actin, myosin II, anillin and pnut, one of the septins, showed wild-type distribution in the cleavage furrow during late anaphase and early telophase (Hime et al., 1996; data not shown).

In wild type, phosphotyrosine (P-tyr) epitopes accumulate at the site of the arrested cleavage furrow late in cytokinesis in dividing male germ cells and persist to mark the ring canal walls after both the mitotic and meiotic divisions (Hime et al., 1996). P-tyr epitopes initially appear during late anaphase and early telophase as punctate spots in the partially constricted cleavage furrow. These spots gradually coalesce to form a tight band coincident with F-actin in the constricted contractile ring (Fig. 3A, arrow; Hime et al., 1996). Of 63 wild-type telophase cells examined, 57 (91%) showed P-tyr epitopes associated with the F-actin contractile ring. The remaining six cells, although in telophase, may have been at too early a stage of constriction for accumulation of P-tyr epitopes. In 48/63 (76%) of wild-type cells, P-tyr epitopes were present in a smooth ring that colocalized with the F-actin contractile ring. In fwd mutant telophase cells, by contrast, only 35/64 (55%) of the contractile rings had associated P-tyr epitopes and, for most of these (29/64 cells; 45%), the P-tyr staining was usually only a few dots (Fig. 3D, arrowhead; Fig. 3G, arrow). A smooth ring of P-tyr epitopes was observed in only six fwd cells (9%; c.f. Fig. 3D, arrow). Rings of P-tyr failed to accumulate even in fwd mutant dividing spermatocytes with tightly constricted contractile rings (Fig. 3D, arrowhead). In contrast, intercellular bridges that had formed successfully during the preceding mitotic divisions in the mutants contained smooth rings of P-tyr epitopes that, like mature ring canals in wild type, lacked F-actin (Fig. 3G, arrowheads).

fwd function is required for normal F-actin distribution in the constricted contractile ring. In wild-type cells undergoing meiosis, F-actin and myosin II remained tightly colocalized in the constricting contractile ring (Fig. 4A, arrow; Hime et al., 1996) However, in fwd, aberrant accumulations of F-actin appeared in the vicinity of the constricted contractile ring in late telophase (Fig. 4D, arrow; see also 3D, arrowhead). In these abnormal contractile rings, only a small fraction of the total F-actin colocalized with myosin II (Fig. 4D). The frequency of the defect in F-actin distribution was evaluated by examining the contractile rings of wild-type and fwd late anaphase and telophase cells stained for both F-actin and P-tyr (as in Fig. 3). In only two cases (3%) did the F-actin in the contractile ring appear delocalized in wild type. In contrast, 20/64 (31%) fwd cells exhibited aberrant actin accumulations similar to those shown in Figs 3D, 4D. Thus fwd mutant cells exhibited a dramatic increase in the frequency of disrupted F-actin organization in constricted contractile rings relative to wild type.

The fwd gene was initially localized to polytene interval 61A-C1 by recombination and deficiency mapping (Fig. 5A; recombination data not shown). An interval of approximately 27 kilobases (kb) containing sequences required for fwd function was defined by molecular mapping of the distal (telomeric) breakpoint of Df(3L)7C and the proximal (centromeric) breakpoint of Df(3L)17E, and a series of overlapping genomic phage representing the deficiency interval was isolated and mapped (Fig. 5). Southern blot analysis of wild-type and mutant genomic DNA using a subset of fragments across the 27 kb interval as probes revealed several restriction fragment length polymorphisms (RFLPs) within a 1.2 kb SalI-BamHI fragment (Fig. 5B, gray bar) consistent with a complex rearrangement in the fwd¹ allele (data not shown).

Sequence analysis of a total of 9.6 kb of cloned wild-type genomic DNA around the region of the fwd¹ RFLPs (thick bar, Fig. 5B) revealed a 2.6 kb contiguous open reading frame (ORF) with a significant degree of homology (e^–50) to human phosphatidylinositol 4-kinase β (PI4Kβ), including part of the catalytic domain. A predicted protein sequence for fwd (Fig. 6) was obtained by conceptual translation of a cDNA identified from an EST that matched the genomic sequence (see Materials and Methods). To confirm that the PI4K gene corresponds to fwd, we sequenced 6.1 kb of genomic DNA, including nearly the entire coding region of the gene, from each of the four EMS-induced fwd alleles (see Materials and Methods). The fwd², fwd³ and fwd⁴ mutations introduced stop codons into the PI4K ORF, whereas the fwd⁵ allele carried a Ser→Phe missense mutation at a conserved residue in the PI4K catalytic domain (Fig. 6; mutated residues indicated in bold).

The fwd gene may encode both germline-dependent and somatic transcripts. Northern blot analysis of mRNA from whole adult flies using a 1.3 kb BamHI fragment containing 1.2 kb (418 amino acids) of the predicted PI4K ORF as a probe (*, Fig. 5B; outlined region, Fig. 6) revealed a 4.4 kb transcript in males (Fig. 7A). This transcript was present at a lower relative abundance in females (Fig. 7A, compare lanes 1 and 2), and was not detected in males or females lacking a germline (Fig. 7A, lanes 3-4). In addition, a larger germline-independent transcript of approximately 6.5-7.0 kb was detected in both males and females (Fig. 7A). On testis northern blots, the higher molecular weight transcript was not observed. The germline-dependent transcript resolved into a doublet of 4.0 and 4.4 kb that was not detected in testis RNA from fwd¹/Df males, suggesting these transcripts are products of the fwd gene (data not shown).

As expected from the requirement for fwd function during male meiosis, transcription of fwd was detected in primary spermatocytes, late primary spermatocytes, meiotically dividing cells and early round haploid spermatids by in situ hybridization of RNA in whole adult testes (Fig. 7B,C; see Materials and Methods). fwd mRNA was detected at lower abundance in the spermatogonial cells at the tip of the testis, but was not detected in bundles of elongating spermatids (data not shown). fwd transcripts were nearly undetectable in testes from fwd¹/Df(3L)7C males (Fig. 7D), consistent with results from the northern analysis (see above), and were not detected in testes from flies deleted for the entire fwd gene (Df(3L)17E/Df(3L)7C, Fig. 7E).

Comparison of the predicted protein sequence of the fwd PI4Kβ with sequences of PI4Kβs from other species revealed a striking degree of homology in three highly conserved domains in the carboxy (C) terminal half of the predicted protein (Fig. 8A): (1) the PIK (phosphoinositide kinase), also called the LKU (lipid kinase unique), domain, (2) a domain specific to PI4Kβ family members that we have termed the BSM (β signature motif), and (3) the catalytic (kinase) domain. Pairwise comparisons between the fwd PI4Kβ and its homologs from other species revealed that the fwd protein is most closely related to those from human (Fig. 8A), rat and cow (not shown). The PIK domain is highly conserved (Fig. 8A,B) and is homologous to PIK domains from phosphoinositide 3-kinases and PI4Kαs as well as PI4Kβs (data not shown). The novel BSM, noted previously by Hsuan et al. (1998) and Xue et al. (1999), is also highly conserved (Fig. 8A,C). The BSM lies between the PIK and catalytic domains in all PI4Kβ sequences in the database (Fig. 8A) and appears to be unique to members of the PI4Kβ subfamily of lipid kinases, as it was not found in phosphoinositide 3-kinases or PI4αs (Hsuan et al., 1998; Xue et al., 1999), nor in any other proteins currently in the database. The most striking conservation between the fwd protein and PI4Kβs from other species was in the catalytic domain (Fig. 8A,D). The overall structure of many lipid and protein kinases is similar, and the predicted fwd kinase domain contains all residues thought, by analogy to protein kinases, to be required for lipid kinase catalytic function (Fig. 8D, asterisks; Gehrmann and Heilmeyer, 1998).

Identification of the fwd protein as a PI4Kβ homolog indicates an important role for phospholipids in regulating the final stages of cytokinesis. The principal known biochemical function of PI 4-kinases is to convert phosphatidylinositol (PI) to phosphatidylinositol 4-phosphate, or PIP. Subsequent phosphorylation of PIP by PI4 5-kinases results in the formation of the regulatory molecule PIP2. PIP and PIP2 can also be phosphorylated by phosphoinositide 3-kinases to form phosphatidylinositol 3,4-bisphosphate and phosphatidylinositol 3,4,5-trisphoshate (PIP3). PIP2, best known as the precursor of the second messengers inositol trisphosphate and diacylglycerol, has also been shown to interact directly with a number of proteins in the cell, regulating both their localization and activities (see below). Phosphoinositides make up only a small fraction of the membrane lipid molecules in a cell. Local synthesis of phosphoinositides is thought to regulate particular subcellular processes (reviewed in Godi et al., 1999). Due to their subcellular compartmentalization (see below), PI4Ks can be thought of as localized gatekeepers for PIP and PIP2 accumulation.

Although PIK1, the budding yeast homolog of fwd, is an essential gene (Flanagan et al., 1993; Garcia-Bustos et al., 1994), fwd is not. Flies carrying overlapping deletions that remove the entire coding region of fwd are viable and female fertile. This is particularly surprising given that fwd encodes the sole predicted PI4Kβ in the Drosophila genome. The fwd protein may be redundant with other PI4Ks. Indeed, most eukaryotes possess multiple PI4Ks (reviewed by Balla, 1998; Fruman et al., 1998; Hsuan et al., 1998), including type II (identified only by biochemical activity) as well as type III (now identified by the two cloned isoforms, PI4Kα and PI4Kβ). The domain structure of the mammalian PI4Kα and PI4Kβ proteins is different (in PI4Kα, the PIK domain and the catalytic domain are separated by a pleckstrin homology (PH) domain rather than the BSM; reviewed by Hsuan et al., 1998), yet both classes of proteins carry out similar biochemical activities (reviewed in Fruman et al., 1998). PI4Kα and PI4Kβ have overlapping but distinct subcellular localization patterns in human tissue culture cells (Wong et al., 1997), suggesting that the two isoforms might be redundant for some activities. In Drosophila, wild-type function of the fwd PI4Kβ could be required for incomplete cytokinesis in the male germline either because the single predicted PI4Kα is not expressed in spermatocytes or because regulatory properties or subcellular distributions distinguish the fwd protein from the still uncharacterized PI4Kα.

Precedents from other organisms suggest that loss of fwd function could lead to defects in membrane trafficking and phosphoinositide accumulation required for normal resolution of cytokinesis. Vesicle fusion during cleavage furrow formation in Xenopus oocytes and Caenorhabditis elegans embryos (Drechsel et al., 1997; Danilchik et al., 1998; Jantsch-Plunger and Glotzer, 1999) provides protein and phospholipid to the enlarging plasma membrane surface as cytokinesis progresses. Yeast and human homologs of fwd PI4Kβ proteins are implicated in secretion and membrane vesicle trafficking, consistent with accumulating evidence that PIP2 participates in these processes (reviewed by Balla, 1998; Hsuan et al., 1998). Yeast two-hybrid experiments and double mutant analysis detected both physical and functional interactions between PIK1 and FRQ1 (Hendricks et al., 1999), the yeast homolog of Drosophila frequenin, which modulates synaptic vesicle fusion events at nerve termini (Pongs et al., 1993). The human homolog of fwd, PI4Kβ, requires ARF, a small GTPase involved in budding of vesicles from the Golgi, for its Golgi localization and activity (Rothman and Wieland, 1996; Schekman and Orci, 1996; Nakagawa et al., 1996; Wong et al., 1997; Godi et al., 1999). Because dominant negative PI4Kβ causes Golgi disorganization in human tissue culture cells (Godi et al., 1999), it has been proposed that local distribution of phosphoinositides activates ARF and mediates membrane skeleton reorganization to permit formation and secretion of Golgi-derived vesicles (Terui et al., 1994; Godi et al., 1999; Lorra and Huttner, 1999). These vesicles, in turn, could be targeted to the cleavage furrow during cytokinesis.

fwd may stabilize interactions between contractile ring proteins and the plasma membrane. PIP2 and PIP3 bind proteins with PH domains (Harlan et al., 1994; Ma et al., 1997; Rameh et al., 1997; Varnai and Balla, 1998; Varnai et al., 1999). fwd-dependent local accumulation of PIP2 or PIP3 could anchor PH domain-containing contractile ring proteins, such as anillin (Field and Alberts, 1995), to the plasma membrane. Likewise, septins, which contain a conserved polybasic domain that binds PIP2 and PIP3 (Zhang et al., 1999), may require fwd to maintain their cleavage furrow association. Disruption of the membrane binding of these actin-associated proteins could lead to disorganization of F-actin in fwd mutant flies. Because of its potential to regulate both vesicle trafficking and membrane-cytoskeletal interactions, the fwd PI4K may coordinate contractile ring function with membrane addition during cytokinesis.

Activity of fwd is required for appearance of phosphotyrosine epitopes in the constricting cleavage furrow. Intriguingly, distribution of PH domain-containing src-related tyrosine kinases such as Btk appears to be controlled by PIP2 or PIP3 localization (Harlan et al., 1994; Varnai and Balla, 1998; Varnai et al., 1999). The failure of P-tyr epitopes to accumulate on contractile rings in fwd mutant spermatocytes may reflect the absence of phosphoinositides that serve as essential localization signals for the relevant tyrosine kinase(s). Alternatively, the defect in P-tyr accumulation in fwd could be due to failure to bring the kinase and its substrate together. For example, the kinase could be localized to the membrane and the substrate to the contractile ring. If these two structures failed to maintain contact during constriction, no tyrosine phosphorylation would take place. fwd-dependent tyrosine phosphorylation of a target protein could be a key regulatory event governing ring canal stabilization in the male germline of Drosophila.

The authors thank Lynn Cooley and Allan Spradling for the fwd¹ allele, Nurit Wolf for her excellent technical assistance in the initial characterization of fwd¹, Jim Posakony for deficiency stocks, Bill Gelbart for the Δ2-3 line, Dan Kiehart for the anti-myosin II antibody, Justen Andrew and Brian Oliver for the BS testis cDNA library, and members of the Baker and Schubiger labs for generously allowing us to use their microscopes. J. A. B. gratefully acknowledges Barbara Wakimoto for advice and encouragement, and for providing facilities to carry out the latter stages of these experiments, Craig Waters for assistance with the P-element mutagenesis screen, Salli Tazuke for her Northern blot, Lekha Devarayalu of the Biochemistry Sequencing Facility at the University of Washington for assistance with genomic sequencing and members of the Fuller and Wakimoto labs for excellent discussions. Many kind thanks to John Ashkenas, Lynn Cooley, Rebecca Farkas, Karen Hales, Rachel Meyers, Barbara Wakimoto and Philippa Webster for helpful comments on the manuscript. J. A. B. was supported by an NIH National Research Service Award #5-F32-HD07728, a Katharine McCormick Fellowship from Stanford University and a Special Fellowship of the Leukemia and Lymphoma Society. G. R. H. was supported by C. J. Martin Fellowship #937316 from the National Health and Medical Research Council of Australia and a Walter V. and Idun Y. Berry Fellowship from Stanford University. M. S.-S. was an American Heart Association-California Affiliate summer student. This work was funded by NIH grants RO1-HD18127 and RO1-HD29194 to M. T. F.