During Drosophila oogenesis, defective or unwanted egg chambers are eliminated during mid-oogenesis by programmed cell death. In addition, final cytoplasm transport from nurse cells to the oocyte depends upon apoptosis of the nurse cells. To study the regulation of germline apoptosis, we analyzed the midway mutant, in which egg chambers undergo premature nurse cell death and degeneration. The midway gene encodes a protein similar to mammalian acyl coenzyme A: diacylglycerol acyltransferase (DGAT), which converts diacylglycerol (DAG) into triacylglycerol (TAG). midway mutant egg chambers contain severely reduced levels of neutral lipids in the germline. Expression of midway in insect cells results in high levels of DGAT activity in vitro. These results show that midway encodes a functional DGAT and that changes in acylglycerol lipid metabolism disrupt normal egg chamber development in Drosophila.
RESEARCH on Drosophila oogenesis has led to numerous advances in our understanding of fundamental mechanisms of development including stem cell maintenance, cell fate determination, intercellular communication, cytoskeletal regulation, and signaling pathways (Spradling 1993; Van Buskirk and Schüpbach 1999). During oogenesis, developmental cues are generated within each egg chamber that control the co-ordinated activities of the somatic follicle cells and the germline nurse cells and oocyte. In addition to egg chamber-autonomous information, the progression of oogenesis is influenced by environmental factors such as food availability. Stage 8 of oogenesis, which is just prior to the initiation of vitellogenesis, is particularly sensitive to environmental conditions. Underfed females accumulate many previtellogenic egg chambers and produce few if any mature eggs. Stage 8 may also be a mid-oogenesis checkpoint for monitoring general egg chamber health; stage 8 egg chambers in wild-type females occasionally degenerate (Giorgi and Deri 1976). The removal of defective or doomed egg chambers at stage 8 could allow the female to conserve the energy required for vitellogenesis. The mechanism by which egg chambers are selected to die during mid-oogenesis is largely unknown, but recent findings point to the involvement of the ecdysone signaling hierarchy. Disruption of dare, a gene encoding an enzyme required for steroidogenesis, the ecdysone receptor (EcR), and E75, an ecdysone early response gene, all cause the developmental arrest and subsequent degeneration of egg chambers during stage 8. It is possible that this hormone response pathway is a link between environmental conditions and oogenesis progression as is the case in mosquitoes.
Cell death also plays an essential role in the development of every Drosophila egg. During oogenesis, the 15 nurse cells of each egg chamber are dedicated to the synthesis of maternal components, which are transported to the oocyte through intercellular bridges called ring canals (Mahajan-Miklos and Cooley 1994a). Several classes of molecules are selectively transported from the nurse cells to the oocyte during the first 10 stages of oogenesis. However, near the end of oogenesis during stage 11, the nurse cells contract to force their remaining cytoplasm into the oocyte, resulting in the rapid growth of the oocyte. During stage 10, nurse cells prepare for fast cytoplasm transport with a massive rearrangement of filamentous actin to form arrays of actin bundles surrounding each nucleus. Actin bundle formation is followed by nurse cell nuclear membrane permeabilization and nurse cell contraction (Matovaet al. 1999). Once the nurse cells have successfully transferred their cytoplasm into the oocyte, the nurse cell remnants are cleared from the egg chamber by apoptosis (Cavaliereet al. 1998; Foley and Cooley 1998). Several lines of evidence suggest that programmed cell death does not simply follow rapid transport but actually serves to trigger this morphological event. For instance, electron microscopy and the use of molecular markers show that nurse cells exhibit symptoms of apoptosis before the completion of rapid cytoplasm transport (Guildet al. 1997; Chenet al. 1998; Varkeyet al. 1999). Furthermore, germline clones of dcp-1 exhibit a block in final transport, characterized by a lack of actin bundle formation, nuclear envelope permeabilization, and nurse cell contraction (McCall and Steller 1998). This genetic analysis of caspase function suggests that activation of the programmed cell death pathway is essential for rapid cytoplasm transport.
We sought to identify factors that influence an egg chamber's decision to develop past the mid-oogenesis checkpoint or to undergo apoptosis. This was done in the hope that characterizing such factors would provide a link between the molecular mechanisms that control programmed cell death during mid-oogenesis and those that trigger rapid cytoplasm transport. We focused our efforts on cloning the midway (mdy) gene because of its loss-of-function phenotype. midway mutant egg chambers display precocious nurse cell actin bundling and nuclear membrane permeabilization phenotypes prior to their degeneration during mid-oogenesis. Aspects of this phenotype are reminiscent of the morphological changes that normally accompany the rapid phase of cytoplasm transport during late oogenesis. Characterization of the midway locus reveals that the gene encodes a protein with a high degree of similarity to acyl coenzyme A: diacylglycerol acyltransferase (DGAT). Mammalian and plant DGATs catalyze the conversion of diacylglycerol (DAG) into triacylglycerol (TAG). Loss of midway results in a severe reduction of lipid esters in nurse cells and oocyte. Furthermore, expression of midway in cultured insect increases DGAT activity, demonstrating that the Midway protein can function as a DGAT. These results show that changes in glycerol lipid levels can cause nurse cells to undergo programmed cell death.
MATERIALS AND METHODS
Drosophila stocks: All stocks were maintained at 25° using standard culturing conditions. Females were fed yeast paste 24–36 hr before ovary dissection to stimulate oogenesis. The mutations mdyqx25 and mdyrf48 were induced in an EMS mutagenesis screen (Schüpbach and Wieschaus 1991). The mutations l(2)k03902 and ep(2)2592 were isolated by P-element mutagenesis (Rorthet al. 1998; Spradlinget al. 1999). Canton-S and cn bw were used as wild-type controls. es(3)79 was used as the germline-specific enhancer trap (Cooleyet al. 1992).
Staining procedures: Ovaries were dissected in IMADS (Singleton and Woodruff 1994) and fixed with 100 μl Devit buffer [6% formaldehyde, 16.7 mm KH2PO4/K2HPO4 (pH 6.8), 75 mm KCl, 25 mm NaCl, 3.3 mm MgCl2] and 600 μl Heptane (Cooleyet al. 1992). Ovaries were then washed three times with PBT (1× PBS, 0.3% Triton-X-100, 0.5% BSA). For immunofluorescence, ovaries were incubated overnight at 4° with undiluted anti-Quail (6B9) monoclonal antibody (Mahajan-Miklos and Cooley 1994b). The ovaries were then washed four times with PBT and incubated overnight at 4° with an anti-mouse secondary antibody conjugated to FITC (Pierce, Rockford, IL) and 3 units of either rhodamine-phalloidin (Molecular Probes, Eugene, OR) or alexafluor-488-phalloidin (Molecular Probes). The TdT-mediated dUTP nick end-labeling (TUNEL) assay was performed using a kit from Boehringer Mannheim (Indianapolis) and according to Soller et al. (1999). Nile red staining was carried out according to Greenspan et al. (1985). β-Galactosidase staining was performed according to Cooley et al. (1992). In situ hybridizations were done using a 2.4-kb antisense RNA probe corresponding to the midway coding region (Buszczaket al. 1999). A sense strand probe was also synthesized and used as a negative control.
Molecular biology: Plasmid rescue was performed as described in Xue and Cooley (1993). Phage DNA was isolated using standard techniques. Several cDNA clones, corresponding to the Drosophila expressed sequence tags (ESTs) LD14270, LD33852, LD47766, LD13711, and LD11422 (Berkeley Drosophila Genome Project/HHMI EST Project, unpublished data), were obtained from Research Genetics (Huntsville, AL). Cas and midway cDNAs were isolated from a library enriched in Drosophila ovarian cDNAs, sequenced (Yale KECK facility), and analyzed as described in Xue and Cooley (1993). Genomic DNA was isolated from adult flies using a DNeasy tissue kit (QIAGEN, Valencia, CA). cn bw was the parental strain in the original EMS mutagenesis and served as a wild-type control. DNA fragments corresponding to the Cas and midway coding regions were independently amplified by PCR at least three times and subsequently sequenced. GenBank database accession numbers of related sequences used are as follows: human DGAT, NP036211; mouse DGAT, NP034176; Caenorhabditis elegans DGAT, AAF82410; human ACAT1, NP03092; human ACAT2, NP003569; Drosophila CG8112, AAF54271; and Saccharomyces cerevisiae SAT1, AAC49441.
Expression of recombinant midway and DGAT assays: midway coding sequence with an N-terminal FLAG epitope (IBI/Kodak, New Haven, CT; MGDYKDDDDG) was subcloned into pFastBac (Life Technologies, Grand Island, NY). Recombinant baculoviruses were obtained using the Bac-to-Bac expression system (Life Technologies) and amplified in SF9 cells [cultured in SF900 medium (Life Technologies) and 7.5% fetal calf serum].
High Five cells were infected with recombinant baculovirus carrying the midway gene or a control gene (human ACAT1; Changet al. 1997) for 48 hr. Membrane fractions from uninfected or infected cells were isolated. For ACAT and DGAT activity assays, 100 μg of membrane proteins from uninfected or infected cell extracts in 100 mm Tris-HCl pH 7.2, 250 mm sucrose, 1 mm EDTA, and 3 mm MgCl2 in a total volume of 100 μl was used. DAG, cholesterol, or solvent only was added as a substrate from stock ethanolic solution (16 mm) to a final concentration of 20 μm (Figure 5C) or 100 μm (Figure 5, A and B). The final ethanol concentration in each assay tube was kept at 0.6%. DAG was added in the form of DAG/phosphatidylcholine (PC) liposomes at a DAG:PC ratio of 1:100 (Figure 5C). The assays were started by adding 5 nmol of [3H]oleoyl-CoA/BSA (43.5 mCi/mol) to 50 μm final concentration. The enzyme reactions were performed at 37° for 10 min (Figure 5, A and B) or the indicated time (Figure 5C) and were terminated by adding chloroform/methanol (2:1). The lipid products were extracted and separated on TLC. The [3H]triacylglycerol and the [3H]cholesterol bands were scraped off and quantified by scintillation counting.
midway egg chambers undergo premature apoptosis: midway alleles were first identified in a screen for female sterile mutations on the second chromosome and were reported to result in egg chamber degeneration during stages 8 and 9 of oogenesis (Schüpbach and Wieschaus 1991). We stained wild-type and midway egg chambers with propidium iodide to visualize DNA and alexafluor-488-conjugated phalloidin to visualize the F-actin cytoskeleton. Labeling egg chambers with propidium iodide revealed nuclear condensation occurred in midway egg chambers during stage 9 (Figure 1C). In wild-type egg chambers F-actin localized subcortically in both germline and follicle cells and to ring canals in the nurse cells from stages 1–10A (Figure 1D). At stage 10B, a dramatic rearrangement of F-actin resulted in the formation of prominent actin bundles in the nurse cell cytoplasm (Figure 1E). Actin bundles were not observed prior to stage 10B in wild-type egg chambers. In contrast, midway egg chambers frequently contained actin bundles in the nurse cell cytoplasm during stages 8 and 9 (Figure 1, F and G). Interestingly, these bundles had a striated appearance similar to actin bundles present in wild-type stage 10 egg chambers (Riparbelli and Callaini 1995; Guildet al. 1997). The quail gene encodes a villin-like protein that colocalizes with cytoplasmic actin bundles in wild-type nurse cells during stage 10B of oogenesis (Mahajan-Miklos and Cooley 1994b). Immunofluorescence revealed that Quail protein was associated with the actin bundles present in midway stage 8 and 9 nurse cells (Figure 1H). This observation suggested that the bundles detected in stage 8 and 9 midway egg chambers were similar in composition to the bundles that normally form during stage 10B of oogenesis.
The behavior of stage 8 and 9 midway nurse cell nuclei also resembled stage 10 wild-type nurse cell nuclei. The nurse cell nuclear envelope permeabilization that occurs during stage 10B was assayed using a germline enhancer trap that expresses β-galactosidase (β-gal) with a nuclear localization signal (Cooleyet al. 1992). In wild-type egg chambers, β-gal from the enhancer trap remained in the nurse cell nuclei during stage 9 (Figure 2A) and did not begin to leak into the cytoplasm until stage 10B of oogenesis (data not shown). In a midway background, the β-gal marker leaked out of the nurse cell nuclei during stages 8 and 9 of oogenesis (Figure 2B).
The rapid phase of cytoplasm transport of Drosophila egg chambers occurs in response to an apoptotic signal (Buszczak and Cooley 2000). The TUNEL assay was used to examine whether midway egg chambers were undergoing programmed cell death. Wild-type egg chambers did not exhibit TUNEL during stages 8 and 9 of oogenesis (Figure 2C). However, midway egg chambers were strongly labeled prior to their degeneration (Figure 2D). This result demonstrated that mutations in the midway gene caused egg chambers to undergo programmed cell death during mid-oogenesis. After actin bundles form and the nuclear envelopes become permeable, wild-type nurse cells contract and force their cytoplasm into the oocyte. midway nurse cells, on the other hand, did not appear to contract; there was no movement of β-gal into the oocyte (Figure 2B). Thus, midway egg chambers appeared to prematurely undergo the events similar to those that occur in wild-type egg chambers leading up to but not through nurse cell contraction.
midway encodes an acyl coenzyme A: diacylglycerol acyl-transferase: The two midway alleles, mdyqx25 and mdyrf48, were mapped to region 36A8-E2 on the second chromosome (Schüpbach and Wieschaus 1991). Complementation tests using deficiencies in the region were used to further localize midway to 36A8-36B2. Mutations induced by P elements mapping to this region were tested for noncomplementation with midway. One, l(2)k03902, did not complement either of the midway alleles. Trans-heterozygotes of l(2)k03902 with either midway allele were female sterile due to egg chamber degeneration during mid-oogenesis. Both the lethality of l(2)k03902 and female sterility of l(2)k03902/mdy were reverted by excision of the P element. Sequence analysis of the l(2)k03902 insertion site revealed that the P element was inserted in the 5′-untranslated region (UTR) of a gene encoding a protein with a high degree of similarity to the mammalian protein cellular apoptosis susceptibility factor (CAS; Figure 3A; data not shown). Hereafter, this gene is referred to as Drosophila Cas (Cas). A second transposable element, ep(2)2592, was inserted near l(2)k03902 (Figure 3A). Genetic analysis revealed that ep(2)2592 behaved similarly to l(2)k03902. ep(2)2592 did not complement either mdyqx25 or mdyrf48 and was lethal in combination with l(2)k03902.
To determine if midway encoded Cas, genomic DNA corresponding to the Cas coding region was sequenced from the cn bw parental strain and mdyqx25 and mdyrf48 homozygotes. No molecular lesions or polymorphisms were found in the Cas coding region in either allele. Both Northern analysis using a Cas-specific probe and Western analysis using a polyclonal antibody generated against a GST-CAS fusion protein revealed that there was no reduction of transcript or protein levels in either of the midway alleles (data not shown). Furthermore, recently isolated point mutations in Cas complemented the mdyqx25 and mdyrf48 alleles but did not complement l(2)k03902 (Tekotteet al. 2002). These data suggested that l(2)k03902 and ep(2)2592 affected both Cas and a second gene disrupted by midway mutations.
Sequence analysis of the Cas genomic region (Berkeley Drosophila Genome Project, unpublished data) using a combination of commercially available (Research Genetics) and cloned (this study) cDNAs revealed that Cas was contained within an intron of another gene (Figure 3A). Homology searches suggested that this second gene encoded a protein with a high degree of similarity to members of the acyl coenzyme A: cholesterol acyltransferase (ACAT) family of enzymes (Figure 3B). Members of this family of acyltransferases fall into two classes on the basis of the substrates they use: those that use cholesterol (ACATs; Changet al. 1997) and those that use diacylglycerol (DGATs; Caseset al. 1998). The predicted protein was most similar (38.3% identical) to human DGAT, also referred to as ACAT-related gene product-1 (Oelkerset al. 1998). Molecular lesions in the coding region of this DGAT-like gene were found in both EMS-induced midway alleles. mdyqx25 contained a nonsense mutation that truncated the predicted protein by 66 amino acids. The mdyrf48 allele had a missense mutation that resulted in a valine-to-aspartic-acid change (Figure 3A). Identification of these molecular lesions provided strong evidence that the DGAT-like gene was disrupted by midway mutations.
The sequencing of genomic DNA, EST, and cDNA clones revealed that two different midway transcripts arose as a result of alternative 5′ exon utilization (Figure 3A). Several ovarian cDNA clones were derived from transcripts initiating with the more upstream of the two alternative 5′ exons, which contained several stop codons. Clones derived from the other midway transcript did not contain stop codons in their 5′ exons. A methionine in the first common exon was a good candidate for a translational initiation site, based on consensus sequence (Figure 3A), but we cannot rule out the possibility of other upstream initiation sites.
Searching the Drosophila genomic sequence database revealed the existence of a second Drosophila gene (CG8112) that was predicted to encode another ACAT family member. The predicted protein encoded by this gene was related to human ACAT1 and ACAT2 more closely than to human DGAT (Figure 3C).
In situ hybridization showed that midway was expressed specifically in the germline at low levels beginning in stages 4 and 5. Expression of midway increased dramatically during stage 9 of oogenesis (Figure 4A), the same stage when most midway mutant egg chambers degenerated.
Neutral lipid accumulation is disrupted in mdy mutants: ACAT family members catalyze the esterification of long-chain fatty acids to hydroxyl groups of membrane-embedded targets including cholesterol and diacylglycerol (Hofmann 2000). To determine whether midway mutations disrupt neutral lipid metabolism, wild-type and midway mutant egg chambers were stained with Nile red. Nile red strongly fluoresces in the presence of triacylglycerol and sterol esters (Greenspanet al. 1985). Staining of wild-type egg chambers revealed the presence of Nile red positive neutral lipids. The level of detectable conjugates greatly increased during stage 9 of oogenesis (Figure 4B). This increase coincided with the increase of midway transcription (Figure 4A). midway mutant egg chambers contained very few Nile red positive conjugates in the germline, while follicle cell levels were unchanged (Figure 4C). Comparison of Nile red staining of wild-type and midway egg chambers suggested that synthesis of triacylglycerol and/or sterol esters had been disrupted in the mutant nurse cells.
Substrate specificity of midway: To determine the substrate specificity of Midway protein, an N-terminally FLAG-tagged version of midway cDNA was expressed in insect cells using a baculovirus expression system. Cells infected with virus containing this cDNA expressed a 69-kD protein (data not shown), in close agreement with the predicted size of the Midway protein. Midway-expressing cells did not have any detectable cholesterol esterification activity as compared with ACAT virus-infected cells (Figure 5A). However, midway virus-infected cells were able to incorporate more oleoyl CoA into triacylglycerol relative to control and ACAT1 virus-infected cells (Figure 5, B and C). These results show that Midway protein has DGAT activity.
We have shown that the Drosophila midway gene encodes a DGAT enzyme that converts DAG into TAG, the major lipid-based energy reserve in eukaryotic cells. The predicted Midway protein is highly homologous to DGAT family members in its carboxy-terminal two-thirds and contains the FY.DWWN motif that is invariant among identified ACAT and DGAT molecules (Figure 3B). Interestingly, our searches of the Drosophila genomic and EST databases revealed another open reading frame, CG8112, that is predicted to encode a protein closely related to human ACAT1. Based on this homology, CG8112 is likely to be directly involved in cholesterol metabolism. Confirmation of this prediction requires genetic and biochemical analysis of CG8112.
The phenotypes we have observed for midway alleles appear specific to the female germline. This raises the question of whether the available midway mutations represent null alleles. Sequence analysis predicts that mdyqx25 encodes a truncation product while the mdyrf48 allele encodes a protein that differs from the wild-type protein by a single residue in its C terminus. The phenotypes of these alleles do not change dramatically over a noncomplementing deficiency, suggesting that the midway alleles act as severe hypomorphs. In the ovary, midway expression appears to be limited to nurse cells. Furthermore, the defects in lipid metabolism manifest in the germline. Additional work is required to determine whether midway functions outside the female germline.
Phenotypic characterization and in vitro DGAT assays demonstrate that Midway functions as a DGAT and is required for the formation of neutral lipid esters in the ovary. We used Nile red (Greenspanet al. 1985; Cadiganet al. 1989; Yanget al. 1996) to show that accumulation of neutral lipids was severely diminished by loss of midway. Experiments using insect cells infected with recombinant midway virus show that Midway protein exhibits DGAT activity (Figure 5), in agreement with its sequence similarity to other DGAT family members. Midway does not appear to use cholesterol as a substrate (Figure 5A).
This study was initiated to identify potential regulators of programmed cell death in the Drosophila female germline. Disruption of midway leads to nurse cell death during stages 8 and 9 of oogenesis. Prior to degenerating, midway egg chambers exhibit premature nurse cell actin bundle formation and nuclear envelope permeabilization. This phenotype is reminiscent of the cytological changes that accompany rapid cytoplasm transport during the later stages of oogenesis (Mahajan-Miklos and Cooley 1994a). Previous results have suggested that actin bundle formation, nurse cell nuclear envelope permeabilization, and nurse cell contraction are all triggered by caspase activation (McCall and Steller 1998). Analysis of the midway mutant phenotype suggests that initiation of premature nurse cell apoptosis can also result in, at least in part, these same morphological changes earlier in oogenesis. Mutations in other genes, including diminutive (dm; Gallantet al. 1996) and RpII215 (Perrimonet al. 1989), also cause egg chamber degeneration during mid-oogenesis, but no premature actin bundle phenotype has been reported for these mutants. Interestingly, expression of cell death genes in the follicle cells can cause nurse cells to undergo apoptosis during stages 8 and 9 (Chao and Nagoshi 1999). These nurse cells often contain small actin bundles in their cytoplasm. The link between the midway phenotype and cell death genes will have to be investigated further.
Previous results suggest that the steroid hormone ecdysone provides temporal signals needed for egg chamber development during mid-oogenesis (Buszczaket al. 1999). Disruption of the ecdysone signaling pathway in the germline results in phenotypes slightly different than those seen in midway egg chambers. For instance, loss of either EcR or E75 in the germline results in egg chamber degeneration during stage 8 of oogenesis, while midway egg chambers typically develop to stage 9 before undergoing apoptosis. Furthermore, EcR and E75 germline clones do not exhibit the premature actin bundling phenotype observed in midway egg chambers. These phenotypic differences suggest that the ecdysone response hierarchy and acylglycerol synthesis pathway control distinct aspects of egg chamber development during mid-oogenesis.
Disruption of TAG formation likely accounts for the premature nurse cell apoptosis phenotype observed in midway egg chambers. Loss-of-function phenotypes for DGAT molecules have been analyzed in mice and Arabidopsis. Mice that lack DGAT are fully viable and can still synthesize some TAG (Chen and Farese 2000; Smithet al. 2000). However, DGAT mutant mice do have altered triglyceride metabolism in specific tissues including the mammary gland, leading to defective lactation (Chen and Farese 2000; Smithet al. 2000). A mutation in the Arabidopsis TAG1 gene leads to reduced DGAT activity, altered seed fatty acid composition, and delayed seed development (Zouet al. 1999). The characterization of DGAT loss-of-function mutations in mice and Arabidopsis has not yet revealed an ectopic apoptosis or cell lethal phenotype. This leaves open the question of why midway egg chambers undergo premature apoptosis. One possibility is that TAG formation is required for the development of a mature egg. In this model, egg chambers that do not synthesize enough TAG for the production of viable progeny are unable to pass through the mid-oogenesis checkpoint because of a disruption in general metabolism and thus undergo apoptosis. In this way, egg chambers could use acylglycerol levels to monitor their own metabolic health.
Another possibility is that loss of midway leads to an increase of free, unesterified DAG. DAG acts as a second messenger and is a well-known activator of protein kinase C (PKC). Different PKC isoenzymes have been implicated in the regulation of programmed cell death in other systems (Crosset al. 2000). In midway egg chambers, increased levels of DAG may lead to increased PKC activity, which in turn could trigger premature nurse cell death. Aspects of the midway phenotype are reminiscent of the cell biological changes that occur in normal stage 10 nurse cells. These observations suggest the possibility that DAG may be a signal involved in the initiation of normal nurse cell apoptosis. While further work will be required to determine the exact cause of the midway phenotype, characterization of the midway locus has revealed a potentially interesting role for acylglycerol metabolism in the regulation of germ cell apoptosis.
We thank Trudi Schüpbach for midway stocks, Ruth Steward for lethal alleles mapping to 36A8-36E2, Laura Mitic for help in generating the midway baculovirus, and Marc Freeman and members of the Cooley lab for helpful comments on the manuscript. This work was supported by grants from the National Institutes of Health to L.C. (GM-43301) and T.Y.C (HL36709), and from the National Science Foundation to W.A.S. (IBN9205565).
Sequence data from this article have been deposited with the EMBL/GenBank Data Libraries under accession nos. AF468649 and AF468650.
Communicating editor: T. Schüpbach
- Received November 16, 2001.
- Accepted January 28, 2002.
- Copyright © 2002 by the Genetics Society of America