Drosophila variable nurse cells encodes Arrest defective 1 (ARD1), the catalytic subunit of the major N-terminal acetyltransferase complex

Identification of a lethal vnc allele

The vnc locus was first identified by Rees and co-workers in the course of an X-ray mutagenesis screen for female sterile mutations mapping to the 67A-D genomic interval (Rees, 1990; Matthews et al., 1993). A total of 15 hypomorphic alleles were isolated, of which only two, vnc² and vnc¹⁴, are still extant. While vnc mutant males were viable and fertile, homozygous females were reported to have egg chambers with abnormal numbers of nurse cells (Rees, 1990). Based on the predicted map position and partial phenotypic overlap with dynein pathway mutants, vnc alleles were potential candidates for mutations in the nudE gene that encodes a component of the LIS-1/cytoplasmic dynein motor complex (Kardon and Vale, 2009). To more precisely map the vnc locus, we tested all available lethal or sterile mutations in the 67A2-D14 region and determined if any failed to complement the fertility defects in vnc mutants. We found that only l(3)67BDk, one of 23 previously reported lethal loci identified in ethyl methanesulfonate and diepoxybutane-based mutant screens of the 67B-D region (Leicht and Bonner, 1988), was female sterile in trans to both vnc alleles. These results as well as molecular analyses (see below) indicated that l(3)67BDk corresponds to a lethal mutation in vnc, henceforth refered to as vnc^BDk. Identification of this lethal allele simplified molecular mapping and characterization of the gene.

Egg chamber formation is affected in vnc mutants

To control for potential background mutations, we examined females transheterozygous for the independently isolated vnc^BDk and vnc¹⁴ alleles, as well as vnc¹⁴ homozygotes. Females with both genotypes produce fewer eggs than wild type, the majority of which are shortened along the anterior/posterior axis and collapsed. In addition eggs laid by vnc mutant females frequently (24.6% for vnc¹⁴/vnc¹⁴; 28.2% for vnc¹⁴/vnc^BDk) show defects in the shape or position of the dorsal appendages (DAs), such as appendages that are closely apposed or fused (Fig. 1A-C and Table 1, row 8) suggestive of ventralization. In a few cases we observed eggs with supernumerary DAs (Fig. 1D). Examination of dissected ovaries revealed several additional defects. Oocyte growth in vnc mutants appears essentially normal until stage 10 when the oocyte occupies 50% of the volume of the egg chamber. However many egg chambers (44.4% vnc¹⁴/vnc¹⁴; 38.5% vnc¹⁴/vnc^BDk) that appeared to be at stage 11 or older based on follicle cell morphology, contained unusually small oocytes and persistent nurse cells (Fig. 1E; Table 1, row 7), phenotypes that typically result from a failure of nurse cells to transfer or dump their contents into the oocyte at stage 11 (Cooley and Theurkauf, 1994).

Wild-type egg chambers contain 15 nurse cells and an oocyte surrounded by a layer of somatically derived follicle cells (Fig. 1F). In contrast, the number of nurse cells in ovaries from vnc females varied widely from 1 to more than 60 (Fig. 1G, H). Visualization of nurse cell nuclei by staining with the DNA dye DAPI revealed that 50% of vnc¹⁴ and 13.3% of vnc¹⁴/vnc^BDk mutant egg chambers contained abnormal numbers of germline cells (Table 1, rows 4 and 5 summed). The diverse morphological defects observed suggest that vnc activity has important and pleiotropic roles in egg chamber formation.

vnc ovaries have defects in follicle cell encapsulation and cystocyte mitoses

Defects in at least two distinct processes are known to give rise to egg chambers with excess germline cells. First, instead of surrounding a single 16-cell cyst, follicle cells in mutant ovaries can encapsulate several cysts or partial cysts. Compound egg chambers generated through this process include additional nurse cells and oocytes derived from independent cysts that have undergone no more than four divisions, and therefore all cells have four or fewer ring canals. Further, since misincorporated germline cells in compound egg chambers may have more than one ring canal, the number of ring canals may exceed the number of nurse cell nuclei. Egg chambers with excess germline cells can also be formed when germline cells undergo additional rounds of mitosis. If the extra division occurs in a cystocyte that already has four ring canals, one of the daughter cells will have five ring canals. Since each germline division produces both an extra cystocyte and an additional ring canal that marks the position of incomplete cytokinesis, these extra divisions generate cysts containing equal numbers of nurse cells and ring canals.

To determine whether egg chambers with extra nurse cells arose from one or both of these processes, we co-stained ovaries from vnc¹⁴ homozygotes with antisera against HTS-RC that specifically stains ring canals and against ORB which accumulates in specified oocytes (Lantz et al., 1994; Huynh and St Johnston, 2000). In 26.4% of vnc¹⁴ and a lower fraction (8.2%) of vnc¹⁴/vnc^BDk egg chambers we observed several foci of ORB staining, indicating the presence of multiple specified oocytes (Table 1, row 5). Analysis of HTS-RC staining showed that each of these oocytes was associated with no more than four ring canals, suggesting they had developed in independent cysts and had been mispackaged (Fig. 2A, B).

In addition to supernumerary germline cells resulting from encapsulation defects, a significant fraction (12.3%) of abnormal vnc¹⁴ egg chambers were clearly generated through extra mitoses (Table 1, row 4 and legend). For example, the egg chamber in Fig. 2C has 30 nurse cells and a single oocyte that is positioned normally at the posterior. The oocyte is associated with 5 ring canals consistent with it having undergone an extra round of cell division. However, a total of only 30 ring canals are present rather than the 31 expected if all germline cells had undergone an extra round of division. This suggests that an extra division occurred in all but one germline cell. Thus unlike normal cystocyte divisions, the extra divisions were not perfectly coordinated for all cells within the cyst. In other egg chambers only a few additional germline cells were present. For example, the egg chamber in Fig. 2D contains 16 nurse cells and 16 ring canals rather than 15 of each, consistent with only one germline cell within the cysts having undergone an extra division. Importantly, only four ring canals are present on the oocyte, indicating that it did not undergo an additional cell division. We also observed occasional egg chambers with binucleate oocytes (Fig. 2E, F). Some of these contained varying numbers of additional nurse cells (Fig. 2E) underscoring the fact that mitosis appears to be desynchronized in vnc mutants; while others contained the normal number of nurse cells suggesting that only the oocyte had undergone an additional asynchronous division unaccompanied by cytokinesis (Fig. 2F). The occurrence of binucleate oocytes could explain the presence of supernumerary DAs in eggs laid by vnc mutant mothers (see Fig. 1D).

In contrast to the extra-nurse-cell phenotype, 11.3% of vnc¹⁴ egg chambers contained fewer germline cells than wild type (Table 1, row 4 and legend). These egg chambers usually had equal numbers of ring canals and nurse cells suggesting an underlying failure of one or more germline cystocyte cells to divide, rather than encapsulation defects. For example, 7 nurse cells, 7 ring canals and a single oocyte are present in the egg chamber in Fig. 2G compatible with a failure of all cells in the cyst to complete the last division. In other instances egg chambers lack one or more nurse cells, consistent with a failure of individual germ line cells to divide (Fig 2H). Thus it appears that the variable-nurse-cell-number phenotype results from both aberrant cyst encapsulation by follicle cells, and too many or too few cystocyte divisions.

vnc mutants display altered fusome morphology

Since impaired fusome function can lead to failure of cystocyte division synchrony, abnormal cyst formation and variable cystocyte number (McGrail and Hays, 1997; Liu et al., 1999; Mathe et al., 2003), we examined fusome morphology in ovarioles stained with anti-Adducin antisera 1B1 (Zaccai and Lipshitz, 1996). Wild type cysts contain intact, smoothly branched fusomes (Fig. 3A). In vnc mutant cysts several defects were observed, including a reduction in the amount of fusome material (Fig. 3B-E′). In addition, the morphology of the fusome was altered. Mutant cysts often contained thinner and fragmented fusomes (Fig. 3B, B′). Some cysts contained large spherical or barbell shaped fusomes (Fig. 3C-D′), a phenotype frequently observed when cystoblasts fail to divide (McGrail and Hays, 1997; Liu et al., 1999; Mathe et al., 2003). In severe examples mutant vnc germaria can be seen that contain a few spherical fusomes and cysts with reduced germline cells (Fig. 3D, D′). Finally, in some vnc germaria fusomes persist as late as region 3 (Fig. 3E, E′). This is in contrast to wild type where fusomes disassemble and are no longer visible by the middle of region 2a (see Fig. 3A, A′; Lin et al., 1994). In sum, fusome defects were seen in 46.7% of vnc¹⁴ and 28.3% of vnc¹⁴/vnc^BDk germaria (Table 1, row 1). Consistent with the presence of some 15-nurse-cell egg chambers at later stages, we also observed cysts with apparently normal fusomes (data not shown).

In addition to enabling the coordination of cystocyte divisions, fusomes are also important for oocyte specification (Lin and Spradling, 1995). In wild type ovaries, the cystocyte that retains the most fusome material is specified as the oocyte (Lin et al., 1994; de Cuevas and Spradling, 1998), while mutants with disrupted fusome morphology often fail to specify or misspecify oocytes (McKearin and Ohlstein, 1995; McGrail and Hays, 1997; McKearin, 1997; Liu et al., 1999). To test whether alterations in vnc activity affect oocyte specification, we stained germaria with antisera against the oocyte-specific molecular markers ORB, Egalitarian (EGL) and C(3)G. In wild type the RNA binding protein ORB is initially expressed in all cystocytes, but is subsequently restricted to just two pro-oocytes by germarial region 2a and further refined to the oocyte by region 2b (Fig. 4A; Lantz et al., 1994). ORB localization appeared normal in the majority of cases although in 10.5% of vnc¹⁴ and 11.4% of vnc¹⁴/vnc^BDk germaria, staining was greatly reduced or absent in region 1-2a (Fig. 4B; Table 1 row 3). A similar loss of staining was also observed for EGL, another early cystocyte marker whose expression is progressively restricted to the oocyte by region 2b (data not shown). Despite the lack of ORB staining in a subset of early ovarioles, oocyte-specific ORB staining was generally detectable in all egg chambers after stage 2 (Fig. 4B). This result suggests that the restriction of ORB and EGL to the oocyte may be delayed in a subset of vnc cysts or that these cysts fail to mature.

Staining for C(3)G, a synaptonemal complex (SC) component (Page and Hawley, 2001) yielded similar results. Formation of the SC, a precursor to meiosis, initially occurs in two to four cells per cyst in region 2a (Fig. 4C). However only the specified oocyte retains the SC through later stages and eventually completes meiosis (Fig. 4C). We found that in vnc ovaries C(3)G staining was unaffected in the majority of germaria. However in a subset of ovarioles progressive restriction of the SC into a single cell was delayed, and C(3)G staining could be seen in up to four cells as late as germarial stage 3 (Fig. 4D). In other mutant cysts C(3)G staining was entirely absent after region 2a, suggesting failure to maintain the SC in the oocyte (Fig. 4E). In total, aberrant C(3)G localization was detected in 12.8% and 13.2% of germaria from vnc¹⁴ and vnc¹⁴/vnc^BDk females respectively (Table 1, row 2). The presence of both persistent expression and loss of C(3)G staining in vnc mutants could be due to improper distribution of oocyte-determining components: cystocytes that aberrantly receive determinants above a critical threshold would fail to restrict the SC, while oocytes with below-threshold levels of these components would fail to maintain it. We conclude that the wide spectrum of fusome defects in vnc mutants are likely to contribute to the aberrant germline cell number observed in vnc mutant egg chambers through effects on the synchrony of cystocyte divisions.

Mutations in vnc affect chromatin reorganization in nurse cell nuclei

During oogenesis, nurse cell chromosomes undergo stereotypic conformational changes tightly correlated with the endocycle (Dej and Spradling, 1999). For the first four endocycles homologous chromosomes are joined together and nurse cell chromatin displays a polytene-like banded appearance. In endocycle 5, homolog pairing weakens and chromosomes assume a “five-blob” conformation, in which each individual chromosome arm is still distinguishable (Fig. 5A, S5 inset). At the end of this endocycle, the blob-like chromatin masses decondense and chromosomes become evenly dispersed as individual chromatid pairs throughout the nucleus (Fig. 5A, S6 inset). In vnc¹⁴ mutant egg chambers this dispersal process is severely disrupted with 70.3% of the nurse cells beyond stage 5 retaining the five-blob conformation (Fig 5B). In contrast, only 3.2% of wild type nurse cell nuclei display this organization beyond stage 5 (Table 1, row 6). A similar penetrance is observed in vnc¹⁴/vnc^BDk heteroallelic egg chambers (72.8%). The condensed five-blob chromosome conformation can persist in vnc mutant egg chambers beyond stage 10 (Fig. 5C). The dispersion of nurse-cell chromatin during endocycle 5 is thought to result from entry of the polyploid nurse cell nuclei into a truncated mitotic phase that begins before completion of chromosome replication (Dej and Spradling, 1999). By contrast, nurse cell nuclei prior to endocycle 5 undergo complete rounds of S-phase replication without entering M phase. The persistence of the five-blob conformation in vnc mutants suggests that entry into the truncated mitotic phase may be blocked, allowing completion of S-phase replication and retention of the 5-blob chromatin conformation without dispersion well beyond S5 (Fig. 5C). Thus, while vnc activity does not appear to affect S-phase DNA replication, it may impact entry into the early stages of mitosis.

vnc activity is required for somatic cell growth and division

To investigate whether there is a broader requirement for vnc in other stages of development, we examined the phenotypes of homozygous vnc^BDk animals. Homozygotes for vnc^BDk perish during the second larval instar, similar to hemizygotes in trans to a small deficiency Df(3L)vnc^MK (see below). The comparable lethal phases argue that vnc^BDk likely represents a strong loss-of-function or null allele. We were unable to determine the contribution of maternally provided vnc RNA and protein to embryonic development since the few eggs laid by vnc² and vnc¹⁴ homozygotes are unfertilized and germline clones mutant for the vnc^BDk allele could not be recovered. To examine the requirement for vnc activity in larval tissues we used FRT mediated recombination to generate clones of mutant cells in wing imaginal discs. Late induction of mitotic recombination yielded small clones homozygous for vnc^BDk that were marked by the absence of GFP (Fig. 6, asterisks), paired with significantly larger twin spots of homozygous wildtype cells. Clones induced at earlier stages gave rise to large wild type twinspots, without any accompanying mutant clones (Fig. 6, arrows). The inability to recover large mutant clones suggests that vnc activity is required for cell proliferation or survival.

vnc encodes ARD1, the catalytic subunit of an N-terminal acetyltransferase

As a first step towards determining the molecular identity of vnc, we used high-resolution meiotic mapping to determine the gene location relative to P-element transposons with molecularly defined insertion points (Zhai et al., 2003). We obtained five Pw⁺ transposons inserted in the 67A-D region (Fig. 7A). Each stock was individually crossed to vnc^BDk to produce daughters in which meiotic recombination between the P-element and vnc^BDk could occur. These females were crossed to hemizygotes for Df(3L)AC1, a deficiency that fails to complement vnc, balanced with a temperature-sensitive lethal third chromosome balancer. Progeny were raised under restrictive conditions to eliminate balancer animals. As a result, only two classes of viable progeny were possible: red-eyed non-recombinants carrying the P-element and white-eyed recombinants lacking the P-element (see Materials and Methods; Zhai et al., 2003). By scoring ~ 5000 progeny per cross we were able to determine the map distance between each P-element and vnc^BDk with a resolution of approximately 0.1 cM. These data indicated that vnc^BDk is approximately 0.05 cM or 55 kb distant from the closest P-element, P{SUPor-P}dpr6^KG07985, inserted in the first intron of the defective proboscis response 6 (dpr6) gene (Fig. 7A).

To further refine the location of vnc, we mobilized this P-element and screened for small deletions generated through imprecise excision. Complementation analysis with vnc^BDk yielded two lines, vnc^M2 and vnc^MK that failed to complement the sterility defect of vnc². The extent of the lesions caused by the imprecise excisions was determined using PCR. We found that both excision alleles contained deletions that extended distally from the P-element insertion site. The deletion vnc^MK extended 4 kb upstream of dpr6 and disrupted five genes: dpr6, CG6685, Nc, CG6674 and Ard1, identifying them as candidate open reading frames for vnc. Sequencing of genomic DNA isolated from vnc², vnc¹⁴ and vnc^BDk homozygous embryos or adults in the region of candidate genes revealed molecular lesions in the Ard1 gene in all three cases. Both vnc² and vnc¹⁴ animals contain a 5.2 kb copia element inserted at an identical location in the only predicted intron of Ard1 (Fig. 7A), while in vnc^BDk the Ard1 coding region is interrupted by a 20 bp deletion and 14 bp of novel sequence generating a truncated protein predicted to lack the entire ARD1 C-terminal region as well as a portion of the acetyltransferase domain (Fig. 7A-C). Consistent with vnc corresponding to Ard1, vnc^BDk failed to complement P{EPgy2}Ard1^EY23147, a lethal mutation linked to a P-element inserted 5′to the Ard1 open reading frame.

The Ard1 locus is predicted to encode the Drosophila ortholog of human ARD1a (also referred to as Naa10p), the catalytic subunit of NatA, a highly conserved N-terminal acetyltransferase (Polevoda et al., 2009). In yeast and humans, the NatA complex, consisting of ARD1 and an auxiliary subunit Nat1/Naa15p, is anchored to ribosomes and acetylates nascent peptides (Gautschi et al., 2003). ARD1 proteins show extensive sequence conservation and the Drosophila homolog shares 69.6% and 47.1% identical residues with human ARD1a and yeast ARD1 respectively (Fig. 7B, C). Like other acetyltransferase catalytic subunits, ARD1 contains an acetyltransferase enzymatic domain (residues 52-139) that harbors an acetyl-CoA binding site (RRLGLA; (Schultz and Pils, 2002). A predicted nuclear localization signal (KRSYRR; (Chelsky et al., 1989) is also present within the acetyltransferase domain.

To confirm that the vnc phenotypes are indeed caused by lesions in Ard1, we generated a genomic rescue construct containing the Ard1 coding region and 1.3 kb of flanking sequences (see Fig. 7A). This transgene successfully restored vnc¹⁴ and vnc¹⁴/vnc^BDk mutants to full viability and fertility. Analysis of eggs and egg chambers revealed that in the presence of the transgene the frequency of all oogenesis defects assayed was reduced to near wild type levels (Fig. 5C, D; Table 1), unambiguously demonstrating that the vnc locus corresponds to the Ard1 gene.

Since the copia insertion in vnc² and vnc¹⁴ is in the Ard1 intron and does not alter the protein sequence, we hypothesized that the lesion might affect Ard1 expression in the ovary. To examine ARD1 levels in vnc mutants, we first determined whether polyclonal antisera against human ARD1 could recognize the homologous fly protein. Total protein extracts from wild type and vnc allelic combinations were separated on SDS poly-acrylamide gels, transferred to nitrocellulose filters, and the western blots probed with anti-human ARD1. A single band running at ~22 kDa was detected in larval extracts from wildtype and Df(3L)vnc^MK/+ hemizygotes consistent with the 22.3 kDa molecular weight of the conceptually translated protein (Fig. 7D left panel, lanes 1 and 3). No signal was seen in vnc^BDk/Df(3L)vnc^MK mutant extracts (Fig. 7D left panel, middle lane). The presence of a single appropriately sized band in wild type and its absence from mutant extracts argues that the antisera specifically recognize Drosophila ARD1. The failure to detect the truncated 17 kDa polypeptide encoded by vnc^BDk allele could be due to the absence of critical antigenic epitopes or because of instability of the mutant protein, consistent with the strong loss-of-function phenotype. Comparison of ARD1 expression levels in ovaries from wild type and a hypomorphic vnc mutant revealed that ARD1 levels in vnc^BDk/vnc¹⁴ and vnc¹⁴/vnc¹⁴ ovaries are indeed dramatically decreased compared to wild type controls (Fig. 7D, right panel).

Microarray data and RNA-seq analysis from the modENCODE project indicate that Drosophila Ard1 is widely transcribed (Chintapalli et al., 2007) and the NatA complex is predicted to be active in most tissues. Therefore the fact that phenotypic defects in vnc² and vnc¹⁴ mutants were restricted to oogeneisis could reflect a more stringent requirement for N-terminal acetylation during oogenesis, or alternatively, result from a specific reduction in Ard1 levels or activity in ovaries. To distinguish between these possibilities we used quantitative PCR to compare Ard1 mRNA levels in ovaries and other tissues from wild type and vnc mutant adults. We found that Ard1 transcript levels in mutant ovaries decreased, while Ard1 transcripts in other tissues were marginally increased. In fact, a four-fold reduction in transcript levels was seen in ovaries from vnc¹⁴ homozygotes mutant and an eight-fold decrease in vnc¹⁴/vnc^BDk ovaries (Fig. 7E). At the same time we observed a two-fold increase for Ard1 transcripts in vnc¹⁴/vnc¹⁴ female carcasses and a 50 percent increase in vnc¹⁴/vnc^BDk carcasses (Fig. 7E). Based on these results we conclude that vnc² and vnc¹⁴ mutants are viable and male fertile but show pleiotropic phenotypes in oogenesis due to ovary-specific reduction in Ard1 expression.

The variable nurse cell phenotype correlates with defects in fusome morphology

Wild type cystoblasts undergo four synchronous rounds of cell division to produce 16 cystocytes that remain cytoplasmically interconnected through ring canals and the fusome, a vesicle-rich structure that consists of modified endoplasmic reticulum (Snapp et al., 2004). The fusome endomembranes form a continuous connection through the ring canals that enables oocyte specification and cell division synchrony. Mutations in genes such as hts and α-spectrin that cause fusome disruption result in reduced numbers of cystocyte divisions, failure of division synchrony and defects in oocyte specification (Lin et al., 1994; de Cuevas et al., 1996). During mitotic divisions the cystoblast that retains the most fusome material differentiates as the oocyte (reviewed in Huynh, 2006), whereas synchrony of cystocyte divisions is mediated through coordinated degradation of fusome-associated cyclins (Lilly et al., 2000) Thus germline expression of ubiquitin-protease resistant CycA induces an extra round of cystocyte division, creating egg chambers with 32 cystocytes. Likewise, hypomorphic mutations in UbcD1 which encodes an E2 Ubquitin conjugating enzyme, produces 32-cell cysts with highly branched fusomes.

While some ovarioles in vnc mutants contain normal-looking fusomes, others display a wide variety of fusomal defects, including cysts lacking fusomes, cysts with abnormally thin, threadlike fusomes, and fragmentary or spherical fusomes (see Fig. 3). Moreover, while fusomes are lost in wild type cysts by germarial region 2a, some vnc mutant ovarioles showed persistent fusome staining as late as region 3. These diverse fusome defects could explain the wide variety of phenotypes observed in late stage vnc egg chambers (see Table 1). For example, fragmentation or loss of the fusome would be expected to lead to loss of cystocyte division synchrony (Lin et al., 1994; de Cuevas et al., 1996), while abnormal persistence of the fusome beyond region 2a could contribute to the additional cystoblast divisions that give rise to the egg chambers with large numbers of germline cells.

N-terminal acetylation mutants and chromatin defects

Among the diverse phenotypes encountered in vnc mutants, defects in nurse cell chromatin morphology are detected with the highest penetrance (see Table 1). In wild type egg chambers polytene nurse cell chromosomes assume a 5-blob morphology at endocycle 5 that disperses by the end of the endocycle (Dej and Spradling, 1999). In vnc mutant nurse cells, chromosomes also adopt the 5-blob conformation by stage 5, but fail to disperse and retain the condensed morphology through subsequent endocycles (see Fig. 5). However DAPI stained nurse cell nuclei fluoresce with increased intensity in later stage egg chambers indicating that despite the failure of chromosomes to disperse, multiple subsequent rounds of replication occur. The dispersal of nurse cell chromatin after endocycle 5 shares several characteristics with entry into a normal M phase and has been suggested to constitute a pseudomitosis (Dej and Spradling, 1999). In support of this idea, morula, a component of the Anaphase Promoting Complex/Cyclosome (APC/C) which facilitates completion of mitosis in diploid cells, is first required in nurse cells at this stage to prevent inappropriate reversion to a full-blown mitotic state that includes chromatin condensation and spindle formation (Reed and Orr-Weaver, 1997; Kashevsky et al., 2002). The failure of chromatin dispersal in vnc mutant nurse cells may reflect disruption of this mitosis-like phase. This defect, coupled with continued S-phase DNA replication suggests that some aspects of the progression into mitosis could be highly sensitive to decreased ARD1 levels. If this requirement extends to somatic cells, it would be predicted to impact cell proliferation. Consistent with this expectation, vnc^BDk mutant clones in wing discs failed to proliferate beyond three cells (see Fig. 6). The finding that knockdown of Ard1 inhibits cell division in the HepG2 human liver cell line (Fisher et al., 2005) supports the idea that ARD1 plays an important role in cell proliferation.

In Drosophila the Naa50p/Nat5p ortholog Separation Anxiety (SAN) has been shown to co-purify with ARD1 and NatA in embryonic extracts (Williams et al., 2003). SAN is also predicted to have N-terminal acetyltransferase activity, but modifies a substrate (Met-Leu) not recognized by ARD1 (Polevoda et al., 2009). Mutations in san cause defects in chromosome segregation at the embryonic syncitial blastoderm stage (Pimenta-Marques et al., 2008) and in sister chromatid cohesion in Drosphila neuroblasts (Williams et al., 2003). The premature separation of sister chromatids in san mutants suggest that SAN normally promotes centromeric sister chromatid cohesion, in contrast to ARD1 that may enhance chromatid separation based on the failure of nurse cell chromosomes to disperse in Ard1 mutants. Thus Ard1 and san mutants appear to display opposite defects, albeit in distinct tissues. Since ARD1 and SAN co-purify, it is possible that their activities are linked. However, even though SAN is expressed in nurse cells (and oocytes), san mutant germline clones show no detectable phenotypes (Pimenta-Marques et al. 2008), and vnc phenotypes are unaffected by heterozygosity for san (YW and RW unpublished data) arguing against the involvement of SAN in nurse cell separation. In addition, knockdowns in san homologs do not affect NatA-mediated acetylation in lower eukaryotes (Gautschi et al., 2003). Conversely, although knockdown of human SAN caused chromosome cohesion defects in HeLa cells, knockdowns of Nat1p did not result in mitotic defects, suggesting that SAN can support cohesion in the absence of the rest of the NatA complex (Hou et al., 2007). The absence of oogenesis defects in Drosophila san mutants could be due to redundancy with some other factor in this tissue.

vnc mutants show phenotypic overlap with potential ARD1 substrates

The pleiotropic phenotype of vnc mutants overlaps with the phenotypes of many genes that play key roles in oogenesis. Thus mutations that affect HTS disrupt fusome organization while alterations in CycA, CycB and CycE activity and mutations in genes such as enc, cullin 1/lin19, effete/ubcD1, ubcD2, twin, half-pint/polyU binding factor 68 (pUf68), Lis-1, Dhc64C and γ-tubulin result in cysts with too many or too few germline cells. Encapsulation defects similar to those seen in vnc ovarioles are found in mutants that perturb Notch and Jak-Stat signaling pathways as well as when follicle cell precursor numbers are reduced (reviewed in Horne-Badovinac and Bilder, 2005), while binucleate oocytes are also encountered in hypomorphic mutants for spaghetti squash (sqh) that encodes the myosin light chain (Roth et al., 1999). Defects in genes that affect gurken (grk) localization, translation or signaling, with consequent effects on chorionic patterning result in alterations in DA morphology and placement reminiscent of those seen in vnc eggs. Finally, mutations in genes such as cup, ovarian tumor, spoon, Hrb27C, fs(2)B, pUf68, squid, Survival motor neuron (Smn) and rhino result in defects in nurse cell chromosome dispersion similar to vnc alleles. Rhino encodes a germline-specific homolog of Heterochromatin Protein 1 (HP1; Volpe et al., 2001) and is required for piRNA production and transposon silencing (Klattenhoff et al., 2009) while the other loci appear to be involved in regulation of mRNA splicing or translation and interact either genetically or functionally, suggesting they may affect a common pathway (Goodrich et al., 2004; Neuman-Silberberg, 2007; Lee et al., 2009). Many of these genes encode polypeptides that are potentially direct targets for ARD1 mediated N-terminal acetylation based on their predicted sequence (Arnesen et al., 2009 and supplementary Table S1). Other genes that show phenotypic overlap with vnc mutants could be affected indirectly, by reduction in ARD1 levels. The identification of in vivo targets for ARD1 and determining the functional consequences of this protein modification are important areas for further investigation.

Drosophila stocks

All fly strains were from the Bloomington Stock Center and were raised on standard media at 25°C, with constant humidity. Stocks used included vnc² kni^ri-1 e¹/ TM3 Sb¹, vnc¹⁴ kni^ri-1 e¹/TM3 Sb¹, l(3)67BDk¹ kni^ri-1 p^p/ TM3 Sb¹, y¹ w^67c23; P{EPgy2}Ard1^EY23147 and Df(3L)AC1 rn^roe-1 p^p/ TM3 Sb¹ a deficiency for 67BD region. The P-element lines y¹ w^67c23; P{SUPor-P}KG02042 ry⁵⁰⁶, y¹ w^67c23; P{SUPor-P}CG3204^KG0312 ry⁵⁰⁶, y¹ w^67c23; P{SUPor-P}CG8177^KG01611 ry⁵⁰⁶, y¹ w^67c23; P{SUPor-P}dpr6^KG07985 ry⁵⁰⁶, and y¹ w^67c23; P{SUPor-P}KG03548 ry⁵⁰⁶ were used for recombination mapping.

High resolution recombination mapping

The vnc locus was mapped with a resolution of +/− 50 kb by determining the recombination distance (RD) between the lethal vnc^BDk allele (alternatively called l(3)67BDk¹) and five P transposon lines with molecularly defined insertion sites in the 67BD region (see above and Table 2). Virgin female vnc^BDk kni^ri-1 p^p/ TM3 Sb¹ adults were crossed to males from each P insertion strain. The female progeny vnc^BDk/P{w⁺} were mated to Df(3L)AC1/TM3 hs-hid males. Four day old larvae were subjected to a two hour heat shock at 37°C. The only progeny expected to eclose are red eyed P{w⁺}/Df(3L)AC1 and white eyed flies resulting from recombination between the P{w⁺} insertion and vnc. The recombination distance (RD) in centi-Morgans was deduced by determining the percentage of white eyed progeny (Zhai et al., 2003). For each P insertion, roughly 5000 F2 offspring were scored for eye color (see Table 2). The projected molecular distance (PMD) of the vnc^BDk lesion from individual P insertions was calculated based on the known molecular position of the inserts and the recombination distance (RD) to the lesion as described (Zhai et al., 2003).

Mobilization of P element to induce imprecise excision

Imprecise excision of dpr6^KG07985 a Pw⁺ transposon inserted at nucleotide 9968287, 1321 bp 3′ of the predicted Ard1 stop codon, was used to generate two small deficiencies that delete vnc, Df(3L)vnc^M2/TM3 Sb and Df(3L)vnc^MK/TM3 Sb. Flies homozygous for the transposon are viable and fertile. Female P{w⁺} flies homozygous for the insert were mated to males carrying the Δ2-3 transposase source. White eyed male progeny were selected and used to establish stocks that were tested for sterility and lethality. Heterozygous adults from lethal lines were tested for their ability to complement the fertility defects in vnc² and vnc¹⁴ mutants. Two independent lines M2 and MK were sterile in trans to both vnc² and vnc¹⁴. PCR analysis of genomic DNA isolated from homozygous mutant embryos revealed that both lines contain small deletions extending from the insertion site that completely remove the Ard1 locus as well as the adjacent Nc open reading frame. These two stocks were named Df(3L)vnc^M2 and Df(3L)vnc^MK.

Molecular characterization of vnc

The molecular lesions in vnc², vnc¹⁴ and vnc^BDk, were characterized by directly sequencing PCR products generated from genomic DNA isolated from homozygous mutant animals. Primers used include: CG6685F: 5′-GGTTGGGGCATTTGGCATAG-3′; CG6685R: 5′-GCAGCGTTTATTTGACCACCAG-3′; Nedd2F: 5′-CGCTCGTTGGTCCTTTTGTTTC-3′; Nedd2R: 5′-CGGGAATGATTTCGTTTATCGC-3′; CG6674F: 5′-TTTGGCTTATGCGGCTTTTG-3′; CG6674R: 5′-TTACCAGGTGACAGACCATCC-3′; Ard1 Cloning F: 5′-CACCCGTTTGGCTGAAAATTCCTGC-3′; Ard1 Cloning R2: 5′-CATCAATAAAGTATGTTTACCGTTGGA-3′; Ard1 RT PCR1726F: 5′-CTTAGTGTGGAAGAAGTGTGC-3′; Ard1ORFF1: 5′-AATCAACAGCCACACCACCACG-3′; Ard1ORFF2: 5′-TAGCGGTCAAGCGTTCCTATCG-3′; Ard1ORFR2: 5′-TTGGTGTAGAGGTTCAGGGCAG-3′; Ard1ORFR1: 5′-CAACACGACACCTACAACGATGG-3′; DprF1: 5′-CGCCGTTCTTAAACTCAGTTCG-3′; DprSeqF1: 5′-GGAAAAAACGCCCTGCC-3′; DprSeqR1: 5′-GCACTTTCCCTCAACAC-3′; Ard1CloningR2: 5′-CATCAATAAAGTATGTTTACCGTTGGA-3′; DprSeqF2: 5′-CTCGCTAATTCAACCAC-3′; DprSeqR2: 5′-GTCTTTGCTTGCACTTG-3′; DprSeqF3: 5′-GAAGCGGGAGAAAGCTC-3′; Dpr61493R: 5′-CCCGAGAAATTGTTGGCG-3′; CG34356F3: 5′-CAACACAATCAACACAAGAACACG-3′.

Transgenic rescue

A genomic rescue construct was generated by amplifying a 3102 bp fragment encompassing the Ard1 transcribed region and 1.3 Kb of flanking sequence from wild type genomic DNA using primers Ard1 cloning F (5′-CACCCGTTTGGCTGAAAATTCCTGC-3′) and Ard1 cloning R2 (5-′CATCAATAAAGTATGTTTACCGTTGGA-3′). The PCR product was cloned into the Gateway™ entry vector pENTRD. After sequence verification the insert was transferred to pUASTattB-HW, a transformation vector containing φC31 attB sites and a Gateway™ destination vector cassette (Turan and Warrior, unpublished). Transgenic lines were recovered with the Ard1 rescue construct inserted on the X and second chromosomes at the φX-2A and φX-22A sites (Bischof et al., 2007).

Quantitative PCR

Total RNA (2 μg) isolated from ovaries or female carcasses (RNeasy Mini Kit, Qiagen, Inc.) was used as substrate in a 20 μl reverse transcription reaction with an oligo(dT)16 primer (Omniscript Kit, Qiagen, Inc.). The first-strand cDNA (200 ng) served as template for quantitative RT-PCR performed with a DNA Engine Opticon™ continuous fluorescence detection system (MJ Research, Inc.). Each Q-PCR was performed four times using two independent biological samples. PCR specificity was confirmed by monitoring PCR product size and by melting curve analysis at each data point. The mRNA copy numbers for Ard1 were standardized against rp49 mRNA in each sample. Replicates from the same biological samples were first averaged for ΔCt calculation (ΔCt = Ct_wt − Ct_mt; C, cycle; t, threshold). Fold differences in Ard1 transcript levels were calculated as 2^ΔCt. Standard deviations were further derived from two averaged fold differences. Primers were as follows: Ard1: Forward 5′ TAGCGGTCAAGCGTTCCTATCG 3′ Reverse 5′-TTGGTGTAGAGGTTCAGGGCAG-3′; Rp49: Forward 5′-AGCGCACCAAGCACTTCATC-3′ Reverse 5′-GACGCACTCTGTTGTCGATACC-3′.

Generation of somatic clones

The vnc^BDk allele was used to generate vnc^BDk FRT-2A lines by recombination. Imaginal disc clones were generated by crossing vnc^BDk FRT-2A male flies to females homozygous for hs>FLP; 2xUbi-GFPnls FRT-2A. Progeny were heat-shocked for 2 hours at 37°C for 3 consecutive days and reared at room temperature until dissection.

Immunohistochemistry and western blotting

Ovaries were dissected, fixed, and stained as described previously (Lei and Warrior, 2000). Antibodies used: mouse anti-ORB (1:50; 4H8; Developmental Studies Hybridoma Bank), mouse anti-Lamin (1:50; ADL67.10; DSHB), mouse anti-C(3)G (1:500; 1A8-1G2; gift from R.S. Hawley), rabbit anti-C(3)G (1:3000; gift from M.A. Lilly), Rhodamine-Phalloidin (1:50; Molecular Probes), Alexa 488 goat anti-mouse (1:500; Molecular Probes), and DAPI (1:2000). The hybridomas 4H8, ADL67.10, 1B1 and JLA20 developed by P. Schedl, P. A. Fisher, H. D. Lipshitz and J. J-C Lin respectively were obtained from the Developmental Studies Hybridoma Bank developed under the auspices of the NICHD and maintained by The University of Iowa, Department of Biology, Iowa City. All images were acquired on Zeiss LSM510-META NLO confocal microscope and processed with Adobe Photoshop 7.0. Images were taken in Z-sections and projected as needed. Western blots were performed as previously described (Duncan and Warrior, 2002), with rabbit anti-human ARD1 (1:300; sc-33820; Santa Cruz Biotechnology), mouse anti-Actin (1:1,000; JLA20; DSHB) and HRP-conjugated goat anti-rabbit or anti-mouse secondary antibodies (1: 10,000; Jackson ImmunoResearch).