The Drosophila pipsqueak gene encodes a nuclear BTB-domain-containing protein required early in oogenesis

Proper development of the Drosophila early egg chamber is required to generate polarity in the egg and embryo. During the middle stages of oogenesis, RNAs and proteins required for germ cell determination and for the generation of embryonic abdominal segments are localized to the posterior pole of the oocyte (reviewed by St. Johnston, 1993). The localization of the germ cell determinant(s) and the mRNA encoding Nanos, the posterior morphogen, is accomplished by a common mechanism, which requires the activity of at least 9 maternal-effect genes. Mothers homozygous for mutations in any of these genes produce embryos that lack abdominal segments and pole cells (the germline precursors).

The pipsqueak (psq) gene has been shown to be required maternally for correct abdominal segmentation and pole cell formation in the embryo (Siegel et al., 1993). It has been suggested that the psq embryonic defect in posterior patterning might be due to insufficient expression of the maternal posterior group gene vasa in psq mutants (Siegel et al., 1993). In addition, we have shown that certain psq alleles also exhibit dorsal/ventral defects in eggshell structure (Horowitz and Berg, 1995).

Establishment of the dorsal/ventral axis is set into motion with the localization of gurken (grk) mRNA in the developing oocyte. The Grk protein serves as a signal to determine follicle cell fate at two distinct times during oogenesis (reviewed by Anderson, 1995). During mid-oogenesis, Grk signals to the follicle cells at the posterior of the egg chamber, inducing them to adopt a posterior fate (Roth et al., 1995; González-Reyes et al., 1995). Later, in stages 8-10 of oogenesis, Grk is localized as a cap above the oocyte nucleus (Roth et al., 1995) where it signals to adjacent follicle cells to determine their fate as dorsal. This signaling cascade culminates in the proper establishment of dorsal/ventral polarity in the embryo, and in the formation of the two dorsolateral respiratory appendages of the eggshell (reviewed by Schüpbach and Roth, 1994; Chasan and Anderson, 1993).

We report here the cloning and molecular characterization of the psq gene. Our analyses indicate that psq is a complex and large gene encoding multiple transcripts and protein isoforms. The PsqA isoform contains a BTB domain; these domains have been shown to mediate protein-protein interactions (Bardwell and Treisman, 1994; Chen et al., 1995). psq mutants that produce a truncated protein harboring this domain exhibit defects in the dorsal/ventral patterning of the eggshell and embryo.

P[lacZ; ry⁺] (PZ) insertional mutations at the psq locus (alleles psqy²⁴⁰³, psqy⁰¹¹⁵, psqy⁸¹⁰⁹ and psq⁰⁴⁸²) were generated in a screen described in Karpen and Spradling (1992); psq is referred to as zeppelin in Spradling (1993). Excision lines psq^RF9 and psq^RF13 were created from allele psq²⁴⁰³ as described previously (Horowitz and Berg, 1995). The deficiency, Df(2R)psq-lolaΔ18 (Δ18) was generated by excising the two PZ elements resident in allele psq⁰⁴⁸² such an excision is likely to delete all of the DNA between the two P elements (Cooley et al., 1990). The Δ18 deficiency fails to complement both psq and a nearby lethal (lola, E. Giniger, personal communication); Southern blot analysis (data not shown) demonstrates that the Δ18 deficiency has lost DNA beginning at the PZ element within psq and extending through the length of our cloned DNA 5′ to the psq gene (see Fig. 6).

The EMS allele psq^HK38 (Schüpbach and Wieschaus, 1991) was originally recovered as a second mutation on a cappucino mutant chromosome. Our studies employed a recombinant chromosome lacking the cappucino mutation.

The P[lacZ, w⁺] (PW) line psq^fs1 and the excision allele psq¹⁻³⁰ are described by Siegel et al. (1993). These authors refer to psq^fs1 as psq^P1 and psq¹⁻³⁰ as psq^XI−30 Transformant lines were produced by germline transformation carried out according to Spradling (1986).

Cuticles were prepared according to Wieschaus and Nüsslein-Volhard (1986). Ovaries and individual egg chambers were prepared according to Verheyen and Cooley (1994). In situ hybridizations were carried out as described previously (Gillespie and Berg, 1995). Whole-mount analyses using antibodies to the Psq protein were carried out as follows: ovaries from 2-to 3-day-old females were dissected in modified EBR and fixed in devitellinizing buffer for 10 minutes (Verheyen and Cooley, 1994). All washes were carried out for 5 minutes unless otherwise specified. Ovaries were rinsed three times in 1× PBS and subsequently extracted in 1% saponin in 1× PBS for one hour at room temperature. The ovaries were rinsed twice in 0.1% saponin in PBS and once in PBTB (1× PBS, 0.1% Triton-X-100, 0.2% BSA), then incubated for 30 minutes in 5% normal goat serum (NGS) in PBTB. Rabbit serum was diluted (1:500) into PBTB with 5% NGS and overnight incubation was carried out at 4°C. The tissue was rinsed four times for 15 minutes each in PBTB and incubated with a 1:100 dilution of BODIPY-conjugated goat anti-rabbit secondary antibody (Molecular Probes) in PBTB overnight at 4°C. The samples were subsequently rinsed four times for 15 minutes each in PBTB and mounted as described in Theurkauf et al. (1992). Fluorescence was observed on a Bio-Rad MRC-600 confocal microscope, utilizing the Bio-Rad COMOS program. Images were processed using the NIH Image program.

Preparation of DNA and RNA, and northern and Southern blot analyses were described previously (Horowitz and Berg, 1995).

Genomic DNA flanking the PZ psq²⁴⁰³ insertion and the PW psq^fs1 insertion were kindly provided by Haifan Lin and Vivian Siegel, respectively. Genomic DNA spanning the exons downsteam of the PZ insertions was isolated by probing a cosmid library (gift of Marc Champe, Genentech, San Francisco) with the 1.0 kb partial psq cDNA, 13-13, described previously (Horowitz and Berg, 1995). The 5.1 kb psq-1 cDNA, pHPT7-9, was isolated from an ovarian cDNA library (Stroumbakis et al., 1994) using a 300 bp 5′ UTR probe (see Fig. 6). Alternative cDNAs pHH62 and pHH63 (psq-2) were isolated by rescreening the library with the complete pHPT7-9 cDNA.

DNA sequencing was performed using the dideoxynucleotide chain termination method with the Sequenase kit (US Biochemical). Bases 615-624 (TATACTGCAG, Fig. 4A) were not present in pHPT7-9 but were found in both cDNA 13-13 and genomic DNA (TATACTG-GAG). The most likely cause of this deletion is an error by the reverse transcriptase during the cDNA library construction: this region contains direct repeats and significant dyad symmetry.

Determination of genomic structure was performed by a combination of restriction analysis of cloned DNA, Southern blot analysis of whole fly DNA and DNA sequencing. For all introns, sequence determination of both the 5′ and 3′ junctions in the genomic DNA permitted unambiguous identification of the position of each intron relative to the cDNA (see Fig. 4A). We identified introns between the following pairs of nucleotides in the psq-1 cDNA (numbering as in Fig. 4A): (614, 615), (883, 884), (1048,1049), (2014, 2015), (2153, 2154), (2693, 2694), (3332, 3333), (3435, 3436), and (3619, 3620).

Nucleotide pairs presented above in bold indicate positions at which the introns were sequenced in their entirety. The presence of the intron at (883, 884) was inferred based on Southern analysis of genomic DNA and homology to consensus sequences; the precise position of this intron was determined by Zollman et al. (1994). The genomic structure that we have determined differs somewhat from that presented by Weber et al. (1995).

Sequence was compiled using the IntelliGenetics program. Database searches were carried out using the BLAST program from the NIH (Altschul et al. 1990).

The sequences of the three classes of embryonic psq transcripts identified by Weber et al. (1995) are distinct from the ovarian transcripts that we present here. The primary differences are as follows: The 5′ ends of all the embryonic transcripts differ from the ovarian transcripts psq-1 and psq-2. The first 210 nucleotides of the E1A exon of Weber et al. (1995) are unique, but are followed by sequence starting at position 402 of the psq-1 transcript (numbering as in Fig. 4A). Our genomic sequencing indicates that nucleotides 1-614 of psq-1 are present in a single exon, suggesting that the 5′ end of E1A is derived from sequence 5′ to the psq-1 transcript. The sequences of exons E1B and E1C of Weber et al. (1995) are not present in any of the ovarian transcripts that we have identified. The 3′ ends of the embryonic transcripts are extended by approximately 1 to 1.2 kb compared to the ovarian transcripts psq-2 and psq-1, respectively.

The predicted proteins from all embryonic transcripts differ from those of the ovarian transcripts: the predicted embryonic proteins have 24 unique amino acids C-terminal to amino acid 1062 of the protein encoded by psq-1 (see Fig. 4A). Comparison with our sequence shows that the embryonic transcripts lack the sequence from 3920 to 4096 of psq-1, suggesting that an additional splicing event occurs in the embryonic tissues, permitting translation to proceed somewhat further than predicted for the ovarian proteins.

An XbaI fragment containing sequences from position 1 to 4619 of cDNA pHPT7-9 was cloned into the unique XbaI site of the transformation vector pCaSpeRhs (Thummel and Pirrotta, 1991), which uses the w⁺ gene as a marker.

A 570 bp EcoRI-SalI fragment isolated from a RT-PCR product of the psq-l(3)S12 truncation mRNA (see Fig. 4A, Horowitz and Berg, 1995) was cloned into pCaSpeRhs. This fragment contains the entire coding region of the predicted fusion protein, as well as 47 bp of psq 5′ UTR and 29 bp of sequence 3′ to the predicted l(3)S12 stop codon (Dutton and Chovnick, 1991).

A BamHI fragment of psq extending from nucleotide 864 to 5162 of the pHPT7-9 cDNA (amino acids 45-1065) was cloned into the BamHI and BglII sites of expression vector pQE-13 (Qiagen).

A fragment extending from the NcoI site (position 2038 in Fig. 4A) to the 3′ end (NotI site) of the psq-2 cDNA pHH63 (see Results) was cloned into the NcoI and NotI sites of the expression vector, pET-28C (Novagen). Expression of pHH70 produces a protein of 69×10³M_r derived entirely of psq sequence (amino acids 436-1065 in Fig. 4).

This is identical to pHH70, except that the fragment was derived from cDNA pHH62 (see Results), and thus harbors a deletion of amino acids 719-736 (Fig. 4), producing a protein of 67×10³M_r.

A 6xHis-Psq fusion protein generated from pHH14 was isolated from

E. coli inclusion bodies. Protein purified by preparative SDS-polyacrylamide gel electrophoresis was used to immunize rabbits (R&R Rabbitry, Stanwood, WA). Antisera from two different rabbits (AS1 and AS2) showed slightly different immunoreactivity on westerns (see Results).

Drosophila ovary lysates were prepared by homogenizing ovaries in buffer (50 mM Tris, pH 7.5, 3 mM EDTA, 1% NP-40, 0.1% SDS, 1 mM N-ethylmaleimide, 2 μg/ml leupeptin, 200 KIU/ml Traysol, 100 μg/ml PMSF), and protein concentration was determined using the Bio-Rad Protein Assay reagent. Approximately 5 μg of protein were loaded per lane, subjected to SDS-polyacrylamide gel electrophoresis, and transfered to nitrocellulose. Membranes were probed with polyclonal antiserum at a 1:5000 dilution using the ECL Western blotting analysis system (Amersham).

The strain referred to as X-3; psq²⁴⁰³ has the genotype w¹¹¹⁸ P[hs-psq]; cn psq²⁴⁰³; +/− ry⁵⁰⁶. w¹¹¹⁸flies were used as a control. 3-to 5-day-old females (n=40-50) were aged with males for 2 days in embryo collection bottles at 25°C. Flies were subjected to a 1 hour heat pulse at 37°C in prewarmed vials, after which new males were provided. Timed collections on apple juice/agar plates were performed and hatch rates determined 48 hours postcollection. Fig. 7 shows combined data from two experiments; the curves were plotted by splining the data. For examination of heat-shock-induced Psq proteins, shocked flies were permitted to recover with new males at 25°C for 2 hours, after which ovaries were isolated and protein extracts prepared as described above.

Six strains containing insertions of P[hs-psq-l(3)S12] in a w¹¹¹⁸ back-ground were analyzed. w¹¹¹⁸ was used as a control. Approximately 20 1-to 2-day-old female flies from each line were heat-shocked in prewarmed vials at 37°C for 1 hour, then transferred into fresh vials containing wild-type males and allowed to lay eggs for approximately 12 hours. Flies were then transferred into fresh vials for another 12 hour collection. Each day for 7 days the males were removed and the females heat-shocked as before. The number of eggs laid in each vial was counted immediately after the transfer of adults to fresh vials.

Mutations at the psq locus cause a variety of defects in oogenesis and embryogenesis (see Table 1 for a summary of alleles). The phenotypes fall into three classes. The weak allele psq^fs1 disrupts posterior segmentation and pole cell formation during embryogenesis (Siegel et al., 1993). The intermediate alleles psq^HK38 and psq²⁴⁰³ exhibit dorsal/ventral defects in the eggshell and embryo in addition to the posterior axis defects described above (Figs 1, 2). Both of these intermediate alleles display a range of phenotypes; psq^HK38 is significantly weaker than psq²⁴⁰³ (Figs 1, 2). Flies homozygous or transheterozygous for the strong alleles psq⁸¹⁰⁹ and psq⁰¹¹⁵ have reduced viability, produce only a few, strongly dorsalized late stage eggs (Fig. 2), and have defects in early oogenesis. As expected, psq⁸¹⁰⁹/psq²⁴⁰³transheterozygotes produce eggs with stronger dorsal/ventral eggshell defects than psq²⁴⁰³ homozygotes (Fig. 2). As is true for most mutants with eggshell defects, the number of embryos fertilized in all of the psq strains decreases with the increase in severity of the eggshell phenotypes.

Two other alleles do not fall readily into this classification scheme. The PZ insertion allele psq⁰⁴⁸² is lethal, but acts as a weakly moderate allele in trans to all other alleles. psq⁰⁴⁸² also fails to complement mutations in a nearby lethal locus (lola, E. Giniger, personal communication) that is a hotspot for P element insertion. In a complementation analysis between alleles of this lethal locus and all the psq alleles discussed here, only psq⁰⁴⁸² failed to complement mutations in both genes. Southern blot analysis demonstrates that psq⁰⁴⁸² contains two P elements (data not shown), suggesting that both psq and the nearby hotspot locus contain PZ insertions.

A second allele that cannot be classified in our simplified groupings is the excision line psq^1-30 generated by mobilization of the PW element in psq^fs1 (Siegel et al., 1993). As described below, this allele contains a deletion of the 5′ end of the psq gene. The psq^1-30 allele in homozygous flies produces defects early in oogenesis (Siegel et al., 1993), yet acts like a simple posterior group allele in trans to all other psq alleles. We have generated a larger deletion, Δ18 (see Materials and Methods) which encompasses and extends beyond deletion 1-30. psq^1-30/Δ18 trans-heterozygotes produce embryos with a classic posterior group phenotype (Fig. 1); no defects in early oogenesis or in eggshell structures (Fig. 2) are observed. These transheterozygotes are null for the psq 5.1 kb transcript (see below).

To facilitate our understanding of the psq mutant phenotypes, we have conducted a molecular analysis of the gene. We previously reported the identification of a 5.1 kb psq transcript in wild-type females and the isolation of a 1.0 kb partial cDNA, 13-13, representing this message (Horowitz and Berg, 1995); we now call the 5.1 kb transcript psq-1. Using a 300 bp fragment from the 5′ UTR of psq-1 as a probe to northerns of RNA from wild-type adult males, females and isolated ovaries, two transcripts are revealed, the 5.1 kb psq-1 and a 6 kb message (Fig. 3). psq-1 is highly abundant in females but below the level of detection in males, while the 6 kb transcript appears to be present equally in both sexes. It should be noted that probes further 3′ in the psq-1 message detect additional transcripts by northern analysis (see below).

We have found that the psq-1 transcript is reduced in amount in females from all the PZ lines and is replaced by an abundant 1.6 kb transcript (Horowitz and Berg, 1995; and see 2403 and 0115/CyO in Fig. 3A). In a previous study, we demonstrated that this product results from an aberrant splicing event (Horowitz and Berg, 1995). The PZ elements in all four PZ mutant lines lie within a large intron of the psq gene. Sequences present within the partial l(3)S12 gene residing on the PZ element provide a splice acceptor site and termination signal, causing the psq-initiated transcript to be spliced to l(3)S12 sequence in the PZ element, then truncated. This truncated mRNA encodes a putative Psq-L(3)S12 fusion protein (Horowitz and Berg, 1995). Fertile revertants (such as psq^RF13 and psq^RF9) generated by inducing excision of the PZ element in psq²⁴⁰³ delete part or all of the l(3)S12 gene’s splicing and termination signals and restore the wild-type psq-1 transcript (Fig. 3A, and Horowitz and Berg, 1995).

In the current study, we have extended our analysis of the psq gene by examining transcripts produced by other, non-PZ alleles of psq^fs1. In psqfemales, psq-1 is absent and is replaced by a transcript of approximately 2.5 kb (Fig. 3A). Sequence analysis reveals that the PW element of psq^fs1 is inserted in a 3.2 kb intron in the 5′ UTR of the psq-1 message. The new 2.5 kb transcript observed in psq^fs1 females most likely results from psq-initiated transcription that terminates within the PW element, as the transcript does not hybridize to cDNA sequences 3′ to the position of the PW insertion. RNA from the excision strain psq^1-30 produces no signal on the northern blot presented in Fig. 3A because the sequences hybridizing to the probe have been deleted from the genomic DNA. These results indicate that complete absence of the psq-1 transcript results in a weaker phenotype than observed for the PZ psq alleles, suggesting that the fusion protein produced in the PZ lines may be deleterious (see below). RNA from females harboring the EMS-induced allele, psq^HK38 show no apparent change in size or abundance of psq-1 transcript (data not shown).

The expression pattern of psq-1 on a developmental northern blot (Fig. 3B) conforms to the classic pattern observed for maternal mRNAs which are deposited into the developing oocyte. The message is abundant in ovaries and 0-2 hour embryos, but decreases to levels below detection in older embryos, larvae and pupae. Fig. 3B also demonstrates the expression of the 6 kb message described above, which is seen at varying levels throughout development.

We examined the expression of psq mRNA during oogenesis by in situ hybridization of the 5′ UTR psq-1 probe to whole ovaries. psq mRNA is expressed throughout oogenesis. Upon transport into the oocyte at stage 11, the message is uniformly distributed (data not shown).

The 5′ UTR psq-1 probe from cDNA 13-13 was used to screen an ovarian cDNA library (Stroumbakis et al., 1994). Sequence analysis of the longest clone, pHPT7-9, is presented in Fig. 4A. The pHPT7-9 cDNA is 5162 nucleotides in length and contains an open reading frame encoding a protein of 1065 amino acids (M_r 115×10³). The N terminus of the predicted Psq protein contains a BTB (or POZ) domain (underlined in Fig. 4A), a motif that has been shown to function in protein-protein interactions (Bardwell and Treisman, 1994; Chen et al., 1995). The BTB domain of psq has been cloned independently by virtue of its homology to other BTB domains in a study designed to isolate members of the BTB family of proteins (BTB-V in Zollman et al., 1994).

The Psq protein contains a number of repeated amino acid sequences, including multiple regions enriched for serine/glycine, proline or alanine residues, several opa repeats (Wharton et al., 1985), and a domain of histidine residues alternating with other amino acids (mostly glutamate or glycine). In addition, we have identified a 52-55 amino acid sequence, which we call the psq motif, that is repeated in tandem four times at the C terminus of the protein (Fig. 5). The dTKR gene of Drosophila (Haller et al., 1987) contains a single copy of the psq motif; the residues most highly conserved within the psq repeats are also conserved in TKR. A search of the sequence database revealed the presence of a variety of protein sequences from species as diverse as bacteria and humans with significant homology to the psq motif. The most highly conserved sequences are aligned in Fig. 5.

Using a 1.2 kb fragment derived from the 3′ UTR of psq-1 as a probe to northern blots of RNA from wild-type females, we observed a heterogeneous band corresponding to an additional transcript(s) of approximately 4.4 kb in length. Of the mutant lines tested (psq²⁴⁰³, psq^HK38, psq^fs1 and psq^1-30), all retained the 4.4 kb transcript(s) (data not shown). The presence of the 4.4 kb transcript(s) in psq⁰¹¹⁵ or psq⁸¹⁰⁹ homozygous flies was not tested due to the difficulty in obtaining suffficient flies from which to prepare RNA.

We used the entire psq-1 cDNA as a probe to rescreen the ovary cDNA library in search of alternative psq transcripts. One new cDNA, pHH63, contains an insert of approximately 4.4 kb and may correspond to the 4.4 kb transcript described above. We call the message encoded by pHH63 psq-2. psq-2 has 127 bp of unique sequence at the 5′ end (Fig. 4B) which is spliced to the fourth exon (position 1049, Fig. 4A) of the psq-1 sequence. The unique 5′ end of psq-2 is derived from within the large intron of psq-1 (see Fig. 6). Notably, the psq^RF13 deletion line retains the sequences encoding the 5′ end of psq-2. The psq-2 transcript also differs from psq-1 by the presence of an additional 192 bp of sequence at the 3′ end (Fig. 4B). This novel 3′ sequence contains a consensus polyadenylation signal 29 bp 5′ to the poly(A) tail of the cDNA. Extensive restriction analysis suggests that, with the exception of the alternative 5′ end and the extended 3′ end, the sequence of this cDNA is identical to bases 1049 to 5146 of psq-1. The best candidates for translation initiation are four in-frame methionine codons clustered within a 17 amino acid stretch of sequence, positioned in the same frame as the conceptual protein encoded by psq-1 (asterisks, Fig. 4A). Translation initiated at one of these methionines is predicted to produce a protein ranging in length from 630 to 646 amino acids and which is identical to the carboxy-terminal two-thirds of the psq-1-encoded protein. Interestingly, the psq-2-encoded protein would completely lack the BTB domain found in the larger protein (see Fig. 6).

Another cDNA, pHH62, which contains an approximately 4 kb insert, was found to contain yet another transcript variant. This cDNA is distinguished from the psq-1 transcript by the presence of a 54 base-pair deletion extending from position 2889 to 2942. As the sequence flanking the ends of the deleted region provide excellent matches to the consensus splice donor and acceptor sequences (Mount et al., 1992), this deletion most likely represents an alternative splicing event. Such a splice perfectly deletes 18 amino acids but retains the open reading frame of the protein (Fig. 4A, bracketed amino acids). These deleted amino acids are located in the first of the four psq motif repeats described in Fig. 5. The 3′ end of pHH62 contains the same extended 3′ UTR found in psq-2. The 5′ end of the pHH62 cDNA begins within the coding region of the psq-1 sequence, at position 944, and thus it is unlikely that this represents the true 5′ end of this transcript. If the 5′ end of the message is the same as that for psq-1, a protein of 113 x 10M_r, slightly smaller than that encoded by psq-1, should be produced.

It is clear that psq is a complex gene; there may be additional alternative ovarian transcripts for which we have not yet isolated cDNAs. We have chosen to focus on psq-1, which encodes a protein affected in all of the mutants that we have characterized (see below). Additional experiments reveal that mutation of psq-1 is the primary source of the defects observed in all of the mutants discussed here (see below).

To verify that cDNA pHPT7-9 (psq-1) corresponds to the psq gene, we tested whether expression of the protein encoded by this cDNA could rescue psq mutants. We introduced a fragment of the cloned cDNA under the control of the hsp70 heat-shock promoter into various mutant lines and asked if expression of the transgene can restore fertility and/or wild-type eggshell phenotypes to the mutant females.

psq²⁴⁰³ females carrying an X-linked copy of the P[hs-psq] transgene (X-3; psq²⁴⁰³) lay eggs at a frequency similar to that observed for the w¹¹¹⁸ control. Up to 90% of the eggs laid in the absence of heat shock possess two dorsal appendages (see Fig. 2E), indicating that the low level expression of the hsp70 promoter in the absence of heat shock is sufficient to rescue the eggshell defect in this mutant line. These eggs (with the exception of a few rare escapers) do not hatch.

Heat-shock induction of the transgene in psq²⁴⁰³ females restores fertility. We conducted experiments to determine the oogenic stage at which psq is first required by subjecting X-3; psq²⁴⁰³ mutant females to heat shock and determining the hatching frequency of eggs laid at defined intervals following induction. A peak in the hatching frequency (as high as 70-80%) occurs approximately 80 hours post heat shock (Fig. 7). This timing corresponds to a requirement for the Psq protein prior to stage one of oogenesis (King, 1970; Mahowald and Kambysellis, 1980; Lin and Spradling, 1993). The lower levels of rescue observed prior to the peak at 80 hours post heat shock may reflect an additional requirement for psq function at multiple times throughout oogenesis.

Heat-shock induction of psq⁰¹¹⁵ females carrying the p[hs-psq] transgene results in the production of eggs with two dorsal appendages in over 60% of the eggs laid, but none of these eggs hatch (see below).

In order to characterize the protein products of the psq gene, polyclonal antibodies were raised against bacterially expressed Psq protein. These antibodies, AS1 and AS2, recognize slightly different substrates on western blots of ovarian extracts.

On a western blot, immune serum AS1 recognizes at least two Psq protein species in wild-type ovaries (Fig. 8A). PsqA migrates as a broad band corresponding to a M_r of approximately 150×10. This species is found in protein extracts prepared from the ovaries of wild-type females (Fig. 8A), but is absent in mutants psq^fs1, psq^fs1/psq^1-30, psq²⁴⁰³ or psq⁰¹¹⁵ (Fig. 8A). Ovary extracts of the EMS-induced allele, psq^HK38, show a weak but reproducible signal at the fastest migrating position of the PsqA band (Fig. 8A). The PsqA band is restored in fertile revertants of psq²⁴⁰³ (Fig. 8A, RF13 lane) and in flies carrying the heat-shock psq-1 rescue construct after treatment at 37°C (Fig. 8B, X-3; 2403 + lane). These findings demonstrate that protein PsqA is encoded by transcript(s) from cDNA pHPT7-9 (psq-1). The diffuse nature of the protein A signal could result from post-translational modification of the protein (e.g. glycosylation and/or cleavage). It is also possible that the band contains multiple protein species deriving from alternatively spliced messages. As described above, cDNA pHH62 is most likely a partial copy of its cognate message. If the true 5′ end of pHH62 extends to the same initiator methionine as that of the psq-1 mRNA, then the deletion of 18 amino acids of sequence observed in this cDNA would produce a slightly smaller protein. The observation that expression of the P[hs-psq] construct in psq mutant flies restores a broad band of PsqA is consistent with either modification of a single protein or the possibility that more than one species of mRNA, and consequently more than one species of protein, are produced from this construct.

Several additional protein species are recognized by the AS1 immune serum. One of these, PsqB, migrates with an apparent M_r of 80×10, and could be encoded by the psq-2 transcript. Expression in bacteria of a protein initiating at methionine 436 (Fig. 4A), one of the potential start sites in psq-2, produces a protein with the same relative mobility as PsqB (Fig. 8D, pHH70 lane). PsqB is absent in extracts prepared from ovaries of psq⁰¹¹⁵ and psq⁰¹¹⁵/psq⁸¹⁰⁹ females (Fig. 8A and data not shown), but is unaffected in any of the other mutants tested. Expression of the P[hs-psq] construct in psq⁰¹¹⁵ females produces the broad PsqA band, but does not restore the PsqB band (Fig. 8C, X-3; 0115 + lane). The absence of PsqB in psq⁰¹¹⁵ females expressing the P[hs-psq] construct may explain our observation that fertility is not restored in these flies.

On a western blot, the AS1 immune serum also recognizes a third protein, which is not affected in any of the psq mutants examined. AS2 immune serum does not recognize this protein (Fig. 8E); we do not know whether this protein is encoded by the psq locus.

AS2 immune serum recognizes PsqA and only one additional protein species (Fig. 8E). We used this antiserum to determine the localization of PsqA in whole ovaries by comparing the patterns observed in wild-type and psq²⁴⁰³ females. The results are shown in Fig. 9A-E. PsqA is abundant early in oogenesis, and is first detected at the anterior tip of the germarium, suggesting a role in the germline stem cells. PsqA is particularly abundant in the nuclei of the germ cells in region 2a, and is observed in the pro-follicular layer as well as the follicle cells of later stage egg chambers. The protein remains nuclear in both soma and germline throughout oogenesis. There is particularly strong staining in the anterior and posterior polar follicle cells of egg chambers during stages 5 and 6. A very distinct pattern is seen in the oocyte nucleus as two small and discrete spots (Fig. 9C). The staining observed in psq²⁴⁰³ ovaries is greatly reduced in comparison to the wild type (Fig. 9E), indicating that the patterns described above predominantly reflect the distribution of PsqA.

The eggshell and cuticle defects that we observed with some psq mutant alleles suggested that dorsal/ventral patterning is disrupted in these mutants. To determine if this effect is mediated by the gurken signal transduction pathway, we examined the localization of grk mRNA in ovaries from psq mutants. In ovaries from wild-type females, grk mRNA is localized to the oocyte in early egg chambers and by stage 8 is seen as a tight cap on the dorsal side of the oocyte nucleus (Fig. 10A). In egg chambers from psq²⁴⁰³ homozygous females and psq²⁴⁰³/psq⁸¹⁰⁹ transheterozygous females, the amount of grk mRNA is reduced. In psq²⁴⁰³ mutants, grk mRNA is localized correctly in 90% (n=46) of stage 10 egg chambers (Fig. 10B). In the remaining 10% of egg chambers, grk mRNA is found as a diffuse band in the dorsal, anterior region of the oocyte, or fails to be localized at all. In psq²⁴⁰³/psq⁸¹⁰⁹mutants, grk message is mislocalized in 24% (n=29) of stage 10 egg chambers. In these cases, message is either uniformly distributed throughout the oocyte or present in a ring at the anterior end (Fig. 10C). The combination of reduced levels and mislocalization of grk mRNA could lead to the unusual eggshell phenotypes that we observe in PZ psq mutants (Fig. 2). It is also possible that some contribution to the mutant phenotype is due to defects downstream of grk in the follicle cells.

Analysis of the mutant phenotypes associated with the PZ alleles suggests that the aberrant splicing of psq into the l(3)S12 sequences in these lines creates a fusion protein with negative effects on oogenesis. To test this possibility, we expressed the predicted Psq-L(3)S12 fusion protein in transgenic flies. We generated a construct, P[hs-psq-l(3)S12], in which a mRNA encoding the fusion protein could be expressed by heat-shock induction from the hsp70 promoter. Six independent transgenic lines containing homozygous insertions of this construct were placed on a daily heat-shock regime and their egg-laying abilities compared to w¹¹¹⁸ control animals. Three and a half days following the initial heat shock, all six transgenic lines demonstrate a dramatic drop in the number of eggs laid (Fig. 11, and data not shown), while control animals continue to lay eggs at a high rate. Dissection of the ovaries from these animals reveals a decreased number of egg chambers at all stages. In addition, a significant proportion of late stage egg chambers are short in length, with partially dorsalized eggshell structures. Thus, overexpression of the Psq-L(3)S12 fusion protein is sufficient to render flies sterile, phenocopying weak PZ alleles.

We report here the cloning and molecular characterization of the Drosophila psq gene, which encodes a novel nuclear protein involved in axis definition. Our results demonstrate that psq is a complex gene encoding multiple differentially spliced transcripts and several protein isoforms. In the course of submitting this manuscript, a characterization of the embryonic psq transcripts appeared in the literature (Weber et al., 1995). A sequence comparison of the ovarian and embryonic psq transcripts reveals that they have distinct 5′ and 3′ ends, and that the proteins encoded by transcripts from the two tissues differ slightly at their C termini (see Materials and Methods for details).

We have focused our efforts on the characterization of psq-1, the mRNA that encodes PsqA. This protein, which is drastically reduced or absent in all of the psq mutants that we have examined, has a number of interesting features. At the amino terminus, PsqA contains a BTB domain (Godt et al. 1993; also referred to as a POZ domain by Bardwell and Treisman, 1994), a motif that has been shown to function in protein-protein interactions. Although BTB domains are often found near the N terminus of Cys₂-His₂ zinc finger proteins, PsqA does not appear to contain a zinc finger. Downstream of the BTB domain, PsqA contains 34 alternating histidine residues, (HX)_n, a motif that is present in a number of other Drosophila proteins, primarily transcription factors. It has been proposed that these histidine repeats could mediate protein-protein interactions by coordinating metal ions to form a ‘histidine-metal zipper’ between two proteins containing the repeats (Janknecht et al., 1991). The presence of two potential protein-protein interaction domains suggests that PsqA monomers may interact with each other or with heterologous protein species. Additionally, PsqA contains four tandem copies of a conserved sequence of unknown function at its carboxy terminus; we call this sequence the psq motif.

We have identified another psq transcript, psq-2, which encodes a second isoform of the Psq protein lacking both the BTB and (HX)_n domains but retaining the psq motif repeats. This isoform is unlikely to be involved in the same type of protein-protein interactions predicted for PsqA. The finding of Bardwell and Treisman (1994) that a BTB-containing zinc finger protein was more competent to bind DNA when the BTB domain was removed raises the possibility that the isoform encoded by psq-2 could be an activated form of the Psq protein.

We previously hypothesized that the PZ fusion protein might act in a dominant-negative fashion, possibly by disrupting normal protein-protein interactions mediated by the BTB domain of the protein (Horowitz and Berg, 1995). This prediction was based on genetic analyses indicating that the production of a truncated protein is more harmful than production of no protein at all. Our observation that the psq²⁴⁰³ allele shows eggshell defects whereas the psq^fs1 and psq^1-30 alleles do not, despite the fact that all three mutants display the same loss of PsqA using our antibodies, provides additional evidence that the fusion protein is deleterious.

Our experiments with P[hs-psq-l(3)S12] reveal that expression of an isolated BTB domain can behave in a dominant manner. Expression of this construct in wild-type flies leads to marked reduction in egg-laying. Furthermore, the presence of two copies of the transgene causes a more severe effect on egg production than a single copy (data not shown). These observations, along with the results of the rescue experiments (see below), suggest that the dominant nature of the PZ mutation is sensitive to the ratio of the PZ and wild-type forms of the Psq protein, a notion that we previously forwarded, based on the absence of eggshell defects in PZ/+ flies (Horowitz and Berg, 1995).

As no PsqA can be detected in any of the psq PZ alleles, the eggshell defects that we observe in these lines cannot be due to interaction of the PZ fusion protein Psq-L(3)S12 with PsqA. However, if PsqA normally interacts with a heterologous BTB-containing protein, the presence in the PZ mutants of the PsqA BTB domain without its associated carboxy-terminal sequences may result in inappropriate activation or inactivation of the heterologous protein.

Recent work by two groups lends support to this model for Psq-L(3)S12 action. Bardwell and Treisman (1994) and Chen et al. (1995) have demonstrated that the BTB domain can promote dimerization or multimerization of proteins in either homomeric or heteromeric complexes in vitro. Interestingly, the presence of the BTB domain is associated with a reduction in DNA binding by the associated zinc finger region in several zinc finger BTB proteins tested (Bardwell and Treisman, 1994). Of particular relevance to our work is the finding that coexpression of a protein fragment containing only the BTB domain along with the full-length BTB-containing protein relieves the BTB-mediated inhibition of DNA-binding in vitro (Bardwell and Treisman, 1994). Thus, interaction of the BTB domain present in the psq PZ fusion protein with that of a heterologous BTB-containing DNA-binding protein might inappropriately activate the binding activity of that protein.

In contrast, expression of a BTB domain without its associated C terminus (eg. a BTB domain fused to heterologous sequence) may inactivate proteins that interact with the mutant BTB protein. Several examples of human cancers associated with translocations that may involve the joining of a BTB domain from one protein to another protein have been identified (see, for example, Chen et al., 1993). This observation has led to the suggestion that transformation could arise by a dominant-negative mechanism involving sequestration of partner BTB domain proteins in inactive complexes (Bardwell and Treisman, 1994; Chen et al., 1995). Our results provide the first direct evidence supporting the hypothesis that an abnormal BTB-containing protein can exert a dominant effect in vivo.

The dominant effect of the fusion protein is apparent in the dorsal/ventral defects of the PZ mutants. Because proper localization of grk message is required for dorsal/ventral polarity (reviewed by Schüpbach and Roth, 1994), we examined grk mRNA localization in the PZ lines. We find that PZ mutants with more severe eggshell defects display a higher frequency of grk mRNA mislocalization. Interestingly, in wild-type egg chambers, we see an abundance of PsqA in polar cells at stages 6-7, a time when these cells signal to the oocyte. This signaling results in a repolarization of the microtubules in the oocyte, a process required for proper anterior/posterior and dorsal/ventral axis definition (Ruohola et al., 1991; Lane and Kalderon, 1994). It is possible that the mislocalization of grk mRNA in psq PZ mutants is due to disruption of the signaling between the polar follicle cells and the oocyte. In addition, these mutants may affect processes that occur in the dorsal follicle cells during later stages, which could contribute to the dorsal/ventral defects observed. These effects may be produced by interaction of the mutant PZ protein with factors normally associated with the wild-type PsqA. Alternatively, the PZ fusion protein may be interacting spuriously with proteins involved in these signaling cascades.

Our heat-shock rescue experiments suggest that psq is required very early, prior to stage 1 in oogenesis. These results are consistent with the early oogenesis defects observed in the most severe psq mutants and with the abundant level of Psq protein detected in the 16-cell cysts in region 2 of the germarium. Interestingly, the rate of hatching of heat-shocked X-3; psqflies does not drop back to the pre-shock value, even after 160 hours post heat shock (data not shown). Rather, hatching appears to plateau at a value intermediate to the pre-shock and peak value of hatching observed. This may indicate that a permanent change is effected by expression of psq and that continued expression of the gene is not necessary to maintain fertility. Our finding that PsqA is present in the very tip of the germarium, and thus may be expressed in the stem cells, is consistent with this scenario. Alternatively, the PsqA protein produced upon heat shock may be quite stable, such that fertility drops off slowly after the heat-shock peak is achieved.

Antibody staining of ovaries shows that PsqA is a nuclear protein, expressed in both the somatically derived follicle cells and in the germline. The apparent association of the PsqA protein with chromatin in the nurse cells, and the specific fluorescence in distinct dots in the oocyte nucleus, argue that PsqA protein is associated, directly or indirectly, with DNA. These observations, along with the results discussed below, suggest that PsqA plays a role in regulating gene expression in oogenesis.

Previous work (Siegel et al., 1993) has indicated that ovaries from psq mutant females have reduced levels of vasa mRNA and protein, implicating psq in the regulation of vasa expression. Expression of vasa is initially observed in the germarium; we find that PsqA is both abundant in the early germarium and required very early in oogenesis, correlating with the presence of Vasa at this time.

High levels of PsqA are observed in the posterior polar cells at stages 5-6, a time when these cells are purported to be involved in receiving and responding to a signal from the posteriorly localized Grk (reviewed by Anderson, 1995). It is possible that PsqA is involved in regulating gene expression as part of this complex signaling process.

The striking and specific pattern of PsqA localization in the oocyte nucleus at stages 6-10 (seen as two small bright spots of fluorescence in Fig. 9C) suggests that PsqA may bind, either directly or indirectly, to a specific site on the DNA in the oocyte nucleus. It is intriguing to consider the possibility that psq may be directly affecting the expression of a particular locus in the oocyte nucleus.

Finally, psq could also play a more general role in regulating gene expression by affecting chromatin structure. Our results indicate that PsqA protein is associated with the nurse cell chromatin; Siegel et al. (1993) found that psq mutants failed to undergo the normal decondensation of nurse cell DNA at stage 5. Interestingly, two other BTB-containing proteins, E(var)3-93D and GAGA, have been implicated in the modulation of chromatin structure, as well (Dorn et al., 1993; Croston et al., 1991). It is possible that the BTB-containing fusion protein produced in the psq PZ mutants may disrupt the protein-protein interactions of these or other BTB-containing modulators of chromatin structure, leading to some aspects of the mutant phenotypes we observe in these lines.

In conclusion, we have shown that psq encodes a nuclear protein which may be involved in the control of gene expression in oogenesis. The structure of PsqA and the nature of the psq PZ fusion protein suggest that these proteins interact directly with other proteins. These interactions could be involved in regulating the expression of vasa, or other genes required to establish polarity. Future experiments designed to identify the factors with which PsqA interacts should provide valuable insight into psq’s function.

We thank Haifan Lin, Vivian Siegel and Trudi Schüpbach for providing mutant strains and DNA. We acknowledge the expert technical assistance of Doreen Gillespie in whole-mount in situ hybridization and immunocytochemistry and express our gratitude to Philippa Webster and Alison Volpe for critical reading of the manuscript. We are especially indebted to members of the Berg lab for help in preparation of this manuscript and for insightful discussions: Doreen Gillespie, Philippa Webster, Kim Rittenhouse, Alison Volpe, Jon Schnorr, and Rebecca Botham, and to Michael Gelb and Mark Kot. This work was supported by the March of Dimes Basil O’Connor Research Program and NIH (GM45248).

The nucleotide sequence data reported in this paper are available from the GenBank Nucleotide Sequence Databases under the accession numbers U48358 (psq-1) and U48402 (psq-2).