A Developmentally Regulated Splice Variant from the Complexlola Locus Encoding Multiple Different Zinc Finger Domain Proteins Interacts with the Chromosomal Kinase JIL-1*

Using a yeast two-hybrid screen we have identified a novel isoform of the lola locus, Lola zf5, that interacts with the chromosomal kinase JIL-1. We characterized thelola locus and provide evidence that it is a complex locus from which at least 17 different splice variants are likely to be generated. Fifteen of these each have a different zinc finger domain, whereas two are without. This potential for expression of multiple gene products suggests that they serve diverse functional roles in different developmental contexts. By Northern and Western blot analyses we demonstrate that the expression of Lola zf5 is developmentally regulated and that it is restricted to early embryogenesis. Immunocytochemical labeling with a Lola zf5-specific antibody of Drosophila embryos indicates that Lola zf5 is localized to nuclei. Furthermore, by creating double-mutant flies we show that a reduction of Lola protein levels resulting from mutations in the lola locus acts as a dominant modifier of a hypomorphic JIL-1 allele leading to an increase in embryonic viability. Thus, genetic interaction assays provide direct evidence that gene products from the lola locus function within the same pathway as the chromosomal kinase JIL-1.

male X chromosome for dosage compensation in flies (reviewed in Ref. 7). This enhanced transcription is thought to arise from MSL complex-induced histone H4 acetylation generating a more open chromatin structure (8). The increased histone H3 Ser-10 phosphorylation levels that JIL-1 promotes on the male X may also play a role in maintaining a more open and active chromatin structure (5,6). However, it is not known whether physiological substrates of JIL-1 may include other proteins such as transcription factors or whether there are proteins directly regulating the function of JIL-1. To identify proteins that interact with JIL-1 we carried out yeast two-hybrid screens using different JIL-1 regions as baits. Here we report that a novel splice form from the lola locus, which we have named Lola zf5, was identified in such a screen to interact with the first kinase domain (KDI) of JIL-1.
The lola locus was first characterized as a mutation affecting longitudinal axon growth within the central nervous system (9). Two isoforms from the lola locus have been previously described, Lola long and Lola short (10). Lola long contains two zinc finger motifs and is a transcription factor with DNA binding activity (10,11). The lola locus mediates decreased copia retrotransposon mRNA expression in the central nervous system while upregulating its expression in gonads (11). In addition, lola is required for proper expression of the axonal guidance proteins Robo and slit in the central nervous system (12). Although both Lola isoforms share a BTB/POZ domain (13)(14)(15), Lola short contains no zinc finger domains (10). BTB domains are known to mediate dimerization (16,17), which includes the ability to promote heterophilic interactions of different BTB domain-containing isoforms (15,18,19). BTB domain-containing zinc finger proteins have been strongly implicated in regulation of chromatin structure and gene expression (20). For example, human B cell lymphoma (BCL-6) and promyelocytic leukemia zinc finger oncoproteins have been shown to act as transcriptional repressors (21)(22)(23). Specific recruitment of repressor complexes to target promoters occurs through binding of corepressors to the hydrophobic BTB dimer pocket (24). Corepressor binding recruits a complex containing a histone deacetylase (25,26) that represses transcription by inducing condensed chromatin architecture (reviewed in Ref. 27).
In this study we characterized the genomic organization of the lola locus and show that it is a complex locus from which at least 17 different splice variants are likely to be transcribed. All isoforms from the lola locus share a common BTB domain, whereas 15 of the splice variants each contain a different zinc finger domain. We show that one of these variants, Lola zf5, physically interacts with the JIL-1 kinase and that its expres-sion is developmentally restricted to early embryogenesis. Furthermore, we show that a P element insertion in the lola locus enhances the viability of a hypomorphic JIL-1 allele, indicating opposing functions of the JIL-1 and lola loci. Although only Lola long and Lola short have so far been studied in detail, it has long been known that P insertion mutations in lola that prevent the expression of some if not all the isoforms have very complex phenotypes (10 -12, 28). Our findings suggest that this complexity may derive from the expression pattern of multiple gene products from the lola locus that are likely to serve diverse functional roles in different cellular and developmental contexts.

MATERIALS AND METHODS
Drosophila Stocks-Fly stocks were maintained according to standard protocols (29). Oregon-R was used for wild type preparations. JIL-1 EP(3)3657 and JIL-1 z2 alleles have been previously described (6). Balancer chromosomes and mutant alleles are described in (30). The lola 00642 mutant stock cn 1 P{ry ϩt7.2 ϭ PZ}lola 00642 /CyO; ry 506 was obtained from the Bloomington Drosophila Stock Center. Hatch rates were determined by counting the number of eggs laid on standard apple juice/agar plates and then counting the number of unhatched eggs at 22 h and again at 48 h after egg laying. All genetic crosses and interaction assays were conducted at 23°C.
Identification and Molecular Characterization of Lola zf5-JIL-1 cDNA sequence encoding a 304-amino acid fragment (Tyr 251 -Glu 554 ) comprising the first kinase domain (JIL-1 KDI) of JIL-1 was subcloned in-frame into the yeast two-hybrid bait vector pGBKT7 (Clontech) using standard methods (31) and verified by sequencing (Iowa State University Sequencing Facility). The JIL-1 KDI bait was used to screen the Clontech Matchmaker TM 0 -21 h embryonic Canton-S yeast two-hybrid cDNA library according to the manufacturer's instructions. A positive cDNA clone KDIJ1 was isolated, retransformed into yeast cells containing the JIL-1 KDI bait to verify the interaction, and sequenced. Homology searches identified SW59 (GenBank TM Z97377) as well as the lola locus. We obtained the SW59 cDNA from Dr. D. Zhao (University of Edinburgh) and lola ESTs (LD28033, LD33478, and LD17361) from ResGen Invitrogen and assembled the full-length Lola zf5 coding sequence. We note a few differences between Z97377 and the Lola zf5 full-length cDNA assembled in this study: 1) 62 nucleotides at the most 5Ј-end of Z97377 may be a library construction artifact, because they are not present in other Lola zf5 EST clones or the reported lola genomic sequence (32); 2) six gaps are present between Z97377 and our KDIJ1 fragment. At all of the gaps, KDIJ1 cDNA sequence is 100% identical to the available ESTs and genomic sequence (33,32). Sequencing of EST clones LD28033 and LD33478 support the presence of two Lola zf5 splice isoforms using alternative 5Ј-UTR sequences (5Ј-a and 5Ј-c, respectively) but we have not confirmed use of 5Ј-b and 5Ј-d UTR alternative exons for Lola zf5, as predicted in the November 30th, 2002 genome project update.
Antibody Generation and Antibody Affinity Purification-Rabbit anti-Lola common region polyclonal antibody was a generous gift of Dr. Edward Giniger and has been previously characterized (10). Hope and Odin rabbit anti-JIL-1 polyclonal antibodies were described in Jin et al. (4). Affinity purification of anti-Lola and anti-JIL-1 polyclonal antibodies was as described in Giniger et al. (10) using lacZ-Lola and GST-JIL-1 fusion proteins, respectively. To generate Lola zf5-specific monoclonal antibody (mAb) 7F1, the KDIJ1 cDNA fragment (encoding Lola zf5 amino acid residues 427-748) was cloned in the correct reading frame into pGEX4T-1 (Amersham Biosciences), verified by sequencing, and GST-zf5 fusion protein was induced in Escherichia coli according to standard protocols (Amersham Biosciences). Injection of GST-zf5 fusion protein into BALB/C mice, and generation of monoclonal hybridoma lines was performed by the Iowa State University Hybridoma Facility according to standard protocols (34).
Immunohistochemistry-Embryos were dechorionated in 50% Chlorox solution, washed with 0.7 M NaCl/0.2% Triton X-100 and fixed in a 1:1 heptane:fixative mixture for 20 min with vigorous shaking at room temperature. The fixative was either 4% paraformaldehyde in phosphate-buffered saline (PBS) or Bouin's Fluid (0.66% picric acid, 9.5% formalin, 4.7% acetic acid). Vitelline membranes were then removed by shaking embryos in heptane-methanol (35) at room temperature for 30 s. Embryos were blocked in PBS with 1% normal goat serum (Cappel) and 0.4% Triton X-100 and incubated overnight in mAb 7F1 primary antibody diluted in blocking buffer. Embryos were washed in PBS with 0.4% Triton X-100, incubated for 2.5 h with TRITC-conjugated goat anti-mouse secondary antibody (1:200) (Cappell), washed in PBS with 0.4% Triton X-100 followed by a PBS-only wash. For visualization of DNA the antibody-labeled embryos were incubated in 0.2 g/ml Hoechst 33258 (Molecular Probes) in PBS for 10 min. The final preparations were mounted in glycerol with 5% n-propyl gallate and viewed with a 40ϫ NeoFluor objective on a Zeiss Axioskop equipped with filter sets optimized and selective for rhodamine and UV detection. Digital images were obtained using a Spot-cooled charge-coupled device camera (Diagnostic Instruments).
Northern and Western Blot Analysis-Approximately 1 g of wild type Oregon-R animals from different stages was collected and ground under liquid nitrogen, and total mRNA was purified using the Poly(A) ϩ mRNA purification kit (Ambion). 5 g of poly(A) ϩ mRNA from each stage was fractionated on 1% agarose formaldehyde gels, transferred to Duralon-UV TM nylon membrane (Stratagene), and hybridized with 32 P-labeled probe overnight at 65°C according to standard high stringency protocols (31). Lola zf5 isoform-specific probes were generated by purifying 1.8 kb of unique 3Ј-end Lola zf5 cDNA sequence using a QiaQuick gel extraction kit (Qiagen) and synthesizing random primer 32 P-labeled probe using the Prime-A-Gene kit (Promega) according to the manufacturer's instructions. As a loading control, a cDNA fragment of RP49 (ribosomal protein L32) (36) was PCR-amplified from the Clontech yeast two-hybrid cDNA library described above and confirmed by sequencing. After stripping the Lola zf5 signal, labeled RP49 cDNA was used to probe the same membrane to normalize mRNA loading levels.
In Vitro Protein Interaction and Co-immunoprecipitation Assays-Approximately 5 g of GST-KDIJ1 (Lola zf5 C terminus) fusion protein or GST protein alone was coupled to glutathione-agarose beads (Sigma) and incubated with 0.5 ml of S2 cell lysates (3 ϫ 10 6 cells) overnight at 4°C. The beads were pelleted at low speed and washed three times with 1 ml of immunoprecipitation buffer (20 mM Tris-HCl, 150 mM NaCl, 10 mM EDTA, 1 mM EGTA, 0.1% Triton X-100, 0.1% Nonidet P-40, 2 mM Na 3 VO 4 , 1 mM phenylmethylsulfonyl fluoride, and 1.5 g of aprotinin, pH 8.0) for 10 min at 4°C. The proteins retained on the beads were analyzed by Western blot analysis using affinity-purified Hope anti-JIL-1 polyclonal antibody.
For co-immunoprecipitation experiments, anti-JIL-1, anti-Lola zf5, or control normal rabbit antibodies were coupled to protein G beads as follows: 30 l of Odin anti-JIL-1 serum, 30 l of normal rabbit control serum, or 1 ml of 7F1 hybridoma supernatant was coupled to 25 l of protein G-Sepharose beads (Amersham Biosciences) for 2 h at 4°C on a rotating wheel in 50 l of immunoprecipitation buffer. The appropriate antibody-coupled beads were incubated overnight at 4°C with 300 l of 0 -6 h embryonic lysate on a rotating wheel. Beads were washed four times for 10 min each with 1 ml of immunoprecipitation buffer with low speed pelleting of beads between washes. The resulting bead-bound immunocomplexes were analyzed by SDS-PAGE and Western blotting according to standard techniques as described in Jin et al. (4) using Hope antibody to detect JIL-1, mAb 7F1 to detect Lola zf5, or Lola polyclonal antisera against the Lola common core domain (10) to detect all Lola isoforms.
Bioinformatics-Lola genomic DNA sequence corresponding to nucleotides 118,125-182,622 of Drosophila melanogaster genomic scaffold AE003829.3 (GI: 21627529, updated on September 20, 2002) (32) was used to search for zinc finger motifs. Consensus zinc finger motifs include CXXC, HXXXXC, and HXXXXH (X represents any amino acid) using the "Find" function under "Edit" menu of Microsoft Word 98 software. In most cases, when an open reading frame (ORF) contained two of the three consensus sequences, it was considered a potential zinc finger motif and further analyzed. PCR primers were designed to amplify Lola sequences containing the putative cDNA fragments specific to the predicted zinc finger motifs using standard PCR protocols (31). The forward PCR primer ZnF5P (5Ј-GGATGAACTTGGACTAATGGC-3Ј) consists of Lola common core sense sequence derived from exon IV, whereas reverse PCR primers were designed to be isoform-specific, with individual reverse primers comprised of antisense sequence based on the 3Ј-end of each separate predicted zinc finger motif. Oligonucleotides were designed using the Oligo 5.0 program. cDNA templates for PCR reactions were from the 0-to 21-h cDNA Matchmaker TM yeast twohybrid library (Clontech) or a 3-to 12-h cDNA library (Stratagene). PCR conditions were optimized for each primer pair, respectively, and PCR products were directly sequenced.
The lola genomic sequence was also used to search for homologs in the Drosophila EST databases with FLYBLAST (available at www. fruit fly.org/blast/) (33) and with the Geneseqer program (available at bioinformatics.iastate.edu/cgi-bin/gs.cgi) (37). Returned ESTs were compared with lola genomic DNA sequence using BLAST2 (available at www.ncbi.nlm.nih.gov/blast/bl2seq/bl2.html) at NCBI (National Center for Biotechnology Information) (38). Putative zinc finger motifs were aligned using Clustal W available at www2.ebi.ac.uk/CLUSTALW (39).

RESULTS
The JIL-1 Kinase Interacts with a Novel Isoform of the lola Locus-To identify proteins that have direct interactions with the JIL-1 kinase, we performed a yeast two-hybrid screen of a Canton-S 21-h embryonic library using the first kinase domain of JIL-1 as bait. True positive clones detected in the primary screen were confirmed by ␤-galactosidase two-hybrid interaction assays on filter paper following retransformation of the candidate clones and JIL-1 KDI bait plasmid into the yeast strain AH109 (data not shown). One of the positive clones identified in this way was sequenced, and searches of the Drosophila genome database revealed it to be the COOH-terminal domain of a novel splice variant from the lola locus, which we have named Lola zf5. Subsequently, the full-length cDNA sequence for Lola zf5 was assembled from overlapping ESTs obtained from the Drosophila genome project and is currently available as AY058586 as well as the SW59 clone (Z97377). Fig. 1A shows the amino acid sequence of the predicted open reading frame of a protein of 748 residues with a calculated molecular mass of 79.4 kDa. The BTB domain and zinc finger domains are underlined and boxed, respectively. To further characterize the protein, a Lola zf5-specific monoclonal antibody, 7F1, was generated against a GST fusion protein containing the COOH-terminal region unique to Lola zf5. On immunoblots of embryo protein extracts (0 -6 h) mAb 7F1 detects Lola zf5 as a single band migrating at 105 kDa (Fig.  1B). The Lola zf5 protein is highly acidic with a pI of 5.54 accounting for its anomalous gel migration. Immunocytochemical labeling of early Drosophila embryos with mAb 7F1 revealed the Lola zf5 protein to be localized to nuclei in a pattern similar to that obtained by Hoechst labeling (Fig. 1, C and D).
To further explore the interaction between Lola zf5 and JIL-1 that we observed in the yeast two-hybrid assays we performed pull-down assays with the Lola zf5 COOH-terminal GST fusion protein using protein extracts from the S2 cell line. The Lola zf5-GST fusion protein or a GST-only control were coupled with glutathione-agarose beads, incubated with S2 cell lysate, washed, fractionated by SDS-PAGE, and analyzed by immunoblot analysis using JIL-specific antibody ( Fig. 2A). Whereas the GST-only control showed no pull-down activity, Lola zf5-GST was able to pull down JIL-1 as detected by the JIL-1 antibody. In addition, we performed co-immunoprecipitation experiments using embryonic lysates. For these immunoprecipitation experiments, proteins were extracted from 0 -6 h embryos, immunoprecipitated using either JIL-1-or Lola zf5-specific antibodies, fractionated on SDS-PAGE after the immunoprecipitation, immunoblotted, and probed with antibodies to Lola zf5 and JIL-1, respectively. Fig. 2B shows an immunoprecipitation experiment using Lola zf5 antibody where the immunoprecipitate is detected by JIL-1 antibody as a 160-kDa band that is also present in the embryo lysate. This band was not present in lanes where immunobeads only were used for the immunoprecipitation. Fig. 2C shows the converse experiment: JIL-1 antibody immunoprecipitated a 105-kDa band detected by Lola zf5 antibody that was also present in embryo lysate but not in control immunoprecipitations with immunobeads only. These results strongly indicate that Lola zf5 and the JIL-1 kinase are present in the same protein complex.
The Complex lola Locus Encodes Multiple BTB Domaincontaining Proteins Each with Different Zinc finger Motifs-Two alternatively spliced isoforms of lola, Lola long and Lola short, have been previously characterized (10). Lola long is a sequence-specific DNA-binding protein with C 2 HC and C 2 H 2 FIG. 1. The predicted sequence of the Lola zf5 isoform and Lola zf5 antibody labeling. A, the complete predicted amino acid sequence of Lola zf5. Lola zf5 is a 748-residue protein with a calculated molecular mass of 79.4 kDa. The BTB domain is underlined, and the zinc finger domains are boxed. B, immunoblot of SDS-PAGE fractionated 0-to 6-h embryo extracts labeled with the Lola zf5specific mAb, 7F1. The antibody recognizes a single band migrating at 105 kDa. The migration of molecular mass markers in kilodaltons is indicated to the left. C and D, double labeling of a syncytial embryo with mAb 7F1 and Hoechst. The labeling of mAb 7F1 (C) shows that Lola zf5 is localized in a pattern overlapping with that of the Hoechst labeled nuclei (D). zinc finger motifs, whereas Lola short only has a very short COOH-terminal tail segment without zinc finger motifs (10,11). Fig. 3A shows the domain structure of these two isoforms as compared with Lola zf5. Lola zf5 has an NH 2 -terminal common region shared by all cloned Lola isoforms that is followed by an isoform-specific COOH-terminal region (Fig. 3A). The Lola common region is encoded by four exons and contains a 120-amino acid NH 2 -terminal BTB domain as well as a nuclear localization signal (Fig. 3, A and B). The COOH-terminal domain of Lola zf5 contains tandem C 2 HC and C 2 H 2 zinc finger motifs (Fig. 3A). The position of the Lola zf5 zinc fingers are very close to the common region in contrast to the Lola long isoform in which the zinc fingers are positioned at the COOHterminal end.
Our identification of Lola zf5 as a novel zinc finger-containing protein within the locus prompted us to survey the region for additional exons with zinc finger motifs. Consensus zinc finger motifs include the sequences CXXC, HXXXXC, and HXXXXH. We searched the lola genomic region for long open reading frames (ORFs) containing any of these zinc finger motif consensus sequences. A total of fifteen putative exons containing zinc finger motifs were identified following this strategy as summarized in Fig. 4A. From this analysis we predict that at least 17 splice variants are generated from the lola locus. Fifteen of these each have a different zinc finger domain (zf1-zf15), whereas two use exons without zinc finger motifs. Fig. 4B shows an alignment of the 15 different zinc finger domains within the locus. Two isoforms, Lola zf4 and zf14, have only a single C 2 H 2 zinc finger domain, whereas the remaining splice variants have tandem C 2 HC/C 2 H 2 or C 2 HC/C 2 HC zinc finger domains (Fig. 4B). We propose to name the various novel zinc finger domain-containing Lola isoforms according to the order of the zinc finger domain they contain, hence the name Lola zf5. To verify that these putative isoforms, the majority of which were not predicted by the genome project, were indeed expressed, we used PCR to amplify isoform-specific cDNA fragments from the 0-to 21-h yeast two-hybrid cDNA library based on the assumption that all of the Lola isoforms share the same common regions as the known isoforms. In this way the expression of thirteen out of the fifteen predicted zinc finger-containing exons was confirmed by direct sequencing of such PCR products (Fig. 4A, indicated in black). In addition, we searched the Drosophila EST database with the genomic DNA sequence of each isoform and found further EST support for expression of six of the 13 exons containing zinc finger motifs. We also identified an EST clone (GM27815) representing the second Lola isoform without a zinc finger domain and have named it Lola short-like (Fig. 4A).
The coding regions of most of the Lola isoforms are generated by splicing the four common exons together with a single exon containing the different zinc finger domains. However, by sequencing the ESTs and the PCR amplification products we identified a number of smaller exons without zinc finger motifs that are utilized by Lola zf1 and zf8 (Fig. 4A). In addition, the locus has four different 5Ј-UTRs (5Јa through 5Јd) that may further amplify the diversity of transcripts generated from this locus (Fig. 3B). By sequencing ESTs of the various Lola splice variants, we have obtained evidence that each of the four 5Ј-UTRs has been utilized into at least one transcript (Fig. 3B). Interestingly, we identified two ESTs for Lola zf5 that used different 5Ј-UTRs (5Јa and 5Јc, respectively). Transcripts with different 5Ј-UTRs may allow for the fine regulation of their spatial and temporal expression patterns. Although Lola zf5 is the only example where this alternative utilization has been confirmed, it is likely that other isoforms are regulated by the same splicing mechanisms.
Developmental Expression of Lola zf5-To determine the expression pattern of Lola zf5, we carried out Northern blot analysis using mRNA samples from representative developmental stages (Fig. 5). As a probe we used a Lola zf5 isoformspecific cDNA fragment from the COOH-terminal coding region and 3Ј-UTR. Lola zf5 mRNA migrates as a single band with an approximate molecular size of 3.9 kb on 1% agarosedenaturing gels. Potential differences in size between transcripts using alternative 5Ј-UTRs would not be resolved on these gels. Lola zf5 mRNA was abundant in 0-to 2-h embryos, and this level of transcript was maintained in embryos 2-6 h after egg laying (Fig. 5). However, the Lola zf5 mRNA level began to decrease in 6-to 12-h embryos and could not be detected in postembryonic stages except at a low level in female adults, which may reflect the maternal deposition of mRNA into eggs. These findings correlated well with the results from developmental immunoblots using the Lola zf5-specific mAb 7F1 (Fig. 6A). We detected high levels of Lola zf5 protein in early embryos (0 -12 h); however, the protein level decreased  (ZF1-ZF15). The coding sequence for 17 potential ORFs of Lola isoforms generated from the locus are diagrammed below. The existence of the isoforms depicted in black was supported by Drosophila ESTs from the genome project or by PCR-amplified isoform-specific cDNA fragments. Lola long and Lola short have both been previously characterized (10). In addition, a homolog of Lola zf8 has been identified in D. hydei (11). B, alignment of the 15 different zinc finger domains from the zinc finger containing exons. The conserved coordinating cysteines and histidines are in white typeface outlined in black.
ϳ12 h after egg laying and could not be detected at postembryonic stages, including adult females (Fig. 6A). The lack of Lola zf5 protein in female ovaries suggests that Lola zf5 is not translated from maternally stored transcripts until after fertilization. These results indicate that the functional expression of Lola zf5 is restricted to early embryogenesis.
In previous studies a polyclonal Lola antibody was made to the Lola common region that would be expected to recognize all the Lola isoforms (10). In developmental immunoblots using this Lola polyclonal antibody we detected multiple bands throughout all developmental stages (Fig. 6B). In early embryos (0 -6 h) three major bands, including one migrating at 105 kDa, can be detected on the immunoblots. In 12-to 22-h embryos at least 15 bands can be recognized by the polyclonal Lola antibody suggesting that most of the Lola isoforms are expressed at this stage. Some isoforms also appear to be present in later developmental stages such as third instar larvae as well as adults (Fig. 6B). To test whether the 105-kDa protein detected by the Lola antiserum corresponded to Lola zf5 as we would predict, we performed immunoprecipitation experiments with the mAb 7F1 of protein extracts from 0-to 6-h embryos. The immunoprecipitations were fractionated by SDS-PAGE, immunoblotted, and detected with mAb 7F1 and Lola antiserum, respectively (Fig. 6C). As shown in Fig. 6C Lola antiserum recognizes three major bands, including one of 105 kDa that is also labeled by mAb 7F1 and thus is likely to represent the Lola zf5 protein. Interestingly, the presence of the two additional bands labeled by the Lola antiserum and not present in the control lane suggests that Lola zf5 may be involved in heterodimer formation with other Lola isoforms (Fig. 6C, lane 2).
The Lethal lola 00642 Mutation Is a Dominant Modifier of the Hypomorphic JIL-1 EP(3)3657 Allele-To further study whether JIL-1 and Lola zf5 interact in vivo we explored genetic interactions between mutant alleles of lola and JIL-1 by generating double-mutant individuals containing both lola 00642 and JIL-1 EP(3)3657 . The lola 00642 allele contains a recessive lethal P element insertion and fails to complement the lethality of many P element insertion alleles of lola (40). By PCR amplification of the flanking region of lola 00642 followed by direct sequencing, we found that the insertion site is 438 bp downstream to 5Ј-UTRc and 54 bp upstream to 5Ј-UTRd (Fig. 3B). Because Lola zf5 mRNA contains either the 5Ј-UTRa or the 5Ј-UTRc at the 5Ј-end of the transcript, the insertion of a P element within the Lola zf5 transcription unit is likely to disrupt expression of both splicing alternatives of Lola zf5. The JIL-1 EP(3)3657 allele is a hypomorphic allele that can be maintained in a homozygous stock for only a few generations due to the low hatch rate and recessive semi-lethality (6). The hatch rate of JIL-1 EP(3)3657 homozygous embryos produced by homozygous parents is as low as 4 -7% when compared with the hatch rate of wild type Oregon-R embryos (Ref. 6 and this study). We generated double mutants by crossing a chromosome containing the lola 00642 allele into a homozygous JIL-1 hypomorphic EP(3)3657 background. Interestingly, individuals homozygous for JIL-1 EP(3)3657 that also contain a lola 00642 allele can be maintained indefinitely as a stock. This suggests that heterozygous lola 00642 may function as a dominant suppressor of the JIL-1 EP(3)3657 phenotype. To quantify the extent of rescue, we compared the numbers of JIL-1 EP(3)3657 homozygous progeny with or without lola 00642 from a single cross (Fig. 7A). Although equal numbers of curly-and straight-winged phenotypic classes are expected in matings of lola 00642 /CyO; JIL-1 EP(3)3657 / JIL-1 EP(3)3657 males with ϩ/ϩ; JIL-1 EP(3)3657 /JIL-1 EP(3)3657 females, 2.3 times more flies with straight wings were observed than curly wings (Fig. 7A). In this cross straight-winged flies carry a lola 00642 allele, whereas curly-winged flies do not. The difference in numbers observed for the two classes was statistically significant (p Ͻ 0.005, 2 test). To exclude the possibility that the apparent rescue phenotype is due to an enhancer of JIL-1 EP(3)3657 phenotype on the CyO second chromosome balancer, we also set up a control cross in which the lola 00642 chromosome is not present. In the control cross, we did not observe any statistically significant evidence (p Ͼ 0.1, 2 -test) that the CyO balancer chromosome affects the viability of JIL-1 EP(3)3657 homozygotes (Fig. 7A). The control crosses were per-  (6). However, we did not observe any eclosion of z2 homozygotes (data not shown). These results suggest that either the interaction between lola 00642 and JIL-1 is mild or the genetic interaction depends on the presence of a minimal level of JIL-1 protein.
If the observed genetic rescue is a consequence of normalized relative levels of Lola zf5 and JIL-1, we would expect that some or all of this rescue occurs during embryogenesis, because Lola zf5 is expressed only in early embryos (Fig. 6A). We, therefore, quantified embryonic rescue by determining hatch rates for JIL-1 EP(3)3657 /JIL-1 EP(3)3657 embryos that were either heterozygous for the lola 00642 allele or homozygous for the wild type allele. Homozygous JIL-1 EP(3)3657 /JIL-1 EP(3)3657 embryos produced by homozygous mothers typically hatch at a low (7%) rate (Fig. 7B). Adjusting for the fact that none of the embryos with a lola 00642 /lola 00642 or CyO/CyO genotype hatch, the hatch rate of lola 00642 /CyO; JIL-1 EP(3)3657 /JIL-1 EP(3)3657 embryos is 20% (Fig. 7B). Thus, homozygous JIL-1 EP(3)3657 embryos that were heterozygous for lola 00642 hatched at a statistically significant 2.8-fold greater rate than embryos that did not carry the lola mutation (p Ͻ 0.005, 2 -test). Therefore, the increase of viability observed for lola and JIL-1 double mutants is at least partially due to an increase in the frequency with which such individuals survive embryonic development and hatch.

DISCUSSION
In this study we provide evidence that the JIL-1 tandem kinase molecularly interacts with a novel isoform of the lola locus, Lola zf5. This interaction was first detected in a yeast two-hybrid screen and subsequently confirmed by pull-down and cross immunoprecipitation assays. Furthermore, immunocytochemical labeling of Drosophila embryos shows that Lola zf5 is localized to nuclei. This localization is compatible with a direct interaction with JIL-1, because JIL-1 has been shown to be a nuclear kinase expressed throughout embryogenesis (4). Northern and Western blot analyses show that the expression of Lola zf5 is developmentally regulated and is only expressed during early embryogenesis.
An interesting feature of the lola locus is its complex splicing pattern, and we demonstrate that it contains at least 27 exons. Four of these code for a BTB domain and sequences with a nuclear localization signal common to all Lola splice forms. In addition, fifteen of the exons code for sequences with different zinc finger domains. From this analysis, we predict that a minimum of 17 different protein products are generated by the lola locus, fifteen of which contain zinc finger domains and two that do not. By PCR amplification of cDNAs from 0-to 21-h embryos and sequencing of ESTs, we have obtained confirming evidence that at least 15 different Lola polypeptides are likely to be encoded. We were not able to verify the existence of the two remaining isoforms; however, this could be due to their being only expressed at developmental stages or in tissues that we did not examine. We further provide evidence that the number of transcripts from the locus is enhanced by alternative splicing of four different 5Ј-UTRs. The utilization of different 5Ј-UTRs and exon shuffling may provide a way to finely regulate stage-and tissue-specific expression of multiple gene products from the locus that serve different functional roles. This kind of complex gene organization has previously been observed at other loci. The most extreme example may be the Dscam locus that codes for cell adhesion receptors in the Drosophila nervous system and that potentially can generate more than 38,000 Dscam receptor isoforms (41). Another example related to control of gene expression is the locus of the trithorax group protein mod(mdg4). This locus encodes at least 21 different BTB domain-containing protein isoforms (42). At least one of these isoforms has been shown to associate with Su(Hw) to exert gypsy insulator function preventing enhancerpromoter communication (43). Thus, differential splicing may be a general mechanism for BTB domain proteins to generate functional diversity.
The BTB domain has been shown to promote dimerization and the residues necessary for this function have been identified for the Bab protein (16). Comparison between the Lola and Bab BTB domains show that all these residues are conserved in the Lola BTB domain indicating that it has the capacity for homodimer formation. Furthermore, our immunoprecipitation experiments strongly suggest that Lola zf5 forms heterodimers with other Lola isoforms. Formation of homo-or heterodimers between Lola isoforms, including Lola zf5 is likely to lead to different developmental consequences by modifying DNA binding specificities and/or affinities of the Lola zinc finger isoforms. The majority of BTB-containing zinc finger proteins described thus far have been observed to function as transcriptional repressors, and several of these have been shown to directly bind co-repressor components of histone deacetylase complexes to their dimerized BTB domains (reviewed in Ref. 44). Consistent with such a repressive activity it has been suggested that Lola long may reduce expression of copia transposable elements in the Drosophila nervous system (11). However, Cavarec et al. (11) also observed Lola-mediated positive regulation of copia transcription in the gonads indicating that Lola isoforms also can act as transcriptional activators. In support of this notion, they identified a D. hydei isoform of Lola that binds directly to the copia enhancer and positively regulates transcription in transfected S2 cells. However, this effect was abrogated in a dose-dependent manner when the D. melanogaster Lola long coding sequence was co-transfected. Our analysis shows that the zinc finger domains of the D. hydei Lola isoform are nearly identical to those of D. melanogaster Lola zf8, with only a single conservative substitution of serine to threonine in the second zinc finger domain. Thus, the expression of alternative Lola isoforms may determine whether Lola acts as an activator or a repressor within different tissues and at specific stages during development.
C 2 H 2 zinc fingers are one of the most common DNA binding motifs found in eukaryotic transcription factors and are characterized by a small number of conserved residues that generate the folded ␤␤␣ domain structure necessary to align the residues that coordinate binding activity (reviewed in Ref. 45). The identification of alternatively spliced isoforms of the lola locus containing different zinc finger arrangements in conjunction with the ability of Lola isoforms to generate heterodimers suggests that Lola complexes functioning as transcription factors may recognize a wide range of potential DNA regulatory sequences. However, it should be noted that many of the Lola isoforms have a very unusual arrangement of zinc finger domains that suggest they have the potential to be involved in protein-protein interactions as well. Although most zinc fingers are of the C 2 H 2 class, in each of the Lola isoforms where the zinc finger is present as a tandem array, the first finger is of the unusual C 2 HC class. This atypical zinc finger domain is also found in MYST family histone acetyltransferases such as MOF, where it has been shown to be essential for nucleosome interactions (46). Additionally, this domain has been implicated in binding non-histone proteins (47), RNA (48 -50), and DNA (51,52).
Our genetic experiments suggest that a reduction of protein levels resulting from mutation in the lola locus can act as a dominant modifier of a hypomorphic JIL-1 allele leading to an increase in embryonic viability. However, in these experiments the P element insertion into the lola locus is likely to perturb transcription of many of the Lola isoforms. Thus, we do not know which of the Lola isoforms are responsible for the genetic interaction or whether JIL-1 acts upstream or downstream. Thus, several scenarios for a functional interaction between Lola proteins and JIL-1 that enhances viability can be envisioned. On one hand, JIL-1 may act as a derepressor to counteract a potential repressive function of Lola zf5 or other Lola isoforms on gene expression. Derepression of the gene products from these loci may lead to enhanced viability. On the other hand, Lola zf5 or other Lola isoforms may normally act to down-regulate JIL-1 kinase activity by physically interacting with JIL-1. Consequently, the decrease in JIL-1 kinase activity observed in a hypomorphic mutant background may be alleviated by the reduction of its negative regulator in the lola mutant. In a third scenario, the interaction of JIL-1 with Lola zf5 may indeed enhance transcription at some genes, but the loss in the lola mutant of other Lola isoforms that normally function to down-regulate gene expression in other contexts may counterbalance the reduced JIL-1 activity in the JIL-1 hypomorph. Thus, the interaction between JIL-1 and the lola locus may be highly complex and promises to provide new experimental avenues into exploring the mechanisms of modification of chromatin and/or the regulation of gene expression during early embryogenesis.