Barrier-to-autointegration factor plays crucial roles in cell cycle progression and nuclear organization in Drosophila

The nuclear envelope (NE) is continuous with the endoplasmic reticulum (ER) and forms the boundary between the nucleus and cytoplasm. It also participates in the arrangement of chromosomes in that it provides a stable anchor for chromatin in the interphase nucleus. It consists of three major components, the nuclear lamina, the double nuclear membrane, and the nuclear pore complexes (reviewed by Stuurman et al., 1998).

The nuclear lamina is a stable filamentous meshwork lining the inner nuclear membrane (INM), consisting mainly of lamin polymers; lamins belong to the intermediate filament (IF) protein super-family. In vertebrates, there are two classes of lamins, A- and B-types, distinguished on the bases of their amino acid sequence (reviewed by Stuurman et al., 1998). Both types of nuclear lamins are known to bind directly to chromatin components (reviewed by Stuurman et al., 1998; Gruenbaum et al., 2000). Lamins are expressed in a wide range of cells. Moreover, it has been suggested that they have a general role in the attachment of chromosomes to the NE as well as in the maintenance of interphase nuclear organization. In addition to the lamins, a number of less abundant INM proteins are also associated with the nuclear lamina (reviewed by Gruenbaum et al., 2000; Worman and Courvalin, 2000). These include the lamin B receptor (LBR), lamina-associated polypeptides (LAPs) 1 and 2, emerin and MAN1, all found in a wide range of animal cells, as well as the young arrest (YA) protein and otefin in Drosophila. Research has revealed that different INM proteins are not only bound to specific nuclear lamins but also interact directly with specific chromosomal components (reviewed by Gruenbaum et al., 2000; Worman and Courvalin, 2000; Cohen et al., 2001).

LAP2β, a LAP2 isoform generated by alternative splicing, binds to B-type lamin (Foisner and Gerace, 1993), chromosomes (Foisner and Gerace, 1993; Furukawa et al., 1998), DNA (Furukawa et al., 1997; Cai et al., 2001) and HA95 (Martins et al., 2000). LAP2β also binds specifically to the DNA-bridging protein, BAF (Furukawa, 1999; Shumaker et al., 2001). Moreover, microinjection into cultured cells of protein encompassing the binding domains of LAP2β to either B-type lamin or chromatin, or addition to an in vitro nuclear assembly system, clearly revealed influences on DNA replication as well as nuclear assembly (Yang et al., 1997; Gant et al., 1999). LAP2β accumulates at the surfaces of chromosomes prior to the assembly of B-type lamin at the NE during late anaphase (Foisner and Gerace, 1993; Moir et al., 2000). Thus, in addition to the nuclear lamins, other INM proteins may have functions in both structural arrangement and individual dynamic processes within the cell nucleus.

Recently a particular protein family was described, termed LEM-domain proteins. The LEM domain, originally recognized in LAP2, emerin and MAN1, is a conserved series of about 40 amino acid residues (Lin et al., 2000), also found in otefin and other Drosophila proteins (reviewed by Gruenbaum et al., 2000). Structural characterization of the LEM domain of LAP2β demonstrated one short N-terminal α-helix and two larger parallel α-helices (Lin et al., 2000; Cai et al., 2001; Laguri et al., 2001). Recently, Lee et al. (Lee et al., 2001) demonstrated that the LEM domain of emerin mediates direct binding to BAF, supporting previous work on LAP2β (Furukawa, 1999; Shumaker et al., 2001). Thus there is now abundant evidence to support the notion that BAF binding is a common property of several proteins containing LEM domains.

BAF was first identified as a cellular trans-acting factor involved in protecting reverse-transcribed retroviral DNA against self-destructive integration (Lee and Craigie, 1998). BAF dimers bind to double-stranded DNA (dsDNA), and high concentrations form intermolecular bridges with naked DNA to generate perceptible supramacromolecular complexes (Lee and Craigie, 1998; Cai et al., 1998). A BAF dodecamer was demonstrated to form a discrete higher order nucleoprotein complex in vitro with 21 bp dsDNA (Zheng et al., 2000). The cellular functions of BAF in vivo remain unclear, but the BAF protein clearly co-localizes with cell nuclei during interphase (Furukawa, 1999; Haraguchi et al., 2001; Wang et al., 2002). A similar distribution is also found in normal rat kidney cells expressing GFP-tagged BAF (J. Ellenberg, personal communication). In a particularly high resolution study performed using cultured cells, BAF was found to be enriched during telophase in the central region of the assembling nuclear rim (Haraguchi et al., 2001). BAF depletion by RNA interference (RNAi) in Caenorhabditis elegans causes a defect in chromatin segregation during mitosis (Zheng et al., 2000). Furthermore, BAF has now been demonstrated to be ubiquitously expressed in many mammalian tissues and functions as a transcriptional repressor for the paired-like homeodomain protein Cone-Rod Homeobox (Wang et al., 2002). These findings imply that the interactions between LEM domain proteins and BAF/DNA complexes might regulate chromosomal structure and thereby, function.

To elucidate BAF functions in vivo, genetic manipulation of Drosophila was performed. Characterization of the developmental and cellular phenotypes in baf null mutants provided direct in vivo evidence of its importance in organizing the architecture of both chromosomes and the NE as well as for the progression of the cell cycle.

l(2)05836, l(2)s1883 (both connector of kinase to AP-1 mutants; Cka mutants) and l(2)k10210 mutant flies (Spradling et al., 1999) were obtained from the Bloomington Drosophila Stock Center (Bloomington, IN). The mutant second chromosomes of l(2)k10210/l(2)k10210, baf¹/baf¹ and baf²/baf², balanced over CyO or CyO-P[actin::GFP] were used to help identify mutants. Flies and larvae were grown on standard corn-meal medium at 23°C. baf¹/+, l(2)k10210/l(2)k10210, and w¹/w¹ animals pupated at about 5-6 days after hatching from eggs, while pupation of baf²/baf² and baf¹/baf¹ were each delayed for about 1 day. For BrdU labeling, continuous labeling was performed by feeding a 1 mg/ml BrdU solution for 15-20 hours just before dissection of late third instar larvae (Truman and Bate, 1988). To perform shorter (pulse) labeling, dissected larval brains were directly incubated in PBS containing 20% fetal calf serum (FCS) and 0.1 mg/ml BrdU at 25°C for 1.5 hours.

Polyclonal rabbit anti-BAF antibodies were produced against an overexpressed highly purified thioredoxin-his-tagged (TrH)-Drosophila BAF fusion protein. To generate this BAF fusion protein, polymerase chain reaction was performed with the primers ccg tgc gga tcc ATG TCG GGC ACA TCG CAG AAA and cca gcg ctc gag AAC AGT GAA CAC GGC AAA TGC. After BamHI and XhoI digestion, the BAF fragment was subcloned into the BamHI and XhoI sites of the pET32 vector (Novagen, Madison, WI, USA). The TrH-Drosophila BAF fusion protein was then purified after bacterial expression and lysis using steps including chromatography on Ni⁺⁺-IDA agarose.

For immunostaining, larval tissues were fixed for 25 minutes in 3.7% formaldehyde, permeabilized for 10 minutes in 0.2% Triton X-100, blocked for 1 hour in PBS with 3% goat serum, and reacted for 24 hours at 4°C with antibodies to Drosophila lamin Dm₀ and derivatives [either mouse monoclonal antibody (m-mAb) ADL 84 (Stuurman et al., 1995) or highly specific rabbit antiserum] and/or to Drosophila lamin C [m-mAb LC28 (Riemer et al., 1995)], BrdU [m-mAb G3G4 from the Developmental Studies Hybridoma Bank (DSHB) at the University of Iowa], NPC proteins containing FG-repeats (m-mAb414 from BabCO, USA), cyclin E (rat serum provided by Dr H. Richardson, Peter MacCallum Cancer Institute, Melbourne, Australia), cyclins A and B (for A, m-mAb A12 and for B, m-mAb F2F4; both from the DSHB), histone H2A (rabbit antiserum provided by Dr R. Glaser, Wadsworth Center, Albany, NY, USA) and histone H3 phosphopeptides [for phospho-Ser²⁸, rat mAb HTA28; and for phospho-Ser¹⁰, rabbit antiserum PH10 (Goto et al., 1999)]. Subsequently, larval tissues were incubated with dichlorotriazinyl aminofluorescein (DTAF) or Cy3-conjugated secondary antibodies and mounted in 90% glycerol. DNA was visualized with propidium iodide (PI). Before beginning the immunostaining with the anti-BrdU antibody, larval tissues were treated with 2 N HCl for 30 minutes, and then neutralized in 0.1 M borate for 5 minutes. Immunofluorescence images were captured by confocal microscopy using a Radiance 2000 instrument (Bio-Rad Laboratories). To establish specificity of the anti-BAF antiserum, aliquots were incubated for 12 hours before immunostaining with either thioredoxin-his-tag alone (Fig. 1A TrH tag) or thioredoxin-his-tagged (TrH)-Drosophila BAF fusion protein (Fig. 1A TrH-BAF); they were then used directly for immunostaining (see Fig. 1A). Anti-BAF antisera that were pre-incubated with TrH tag protein were routinely used in immunofluorescence experiments. Antiserum specificity was also established using both standard western blotting and immunostaining to compare immune with pre-immune sera, both from the same animal (data not shown). By western blot analyses, anti-BAF antisera were strongly and specifically immunoreactive with recombinant BAF. When cell extracts were analyzed, a relatively faint 10 kDa band was detected in both egg and larval as well as Schneider tissue culture cell extracts using anti-BAF antibodies. However, a band of ∼42 kDa was apparently detected as well. Both the 10 kDa and the ∼42 kDa species, although weak in intensity, were seen specifically, i.e. they were not seen with pre-immune serum. Similar difficulties with BAF staining on blots have previously been reported by others (e.g. Segura-Totten et al., 2002). By immunofluorescence, pre-immune serum showed only weak non-specific background staining.

To examine the interior structures of brain hemispheres, larval tissues were fixed in Bodian's fixative, dehydrated in ethanol, cleared in xylene, mounted in Canada balsam and directly observed by phase contrast microscopy. For thin section EM, larval tissues were directly fixed with 2% glutaraldehyde in 0.1 M cacodylate buffer on ice. Embedding, sectioning and microphotography were carried out at the Hanaichi Electron Microscopic Laboratory, Inc. (Okazaki, Aichi, Japan).

For genomic analyses, 30 homozygous baf null mutant third instar larvae were directly lysed and homogenized at room temperature in 0.5 ml of buffer H (10 mM Tris, pH 7.5; 60 mM NaCl; 10 mM EDTA; 1% SDS). After addition of proteinase K to a final concentration of 100 μg/ml, homogenates were incubated at 50°C for 1 hour, and then genomic DNA was purified by phenol/chloroform extraction and ethanol precipitation. The baf genomic region of mutants was amplified by polymerase chain reaction (PCR) using the primers GGACGTTATTCTGCGACTGG and ATTGGTTCACGTCGCTCTCA, and PCR products were mapped with several restriction enzymes. The nucleotide sequences of these PCR fragments were subsequently determined at both ends by the dideoxy direct sequencing method (Sanger et al., 1977) (also see instructions of ABI cycle sequencing ready reaction kits) with primers TTCAGCGATGGCTATGT and GCTTACACATAATCCCC for baf¹/baf¹, and GAACTGGATCACCCATT and GCTTACACATAATCCCC for baf²/baf², using an Applied Biosystems 320 Genetic Analyzer (Foster City, CA, USA). For rescue experiments, baf genomic fragments (gBAF) were generated by PCR from the BAC clone, BACR08I01 (Drosophila Genome Center, Berkeley, CA) using the primers GGACGTTATTCTGCGACTGG and ATTGGTTCACGTCGCTCTCA, digested with BamHI and XhoI (Fig. 2A), and subcloned into the BamHI and XhoI sites of the pP{lacW} vector. The resultant plasmid was introduced by P-element-mediated transformation, and independent lines were established, each carrying genomic baf (gBAF) on chromosomes 1 (C1) or 3 (C3-2b or C3-4). Rescue of the baf null mutant phenotype with gBAF was analyzed in the larval CNS from animals of four different genotypes: (1) baf¹/baf¹; gBAF(C1)/gBAF(C1); (2) baf¹/baf¹; gBAF(C3-2b)/gBAF(C3-2b); (3) baf¹/baf¹; gBAF(C3-4/gBAF(C3-4); and (4) baf²/baf²; gBAF(C3-4)/gBAF(C3-4).

Drosophila BAF is a nuclear protein and colocalizes with chromosomes during both interphase and mitosis

To study the subcellular distribution of Drosophila BAF, confocal immunostaining analyses of the CNS, imaginal disc and salivary gland tissues were performed. Specific anti-BAF antiserum was used together with either antibodies highly specific for Drosophila B type lamins [lamin Dm₀ derivatives usually referred to lamin Dm₀ hereafter (see Smith et al., 1987; Smith and Fisher, 1989)] or antibodies specifically raised against histone H3 phosphopeptides (PH10 or HTA28 recognizing the Ser¹⁰ or Ser²⁸ phospho-epitopes, respectively) (Goto et al., 1999). Histone H3 phosphorylation at both Ser¹⁰ and Ser²⁸ is known to occur only coincident with chromosome condensation during mitosis, and throughout the transition from prophase to telophase (Goto et al., 1999; Giet and Glover, 2001).

Confocal microscopy indicated that a significant amount of BAF was nuclear in both CNS and imaginal disc tissues as well as in salivary glands (Fig. 1). Within smaller nuclei, punctate anti-BAF staining was often found (Fig. 1C). Plasma membrane staining was also seen in salivary glands, but this staining was variable and of unclear specificity; it was never detected in other endocyclic larval tissues, the imaginal discs or the CNS (Fig. 1C,D). Incubation of anti-BAF antiserum with the thioredoxin-his-tagged (TrH)-Drosophila BAF fusion protein substantially reduced and/or eliminated nuclear staining; in contrast staining with anti-lamin Dm₀ was relatively unaffected (Fig. 1A). Cytoplasmic (non-nuclear) BAF staining was also somewhat reduced by pre-incubation of anti-BAF antiserum with the TrH-Drosophila BAF fusion protein (see Fig. 1A, salivary glands). Pre-immune serum showed only non-specific background fluorescence (not shown) whereas substantial specific staining was seen with our anti-BAF antiserum.

In interphase salivary gland cell nuclei, BAF exhibited a heterogeneous though clearly chromosomal localization pattern (Fig. 1E). Significant `banding' was seen (Fig. 1E). When CNS and imaginal disc tissues, both containing mitotic cells, were stained with both anti-BAF and HTA28 antibodies, significant BAF staining was observed coincident with HTA28 staining of condensed mitotic chromosomes; cytoplasmic staining was also seen (Fig. 1B,D). BAF cytoplasmic staining increases substantially during M-phase (relative to that of interphase), suggesting that BAF is also solubilized from and released from interphase DNA. BAF solubilization was reported by others (Haraguchi et al., 2001). Only BAF staining disappeared after pre-incubation of both BAF and HTA28 antibodies with TrH-Drosophila BAF (data not shown). Thus, although BAF behavior during M phase is complex, we suggest that a portion of BAF binds to chromosomes throughout the cell cycle.

l(2)k10210 is a recessive lethal strain from the Berkeley Drosophila Genome Gene Disruption Project (Spradling et al., 1999) with an artificial P element (pP{lacW}) inserted approximately 350 bp downstream of the baf termination codon. The P element was confirmed to be single by Southern hybridization (data not shown). To produce baf null mutants, the integrated pP{lacW} (Fig. 2A) was excised by mating with a Drosophila line carrying a Δ2-3 transposase. A total of 49 lines were established based on reversion to white eye color of the red-eye color characteristic of pP{lacW}-bearing flies; three of these reverted to complete viability while the others (46) remained homozygous lethal. The complete reversion in the three lines indicates that the lethal phenotype of l(2)k10210 depends only on the pP{lacW} insertion. To identify baf null mutants, restriction mapping of baf genomic regions from the 46 lethal lines was performed (data not shown). For 2 of these lines, baf¹/baf¹ and baf²/baf², we determined the genomic sequence flanking the P{lacW} insertion site. In each case, a small deletion (∼1 kb and ∼3 kb, respectively) that removed the entire baf gene was found, together with remnants of the P{lacW} inserts (Fig. 2A; remnants indicated by black bars with lengths shown in parentheses). The ∼1 kb deletion of the baf¹ allele was found to remove only BAF, but the ∼3 kb deletion of the baf² allele included the ORF of the Drosophila homolog of mouse pancreatic triacylglycerol lipase (38% identity) in addition to the baf ORF. Furthermore, the major start site for mRNA transcription of a connector of kinase to AP-1 (Cka) protein was found near the P{lacW} integration site in l(2)k10210 animals (Fig. 2A). To establish the relationship to this Cka protein, we determined that l(2)k10210, baf¹ and baf² could all complement available mutants (l(2)05836 and l(2)s1883) of the cka gene (data not shown). We therefore conclude that the l(2)k10210, baf¹ and baf² mutations are all independent of the cka gene.

l(2)k10210/l(2)k10210, baf¹/baf¹ and baf²/baf² animals all exhibited lethality relatively late in development, but at stages which were different for l(2)k10210/l(2)k10210 than for baf¹/baf¹ and baf²/baf². While l(2)k10210/l(2)k10210 animals survived up to the late pupal stage just before eclosion, both baf¹/baf¹ and baf²/baf² were not able to develop beyond the larval-pupal transition, adult structures never being formed (data not shown). It is known that larval growth is driven almost exclusively by increases in cell size with endocyclic DNA replication (replication without mitosis or cell division), and mitosis during larval development is restricted primarily to the imaginal discs and central nervous system (CNS) (Glover, 1989; Gatti and Baker, 1989). In the baf null mutants, endocyclic larval tissues grew normally (data not shown), but the precursors for adult structures were considerably smaller, including the imaginal discs and the CNS of third instar larvae just before pupation (Fig. 2B, for baf¹/baf¹ and baf²/baf², compared with the baf¹/+ heterozygote and l(2)k10210/l(2)k10210). Imaginal discs were completely absent, and cell proliferation and differentiation were not detectable in the thoracic ganglia and the brain hemispheres of the baf¹/baf¹ and baf²/baf² CNS. Typically the brain hemispheres of baf¹/+ control animals form a complex optic lobe anlagen (Fig. 2C). These structures were also observed in the brain hemispheres of l(2)k10210/l(2)k10210 and white¹/white¹ (w¹/w¹) animals (the latter as a wild-type control; data not shown). However, the small hemispheres of baf¹/baf¹ and baf²/baf² larvae did not show any significant complexity (Fig. 2C). Even just before pupation, the CNS structure of baf null mutants resembled that of late first or early second instar larval brain hemispheres isolated from wild-type animals (Hofbauer and Campos-Ortega, 1990). Based on these observations, we would suggest that in Drosophila baf null mutants, maternal BAF becomes insufficient during early larval development (see also Fig. 7). Preliminary results with both Acridine Orange and TUNEL staining of the CNS of baf null mutant animals suggested that there was no increase in apoptosis (data not shown).

To confirm whether the deletion of baf actually causes these phenotypes in baf¹/baf¹ and baf²/baf² larvae, the wild-type baf genomic sequence (Fig. 2A, B*-Xh*DNA fragment; gBAF) was introduced by P-element-mediated germline transformation. In all baf¹/baf¹ and baf²/baf² animals carrying gBAF, development was extended to the late pupa (data not shown), and both the imaginal discs and CNS developed normally (Table 1, gBAF; see also Fig. 6A). After transformation, the stage of lethality as well as the appearance of both the CNS and imaginal discs resembled those seen for the l(2)k10210/l(2)k10210 parent line, indicating that differences of both baf¹/baf¹ and baf²/baf² in comparison to l(2)k10210/l(2)k10210 reflect baf function.

To assess cell cycle progression in baf null mutants, CNS tissues from late third instar larvae were examined. Initially, condensed chromosomes of l(2)k10210/l(2)k10210, baf¹/+ and baf¹/baf¹ larvae were visualized with the PH10 and HTA28 antibodies specifically raised against histone H3 phosphopeptides containing the Ser¹⁰ and Ser²⁸ phospho-epitopes, respectively (Goto et al., 1999). PH10-positive staining was detected throughout the CNS of baf¹/+ control third instar larvae just before pupation (Fig. 3A). Prominent labeling was located within distinct proliferating zones in the ventral regions of thoracic ganglia as well as in the brain hemispheres. PH10 immunostaining of whole mitotic chromosomes was apparent from prophase to telophase, but not in interphase (Fig. 3C). Similar results were found with l(2)k10210/l(2)k10210 (Fig. 3A) as well as with baf¹/baf¹ rescued by gBAF (data not shown). In contrast, in the CNS of baf¹/baf¹, most cells were not recognized by the PH10 antibodies, very little signal being seen overall (Fig. 3B; several specimens are shown). Similar results were obtained when mitoses were observed in DAPI-stained brain squashes of baf¹/baf¹ larvae in which an M-phase block had been introduced by colchicine (data not shown), suggesting that most cells in baf¹/baf¹ larvae never entered M phase. Therefore, the reduced brain size is most likely the result of a stop in cell cycle progression rather than delayed proliferation, and the arrest most likely occurs primarily outside of M phase as most baf¹/baf¹ neuroblasts apparently do not enter prophase. Those few positive mitotic cells that were seen in the baf¹/baf¹ CNS stained with PH10 antibodies displayed grossly abnormal chromosomal morphology (Fig. 3D). These aberrations were also observed in the CNS of early- and mid-third instar baf¹/baf¹ larvae (data not shown), and suggest that M phase does not progress normally in baf null mutants.

In mitotically cycling cells, completion of S phase is essential for nuclear division. As M-phase progression was shown to be almost totally absent in animals lacking the baf gene, BrdU incorporation was examined to evaluate DNA synthesis. Upon continuous labeling for 15-20 hours just before dissection of late third instar larvae, neuroblasts in the CNS of baf¹/+ and l(2)k10210/l(2)k10210 exhibited substantial incorporation of BrdU (Fig. 4A). This was similar to BrdU incorporation seen in w¹/w¹ control larvae. BrdU was distributed in the ventral region of thoracic ganglia as well as in particular regions of the brain hemispheres. These patterns of incorporation were coincident with the cell proliferation patterns reported for the CNS (Truman and Bate, 1988; Hofbauer and Campos-Ortega, 1990). In the case of baf¹/baf¹ animals, BrdU incorporation during late larval development was reduced compared to that in the control baf¹/+ CNS; labeled cells were scattered throughout the CNS, with some concentrations in the brain hemispheres (Fig. 4A,B). To test further the rate of BrdU incorporation in late third instar larvae, dissected baf¹/baf¹ CNS samples were directly incubated (for 1.5 hours) in medium containing BrdU. Incorporation of BrdU was dramatically reduced, but was still detected in individual baf¹/baf¹ nuclei (Fig. 4C).

To monitor cell cycle progression further, CNS cells were studied for expression of cyclins. Immunostaining was performed separately with anti-cyclin A, anti-cyclin B, or anti-cyclin E antibodies (Fig. 5). Both cyclin A and cyclin B act synergistically during the G2-M phase transition (Knoblich and Lehner, 1993), whereas cyclin E is required for progression through S phase of the cell cycle (Knoblich et al., 1994). All were readily detected in the baf¹/+ control CNS (Fig. 5, right). However, although large numbers of baf¹/baf¹ larvae were incubated with anti-cyclin A antibodies (43 larvae), anti-cyclin B antibodies (63 larvae) or anti-cyclin E antibodies (68 larvae), little or no staining was seen in the CNS of any late third instar larvae (Fig. 5, left). Thus cyclins A, B and E all seem to be down-regulated in baf¹/baf¹ larvae.

To summarize, loss of the BAF gene (baf¹/baf¹) apparently leads both to a cell-cycle block before M phase and aberrant BrdU incorporation (see Discussion).

As BAF is known to interact directly with the LEM domain proteins, the loss of BAF would be expected to influence nuclear architecture globally. Therefore the nuclear lamina of CNS cells was investigated with anti-lamin Dm₀ (the Drosophila B-type lamin) antibodies. The Drosophila A-type lamin was not investigated because it was undetectable in larval neuroblasts with a specific antibody (m-mAb LC28; Reimer et al., 1995), and thus may not be expressed in this tissue; A-type lamins were detected in other tissues (data not shown).

Both thoracic and abdominal ganglia as well as brain hemispheres of baf¹/+ control larvae all showed uniform staining when viewed at low magnification (Fig. 6A); distinct nuclear `rim' staining was seen by confocal microscopy (Fig. 6B). Uniform nuclear rim staining was also observed in the CNS of both baf¹/baf¹ rescued with gBAF (Fig. 6A,B) and the l(2)k10210/l(2)k10210 parent line (data not shown). In contrast, conspicuous heterogeneity of intensity was seen when nuclei of the baf¹/baf¹ larval CNS were stained with the anti-lamin Dm₀ antibodies (Fig. 6A). More intensely stained nuclei were distributed throughout the CNS, with some tendency for accumulation. By confocal microscopy, cell heterogeneity was even more conspicuous (compare Fig. 6B [baf¹/+] with C [baf¹/baf¹]). Quantitative analyses of baf¹/baf¹ mutant animals suggested that lamin distribution was obviously unusual in about 30% of brain hemisphere nuclei (n=180). Highly convoluted structures and/or intranuclear accumulation of lamin-containing structures were commonly observed.

Cell heterogeneity with normal and abnormal lamin distribution appears to be a phenotype specific to baf¹/baf¹ CNS tissues. We hypothesized that baf¹/baf¹ mutant embryos were endowed with substantial BAF stores (either protein or mRNA), presumably of maternal origin, and that continuous division of neuroblasts, in contrast to non-dividing differentiated cells, during larval development might cause the depletion of maternal BAF supplies in nuclei. These nuclei appear morphologically abnormal as revealed by aberrant lamin staining. To test whether BAF is specifically absent from these unusual nuclei of the baf¹/baf¹ CNS, double-label confocal immunofluorescence was performed. Wherever lamin staining was grossly aberrant, BAF could not be detected with anti-BAF antiserum (Fig. 7). In the baf¹/baf¹ mutant used for this analysis, residual BAF protein was also detected in endocyclic tissues (data not shown). Thus loss of BAF was directly correlated with abnormal lamin distribution and misshapen nuclei.

Furthermore, in the baf¹/baf¹ CNS, abnormal nuclei appeared with a similar distribution to the pattern of BrdU incorporation. Lamina distortion and BrdU incorporation were directly compared by double immunostaining with anti-lamin Dm₀ and anti-BrdU antibodies in dissected late third instar larval baf¹/baf¹ CNS, which were directly incubated for 1.5 hours in medium containing BrdU. While active DNA synthesis and apparently normal nuclear structure were routinely observed in the baf¹/+ control CNS (Fig. 8), intense BrdU incorporation seen in baf¹/baf¹ nuclei usually appeared to coincide with substantial lamina distortion (see particularly Fig. 8, Merge), in spite of lack of detectable cyclin E (Fig. 5). This observation was certainly unexpected and raises several key questions. In addition, although we assume that the bulk of BrdU incorporation results from DNA replication, repair synthesis cannot be excluded.

The abnormal nuclear lamin staining of the baf¹/baf¹ CNS suggested that there might be other changes in nuclear structure. To evaluate this possibility, the baf¹/baf¹ CNS was further examined both by immunostaining for nuclear pore complex (NPC) proteins and histones, and by transmission electron microscopy (TEM) of ultra-thin sections.

When the baf¹/baf¹ CNS was stained with mAb414 for NPC antigens, in nuclei in which lamin staining appeared normal (bright rim staining with relatively dark central regions), mAb414 labeling was comparable in appearance (Fig. 9). Similar staining was also observed in the baf¹/+ control CNS cells (data not shown). However, when abnormal lamin staining was seen in the baf¹/baf¹ CNS, abnormal mAb 414 staining was seen as well. Abnormalities of both distribution and intensity were observed (Fig. 9). To investigate further, lamin Dm₀ distribution was compared with chromatin distribution, the latter being visualized with anti-histone H2A antibodies. When lamin distribution appeared grossly abnormal in baf¹/baf¹ nuclei, colocalization with chromatin was evident; in addition, these nuclei showed prominent chromatin staining that extended well beyond that of lamin Dm₀ (Fig. 9, arrows). Similar images were also obtained after double staining for lamin Dm₀ and DNA, respectively (data not shown).

The uniform distribution of chromatin seen in baf¹/+ control nuclei (Fig. 10A) by TEM was revealed in many instances in baf¹/baf¹ nuclei to be lost, and heterochromatin-like clumps appeared (Fig. 10B). Moreover, in another baf¹/baf¹ nucleus, the NE was highly folded, particularly in regions where clumped chromatin was most prominent (Fig. 10C-E, white arrowheads). This folding was similar to the heavily convoluted pattern that was observed by immunostaining with anti-lamin Dm₀ antibodies. At the EM level, about 58% of the nuclei of the baf¹/baf¹ mutant brain hemispheres showed clumped chromatin and convoluted NE. Despite dramatic NE distortions, higher magnification revealed that both nuclear pore complexes and a complete double membrane appeared morphologically normal (Fig. 10D,E, black arrowheads).

We have identified that Drosophila lacking the chromosomal baf gene exhibit several phenotypic characteristics that distinguish them from their l(2)k10210/l(2)k10210 parents. This parental line contains a single P-element insertion and its late pupal lethal phenotype was shown to revert to complete viability by remobilization of the insert. Thus its phenotype is dependent solely on this insert and may represent a hypomorph of the baf gene that cannot be complemented by the rescue construct utilized in this study, or may be due to altered expression of one of the neighboring genes.

Two different fly lines deleted for the baf gene are almost identical phenotypically, and gBAF rescues all their specific defects. Thus it is highly likely that the phenotypic characteristics described in this article represent effects specifically due to BAF depletion during normal development. However, it remains a formal possibility that effects seen result from BAF depletion only when combined with the l(2)k10210/l(2)k10210 genotype. In either case, the differences of both mutants with the parental line can be considered to reflect the function of the baf gene.

Confocal immunofluorescence demonstrated that a major fraction of Drosophila BAF localized to nuclei (chromatin) during interphase (Fig. 1A,C,E) and that a portion still binds chromosomes throughout mitosis (Fig. 1B,D). Similar immunolocalization was reported for the endogenous BAF in rat FRSK cells (Furukawa, 1999) and after expression of GFP-tagged BAF, in normal rat kidney cells (Ellenberg, personal communication). Nuclear localization of BAF was also demonstrated in HeLa cells (Haraguchi et al., 2001) and in a Xenopus in vitro assay [(Segura-Totten et al., 2002), also see the online supplement]. Our observation of chromosomal localization of BAF during M phase is apparently at odds with that of others (e.g. Haraguchi et al., 2001) and raises the possibility that during the cell cycle, BAF is regulated somewhat differently between different animals and/or cell types. Drosophila baf null mutants exhibited typical mitotic defect phenotypes with a conspicuous abnormality in interphase nuclear morphology but without a significant accumulation of nuclei blocked in M phase. In addition, in the grossly abnormal nuclei defined as characteristic of baf null mutants by immunofluorescence microscopy with anti-lamin antibodies, BAF could not be detected immunocytochemically with anti-BAF antibodies. Thus, we suggest that BAF may be essential in cell cycle progression via its involvement in nuclear organization.

We have demonstrated that mitotic chromosomal abnormalities occur in baf¹/baf¹ animals. Although cells of the baf¹/baf¹ CNS were never seen undergoing normal mitosis (see Fig. 3), in at least the few cells that enter M phase, irregular chromosomal shapes are easily found by propidium iodide staining as well as with phosphorylation-specific anti-histone H3 antibodies. Furthermore, individual chromosomes, seen commonly in normal cells, were almost never detected (Fig. 3C compare with D). However, we did not detect any anaphase-bridged structures as reported by Zheng et al. (Zheng et al., 2000) for C. elegans. These findings suggest that abnormal chromosome condensation and/or improper segregation can be induced by BAF depletion.

The effects of BAF on mitotic chromosome structure in vitro are quite complex. For example, the addition of small amounts of bacterially expressed BAF to a Xenopus egg extract cell-free nuclear assembly assay was recently shown to enhance chromatin decondensation, whereas addition of much greater amounts caused BAF accumulation at chromosome surfaces and inhibited decondensation (Segura-Totten et al., 2002). Similar defects of chromosome decondensation after mitosis have also been demonstrated in vivo in cells overexpressing GFP-tagged BAF (Ellenberg, personal communication). These effects of BAF on mitotic chromosome structure agree with the mitotic chromosomal abnormalities observed in Drosophila bearing the baf null mutation. Together, these observations imply that the DNA bridging protein BAF has a role in both assembly and disassembly of mitotic chromosomes.

In addition, our genetic analyses demonstrated abnormal interphase chromosomal structures (chromatin `clumps') seen preferentially in Drosophila baf null mutant cells (Fig. 10B). Apparently similar clumps were reported in Drosophila embryos mutated for lamin Dm₀ (Harel et al., 1998). Since anti-phospho-histone H3 antibodies (PH10 and HTA28) barely label CNS cells from baf null mutant larvae, it would appear that these chromatin clumps do not represent chromosomal regions prematurely condensed for mitosis. Rather we suggest that the abnormal clumping observed reflects some sort of normal interphase condensation process, such as heterochromatin formation, that has gone awry.

Although mitotic defects were not the only phenotype observed herein, substantial chromatin localization of BAF was demonstrated both in this work and by others (e.g. Segura-Totten et al., 2002). In addition, recent findings in vertebrate systems showed that BAF concentrates to the NE (Segura-Totten et al., 2002) (J. Ellenberg, personal communication). Together, these and our current observations suggest that the DNA binding and bridging functions of BAF (Lee and Craigie, 1998; Zheng et al., 2000), as well as interactions between chromatin and the NE are all required for normal interphase chromosomal organization.

From the perspective of NE organization, the results presented here show that loss of the baf gene actually leads to significant distortion of the nuclear lamina in the cells of a living animal (Fig. 6). This observation could be anticipated from studies in vitro and in cultured cells, but explicit demonstration nevertheless seems important. Results obtained after manipulation and transfection of lamin genes (Furukawa and Hotta, 1993; Lenz-Bohme et al., 1997; Harel et al., 1998; Sullivan et al., 1999; Liu et al., 2000; Schirmer et al., 2001) indicate that nuclear lamins have primary roles in the determination of nuclear shape. For example, transfection of somatic tissue culture cells with a lamin B1 mutant lacking most of the protein's rod domain (B1Δrod) (Schirmer et al., 2001) and mammalian spermatocyte lamin B3 (Furukawa and Hotta, 1993) are of particular interest. Both lamins lack the domains for interaction with multiple proteins including that which putatively binds LAP2β (Furukawa and Kondo, 1998) and induce distortion of nuclear structure in somatic tissue culture cells. Moreover, fly embryos deficient in lamin Dm₀ were reported to lack the normal attachment of peripheral chromatin to the NE (Harel et al., 1998). Together, these observations suggest that chromatin not only anchors on the NE but also that normal chromatin-NE interactions are needed to establish and/or maintain nuclear lamina structure.

Many LEM domain-containing proteins, including otefin are expressed in Drosophila. In this context, it is noteworthy that BAF interacts with LEM domain-containing proteins in general (Furukawa, 1999; Shumaker et al., 2001; Lee et al., 2001; Haraguchi et al., 2001). As LEM domain-containing proteins are also known to interact directly with nuclear lamins (Foisner and Gerace, 1993; Lee et al., 2001) (reviewed by Worman and Courvalin, 2000), the NE distortion apparently brought on by the loss of BAF (e.g. see Fig. 7) might result from abrogation of structural interactions between BAF/DNA complexes and LEM domain-containing proteins interacting with nuclear lamins. The hypothetical BAF-otefin-lamin Dm₀ interaction is a candidate for this role (see Gruenbaum et al., 2000) in Drosophila brain neuroblasts, particularly since lamin C (the other fly lamin) was not detected in these cells (data not shown). Alternatively, it is certainly possible that BAF interacts with lamin Dm₀ directly and this interaction is needed for normal lamina morphology; recently a BAF-lamin A interaction was demonstrated in vitro (Holaska et al., 2003).

BAF loss causes lethality and morphological changes in the nuclei as expected for the depletion of a protein that mediates interactions between the nuclear envelope and chromatin. Staining for histone H3 phosphorylation showed a reduction rather than an accumulation of mitotic figures (Fig. 3) suggesting that cell cycle defects were likely to be occurring outside of M phase. Interestingly, the few cells that seem to enter mitosis appear to be abnormal during M phase (Fig. 3D) as has been reported from baf RNA interference experiments in C. elegans (Zheng et al., 2000). The difference between the vast majority of cells that arrest outside mitosis and those few that enter may depend on levels of residual maternal BAF. Specifically some cells may retain enough BAF to proceed into M phase, albeit aberrantly. Immunolocalization with our current anti-BAF antiserum has not proved sufficiently sensitive to evaluate this possibility rigorously.

After 1.5 hours of in vitro BrdU labeling of CNS tissues dissected from late third instar larvae, the number of cells in which BrdU incorporation was observed was reduced significantly in the baf¹/baf¹ animals. However, the amount of BrdU incorporation seen in some baf¹/baf¹ cells apparently resembled levels in baf¹/+ cells (Fig. 8). Curiously, BrdU incorporation is most prominent in nuclei with distorted lamin distribution (Fig. 8). These nuclei seem not to be able to progress into M phase. Indeed, distorted nuclei (30% from confocal immunofluorescence observation) in the baf¹/baf¹ brain hemispheres are seen considerably more frequently than cells with grossly abnormal M phase chromosomes. Furthermore, our immunofluorescence studies of baf¹/baf¹ larvae suggest both that these distorted nuclei lack BAF (Fig. 7) and that the S phase inducing cyclin E is down-regulated. We therefore suggest that the BrdU incorporation observed may represent abnormal DNA replication that is a specific feature of the baf loss-of-function phenotype. If this is the case it would suggest that BAF has a role in events leading to or including S phase, and combined with the fact that others have demonstrated that BAF is localized in both the nuclear rim and nucleoplasm (Segura-Totten et al., 2002) and directly interacts with LEM domain proteins, supports the interpretation that BAF is directly involved in interphase nuclear functions including DNA replication. Recent reports that BAF has an additional interphase function in transcriptional repression (Wang et al., 2002) may be relevant in this respect.

In conclusion, the results obtained in this work demonstrate for the first time in vivo that BAF is required for the organization of the nuclear envelope and interphase chromosomes. This requirement probably affects the progression of the cell cycle during Drosophila development.

We thank H. Richardson for providing anti-cyclin E antibodies; R. Glaser for providing anti-histone antibodies; and D. Henderson and H. Siomi for critical reading of the manuscript. This work was supported by grants to K.F. from the Ministry of Education, Science, Sports, Culture and Technology of Japan and The Asahi Glass Foundation, The Inamori Foundation, The Kato Foundation, The Tokai Foundation, The Intelligent Cosmos Foundation and The Utida Energy Science Promotion Foundation.