The Drosophila kismet gene is related to chromatin-remodeling factors and is required for both segmentation and segment identity

The homeotic genes of the Antennapedia (ANTC) and bithorax (BXC) complexes specify the identities of body segments in Drosophila (Duncan, 1987; Kaufman et al., 1990). Homeotic genes encode homeodomain transcription factors that specify the identity of one or more body segments by regulating the transcription of a battery of downstream target genes. Counterparts of the Drosophila homeotic genes (the Hox genes) are present in vertebrates and other metazoans where they play highly conserved roles in the control of cell fate (Gellon and McGinnis, 1998; Maconochie et al., 1996; Manak and Scott, 1994). During the past decade, much of the research on Hox genes in Drosophila and other organisms has been focused on three questions. How are the spatially restricted patterns of homeotic gene transcription established and maintained during development? How do the homeotic transcription factors regulate the transcription of their target genes? What target genes are regulated by each homeotic transcription factor to specify the identities of the individual body segments?

Of these three issues, perhaps the most is known about the regulation of homeotic gene expression. The initial patterns of homeotic gene transcription are established in the early embryo by transcription factors encoded by segmentation genes. During subsequent development, these patterns are maintained by two ubiquitously expressed groups of regulatory proteins: the Polycomb group of repressors and the trithorax group of activators (reviewed in Kennison, 1995; Pirrotta, 1997; Simon, 1995). Mutations in Polycomb group genes cause homeotic transformations due to the ectopic transcription of homeotic genes. Conversely, mutations in many trithorax group genes cause homeotic transformations due to the failure to maintain the transcription of homeotic genes. By maintaining states of transcription established earlier in development by the segmentation genes, Polycomb and trithorax group proteins play critical roles in the specification of body segment identities.

Although the mechanism of action of the Polycomb group proteins is not well understood, a growing body of evidence suggests that they act in concert to repress transcription. Polycomb (PC) physically interacts with other Polycomb group proteins, including Polyhomeotic (PH), Posterior sex combs (PSC) and Sex combs on midleg (SCM) (Franke et al., 1992; Kyba and Brock, 1998a,b; Peterson et al., 1997; Strutt and Paro, 1997). Physical interactions between two other Polycomb group proteins, Enhancer of zeste [E(Z)] and Extra sex combs (ESC) have also been detected (Jones et al., 1998; Tie et al., 1998). Many Polycomb group proteins repress transcription of their target genes via cis-regulatory elements known as Polycomb group response elements (PREs). Complexes of Polycomb group proteins may be targeted to PREs via interactions with the Polycomb group member Pleiohomeotic, which directly binds a DNA sequence conserved in several PREs (Brown et al., 1998).

How do PC and other Polycomb group proteins repress transcription once targeted to a PRE in the vicinity of a homeotic gene? One current view is that Polycomb group proteins render homeotic genes inaccessible to activators or components of the general transcription machinery by altering chromatin structure (reviewed in Pirrotta, 1998). The Polycomb protein contains a short domain, the chromodomain, which is conserved in HP1, a component of Drosophila heterochromatin involved in heritable gene silencing (Eissenberg et al., 1992; Paro and Hogness, 1991). A direct connection between Polycomb group proteins and chromatin structure has not been proven, however, and it is possible that they repress transcription by blocking enhancer-promoter interactions, interfering with the assembly of the preinitiation complex or altering the subnuclear localization of specific genes (McCall and Bender, 1996; Pirrotta, 1998).

Recent studies of a trithorax group gene, brahma (brm), have provided a direct connection between the regulation of homeotic gene expression and chromatin. brm was identified in a screen for dominant suppressors of Pc (Kennison and Tamkun, 1988). The rationale of the screen was that reduction of an activator function could compensate for the reduced PC repressor function. brm is highly related to a yeast transcriptional activator, SWI2/SNF2, that encodes the ATPase subunit of a 2 MDa chromatin-remodeling complex, the SWI/SNF complex (Tamkun et al., 1992). The SWI/SNF complex facilitates the binding of many transcriptional activators to their binding sites in the context of chromatin by causing ATP-dependent alterations in nucleosome structure (reviewed in Pollard and Peterson, 1998; Kingston et al., 1996; Peterson and Tamkun, 1995; Winston and Carlson, 1992). The BRM protein is also highly related to the yeast STH1 protein, the ATPase subunit of RSC, a chromatin-remodeling complex related to SWI/SNF (Cairns et al., 1996). BRM has been purified from embryos and shown to be the ATPase subunit of a complex related to both SWI/SNF and RSC (Papoulas et al., 1998). Based on these findings, it is likely that BRM uses the energy of ATP hydrolysis to counteract the repressive effects of chromatin on ANTC and BXC homeotic genes.

In addition to brm, mutations in eleven other previously unidentified genes were recovered as dominant suppressors of Pc (Kennison and Tamkun, 1988). Here we present the genetic and molecular analysis of one of these genes, kismet (kis). kis mutations strongly suppress the homeotic transformations observed in heterozygous Pc adults. Conversely, kis duplications strongly enhance Pc mutant phenotypes (Kennison and Russell, 1987). kis plays a dual role in development. Loss of zygotic kis function causes homeotic transformations, indicating that kis is a member of the trithorax group of homeotic gene activators. Loss of maternal kis function causes dramatic segmentation defects similar to those caused by mutations in pair-rule genes. Molecular analysis of the kis gene revealed that it encodes multiple, large nuclear proteins that are related to BRM and other chromatin-remodeling factors. These findings suggest that KIS and BRM function by similar mechanisms and reinforce the importance of chromatin-remodeling factors in developmental processes.

Flies were raised on a cornmeal/molasses/yeast/agar medium with either Tegosept or propionic acid at 25°C. Unless otherwise noted, all mutations and chromosome aberrations are described in Lindsley and Zimm (1992). The isolation of the first nine kis alleles (kis^1-9) was described previously (Kennison and Tamkun, 1988). Of these, kis¹ and kis² are the strongest alleles by both the criteria of lethal phase and dominant suppression of Pc⁴. All nine appear to be loss-of-function alleles, since deficiencies that include kis are also strong dominant suppressors of Pc⁴ and males homozyous for kis¹ (or heterozygous for kis¹ and any of the other eight alleles) are normal if they carry a duplication that includes kis⁺ [either Dp(2;Y)L124 or Dp(2;1)L124]. The kis^S allele (previously known as AS760) was identified in Allan Spradling’s laboratory and results from a P-element insertion. The Df(2L)net chromosomes are described in Caggese et al. (1988). Dp(2;1)L124 was generated for the dosage experiments by irradiating C(1)RM, y²su(w^a) w^abb /Dp(2;Y)L124, B^S females with 2000 R of gamma rays, crossing to Oregon R males, and selecting exceptional y²su(w^a) w^aB⁺ sons. One of these sons, which proved to carry a detachment of the attached X chromosome with the duplication for the tip of 2L on the right arm of the X chromosome, was used to balance the Dp(2;1)L124 chromosome.

Clones of homozygous kis tissue were induced by mitotic recombination using either X-irradiation (Lawrence et al., 1986) or the FLP-FRT technique (Golic, 1991; Xu and Rubin, 1993). The Minute technique was used to increase the size of the clones (Morata and Ripoll, 1975). To generate X-ray-induced clones, y; kis¹ck cn bw sp/SM6a virgin females were mated to y f^36a; Dp(1;2)sc¹⁹, y⁺M(2)201/SM6a males. Blastoderm embryos, first instar larvae and third instar larvae from this cross were irradiated with 500 R, 1000 R and 1260 R, respectively, using a Torrex 120D X-ray generator (Astrophysics Research Corporation). kis clones were scored in kis¹ck cn bw sp/Dp(1;2)sc¹⁹, y⁺M(2)201 flies using the markers yellow (Y) and crinkled (ck). y females heterozygous for kis² or kis^S were similarly crossed to y f^36a; Dp(1;2)sc¹⁹, y⁺M(2)201/SM6a males, irradiated and scored for the presence of clones marked with y. As a control, iso1 (y; cn bw sp) virgin females (Brizuela et al., 1994) were crossed to y f^36a; Dp(1;2)sc¹⁹, y⁺M(2)201/SM6a males. Progeny were irradiated as above and the frequency of y clones was compared to the frequency of y kis clones from the experimental crosses. A total of 87 radiation-induced blastoderm clones were observed in 9287 legs (0.94 clones/100 legs). Of these, 7283 legs containing 65 homozygous kis¹ clones were mounted and examined (0.89 clones/100 legs). 22 kis^S clones were observed in 2004 legs (1.10 clones/100 legs).

To use the FLP-FRT technique, the kis¹ and kis^S alleles were recombined onto the P[ry⁺, y⁺]25F P[ry⁺, hs-neo, FRT]40A chromosome (Xu and Rubin, 1993). Virgin y w P[hsFLP]; P[ry⁺, y⁺]25F P[ry⁺, hs-neo, FRT]40A females were mated to either (1) y w¹¹¹⁸; kis^SP[ry⁺, hs-neo, FRT]40A/SM6a males, (2) y w¹¹¹⁸; kis¹P[ry⁺, hs-neo, FRT]40A bw sp/SM6a males, or (3) y w¹¹¹⁸; P[ry⁺, hs-neo, FRT]40A bw sp/SM6a males. Mitotic recombination was induced by heat-shocking blastoderm embryos or third instar larvae at 37°C for 60 and 90 minutes, respectively. Adults were dissected, mounted in Gary’s magic mountant (Lawrence et al., 1986) and examined under the light microscope.

Germline mosaics were generated using the dominant female-sterile technique (Chou and Perrimon, 1992; Chou et al., 1993). Males carrying an insertion of ovo^D1 on the second chromosome (P[w⁺, ovo^D1]/CyO) (Chou et al., 1993) were crossed to females of the genotypes (1) w; kis¹cn bw sp/SM6a, (2) y; kis^S/SM6a, or (3) y Df(1)w67c2. Germline clones were induced by irradiating late first instar larvae with 1000R. Female offspring of the genotypes (1) +/w; kis¹cn bw sp/P[w⁺, ovo^D1], (2)+/y; kis^S/P[w⁺, ovo^D1], or (3)+/ y Df(1)w67c2; +/P[w⁺, ovo^D1] were crossed to males heterozygous for a kis deficiency (w; Df(2L)net-PMC/SM6a). The cuticle of unhatched larvae were examined as described by Ashburner (1989). As a control for leakiness of the ovo^D1 mutation, crosses were set up as above but were not irradiated.

To generate germline mosaics using the FLP-FRT system, females of the genotypes (1) y w; kis^SP[ry⁺, hs-neo, FRT]40A/SM6a, (2) y w; kis¹P[ry⁺, hs-neo, FRT]40A bw sp/SM6a, or (3) y w; P[ry⁺, y⁺]25F P[ry⁺, hs-neo, FRT]40A/CyO were crossed to y w P[ry⁺, hsFLP]¹²; P[w⁺, ovo^D1]^2L-13X13P[hs-neo, ry⁺, FRT]40A/CyO males. First instar larvae from this cross were heat-shocked at 37°C for 90 minutes to induce mitotic recombination. Female progeny of the genotypes (1) y w/y w P[ry⁺, hsFLP]¹²; kis^SP[ry⁺, hs-neo, FRT]40A/P[w⁺, ovo^D1]^2L-13X13P[hs-neo, ry⁺, FRT]40A, (2) y w/y w P[ry⁺, hsFLP]¹²/kis¹P[ry⁺, hs-neo, FRT]40A bw sp/P[w⁺, ovo^D1]^2L-13X13P[hs-neo, ry⁺, FRT]40A, or (3) y w/y w P[ry⁺, hsFLP]¹²; P[ry⁺, y⁺]25F, P[ry⁺, hs-neo, FRT]40A/P[w⁺, ovo^D1]^2L-13X13P[hs-neo, ry⁺, FRT]40A were crossed to w; Df(2L)net-PMC/SM6a males. Embryos produced by females bearing a germline clone were examined as described above.

RNA was isolated from developmentally staged embryos, larvae, pupae and adult flies as described in Tamkun et al. (1992). Standard techniques were used for nucleic-acid blot analyses (Sambrook et al.,1989). Filters were probed with DNA fragments labeled by the random-primer method (Feinberg and Vogelstein, 1983) and washed under high stringency (0.1× SSC, 0.1% SDS, 65°C). To control for even loading, the RNA blots were probed with a radiolabeled fragment from the rp49 gene (O’Connell and Rosbash, 1984). RNA probes for direction of transcription assays were generated and used as described in the 1991 Promega Protocols and Applications Guide.

Genomic and cDNA clones were isolated from phage and cosmid libraries using standard techniques (Elfring et al., 1994; Tamkun et al., 1992). cDNA clones corresponding to the kis mRNA were isolated from an iso-1 λgt11 cDNA library (0-24 hour embryos; Tamkun et al., 1992), an OregonR λgt10 cDNA library (3-10 hour embryos; Poole et al., 1985), and a CantonS λZAP cDNA library (0-24 hour embryos; Stratagene). cDNA fragments were subcloned into plasmid vectors and sequenced as described in Elfring et al. (1994). Both strands of DNA were sequenced for all reported sequences. The cosmid isozakB was isolated by screening an iso-1 genomic cosmid library constructed in a NotBamNot-CoSpeR vector (Tamkun et al., 1992) with the gt10-3 cDNA. This cosmid was introduced into the Drosophila germline using P-element-mediated transformation as described in Tamkun et al. (1991) and tested for the ability to rescue the recessive lethality of kis mutations.

Digoxigenin-labeled DNA fragments were prepared using the Boehringer Mannheim Genius kit and hybridized to salivary gland polytene chromosomes (Engels et al., 1986), embryos (Tautz and Pfeifle, 1989) or third instar larvae (Kramer and Zipursky, 1992). Anti-digoxigenin antibody (Boehringer Mannheim) was used to detect the bound DNA fragments according to the manufacturer’s directions.

SDS-polyacrylamide gel electrophoresis and western blotting were performed as described previously (Tsukiyama et al., 1995). Native embryo extracts and purified BRM complex were prepared and analyzed by western blotting as described in Papoulas et al. (1998). To generate a KIS fusion protein, a 962 bp PstI fragment from the kis cDNA clone Zap1 was cloned into pUR292 (Rüther and Müller-Hill, 1983). A 908 bp BamHI fragment from this construct was then cloned into the pGEX 3X glutathione-S-transferase (GST) fusion vector (Pharmacia). This fragment contains 893 bp of Zap1 encoding amino acids 730-1029 of the KIS protein (Figs 4, 6), and 15 bp of the pUR292 polylinker encoding three additional residues. GST-KIS fusion proteins were expressed in E. coli, purified on glutathione-agarose columns and used to immunize rabbits as described in Harlow and Lane (1988). Antibodies were affinity purified on columns containing either GST or GST-KIS fusion proteins coupled to Affigel 10 or Affigel 15 resins (BioRad) according to the manufacturer’s instructions. Bound antibodies were eluted with 3.5 M MgCl₂ in PBS, dialyzed and assayed by western blotting.

Whole-mount preparations of embryos and third instar larvae were stained with primary antibodies and visualized with secondary antibodies conjugated to horseradish peroxidase (BioRad) as described by Pattatucci and Kaufman (1991). To generate larvae carrying a Pc mutation and four wild-type copies of the kis gene, Basc/Dp(2;1)L124, y w females were crossed to +; Pc⁴p^pe^s/TM6B, Hu Sb e Tb ca; T(Y;2)L124, y⁺B^S males. Larvae of the genotype Dp(2;1)L124, y w/Dp(2;Y)L124, B^s; +/Pc⁴p^pe^s could be distinguished from other progeny by the presence of light-brown mouth parts, colorless Malpighian tubules and the non-Tb phenotype. Rabbit polyclonal antibodies against SCR were a generous gift from Thom Kaufman (Indiana University). Charles Girdham and Pat O’Farrell (University of California, San Francisco) generously provided rabbit polyclonal antibodies against EN. Antibodies against EVE were a gift from Michael Levine (University of California, Berkeley).

Heterozygous kis mutations and deficiencies spanning the kis gene strongly suppress the homeotic transformations observed in heterozygous Pc adults, including the transformation of second and third leg to first leg [due to Sex combs reduced (Scr) derepression], wing to haltere [due to Ultrabithorax (Ubx) derepression] and anterior abdominal segments to more posterior identities [due to Abdominal-B (Abd-B) derepression]. Conversely, kis duplications strongly enhance these transformations (Kennison and Russell, 1987). Since the adult Pc phenotypes are due to the ectopic transcription of homeotic genes, it seemed likely that kis might act antagonistically to Pc to activate the transcription of Scr and other homeotic genes.

To test this possibility, we examined the effects of varying the dosage of the kis gene on the expression of the homeotic gene Scr in the imaginal discs of heterozygous Pc larvae (Fig. 1). Scr specifies the identities of the labial and first thoracic segments and is normally not expressed in either the second or third leg imaginal discs. As previously observed by Pattatucci and Kaufman (1991), individuals heterozygous for a Pc mutation show ectopic expression of SCR proteins in both the second and third leg imaginal discs. The kis² mutation strongly suppresses the ectopic expression of SCR proteins in Pc heterozygotes, while four wild-type copies of the kis gene strongly enhance the ectopic expression. We observed similar results when we examined the levels of Scr RNA by in situ hybridization to imaginal discs (data not shown). These data suggest that kis acts antagonistically to Pc to activate the transcription of Scr and other homeotic genes.

kis is an essential gene; kis homozygotes die as first or second instar larvae with no obvious cuticular defects (data not shown). To examine the roles of kis during later stages of development, we induced mitotic exchange in first or third instar larvae and examined adult cuticles for changes caused by loss of kis function. Loss of kis function in the fifth abdominal segment caused a complete transformation toward a more anterior identity, as shown by the loss of the black pigmentation characteristic of this segment (Fig. 2A). This phenotype is identical to that associated with loss-of-function mutations in the Abd-B gene of the BXC. Identical results were obtained using three different kis alleles (kis¹, kis² and kis^S). The boundaries of homozygous kis mutant tissue coincided with the mutant phenotypes, indicating that kis is cell autonomous. Clones of homozygous kis tissue in other abdominal segments appeared phenotypically normal, although the lack of obvious differences between the second, third and fourth abdominal segments would not allow us to detect subtle transformations between these segments.

Because kis mutations suppress the ectopic expression of Scr in Pc heterozygotes, we anticipated that loss of kis function in the developing leg discs would alter the identity of the first leg. Surprisingly, no clear homeotic transformations were observed in homozygous kis clones in the thorax when they were induced during larval development. We therefore generated clones of mutant kis tissue at the cellular blastoderm stage of embryogenesis. The average sizes and frequencies of kis^S and kis¹ clones induced at the cellular blastoderm stage were not significantly reduced in any body segments relative to the controls, suggesting that kis is not essential for cell viability or division (data not shown). The majority of clones were found in regions of the legs that are similar in each thoracic segment, precluding the detection of homeotic transformations. Several kis^S clones did, however, exhibit morphological abnormalities. Homozygous kis^S clones in the anterior compartment of the first leg display a reduction in landmark first leg features, such as the transverse row of bristles and the number of sex comb teeth (Fig. 2B). More dramatically, a kis^S clone in the first leg also displayed an apical bristle, a landmark feature of the second leg, at the distal end of the tibia (Fig. 2D). This homeotic transformation mimics loss-of-function mutations in Scr and is consistent with a role for kis as an activator of Scr transcription.

Examination of fifteen kis¹ clones and five kis^S clones in the second leg revealed no phenotypic abnormalities, even though they were induced at the cellular blastoderm stage of embryogenesis. In the third leg, twenty kis¹ and nine kis^S clones were examined. Clones of both alleles exhibited abnormal morphologies, particularly in the distal leg structures; the femur and tibia were slightly abnormal while the tarsal segments were severely distorted (Fig. 2F,G). All five tarsi were present but were truncated and exhibited a hooked shape. These abnormalities could be the result of an absence or reduction of portions of one compartment, resulting in the hooked appearance. No homeotic transformations were observed, although these would be difficult to recognize in the distal leg given the degree of pattern disruption. kis clones in other body segments, including the head, thorax, wing, haltere and genitalia, did not display any obvious developmental abnormalities.

A possible explanation for the lack of discernible pattern defects in kis homozygotes is that sufficient maternal kis gene products are present in the embryo to allow normal development in the absence of zygotic kis function. To investigate this possibility, we examined the consequences of eliminating both the maternal and zygotic contributions of kis gene products using germline clonal analysis. Embryos derived from mothers bearing germline clones of kis¹ exhibited a deletion of pattern elements approximating alternate segments (Fig. 3C). The most common defect was a reduction in size of one or more alternate segments (T3, A2, A4, A6 and sometimes A8). The second thoracic segment was also often absent or severely reduced and the head region was grossly altered. In more extreme cases, only one or two patches of denticle were observed. The cephalopharyngeal skeleton was internalized and malformed, which may be a secondary result of the major pattern alterations observed. The absence of the T2 and T3 denticle belts was accompanied by the loss of their associated sense organs, both the Keilin’s organ and the campaniform sensilla, indicating that the segments were actually missing. The first thoracic segment did not appear to be deleted or transformed, since denticle hairs and the T1-associated hairs (beard) were usually present at the anterior end of the larvae. Similar but less extreme phenotypes were observed in embryos from mothers bearing germline clones of kis^S (Fig. 3B). The pattern defects resulting from loss of maternal kis function are most similar to those seen in embryos homozygous for mutations in the pair-rule segmentation gene even-skipped (eve), in which odd-numbered parasegments are deleted (corresponding to the even-numbered abdominal segments). These results indicate that kis plays an unanticipated role in embryonic segmentation.

To test whether loss of maternal kis function alters the expression of segmentation genes, we examined the distribution of several gap and pair-rule proteins in embryos derived from mosaic mothers. We found that the expression of the gap genes Krüppel (Kr), hunchback (hb) and knirps (kni) was normal, as was the expression of the primary pair-rule gene eve (data not shown). However, the expression of the segment-polarity gene engrailed (en) is dramatically altered by the loss of maternal kis function (Fig. 3D-F). en is normally expressed in fourteen stripes and this pattern is dependent on the function of pair-rule genes, including eve. Embryos from kis^S mutant mothers exhibited the fourteen en stripes (Fig. 3E), but the stripe borders were not as defined as in wild-type embryos (Fig. 3D). In the kis¹ mutant embryos, only seven poorly defined en stripes were apparent (Fig. 3F). These patterns of en expression roughly correlate with the cuticular phenotypes. Taken together, these data indicate that maternal kis function is required for the expression or function of one or more segmentation genes, including en. The severity of the segmentation defects caused by loss of maternal kis function made it difficult to determine its role in homeotic gene regulation during embryogenesis.

To further investigate the roles of kis in segmentation and segment identity, we cloned and characterized the kis gene. Deficiency mapping revealed that the kis gene is located in salivary chromosome region 21B6-7 (Kennison and Tamkun, 1988; Lindsley and Zimm, 1992). To isolate DNA from this region, we conducted a chromosome walk in cosmid and phage libraries spanning more than 160 kb of contiguous genomic DNA. To locate the kis gene within our chromosome walk, we used in situ hybridization to map the breakpoints of several terminal deficiencies that lie within the 21B region (Fig. 4; data not shown). The breakpoints of five deficiencies allowed us to map an essential portion of the kis gene within a 17 kb region of our walk between the proximal breakpoints of Df(2L)net-PM47C and Df(2L)net-PMC (coordinates +28 to +45, Fig. 4). The position of the kis gene was more precisely determined by the molecular mapping of a P-element allele, kis^S. Upon providing P-element transposase, the kis^S allele is revertible to wild type, indicating that the P-element insertion disrupts the kis gene (data not shown). Consistent with our deficiency mapping, this insertion maps to a 2.9 kb EcoRI fragment that lies between the proximal breakpoints of Df(2L)net-PM47C and Df(2L)net-PMC (Fig. 4; data not shown).

Because of the lack of relevant chromosomal aberrations, we could not determine the proximal boundary of the kis locus by deficiency mapping. We therefore used a functional assay to further define the boundaries of the kis gene. A cosmid bearing genomic DNA extending from approximately +27 to +61 of our walk was introduced into the Drosophila genome by P-element-mediated transformation. Although the distal end of this DNA lies within Df(2L)net-PM47C (which complements kis mutations), it is unable to rescue the hemizygous or homozygous lethality of several kis alleles. This result suggests that the functional limit of the kis gene extends beyond the map coordinate +61 of our walk.

To identify potential kis transcripts, subcloned genomic fragments from the regions flanking the kis^S insertion allele were used to probe northern blots of RNA isolated from Drosophila embryos. The 2.9 kb EcoRI fragment that spans the site of the kis^S insertion hybridized to an 8.5 kb RNA on northern blots (Fig. 5A,B). All fragments between the kis^S insertion and the breakpoint of Df(2L)net-PM47C detected two major RNAs of 8.5 kb and 17 kb (Fig. 5A). All genomic probes from the region of our walk proximal to kis^S recognized only the larger 17 kb RNA (Fig. 5B). Both the 8.5 and 17 kb RNAs are transcribed in the proximal-to-distal direction, as determined by hybridization of single-stranded RNA probes to blots of Drosophila RNA (Fig. 5B). The 8.5 and 17 kb RNAs could result from the expression of more than one gene, expression of a single gene with multiple promoters, alternative RNA processing or multiple polyadenylation sites.

To distinguish among these possibilities and identify the kis RNA, two embryonic cDNA libraries were screened with genomic probes surrounding the kis^S P-element insertion. By probing blots of Drosophila RNA with isolated cDNA fragments, we identified six cDNA clones (cDNAs gt10-1, gt10-2, gt10-3, gt10-7, gt10-8 and Zap1) that hybridized to both the 8.5 kb and 17 kb RNA (Fig. 4). These overlapping cDNAs were subcloned and their complete sequences determined. Four of the clones (gt10-1, gt10-2, gt10-3 and Zap1) together define an 8238 nucleotide transcript (GenBank AF113847). The sequence of the predicted kis RNA reveals a long 1450 nucleotide 3′ untranslated region. The poly(A) tail is preceded by two consensus polyadenylation signals (AATAAA) at nucleotides 8189 and 8204. This 8238 nucleotide transcript appears to correspond to the 8.5kb RNA and contains a single long ORF (Fig. 6). Using the first AUG as the initiation codon, this 6454 nucleotide ORF extends from nucleotides 316 to 6769. The predicted 2151 residue protein (224.8 kDa and a pI of 5.82) lacks any previously identified sequence motifs indicative of biochemical function (Fig. 6; see below).

The pairing of different 5′ splice sites with a common 3′ splice site generates RNAs encoding proteins bearing a common 2105 amino-acid C terminus. The protein encoded by the 8.5 kb RNA is unique only in its amino-terminal 46 residues (Fig. 6). The amino-terminal extensions encoded by the partial cDNA clones gt10-7 and gt10-8 are 90 and 425 amino acids in length, respectively. Our biochemical studies of the KIS protein (see below) have shown that the 17 kb RNAs encode proteins with molecular masses of greater than 500 kDa, indicating that these N-terminal extensions are quite large. Thus, alternative RNA processing generates kis RNAs encoding at least three large proteins.

To identify sequences related to the 225 kDa KIS protein, we searched the current nucleic acid and protein databases as well as the complete sequence of the Saccharomyces cerevisiae genome using the TBLASTN and BLASTP programs (Altschul et al., 1997). Although no extensive homologies to other proteins were found, three proteins involved in transcriptional activation showed significant similarity (37-46% identity) to kis in a 41 amino-acid segment (Fig. 7), which we call the BRK domain (BRM and KIS). The three other proteins containing this motif are the Drosophila BRM protein and its two putative human homologs, BRG1 (hSNF2α) and hBRM (hSNF2β) (Chiba et al., 1994; Khavari et al., 1993; Muchardt and Yaniv, 1993). Although this region represents only a small fraction of the KIS protein, its presence in BRM and its human homologs is intriguing given that kis and brm were identified in the same genetic screens (Kennison and Tamkun, 1988).

Database searches also identified a partial 6452 bp human cDNA (KIAA0308: GenBank accession AB002306) encoding a large protein that also contains the BRK domain (Fig. 7) as well as an additional 129 amino-acid segment that is highly related (45% identity) to residues 109 to 237 of the KIS protein. The same two domains are present in a predicted Caenorhabditis elegans protein (PID 3158516).

Alternative splicing of the kis RNA thus leads to the production of BRK domain proteins that differ by the presence or absence of long N-terminal extensions. To investigate the significance of this heterogeneity, we compared the developmental expression of the kis RNAs by northern blotting and in situ hybridization. Northern blots of staged Drosophila RNA were hybridized to a genomic DNA fragment that recognizes all known kis transcripts. Both the 8.5 and 17 kb kis RNAs are easily detected in embryos, larvae and pupae, but not adults (Fig. 5A). Hybridization of a cDNA probe (Zap1) that recognizes all kis RNAs to whole-mount preparations of embryos revealed that kis is ubiquitously expressed from the onset of embryogenesis through germ-band extension (Fig. 8). During germ-band retraction, the RNAs gradually become localized to the ventral nerve cord and brain, eventually falling below detectable levels just prior to hatching. Identical patterns were observed using probes that specifically recognize either the 8.5 or 17 kb kis RNA (data not shown).

To monitor the expression and subcellular distribution of the KIS proteins, we raised rabbit polyclonal antisera against a 299 amino-acid segment common to all three of the identified protein isoforms (residues 731-1029, Fig. 6). Our molecular analysis predicted the existence of a 225 kDa protein encoded by the 8.5 kb kis RNA, and at least two larger proteins encoded by the 17 kb kis transcripts (Figs 4, 6). As anticipated, our affinity-purified antibody detected two major polypeptides in Drosophila embryo extracts (Fig. 9A). The smaller of the two most abundant polypeptides has an apparent molecular mass of 230 kDa, which is very close to the predicted mass of the protein encoded by the 8.5 kb kis RNA. The larger polypeptide has an apparent molecular mass greater than 500 kDa and is likely to be the encoded by the 17 kb kis RNA. Consistent with this possibility, the independent analysis of cDNA clones corresponding to the 17 kb kis RNA has shown that it encodes a protein with a predicted molecular mass of 574 kDa (Allan Wong, Marc Therrien, Debbie Morrison and Gerald Rubin, personal communication; see Discussion). Several much weaker signals were also detected on western blots, which could represent either additional protein isoforms or degradation products.

We next used our affinity-purified antibodies to examine the spatial distribution of KIS proteins in whole-mount preparations of Drosophila embryos and third-instar larvae. Consistent with a possible role in transcriptional regulation, the KIS proteins were detected at uniform levels in nuclei from the onset of embryogenesis through germ-band extension (Fig. 9B). The proteins then became localized to the developing central nervous system and were detected in the nuclei of the epidermal cells (data not shown). Levels gradually diminished after dorsal closure, when the intersegmental grooves can first be distinguished. In third-instar larvae, KIS proteins were detected in nuclei of all imaginal discs, with the exception of the labial disc, where little to no protein was detected. KIS proteins were also detected in the polytene nuclei of the larval salivary gland (data not shown).

Since the trithorax group genes brm and kis were identified in the same genetic screens for Pc suppressors, it seemed possible that they might physically interact to regulate the transcription of homeotic genes. We recently purified the BRM complex from Drosophila embryos and characterized its subunit composition (Papoulas et al., 1998). Although KIS was not identified as a major subunit of the BRM complex in that study, substoichiometric amounts of several large polypeptides reproducibly copurified with BRM. To investigate whether any of these peptides might correspond to KIS, we probed western blots of purified BRM complex and whole embryo extracts with antibodies against KIS. Neither the large nor small forms of the KIS protein were present in purified preparations of BRM complex (Fig. 9A). We have also been unable to coimmunoprecipitate the BRM and KIS proteins from Drosophila embryo extracts (data not shown). These findings suggest that the BRM and KIS proteins do not stably interact in the Drosophila embryo.

The genetic interactions between kis and Pc provided the first clue that kis plays an important role in the determination of body segment identity. We showed that kis mutations suppress the adult Pc phenotype by preventing the ectopic transcription of homeotic genes. Thus, kis is a member of the trithorax group of homeotic gene activators. Mosaic analyses revealed that loss of kis function causes homeotic transformations, including the transformation of first leg to second leg and the fifth abdominal segment to a more anterior identity. These phenotypes are identical to those associated with loss-of-function Scr and Abd-B mutations, respectively. Taken together, these findings suggest that kis acts antagonistically to Pc to activate the transcription of both Scr and Abd-B.

It is intriguing that kis mutations alter the fate of only the fifth abdominal segment, since the identities of the fifth through ninth abdominal segments are determined by a single homeotic gene, Abd-B (Celniker et al., 1989; Delorenzi and Bienz, 1990). Variations in the levels of ABD-B protein result in the differences between these abdominal segments, with Abd-B expression being lowest in the fifth abdominal segment (Duncan, 1987; Hopmann et al., 1995). Parasegment-specific cis-regulatory regions, termed infra-abdominal (iab) regions (Lewis, 1978) control Abd-B expression. Each iab region is named for the segment that it affects (iab-5 through iab-9). Mutations in both iab-5 (Celniker et al., 1990) and kis affect the identity of only the fifth abdominal segment, suggesting that the KIS protein may interact specifically with the iab-5 cis-regulatory element of Abd-B.

We suspect that kis interacts not only with Scr and Abd-B, but with other homeotic genes as well. For example, the isolation of kis mutations as enhancers of loss-of-function Deformed (Dfd) mutations (Gellon et al., 1997) suggests that kis is probably also required to activate transcription of this ANTC homeotic gene. Furthermore, kis duplications strongly enhance the transformation of wing to haltere in Pc heterozygotes, a phenotype caused by the ectopic transcription of Ubx in the wing imaginal disc (J. A. K., unpublished data). However, kis mutations do not cause haltere-to-wing transformations due to decreased Ubx transcription. A possible explanation for the lack of homeotic transformations in kis clones in segments other than the prothoracic and fifth abdominal segment is that the mutations used in these studies are not null alleles. kis¹ is a strong loss-of-function mutation. It has not been characterized at the molecular level, however, and may not completely eliminate kis function. It is also possible that sufficient levels of KIS protein persist in homozygous mutant tissue following mitotic recombination to support normal development. Further genetic studies, including the analysis of conditional kis alleles, will be necessary to distinguish between these possibilities.

Germline clonal analysis revealed an unanticipated role for kis in segmentation. Embryos from mosaic kis^S females exhibit a deletion or alteration of every other segment, while mutant embryos from mothers bearing germline clones of the stronger kis¹ allele usually develop only half of the normal number of segments. This variation in phenotypic severity is closely correlated with the extent to which en expression is disrupted. The phenotypes associated with loss of maternal kis function resemble those caused by mutations in pair-rule segmentation genes that cause the deletion of the odd-numbered parasegments. kis thus appears to be necessary for the expression (or function) of one or more pair-rule genes. Recent genetic studies have suggested that kis may also be involved in the Notch signaling pathway (Go and Artavanis-Tsakonas, 1998; Verheyen et al., 1996). Thus it appears that kis plays roles in addition to the regulation of homeotic genes.

What pair-rule genes might require kis for their activity? Based on the kis mutant perhaps the best candidates are eve and hairy (h), both of which are required for the formation of odd-numbered parasegments. Unlike eve, h and most other segmentation genes, kis is uniformly expressed in the early embryo. This raises the possibility that KIS functions as an essential cofactor or modifier of EVE or other pair-rule proteins. It is also possible that loss of kis function might result in pair-rule genes being transcribed outside of their normal expression domains. For example, Cadigan et al. (1994a,b) have shown that ectopic expression of sloppy-paired (slp) results in phenotypes similar to hypomorphic and amorphic eve mutants, and therefore similar to the phenotypes observed in embryos mutant for maternal kis. Additional work will be necessary to determine the molecular basis of the segmentation defects resulting from loss of maternal kis function.

Our molecular studies of kis provide insights into the mechanism of action of the KIS proteins. Alternative RNA processing produces several large nuclear proteins with molecular masses of approximately 225 kDa and ˜600 kDa. These proteins share a common 2105 amino-acid C terminus containing a 41 amino-acid segment – the BRK domain – that is conserved in the trithorax group protein BRM and related chromatin-remodeling factors in humans.

A more direct connection between KIS and chromatin-remodeling factors has recently been provided by independent studies in Dr Gerald Rubin’s laboratory (Allan Wong, Marc Therrien, Debbie Morrison and Gerald Rubin, personal communication). They recovered kis mutations in an unrelated genetic screen and analyzed overlapping cDNA clones corresponding to the 17 kb kis RNA. Their analysis of these clones has confirmed and extended our analysis of the kis RNAs and proteins. The 17 kb kis RNA encodes a protein with a predicted molecular mass of 574 kDa, very similar in size to the largest KIS proteins that we detect by western blotting. This protein contains the same 2105 residue C terminus as the 225 kDa KIS protein.

The N-terminal extension unique to the larger KIS protein is 3217 amino acids in length and contains several interesting functional domains, including an ATPase domain highly related to those found in BRM (44% identity over 478 amino acids) and other SWI2/SNF2 family members and two chromodomains. The ATPase domain of KIS is most highly related (approximately 50% identity) to that of the CHD (chromodomain-helicase domain) proteins (reviewed in Woodage et al., 1997). Both the KIS and CHD proteins contain a single ATPase domain and two chromodomains. Unlike KIS, however, none of the previously identified CHD proteins contain a BRK domain. Although chromodomains are a distinguishing feature of CHD proteins and many other proteins that interact with chromatin, including the Drosophila PC and HP1 proteins (Cavalli and Paro, 1998; Paro and Hogness, 1991; Singh et al., 1991), the function of the chromodomain is currently unknown. It may function as a dimerization or protein-protein interaction domain that targets proteins to the appropriate chromosomal location (Cavalli and Paro, 1998; Messmer et al., 1992; Platero et al., 1995). Both the 225 kDa and 574 kDa forms of the KIS protein contain the BRK domain. Although the BRK domain is conserved in BRM and its human homologs (BRG1 and hBRM), this domain is not present in yeast chromatin-remodeling factors related to BRM, including SWI2/SNF2 and STH1. This suggests that the BRK domain may interact with a component of chromatin unique to higher eukaryotes.

The discovery that alternative RNA processing produces a 225 kDa protein lacking the ATPase domain and chromodomains was quite surprising, since these regions are likely to be critical for the function of the KIS protein. What is the function of the 225 kDa KIS protein? If this protein, which lacks the ATPase domain, retains the ability to interact with other proteins, it may function as a naturally occurring dominant-negative protein that regulates the activity of the larger form. This restraining influence would be particularly useful for regulating the activity of SWI2/SNF2 family members that are unusually abundant or stable. To investigate the functional significance of the alternative processing of the kis RNA, we cloned the kis gene from Drosophila virilis (G. D., I. Z., W. W. and J. W. T., unpublished data), a species that diverged from D. melanogaster more than 60 million years ago. We found that both the 8.5 and 17 kb RNAs are expressed in D. virilis embryos, suggesting that both the 225 and 574 kDa KIS proteins are functionally important.

The presence of an ATPase domain, BRK domain and two chromodomains in the 574 kDa KIS protein strongly suggests that it influences chromatin structure. Do kis and brm play similar roles in chromatin remodeling and development? The KIS and BRM proteins are not highly related outside the ATPase and BRK domain. Furthermore, the KIS protein lacks domain II, an evolutionarily conserved region of the BRM protein that is thought to mediate interactions with the BAP155 subunit of the BRM complex (Elfring et al., 1998; Papoulas et al., 1998). This domain and the C-terminal bromodomain are considered to be distinguishing characteristics of the ATPase subunits of SWI/SNF complexes. Although both BRM and KIS may influence chromatin structure, these differences suggest that the two proteins activate transcription via distinct biochemical mechanisms.

Genetic studies have revealed functional differences between brm and kis. brm is required for oogenesis (Brizuela et al., 1994). Clonal analyses of brm revealed that loss of zygotic brm function decreased cell viability and caused peripheral nervous system defects, including the twinning of mechanosensory bristles and campaniform sensilla (Elfring et al., 1998). Similar defects were not observed in clones of mutant kis tissue, even when we induced clones at the cellular blastoderm stage of embryogenesis. Although brm and kis are required for the activation of Scr and Abd-B, these findings suggest that BRM and KIS do not regulate identical sets of target genes. Additional studies will be required to determine the common and distinct functions of these and other trithorax group genes.

We thank Allan Wong, Marc Therrien, Debbie Morrison and Gerald Rubin for generously sharing their unpublished information on kis. We also thank Kathy Matthews and the Bloomington Drosophila Stock Center, the Umea Drosophila Stock Center, Norbert Perrimon, William Gelbart, Corrado Caggese and Ruggiero Caizzi for providing numerous stocks. Thomas Kaufman, Michael Levine, Sean Carroll, Tom Hays, Patrick O’ Farrell and Charles Girdham generously provided antibodies used in this work. We thank Andrew Chisholm, Yishi Jin and Grant Hartzog for their comments on the manuscript and Bill Sullivan, Clif Poodry and the members of our laboratories for many helpful discussions. The initial phases of this work were conducted at the University of Colorado, Boulder. J. W. T. was supported by a postdoctoral fellowship from the Jane Coffin Childs Memorial Fund while at the University of Colorado. Subsequent work in J. W. T.’s laboratory at the University of California was supported by an American Cancer Society Junior Faculty Award, a Basil O’Connor Starter Scholar Award from the March of Dimes Birth Defects Foundation, and grants from the National Science Foundation (IBN-9004491) and National Institutes of Health (GM49883). Work in M. P. S.’s laboratory was supported by grants from the American Cancer Society (NP-501) and National Institutes of Health (P01 CA70404). J. A. K. would like to thank Igor Dawid and Arthur Levine for their encouragement and support of this work.