Molecular definition of the morphogenetic and regulatory functions and the cis-regulatory elements of the Drosophila Abd-B homeotic gene

The homeotic genes of Drosophila melanogaster act during development to specify the unique identities of each segmental unit (for review see Akam, 1987). The bithorax complex (BX-C) specifies the identities of the posterior thorax and the abdominal segments (Lewis, 1978). There are three lethal complementation groups within the BX-C, designated Ultrabithorax (Ubx), abdominal-A (abd-A) and Abdominal-B (Abd-B) (Sán-chez-Herrero et al. 1985a, b;,Tiong et al. 1985; Casanova et al. 1987). The order of these genes on the chromosome reflects the order of segments along the anterior-posterior axis that are affected by mutations in each gene (Karch et al. 1985; Duncan, 1987). The Ubx gene is required for the proper identity of the third thoracic and first abdominal segments (T3 and Al). The next adjacent gene, abd-A, is required in the second through the seventh abdominal segments (A2-A7). The Abd-B gene is active in the fifth through the ninth abdominal segments, corresponding to parasegments 10– 14 (a parasegment consists of the posterior compartment of one segment and the anterior compartment of the adjacent segment, Martinez-Arias and Lawrence, 1985). Abd-B directs proper development of certain posterior larval and adult structures including genitalia in the adult fly.

The majority of Abd-B mutations fall into one of three classes (Casanova et al. 1986). Class I mutations affect the development of PS 10– 13. Class II mutations cause transformations of PS 14. Mutations of Class III are strictly lethal and affect PS 10– 14. Class I and Class II mutants complement each other, but both fail to complement Class III mutants. It has been proposed that Class I and Class II lesions inactivate two distinct elements of the Abd-B gene and that Class III mutations inactivate both of these elements simultaneously. The activity affected by Class I mutations is called the morphogenetic or m element because it is required to produce the morphological diversity of the region from PS 10 through 13. The activity absent in Class II mutants is called the regulatory or r element.

This element is required for terminalia development and appears to suppress the activity of the Abd-B m element as well as other homeotic genes in PS 14. Class I, II and III mutants will be referred to as m⁻r⁺, m⁺r⁻ and mr⁻ mutants, respectively.

It has recently been shown that the Abd-B gene contains four overlapping transcription units, utilizing at least three separate promoters (DeLorenzi et al. 1988; Sánchez-Herrero and Crosby, 1988; Kuziora and McGinnis, 1988; Celniker et al. 1989; Zavortink and Sakonju, 1989). These mRNAs will henceforth be referred to as class A, B, C (Zavortink and Sakonju, 1989) and gamma transcripts (Kuziora and McGinnis, 1988). Data derived from in situ localization of Abd-B transcripts in wild-type embryos suggest that the m element corresponds to the class A transcript, and that r element functions may be carried out by either the class B or C mRNA species, or both acting together (DeLorenzi et al. 1988; Sánchez-Herrero and Crosby, 1988; Kuziora and McGinnis, 1988; Celniker et al. 1989; Zavortink and Sakonju, 1989). The gamma transcript is expressed in PS 15 (Kuziora and McGinnis, 1988), outside the domain of known Abd-B mutant effects, and will not be considered further in this report. cDNA sequence analysis suggests that the class A transcript encodes a 55 × 10³M_r protein while the class B and class C mRNAs encode a 30×10M_r protein that lacks the amino-terminal portion of the 55×10³AL_r protein (Celniker et al. 1989; Zavortink and Sakonju, 1989). Celniker et al. (1989) have provided evidence that two Abd-B proteins are translated in embryos. We and others have proposed that the SSxlfPAf,. protein provides the m function while the 30 × 10³M_T protein is responsible for the r function (DeLorenzi et al. 1988; Celniker et al. 1989; Zavortink and Sakonju, 1989).

A number of mutations that affect only a portion of the domain of the Abd-B morphogenetic (m) function have been identified (Lewis, 1978; Karch et al. 1985). These alleles, designated infra-abdominal, alter segmental identities in the region from PS 10 through 13. In fra-abdominal-5 (iab-5) mutations affect only PS 10, iab-6 mutations affect PS 10 and 11, and iab-7 mutations cause defects in PS’10, 11 and 12 (Karch et al. 1985; Duncan, 1987). In contrast, the m⁻r⁺ mutants mentioned above affect the entire region from PS 10 through PS 13. Abd-B mutants are unable to complement the iab-5, 6 and 7 mutants in trans. Thus the wild-type iab elements must be present in cis to obtain proper Abd-B activity. Molecular mapping of the DNA lesions associated with various iab alleles showed that breakpoints of a particular category of iab mutant are clustered (Karch et al. 1985; Duncan, 1987), and iab alleles affecting Abd-B are located close to the Abd-B transcription unit. These, and additional observations, have led to the hypothesis that iab mutations cause a disruption of cis-acting regulatory sequences controlling Abd-B expression (Karch et al. 1985; Casanova et al. 1987; Peifer et al. 1987).

Our current understanding of the mRNA and protein products of the Abd-B gene provides a means to test, at the molecular level, the genetic model of Casanova et al. (1986) and the proposal that the iab regions represent cis regulatory elements for Abd-B expression. In situ hybridization and antibody staining of wholemount Drosophila embryos were initially used to characterize the wild-type patterns of Abd-B mRNA and protein expression. We then examined the expression of the m (class A) and r (class B/C) transcripts in genetically m⁻r⁺, m⁺f⁻ and m⁻r⁻ mutant embryos. Our results provide the first conclusive evidence that the expression patterns of the two Abd-B proteins conform to predictions for the morphogenetic and regulatory elements of Casanova et al. (1986). In addition, a study of the mRNA and protein patterns in iab-6, and iab-7 mutants provides evidence that the associated DNA lesions affect the regulation of Abd-B transcription in a parasegment and cell-specific fashion.

Abd-B and iab mutant strains were provided by E. B. Lewis, I. Duncan, and G. Morata and are described in Karch et al. (1985), Casanova et al. (1986), Duncan (1987) and Celniker and Lewis (1987). 38000.11A is associated with an inversion with a breakpoint at genomic position +115 within the BX-C and has been classified as an iab-6 mutant (E. B. Lewis, personal communication). Uab^IrveB9 is an abd-A−Abd-B (m⁺r⁻) mutant isolated by M. Reagan and I. Duncan (personal communication) and is derived from Uab¹ (Lewis, 1978). The abd-A mutation was introduced to eliminate the dominant gain-of-function phenotype associated with the Uab^! breakpoint at map position —14 (Karch et al. 1985); the original m⁺r−phenotype is unaltered in Uab^lrcvb9.

To allow identification of homozygous mutant embryos, most mutant chromosomes were rebalanced over a TM3 chromosome carrying a ftz-lacZ P element insert (provided by S. Smolik-Utlaut). All embryos carrying one or two copies of the balancer will show β-gal expression in a pattern of seven stripes, corresponding to the normal ftz pattern. By double labeling embryos with an anti-β-gal antibody and the anti-Abd-B antibody, homozygous mutant embryos can be identified by the absence of the striped β-gal pattern. Tab or Uab^IrevB9 .chromosomes cannot be balanced over TM3, ftz-lacZ. Tab heterozygous adults are sterile in the absence of a chromosome carrying a duplication of the bithorax complex (Celniker and Lewis, 1987). The Uab^1revB9 chromosome contains markers which are incompatible with the TM3 balancer.

Whole mount in situ hybridization with digoxigenin-labeled probes was carried out according to Tautz and Pfeifle (1989). Probes were prepared using the random primer labeling method or using asymmetric PCR amplification. The PCR method was designed to produce predominantly the desired strand.

The probe that recognizes all Abd-B transcripts corresponds to residue numbers 2933 to 4053 of the B3 cDNA described in Zavortink and Sakonju (1989). The class A probe was derived from residues 508 to 1007 of the class A cDNA (the B3 cDNA clone), a portion transcribed from the class A specific exon. The class B probe was a TaqI–Nael fragment of the E19 cDNA, a region transcribed from the class B specific exon (located at about position +184 on the map shown in Fig. 1). The class C probe corresponds to an Spel–Xmal fragment from the class C specific exon, located at position + 188.

DNA fragments from a class A cDNA clone (B3 cDNA of Zavortink and Sakonju, 1989) were cloned into the appropriate form of the expression vector pWR590 (Guo et al. 1984) to produce in-frame fusions of Abd-B and /3-galactosidase coding sequences. A PvuII fragment extending from position + 1607 to +2485 was used to produce an antibody that recognizes both predicted Abd-B proteins. A PvuII-PstI fragment (+1607 to +1790) was used to generate an. antibody specific for the 55× 10³M_r m protein encoded by the class A mRNA. Fusion proteins were isolated from E. coli and injected into rats at the Pocono Rabbit Farm (Canadensis, PA). Two rats were injected with each of the Abd-B fusion proteins. Antibodies were purified from crude serum according to standard procedures (Carroll and Laughon, 1987). First, anti-j3-gal and anti-bacterial antibodies were removed by passage over column resin (CNBr-activated Sepharose, Pharmacia) coupled to proteins from an extract of E. coli carrying the pWR590 plasmid vector lacking an insert. The partially purified serum was then affinity purified using an anti-β-gal antibody column cross-linked to the larger Abd-B-β-gal fusion protein.

All four rat anti-Ai>d-B antibodies were able to detect Abd-B protein on Western blots and three were able to detect protein in embryos. However, only the common antibody from rat 2 stains embryos with a very low level of background staining.

Drosophila embryo extracts were prepared by homogenization of dechorionated 0 –12 h embryos in SDS gel sample buffer (0.1M DTT, 2% SDS, 0.08M Tris-HCl, pH6.9, 10% glycerol, 0.004 % bromophenyl blue) using 5 pi buffer per mg embryos (dry weight). Extracts were boiled 5 –10min and centrifuged 2 min prior to loading on 8.5 to 10 % SDS-polyacrylamide gels (Krause et al. 1988). E. coli extracts containing full-length Abd-B m or r protein were provided by S. Cumberledge. Proteins were dry blotted (Polyblot, American Bionetics) from gels to nitrocellulose or nylon membranes. Filters were incubated with primary antibodies at a dilution of 1:500 or 1:1000 in non-fat milk solution (20 mM Tris-HCl, pH7.5, 0.5M NaCl, 5% non-fat dry milk). After washing, filters were incubated with goat anti-rat antibodies conjugated with alkaline phosphatase (Jackson Immuno Research Laboratories) at a dilution of 1:10000. The enzyme reaction was carried out using substrates for the standard color reaction (Bio-Rad) or using the chemiluminescent substrate AMPPD (Tropix Inc.) for increased sensitivity. Chemiluminescent products were visualized by exposure of X-ray film. We estimate that the chemiluminescent method is at least 5 times more sensitive than the standard color reaction.

Detection of Abd-B proteins in Drosophila embryos was carried out according to the procedure described in Boulet and Scott (1988). Affinity-purified Abd-B antibodies were used at a dilution of 1:200 to 1:400. Anti-β-galactosidase antibodies (Boulet and Scott, 1988) were used at a dilution of 1:400. To establish the location of parasegment boundaries, embryos were stained with an anti-invected monoclonal antibody (referred to as the anti-engrailed antibody, provided by J. Kassis and D.-H. Huang, used at a dilution of 1:4). This monoclonal antibody recognizes the engrailed protein in the embryo. Fluorescein-conjugated goat anti-rabbit antibodies (Cappel) were used at 1:200. Rhodamine-conjugated and fluorescein-conjugated goat anti-rat antibodies (Jackson Immuno Research Laboratories) were diluted to 1:2000. Goat-anti-mouse antibodies conjugated with rhodamine (Boehringer Mannheim Biochemicals) were used at 1:400. All secondary antibodies were preabsorbed to fixed Drosophila embryos. A fluorescein-coupled anti-HRP antibody (provided by M. Bastiani) was used at a dilution of 1:10000.

Class A, B, and C Abd-B mRNAs are expressed in discrete spatial and temporal patternsSánchez-Herrero and Crosby (1988), DeLorenzi et al. (1988) and Kuziora and McGinnis (1988) have reported the in situ localization of transcripts from the Abd-B locus in wild-type embryos. These studies were carried out prior to the identification of the four transcript classes (Zavortink and Sakonju, 1989) and did not distinguish between transcripts arising from the class B and class C promoters. We have utilized the technique of hybridization to RNA in whole mount Drosophila embryos (Tautz and Pfeifle, 1989) to individually localize class A, B and C transcripts. Fig. 1 depicts the Abd-B gene structure and the locations of probes that distinguish between the class A, B and C mRNA species. To provide a description of expression patterns of transcripts described in Zavortink and Sakonju (1989), a brief summary of our results using wild-type embryos is given below with an emphasis on features not described in the previous studies.

Embryos, stained with a probe that recognizes sequences common to all Abd-B transcripts show that the domain of Abd-B expression encompasses PS 10-14 (data not shown; Sánchez-Herrero and Crosby, 1988; Kuziora and McGinnis, 1988). In cellular blastoderm stage embryos, the class A probe detects a ring of transcript in the primordium of PS 13 as determined by percent egg length (Fig. 2A; Sánchez-Herrero and Crosby, 1988; Kuziora and McGinnis, 1988). Class A transcripts are present at high levels in PS 13 and at lower levels in PS 11, 12 and 14 in a stage 12 embryo (Fig. 2B). A very weak hybridization signal is occasionally detected in PS 10. After germ band retraction, class A transcripts are detected in posterior A7 (anterior PS 13) and surrounding the developing posterior spiracles in A8 (Fig. 2C). In the ventral nervous system, class A transcripts are detected in PS 11 through PS 13 (Fig. 2D). The hybridization signal obtained with the class A probe in PS 14 (Fig. 2B, C, D) is much weaker than that seen with the probe recognizing all Abd-B transcripts. The PS 14 signal could be due to a low level of transcription from the class A promoter or to detection of precursor mRNAs transcribed from the class B or C promoters.

The pattern of transcription from the class B promoter has not been previously described. Using a probe specific for the class B transcripts, a strong hybridization signal is detected in PS 14 and anterior PS 15 of the extended germ band (Fig. 2E). In later embryos, a strong hybridization signal is seen in the terminal region of the ventral nervous system (Fig. 2G) and a much weaker signal is detectable in the epidermis in segment A9 (PS 14; Fig. 2F). Because the epidermal signal in later embryos is very weak, it is possible that this signal reflects hybridization to unspliced precursors of class C mRNAs. Thus, class B transcripts appear to be expressed exclusively or predominantly in the ventral nervous system of later stage embryos.

The class C probe detects a ring of transcripts in cellular blastoderm embryos at a position just posterior to the class A signal (Fig. 2H). Class C transcripts are present at high levels in the epidermis of PS 14 and anterior PS 15 in extended germ band embryos (Fig. 21) and, after germ band shortening, in a narrow band of cells in posterior A8 and anterior A9 (PS 14; Fig. 2J). The level of epidermal staining seen with probe C is much higher than that seen with probe B after germ band retraction (compare Fig. 2K and 2G, arrowheads). Hybridization in PS 14 of the ventral nervous system is also detected with the class C probe (Fig. 2K). The different distribution of class B and C transcripts suggest that they play different functional roles during development.

To provide direct evidence for the presence of the postulated Abd-B m and r proteins in vivo and to examine the distribution of these proteins in developing embryos, antibodies directed against Abd-B-β -galacto-sidase fusion proteins were prepared (see Materials and methods). A PvuII fragment (Fig. 1) was used to produce antibodies recognizing both protein species. A PvuII-PstI fragment (Fig. 1) was used to generate antibodies specific for the predicted Abd-B m protein. Purified fusion proteins synthesized in E. coli were injected into rats, and the antibodies from two rats injected with each antigen were affinity-purified using standard procedures (Materials and methods). The common and m-specific antibodies were tested on Western blots of E. coli extracts expressing full-length Abd-B m (55 × 10³M_r) or full-length Abd-B r (30x103M_r) protein (S. Cumberledge, unpublished). As expected, the common antibody recognizes both m and r proteins, and the m-specific antibody recognizes only the m protein (Fig. 3A).

The Abd-B common and m-specific antibodies were used to probe Western blots of wild-type embryo extracts. The common antibody reacts with a cluster of polypeptides of approximately 80-90 X10³M_r and a single protein of about 40 × 10³ M_r (Fig. 3B). A subset of the polypeptides in the higher relative molecular mass cluster is also detectable in embryonic nuclear extracts (data not shown). Transformant lines expressing the m protein (hsp70-M) or the r protein (hsp70-R) under control of the hsp70 promoter have been generated (A. Boulet, unpublished). To confirm that the protein species detected with our Abd-B antibody correspond to the endogenous m and r proteins, Western blots of extracts from heat-shocked hsp70-M embryos were probed with the common and m-specific antibodies. Both types of antibodies detect .a cluster of proteins migrating at the same position as the cluster seen in untransformed embryos (Fig. 3C). Likewise, the common antibody recognizes a protein in extracts of hsp70-R embryos comigrating with the smaller protein detected in untransformed embryos (Fig. 3C). The presumptive m and r proteins are present at much higher levels in heat-shocked hsp70-M and hsp70-R transformed embryos, respectively, than in heat-shocked control embryos (Fig. 3C). These observations demonstrate that the two clusters of embryonic proteins detected in the Western blots correspond to the predicted m and r proteins. The results also suggest extensive modifications of these proteins in vivo.

Since previous Abd-B protein localization studies used antibodies recognizing both Abd-B proteins, specific expression patterns for the two proteins could not be established (Celniker et al. 1989; DeLorenzi and Bienz, 1990). We have used our antibodies to determine the distribution of Abd-B m and r proteins in wild-type embryos. In stage 12 embryos, the m-specific Abd-B antibody detects a high level of protein in PS 13, but generates little or no signal in PS 14 (Fig. 4D). In embryos of approximately the same developmental stage, the common antibody detects Abd-B protein in PS 13 and PS 14 at approximately equal levels (Fig. 4A). The expected low level of m protein staining in PS 11 and 12 is partially obscured by the high background staining obtained with the m-specific antibody. In later stage embryos, the m-specific antibody detects a relatively high level of protein in PS 11 through PS 13, but little or no protein in PS 14 (Fig. 4E,F). In contrast the common antibody detects a high level of Abd-B protein in PS 14 (Fig. 4B,C) Thus, our data show that the Abd-B protein detected in PS 11-13 represents the 55x10³M_r protein encoded by class A transcripts, while the protein detected in PS 14 comprises mainly, or exclusively, the 30 ×10³ M_r protein translated from class B/C mRNAs.

The genetic model of Casanova et al. (1986) postulated two distinct Abd-B functions, m and r, that are inactivated individually by Class I (m⁻r⁺) or II (m⁺r^-) mutations, respectively, or simultaneously by Class III (m⁻r⁻) mutations. To assign the two genetic functions to Abd-B mRNA and proteins, whole mount in situ hybridization analysis and antibody staining were used to examine m⁻r⁺, m⁺r−and m⁻r⁻embryos.

First, we examined embryos carrying an m⁻r⁻mutation for Abd-B protein expression. When embryos from a heterozygous iab-7^D16 (Karch et al. 1985) stock were stained with the Abd-B antibody, protein was not detectable in approximately one quarter of the embryos. The embryos lacking Abd-B protein also show signs of aberrant development in the posterior regions (data not shown). Thus the complete absence of Abd-B proteins correlates with the loss of both m and r genetic functions.

iab-7^D14, an m⁻r⁺ mutation, is associated with a 411 bp deletion extending from —66 bp to +345 bp relative to the class A transcription initiation site (Fig. 1; Zavortink and Sakonju, 1989). This provides strong evidence that the class A transcript, and therefore the 55X10³M_r protein encoded by this transcript, corresponds to the m element of Casanova et al. (1986). In situ hybridization and antibody staining were used to determine whether mutant iab-7^Dn embryos indeed lack the m, or class A, mRNA and whether this correlates with a lack of protein in PS 10-13. In addition, the expression of the class B and C mRNAs, putatively encoding the r function, should remain unaltered in this mutant. To allow unambiguous identification of homozygous iab-7^D14 embryos, the mutant chromosome was balanced with a chromosome marked with aftz-lacZ P element insert (see Materials and methods). The homozygous embryos can also be recognized by failure of the posterior spiracles to form after germ band retraction (Fig. 5C). Our results show that Abd-B mRNA (Fig. 5A) and protein are absent from PS 10-13 but present at apparently wild-type levels in PS 14 of homozygous iab-7^D14 embryos (Fig. 5B,C,D; compare B to wild-type embryo in H). Thus the m function defined by Casanova et al. (1986) is provided by the class A transcript and its encoded 55x10³M_r protein.

Expression of Abd-B mRNA and protein was also examined in three mutants showing defects in r function. Tab and Uab^lrevB9 have been classified as m⁺r⁻mutants based on phenotype and the ability to complement the m⁻r⁺ allele iab-7^D14 (Casanova et al. 1986; Duncan, personal communication). These mutations affect development of the posterior spiracles and induce a partial transformation of PS 14 into PS 13 in homozygous mutant embryos causing formation of a rudimentary setal belt in A9 (Karch et al. 1985 ; Celniker and Lewis, 1987). iab-7⁶⁵ has been classified as an atypical m⁻r⁻Abd-B allele in which the r function is absent but the m function is not completely destroyed (Casanova et al. 1986; Sánchez-Herrero and Crosby, 1988). In fact, iab-7⁶⁵ is able to partially complement the m⁻r⁺ mutant iab-7^DI4 (A. Boulet, unpublished).

The patterns of Abd-B mRNA and protein expression in homozygous Tab, Uab^IrevB9 and iab-7⁶⁵ embryos are consistent with these mutations having an effect on the amount of r element activity. In embryos undergoing germ band shortening, Abd-B mRNA (data not shown) and protein are present at greatly reduced levels in PS 14 relative to wild-type levels (Fig. 5E,F,G; compare to 5H). The reduction of Abd-B mRNA and protein in PS 14 but not in PS10-13 supports the assignment of the r function to the SOxlCpA/,. protein encoded by class B and/or class C transcript.

Transcription of mRNA encoding the Abd-B m protein is derepressed in the absence of the r function In three mutants with genetic defects in r function, Abd-B expression was reduced in PS 14 but not entirely eliminated. This result may support the proposal of Casanova et al. (1986) that m activity is derepressed in PS 14 in the absence of r function. In the cases of Uab^lrevB9 and Tab, the close proximity of the associated breakpoints to the class B transcription start site raises the possibility that these mutations do not completely disrupt class B transcription (Fig. 1; Zavortink and Sakonju, 1989). In iab-7⁶⁵ embryos, however, the PS 14 mRNAs cannot be transcribed from the class B or C promoters since these transcription units have been disrupted by a breakpoint located about 6 kb upstream from the class A initiation site (Fig. 1; Zavortink and Sakonju, 1989). To directly test the hypothesis of Casanova et al. (1986), class A specific probes were used to determine whether m transcripts are present at increased levels in PS 14 of Uab^,revB9 embryos. Approximately one quarter of the embryos from a heterozygous Uab^,rcvS9 stock showed a higher level of class A transcripts in posterior PS 14 (anterior A9) than seen in wild-type embryos (Fig. 6B; compare to wildtype embryo in A). This evidence supports the hypothesis that class A or m expression is derepressed in PS 14 due to a mutation that affects r expression in PS 14.

The iab mutations were originally thought to disrupt individual ‘abdominal’ genes (Lewis, 1978). More recently, it has been proposed that iab-5, -6 and -7 alleles represent regulatory mutations of the Abd-B gene (Peifer et al. 1987; Casanova et al. 1987); for example, iab-7 lesions would destroy or remove sequences required for Abd-B expression in PS 10-12 and iab-6 and iab-5 lesions would affect PS 10 and 11 or PS 10 expression, respectively. To test whether expression of Abd-B protein and mRNA in iab mutant embryos fits the latter model, several iab mutant lines were examined by in situ hybridization and antibody staining. Again, mutant chromosomes were balanced with a marked chromosome to allow identification of homozygous mutant embryos.

Abd-B mRNA and protein expression were examined in the iab-7 mutant, Abd-B^MX2. This mutation is associated with a breakpoint located approximately 10 kb downstream of the Abd-B transcription units (Fig. 1; Karch et al. 1985). Whole mount in situ hybridization shows that Abd-B mRNA expression in Abd-B^MX2 homozygous embryos is restricted to PS 13-15 (Fig. 7A). Abd-B antibody staining demonstrates that Abd-B protein expression is also limited to PS 13-15 (Fig. 7B). Abd-B protein is not seen in PS 11 and 12 of the extended germ band in Abd-B^MX2 homozygotes. Abd-B protein is also clearly absent from PS 11 and 12 in the ventral nervous system where it is easily detected in wild-type embryos (Fig. 7C). These observations indicate that the regulatory elements affected in iab-7 mutations are required for proper expression of Abd-B in PS 10-12 of stage 12 embryos and the ventral nervous system (VNS) and epidermis of embryos after germ band retraction.

The Abd-B mutant SGA62 has been difficult to classify due to dominant gain-of-function effects caused by the chromosome rearrangement (Duncan, 1987). Karch et al. (1985) refer to SGA62 as an iab-6 allele. However, the SGA62 breakpoint maps between the breakpoints of two alleles, Abd-B^MX1 and Abd-B^MX2 (Karch et al. 1985), reclassified by Duncan (1987) as iab-7. The mRNA and protein patterns observed in SGA62 mutant embryos are essentially identical to those seen in Abd-B^MX2-. Abd-B protein and mRNA is not detected in PS 11 and 12 of homozygous embryos (Fig. 7D; data not shown). Ectopic Abd-B protein expression has not been detected in SGA62 embryos.

The mutations Camel (Karch et al. 1985) and 38000.11A have been classified as iab-6 alleles by Duncan (1987; personal communication). Embryos homozygous for these mutations show very similar patterns of Abd-B mRNA and protein expression (Fig. 7E,F,G and data not shown). The Abd-B gene is expressed in PS 12-15: mRNA and protein are not detectable in PS 11 at stages when expression is easily seen in wild-type embryos.

Embryos homozygous for the iab-5^C7 mutation (Karch et al. 1985) were examined by in situ hybridization and antibody staining. iabS⁰⁷ homozygous embryos show Abd-B protein expression in PS. 12-14 of the ventral nervous system; PS 11 expression is not detectable (Fig. 7H). Based on the absence of mRNA and protein in PS 11, as well as the location of the iab-5^e7 insertion with respect to other iab mutant lesions (Karch et al. 1985), we conclude that iab-5^C7 is an iab-6 rather than an iab-5 allele.

We have not attempted to examine mRNA and protein expression in true iab-5 mutants. Abd-B transcripts can be detected in PS 10 in only some wildtype stage 12 embryos. Therefore, the appearance of this staining is not consistent enough to reach a firm conclusion on the presence or absence of Abd-B expression in PS 10 of iab-5 mutant embryos.

We have examined the transcription patterns of mRNAs transcribed from three promoters of the Abd-B gene using the technique of whole mount in situ hybridization to Drosophila embryos. Previous studies of Abd-B transcript localization (Sánchez-Herrero and Crosby, 1988; DeLorenzi et al. 1988; Kuziora and McGinnis, 1988) were carried out before the three Abd-B transcription start sites had been identified. Our results, obtained with probes specific for each of the three classes of Abd-B mRNA, show that each class is expressed in a unique spatial and temporal pattern. Class A mRNAs are expressed in PS 10-13 while class B and C mRNAs are confined to PS 14 and anterior PS 15. The epidermal staining patterns of class B and class C transcripts are not identical. For example, the class C mRNA is expressed at much higher levels after germ band retraction than the class B mRNA. The sum of the individual class A, B and C patterns is very similar to the pattern seen with a probe that recognizes all Abd-B transcripts.

Analysis of the sequences of Abd-B cDNAs and the patterns of transcript expression have led to the hypothesis that the 55x10³M_r and 30x10³M_rAbd-B proteins carry out morphogenetic (m) and regulatory (r) functions, respectively (Sánchez-Herrero and Crosby, 1988; DeLorenzi et al. 1988; Kuziora and McGinnis, 1988; Celniker et al. 1989; Zavortink and Sakonju, 1989). We have tested the following aspects of this hypothesis: (1) the existence of the predicted m and r proteins in the embryo, (2) the localization of the m and r proteins to specific parasegments and (3) the correspondence between morphogenetic (m) and regulatory (r) mutant effects and expression of the Abd-B proteins.

To determine whether the two predicted Abd-B proteins are indeed expressed in the embryo, antibodies were generated against a 55x10³M_r protein-specific domain and against a domain common to both the 55x10³M_r and 30x10³ proteins. These antibodies recognize proteins of the predicted relative molecular masses in extracts of E. coli carrying Abd-B m (55 × 10³M_r) and r (30 ×10³M_r) constructs. Western blot analysis shows that two Abd-B protein products of different relative molecular masses are synthesized in wild-type Drosophila embryos. The larger protein has an apparent relative molecular mass much greater than that predicted from the DNA sequence. Since m and r proteins synthesized in E. coli migrate close to the expected relative molecular masses, this discrepancy is probably not due to aberrant migration in SDS polyacrylamide gels seen with a number of other homeodomain-containing proteins (Olio and Maniatis, 1987; Jack et al. 1988; Krause et al. 1988). Rather, the larger apparent relative molecular masses of Abd-B proteins on our Western blots may be due to post-translational modifications occurring in embryos. The appearance of proteins of 80 –90 ×10pAT and about 40x10³M_r in heat shocked hsp70-M embryos suggests that the presumptive modifications can occur in most, if not all, cells of the embryo. Celniker et al. (1989) reported detection of two protein products, of 55 × 10³M_r and 30× 10³M_r, in nuclear extracts of wildtype embryos using a monoclonal Abd-B antibody. The 55x10³M_r protein species detected on their Western blots may represent a partially degraded form of the m protein. Alternatively, the difference in the embryonic stages at which the extracts were prepared in the two studies could account for the discrepancy in the apparent relative molecular masses.

The pattern of staining in Drosophila embryos obtained with the antibody specific to the 55× 10³M_r protein clearly shows that this protein is present in PS13,but absent from, or present at very low levels in, PS 14,This result supports the assignment of the morphogenetic function, confined to PS 10-13, to the 55x10³M_r protein. We hereafter will refer to this protein as the m protein. The precise anterior border of m protein expression is difficult to determine because the antibody specific to the m protein gives a high background of nuclear staining in embryos. Since opa repeats, encoding Gin residues, are present throughout the m-specific domain, the high background may be due to cross-reactivity with opa repeats in other proteins. In contrast to the staining pattern obtained with the m protein-specific antibody, the antibody generated against a domain common to the 55xl(r M_T and 30 × 10³M_r proteins detects protein in PS 11 through PS 14. The staining patterns obtained with the antibodies recognizing the common Abd-B domain agree with those described by Celniker et al. (1989) and DeLorenzi and Bienz (1990). Taken together, the patterns seen with our two antibodies indicate that the 30× 10³M_r protein is expressed in PS 14. Thus the 30× 10³M_r protein must be responsible for the regulatory function, and will be referred to as the r protein. Although such an assignment has been suggested previously (Celniker et al. 1989; DeLorenzi and Bienz, 1990), it is only by demonstrating that the m protein is absent from PS 14, or present at very low levels, that the protein detected in PS 14 can be definitively identified as the 30× 10³M_r or r protein.

The results from antibody staining of m⁻r⁺, m⁺r⁻and m⁻r⁻Abd-B mutants support the hypothesis that the Abd-B gene consists of two distinct genetic elements (Casanova et al. 1986) which can be assigned to the m (55× 10³M_r) and r (30× 10³M_r) proteins. The lack of detectable Abd-B protein in m−r−mutant embryos strongly suggests that the phenotype of these mutants represents the null phenotype for the Abd-B gene. iab-7^D14(m⁻r⁺) homozygous mutant embryos clearly lack class A mRNA and m protein in PS 10-13. In these embryos, PS 14 expression of class B and C mRNAs and the r protein are unaffected. Conversely, in m⁺r⁻, mutants, the expression of Abd-B mRNAs and protein is greatly diminished in PS 14, while the expression of class.A mRNA and the m protein is not significantly affected in PS 10-13. These results provide a molecular genetic correlation of different Abd-B mutant classes with the products of the Abd-B gene and substantiate the model postulated by Casanova et al. (1986).

We have also shown that the presence of Abd-B mRNA and protein in PS 14 of embryos mutant for r function is partly or perhaps entirely due to derepress-sion of the class A transcript that encodes m protein. The increase in class A transcription in PS 14 can be interpreted in several ways. First, the class A promoter may be repressed by r protein in PS 14 of wild-type embryos; in the absence of r protein, the class A promoter would be derepressed. Alternatively, r⁻mutations may have removed or disrupted a cis-regulatory element that normally represses transcription of class A mRNA in PS 14. A third possibility is that DNA rearrangements associated with existing r” mutations juxtaposed sequences that activate the class A promoter or a cryptic promoter in PS 14. An effect of the juxtaposed DNA sequence in iab-7⁶⁵ can be observed in the increase of the m protein expression in PS 12. Aberrant expression of the m protein may explain why iab-7⁶⁵ behaves more like an m⁻r⁻, or null, Abd-B allele than an m⁺r⁻mutant.

The r function of Abd-B cannot as yet be assigned to any one of the class B, C or gamma transcript classes; in fact, each may contribute to r element activity. Our results indicate that an r⁻phenotype can be caused by a variety of molecular defects. For example, the lack of all three classes of transcripts, B, C and gamma, or the absence of only one class may produce a phenotype classified as r⁻. It is also possible that derepression of the m protein in PS 14 does not occur in all r⁻mutants. More careful studies of Abd-B mutant phenotypes may reveal differences between r⁻mutants that could be attributed to disruption of one or more of the class B, C and gamma transcription units, and may help to clarify the relationship between the class B, C and gamma transcripts and the genetic function of the r element.

The BX-C has been shown to consist of three lethal complementation groups (Sánchez-Herrero et al. 1985a, b; Tiong et al. 1985). These correspond to the three homeodomain protein-encoding genes, Ubx, abdA and Abd-B. Non-lethal bithorax complex mutations, such as bx, abx, pbx, bxd and iab-2 through iab-7 do not appear to disrupt additional BX-C genes. Mutations in these domains affect spatially distinct subsets of Ubx, abd-A or Abd-B functions, and the protein-coding transcription units must be intact to achieve wild-type bx, abx, pbx, bxd and iab activity. For these reasons, it has been postulated that these regions contain cis-elements that regulate the expression of the proteincoding BX-C genes. Based on mutant expression patterns in the Ubx domain, Peifer et al. (1987) have postulated that the bx, abx, pbx and bxd (or iab) regions consist of enhancers that activate the Ubx (or abd-A and Abd-B) promoter in cells at a specific position within each parasegment. The cell-specific enhancers are postulated to reside within parasegment-specific DNA domains. The parasegment-specific domains are thought to be activated at a particular site along the body axis and in all more posterior regions. The associated enhancers would then promote gene expression in a specific set of cells in each parasegment in which the DNA domain is activated. For the Abd-B gene, the enhancers located in the iab-5, iab-6 and iab-7 regions would be activated in PS 10-13, PS 11-13, and PS 12 and 13, respectively. Our data from the analysis of iab-6 and iab-7 mutants demonstrate the presence of parasegment-specific regulatory elements in the iab region of the Abd-B gene. Sánchez-Herrero and Akam (1989) have obtained similar results to those reported here for the iab mutant Abd-B^MX2. The associated DNA lesions clearly disrupt or effectively remove parasegment-specific regulatory elements that are required for expression of the Abd-B gene in PS 10, 11 and 12. A similar parasegment-specific loss of expression has been observed for the Ubx gene in the mutant abx² (White and Wilcox, 1985). However, our data does not provide any indication that a single enhancer controls expression in a subset of analogous cells in adjacent parasegments. According to the Peifer et al. (1987) model, equivalent cells expressing Abd-B under the control of the iab-6 enhancer in PS 11, 12 and 13 should lack Abd-B products in iab-6 mutant embryos. Instead we observe that Abd-B expression in PS 12 and 13 of iab-6 homozygotes appears to be unaltered while expression in PS 11 is entirely eliminated. Therefore, our results suggest that cell-specific enhancers are either reiterated in successive iab regions or that the iab regions contain only parasegment-specific regulatory elements which activate a common set of cell-specific enhancers.

The phenotypes of the iab mutants in adult flies corresponds fairly well with the patterns of Abd-B protein expression in homozygous iab embryos. Therefore, it seems that regulatory regions required for transcription in embryos in each parasegment lie close to, or are coincident with, the sequences required for the Abd-B expression in. larvae and/or pupae that directs proper development of adult structures.

A combination of genetic studies and molecular analysis has allowed elucidation of the relationship between the genetic functions and complex transcriptional structure of the Abd-B gene. The XlHboxl gene of Xenopus also utilizes separate promoters to produce short and long proteins containing the same homeodomain (Cho et al. 1988). The Hox3.3 genes of mouse and human are closely related to the Xenopus XlHboxl gene (Schughart et al. 1989). Our knowledge of the Abd-B gene may aid in the study of these and other complex genes of higher organisms.

We thank E. B. Lewis, I. Duncan and G. Morata for Abd-B and iab mutant fly strains, S. Smolik-Utlaut for the ftz-lacZ balancer line and J. Mahaffey and T. Kaufman for expression vectors. We thank J. Kassis, D.-H. Huang and M. Bastiani for providing antibodies. Advice on antibody purification and Western blotting from B. Appel and L. Boyd is greatly appreciated. We also thank D. Gard for providing hypersensitized Kodak Technical Pan film, S. Cumberledge for E. coli extracts, and C. Thummel, S. Cumberledge, M. Lamka and B. Appel for comments on the manuscript. Finally, we thank M. Zavortink for construction of the Abd-B fusions and initial fusion protein purifications.

Recently, S. Celniker, S. Sharma, D. Keelan and E. B. Lewis (EMBO J. in press) have also examined Abd-B expression in iab mutants. We thank Ed Lewis for communicating their results prior to publication.