Differential expression of Broad-Complex transcription factors may forecast tissue-specific developmental fates during Drosophila metamorphosis

Holometabolous insects use two morphologically distinct body plans to carry out different life functions. The relatively immobile larval form, with its prominent mouth parts and digestive system and its ability to shed its cuticle, is adapted for feeding and growing. The highly motile adult, with its light weight exoskeleton, prominent wings and reproductive systems, is specialized for dispersal and reproduction. The metamorphic transformation from the larval form into the adult involves sweeping structural modifications in virtually all tissues. Developmental examination of different larval tissues during this process in Drosophila suggests three categories of transformation. First, tissues such as salivary glands, muscle and gut are completely histolyzed (Robertson, 1936; Bodenstein, 1950). Second, other tissues survive into the adult phase but undergo reorganization. For example, the larval fat body disperses early in metamorphosis and then reaggregates in the forming adult (Rizki, 1978), the central nervous system (CNS) undergoes extensive growth and remodeling (Truman, 1990), and the Malpighian tubules undergo slight changes in structural composition (Bodenstein, 1950; Wessing and Eichelberg, 1978). Third, predetermined imaginal tissues differentiate and grow into adult organs and external structures (Robertson, 1936; Bodenstein, 1950; Fristrom and Fristrom, 1993).

This ensemble of tissue-specific changes is thought to be induced by pulses of the steroid hormone ecdysone which are transduced into tissue-specific transcriptional cascades (Ashburner et al., 1974; Burtis et al., 1990; Thummel et al., 1990). In general, this begins with the hormone-dependent activation of a set of transcription factors (early genes) that, in turn, activate sets of tissue-specific effector genes (late genes) thought to be responsible for organ-specific metamorphic outcomes. The molecular processes by which this systemic hormone signal elicits tissue-specific responses are not yet fully understood.

The Broad-Complex (BR-C) is one of the ecdysone-induced early genes and plays key roles in both coordinating the ecdysone response among tissues and in specifying the genetic composition of effector gene sets. Nonpupariating mutations (npr1 alleles) are null mutations that result in developmental arrest and lethality at the onset of metamorphosis indicating that the BR-C is essential for metamorphosis but not for prior developmental events. The npr1 mutations disrupt all BR-C functions (see below) because they are phenotypically identical to BR-C deletion mutants (Belyaeva et al., 1980, 1982; Kiss et al., 1988).

Other BR-C mutations can be grouped into three complementation groups [broad (br), reduced bristles on palpus (rbp), 2Bc] and may encode distinct BR-C functions. Most alleles result in lethality somewhat later in metamorphosis and exhibit developmental defects that are more specific than those caused by npr1 alleles. For example, salivary gland development is profoundly affected in rbp mutant animals. The gland fails to undergo programmed cell death and the tissue persists throughout metamorphosis and can be found in the head and thorax of pharate adults (Restifo and White, 1992). At the molecular level, transcriptional activation of many salivary gland-specific late genes is completely dependent on the rbp⁺ function (Guay and Guild, 1991; Karim et al., 1993). In addition, the evagina-tion of imaginal discs is dependent on the br⁺ function (Kiss et al., 1988) and remodeling of the CNS is dependent on all three BR-C functions (Restifo and White, 1991).

The BR-C is also complicated at the molecular level. The locus is organized into several overlapping but alternatively spliced transcription units (DiBello et al., 1991; C. Bayer and J. Fristrom, personal communication) and is devoted to the synthesis of an extensive family of RNAs (Chao and Guild, 1986; Karim and Thummel, 1992; Karim et al., 1993). Conceptual translation of BR-C cDNA clones indicates that the corresponding RNAs encode a family of related protein isoforms (DiBello et al., 1991; C. Bayer and J. Fristrom, personal communication) that can be distinguished by zinc finger domains. All proteins contain a common amino-terminal 431 amino acid region (designated BRcore) fused to alternatively spliced carboxy-terminal ends containing distinct C₂H₂ zinc finger pairs (designated Z1, Z2, Z3 and Z4). These proteins have been shown to bind DNA in a sequence-specific manner (von Kalm et al., 1994). Most BR-C proteins also contain domains rich in selected amino acids (Q, TNT, NS) located between the BRcore region and the zinc fingers. Domains of this type are common in eukaryotic transcription factors (Mitchell and Tjian, 1989) and coactivators (Hoey et al., 1993; Colgan et al., 1993).

Using specific antibodies, we show that BR-C proteins are localized in nuclei. In salivary glands, BR-C proteins are localized to numerous loci on the polytene chromosomes. Using isoform-specific detection strategies, we show that BR-C isoforms display tissue-specific distributions suggesting a causal mechanism for the generation of tissue-specific metamorphic fates. Finally, analysis of protein expression in BR-C mutants has allowed for the correlation of individual isoforms with specific BR-C genetic functions.

The BR-C mutant alleles used in this study are listed in Table 1. The rbp² and br¹ alleles are viable and were maintained as homozygotes. The br²⁸ allele was maintained in males in combination with the duplication Dp(1;Y)y²67g in a stock with compound-X females. All other alleles were maintained in females in combination with the Binsn balancer X chromosome. Overlapping deficiencies within the Dp(1;Y)Sz280 and the Df(1)S39 chromosomes were used to construct BR-C null trans-heterozygote males (Belyaeva et al., 1980; 1982) deficient for all known BR-C promoters and exons (Chao and Guild, 1986; Belyaeva et al., 1987).

Larvae were raised on standard cornflour food (Elgin and Miller, 1978) containing 0.05% (w/v) bromophenol blue in an incubator maintained at 25°C, 60-70% RH, with a 12 hour/12 hour day/night cycle. As late third instar animals stop eating and begin the wandering stage, the blue dye gradually clears from their intestine (Maroni and Stamey, 1983). This method can be used to stage late third instar larvae (Boyd et al., 1991; Andres and Thummel, 1994). Larvae with completely blue guts (end-to-end) are approximately 18 hours away from pupariation (range 12-24 hours) and have not been exposed to the major late third instar ecdysone titer peak. Larvae with completely clear guts are approximately 4 hours away from pupari-ation (range 1-6 hours) and have been exposed to the major ecdysone titer peak (Andres and Thummel, 1994). Animals were also synchronized at the brief white prepupal stage (Bainbridge and Bownes, 1981) and allowed to develop for the indicated times in humid chambers at 25°C.

Fusion proteins were used as immunogens to produce both polyclonal and monoclonal antibody reagents. Appropriate restriction fragments from BR-C cDNA clones were cloned into the bacterial expression vector pDS-MCS (Schindler et al., 1992) to yield BR-C proteins with amino terminal (His)₆ tags. Plasmids were grown to early log phase (A₆₀₀=0.7-0.9) in E. coli M15 cells (Zamenhof and Villarejo, 1972) transformed with pDMi (lacI^q) and induced with 1 mM IPTG for 2 hours. Proteins were made soluble in 6 M guanidine hydrochloride, applied to a Ni²⁺ NTA resin (Qiagen) and eluted in decreasing pH steps in the presence of 8 M urea. The (His)₆-BRcore, (His)₆-BRcore-Q-Z1 and (His)₆-BRcore-Z2 fusion proteins eluted at pH 6.3. The (His)₆-Z3 fusion protein eluted at pH 4.5. Purified proteins were dialyzed against 10 mM Hepes (pH 7.9), 80 mM KCl, 20% glycerol, 1 mM DTT, 10 μM ZnCl₂, 0.1% Triton X-100. Approximately 1-2 mg of protein was obtained from a 500 ml bacterial culture.

For the production of anti-BRcore polyclonal antibodies, 20 μg aliquots of purified (His)₆-BRcore (Fig. 1) were used to immunize and boost two Balb/c female mice, via alternating subcutaneous and intraperitoneal routes. Polyclonal ascites fluid was obtained by drainage one to two weeks after injection of 5×10⁶ Sp2/0-Ag14 myeloma cells (University of Pennsylvania, Cell Center) into the peritoneal cavity (Harlow and Lane, 1988).

For the production of anti-BRcore, anti-Z1 and anti-Z3 monoclonal antibodies, 20 μg aliquots of purified (His)₆-BRcore-Z2, (His)₆-BRcore-Q-Z1 and (His)₆-Z3 (Fig. 1) were used, respectively, to immunize and boost two 6-week-old Balb/c female mice (each immunogen), via alternating subcutaneous and intraperitoneal routes. The response against the BRcore domain was immunosuppressed in mice immunized with (His)₆-BRcore-Q-Z1 by prior injection of (His)₆-BRcore followed by 40 mg/kg body weight of cyclophosphamide (Matthew and Patterson, 1983). Hybridomas were generated by fusing spleen cells to Sp2/0-Ag14 myeloma cells essentially as described by Lane et al. (1986). Hybridoma supernatants (400-800 per immunogen) were screened by enzyme-linked immunosorbent assay for reactivity to the immunizing fusion protein. Positive clones (approximately 5%) were rescreened by probing western blots containing the fusion proteins. Clones that were again positive (approximately 50%) were tested on western blots against Drosophila protein preparations. Hybridoma lines were cloned by limiting dilution. Screening 697 hybridoma lines from a (His)₆-BRcore-Z2 immunization revealed no clones that specifically recognized the small Z2 domain.

Both monoclonal (25E9) and polyclonal anti-BRcore antibodies recognize a protein containing most of the BRcore domain (aa30-aa423; numbering system of DiBello et al., 1991) but do not recognize protein fragments containing the Z1, Z2 or Z3 domains. These reagents were effective in western blot, whole-mount and polytene chromosome immunostaining assays. The anti-Z1 monoclonal antibody (3C11) recognizes a protein fragment unique to the Z1 isoforms that lies between the BRcore and the finger region (aa432-aa572) but does not recognize the BRcore, Z2 or Z3 domains. This reagent was effective in western blot assays. The anti-Z3 monoclonal antibody (9A7) recognizes a protein fragment that includes the Z3 finger domain (aa516-aa704) but does not recognize the BRcore, Z1 or Z2 domains. This reagent was effective in western blot, whole-mount and chromosome immunostaining assays.

We modified the protocol of Boyd et al. (1991) to immunostain dissected and fixed tissues with anti-BR-C antibodies. In this case, (1) the blocking step was omitted, (2) the anti-BRcore polyclonal ascites fluid was diluted 1/200 in TN3 buffer [50 mM Tris-HCl (pH 7), 0.15 M NaCl, 0.5% Triton X-100, 2% normal goat serum (Sigma)] and incubated at room temperature for 3 hours, and (3) the goat antimouse horseradish peroxidase-conjugated secondary antibodies (Cappel) were diluted 1/500 and incubated at room temperature for 1 hour. Peroxidase activity was detected in a 1 mg/ml diaminobenzi-dine (Sigma) solution. Stained organs were mounted on slides in glycerol, viewed by light microscopy and photographed using Ektachrome 160T color slide film (Kodak).

Protein extracts from whole animals were prepared by homogenizing 3-10 staged larvae or prepupae in 30-100 μl 2× SDS sample buffer [100mM Tris-HCl (pH 6.8), 4% SDS, 0.2% bromophenol blue, 20% glycerol and 200 mM DTT (added fresh)] and a mixture of several protease inhibitors [0.1 mg/ml pepstatin (Sigma), 0.5 mg/ml leupeptin (Sigma), 10 mM PMSF (Sigma)]. The extracts were boiled (5 minutes) and microfuged (5 minutes) prior to electrophoresis. Each gel lane was loaded with protein extract corresponding to 0.5 animal. Protein extracts from dissected tissues were prepared from 6-10 animals essentially as described above except the protease inhibitor and centrifugation steps were omitted. Each gel lane was loaded with protein extract corresponding to 6-8 animals.

Protein extracts were electrophoresed on 7% polyacrylamide SDS-gels essentially as described (Sambrook et al., 1989) and transferred to Immobilon-P membranes (Millipore) with a Genie blotter (Idea Scientific). The membranes were blocked (10% nonfat dried milk, 1× PBS, 0.1% Tween-20) for 1 hour at room temperature or overnight at 4°C. Incubation of monoclonal anti-BR-C hybridoma supernatants (used undiluted) was performed at room temperature for 30-60 minutes. Blots were rinsed 3 times (5 minutes each) in 1× PBS, 0.1% Tween-20. Incubation of horseradish peroxidase-conjugated goat antimouse antibodies (Cappel), 1/5000 dilution in blocking solution, was performed at room temperature for 1 hour. Blots were rinsed as before. Localization of the enzyme-conjugated antibodies was achieved using the Luminol detection system (Amersham) and RX-ray film (Fuji).

Antibody staining was performed following the procedure by Silver et al. (1978) as detailed in Ashburner (1989) with some modifications in incubation times. Salivary glands were dissected from late third instar larvae in medium G containing 0.5% NP-40 and incubated in this medium for 5 minutes at room temperature. Glands were transferred and incubated in formaldehyde fixative for 10-30 seconds, transferred to a cover slip and incubated in squashing solution for 5 minutes. After picking up the glands with an inverted microscope slide, the chromosomes were spread by tapping the coverslip with a pencil tip. The slides were frozen at −70°C until use. The coverslips were flicked off with a razor blade and the slides were rinsed in PBS containing 0.1% Tween-20 (PBT). Chromosomes were incubated with undiluted anti-BRcore monoclonal antibody for 1 hour at room temperature in a humid chamber and then rinsed 3 times (5 minutes each) with PBT. Fluorescein-conjugated goat anti-mouse secondary antibodies were applied at a 1/500 dilution in TN3 buffer. After a 1 hour incubation at room temperature, slides were washed as before and counter stained for 30 seconds with Hoescht 33258 stain (12.5 ng/ml). The slides were rinsed once more and mounted in mounting medium [7.5 ml glycerol, 12 ml 0.2 M Tris-HCl (pH 8.6), 6 ml H₂O, 2.4 g polyvinyl alcohol (Sigma), 3% DABCO (1,4 diazobicyclo-[2,2,2]-octane)] (Harlow and Lane, 1988). Chromosomes were viewed by immunofluoresence microscopy and photographed with T-MAX 400 black and white print film (Kodak).

The BR-C is organized into a series of overlapping transcription units devoted to the synthesis of a family of related proteins. These isoforms can be distinguished by the presence of distinct C-terminal domains each containing a different zinc finger DNA-binding domain (Fig. 1). We expressed portions of these proteins in E. coli and used them as immunogens to generate antibody reagents. We were successful in generating specific polyclonal and monoclonal antibodies directed against the BRcore domain and isoform-specific monoclonal anti-bodies (mAb) directed against the Z1 and Z3 domains.

The specificity of these reagents was tested on western blots containing fractionated preparations of BR-C proteins expressed and purified from E. coli (Fig. 2A). Each of these proteins contains an N-terminal tag of 6 histidines which allows purification using a nickel-chelate column. The anti-BRcore mAb binds to three full-length proteins, (His)₆-BRcore-Q-Z1, (His)₆-BRcore-Z2 and (His)₆-BRcore-NS-Z3, suggesting that this antibody recognizes the common BRcore region. This was confirmed by showing antibody recognition of (His)₆-BRcore protein but not of (His)₆-tagged Z1, Z2 or Z3 proteins (data not shown). The anti-BRcore polyclonal serum exhibits identical specificity by these same criteria (data not shown). The anti-Z1 and anti-Z3 mAbs recognize the (His)_6- BRcore-Q-Z1 and (His)_6-BRcore-NS-Z3 isoforms, respectively. Since neither reagent recognizes the (His)_6-BRcore-Z2 isoform (Fig. 2A) or a (His)₆-BRcore protein (data not shown), we conclude that these reagents exhibit isoform specificity.

These reagents were then tested for their ability to recognize the endogenous Drosophila BR-C proteins. Because the BR-C gene is an early ecdysone inducible gene, we probed western blots containing protein extracted from wild-type late third instar larvae after the rise of ecdysone titer. As negative controls, we used similarly staged BR-C deletion animals [male Df(1)S39/Dp(1;Y)Sz280 transheterozygotes] (not shown) and BR-C null animals (npr1³ hemizygotes) (Fig. 2B). All anti-BR-C monoclonal reagents recognize proteins in the wild-type lanes but do not cross-react with other proteins in the BR-C null lanes. The anti-BRcore mAb recognizes five protein bands labeled according to their apparent molecular weight (p118, p91, p81, p64 and p57). The anti-Z1 mAb recognizes a subset of two protein bands, p91 and p81, while the anti-Z3 mAb recognizes only the p91 band.

We compared the apparent molecular weights of these Drosophila proteins with the apparent molecular weights of BR-C proteins expressed in E. coli and found close correspondence (Figs 1 and 2). The protein bands recognized by the anti-Z1 mAb (p81-Z1 and p91-Z1) correspond to and comigrate with (His)₆-BRcore-Q-Z1 and (His)₆-BRcore-TNT-Q-Z1 (not shown) isoforms produced in E. coli. The protein band recognized by the anti-Z3 mAb (p91-Z3) is similar in size to the (His)₆-BRcore-NS-Z3 isoform produced in E. coli. The anti-BRcore mAb recognizes three additional protein bands. Of these, p64 and p57 are very likely to contain the Z2 domain because both bands are absent in a P-element-mediated BR-C mutation where the insertion occurs in the Z2 exon (see below and Fig. 3). Moreover, p57-Z2 comigrates with the (His)₆-BRcore-Z2 isoform produced in E. coli. The p64-Z2 protein may represent a previously unidentified member of the Z2 isoform family that includes a homopoly-meric region located between the BRcore and the zinc finger regions. The highest molecular weight protein recognized by the anti-BRcore mAb (p118) is not associated with a Z1, Z2 or Z3 finger domain and may represent the newest member of the BR-C protein family, BRcore-Z4, which was recently characterized (C. Bayer and J. Fristrom, personal communication).

In order to correlate individual BR-C functions with specific isoforms, we examined mutants for abnormal isoforms. Protein extracts from late third instar larvae (after the major ecdysone pulse) carrying BR-C mutations representing the npr1, rbp, br and 2Bc complementation groups were displayed on western blots. Specific isoforms were distinguished using a combination of mAb detection reagents and gel mobility criteria.

In agreement with their characterization as null mutations, we found little or no BR-C protein in any npr1 mutation (Fig. 3). The npr1¹ and npr1³ alleles and the BR-C deletion trans-heterozygotes (not shown) produce undetectable levels of Z1, Z2 and Z3 isoforms. The npr1⁶ and npr1⁷ alleles produced significant amounts of apparently truncated products.

The BR-C rbp⁺ function is associated with the Z1 isoforms. Examination of five rbp mutations revealed one, rbp⁵, to be deficient in both Z1 isoforms with the concomitant appearance of a novel truncated product. Similarly, we find the BR-C br⁺ function is associated with the Z2 isoforms. Examination of three br mutations revealed one, br²⁸, which has been characterized as a P element insertion in the Z2 exon (C. Bayer and J. Fristrom, personal communication), to be deficient in the Z2 isoforms with the concomitant appearance of another novel truncated protein. Interestingly, this mutation appears to lead to underexpression of the Z1 and Z3 isoforms, suggesting that the BR-C has autoregulatory features. Examination of a single mutant representative of the 2Bc class showed no clear differences in the level of any BR-C isoform.

Since all BR-C isoforms contain a pair of C₂H₂ zinc fingers (DiBello et al., 1991; C. Bayer and J. Fristrom, personal communication) and bind to DNA in a sequence-specific manner (von Kalm et al., 1994), one expects BR-C proteins to accumulate in nuclei. In agreement with this, we find BR-C proteins localized to the nuclei of numerous tissues during late third instar and early prepupal development. We used anti-BRcore polyclonal antiserum to detect all BR-C isoforms by immunostaining wildtype tissues dissected from late third instar larvae (approximately 18 hours prior to pupariation), white prepupae and white prepupae aged 5 hours (Fig. 4). BR-C proteins were preferentially localized to the nuclei of all tissues tested. Tissues derived from BR-C null (npr1³) late third instar larvae showed no staining and served as a negative control in all cases (Fig. 4A,E,I,M,Q).

We examined larval tissues with different metamorphic fates including tissues that will undergo histolysis in response to ecdysone (salivary glands and gut), tissues that will undergo reorganization (fat body and CNS) and tissues that will differentiate (imaginal discs). In salivary glands, the large polytene nuclei show heavy staining that is most prominent after the major ecdysone pulse (Fig. 4C,D). In addition, cells comprising the imaginal ring, which ultimately gives rise to the adult salivary gland, also contain BR-C protein at all stages examined (Fig. 4B-D). Immunostained nuclei in the CNS can be localized to cells of the brain lobes and the ventral ganglion (Fig. 4N-P). In these CNS whole mounts, moreover, the ring gland is also immunostained (Fig. 4N,O). In the wing imaginal disc, nuclear staining is detected in cells in both the disc proper and the surrounding peripo-dial epithelium (Fig. 4J-L). BR-C proteins were also detected in the fat body (Fig. 4F-H), the gut (Fig. 4R-T), the Malpighian tubules and in the male testis where both spermatogonia and spermatocytes were immunostained (not shown).

Looking at staining pattern and intensity differences among the three time points analyzed, it is clear that BR-C proteins are present before the major pulse of ecdysone and their concentration often increases after this pulse. This increase is especially noticeable in salivary glands (Fig. 4B vs. 4C). This increase in BR-C protein concentration also agrees with western blot experiments (see below) and with previous data showing that BR-C mRNAs are present throughout this developmental period but their concentration increases after the ecdysone pulse in whole animals (Karim and Thummel, 1992; Karim et al., 1993; Andres et al., 1993) and in salivary glands (Chao and Guild, 1986; Huet et al., 1993).

To determine whether nuclear BR-C proteins are localized to chromatin, we immunostained salivary gland polytene chromosomes using an anti-BRcore monoclonal antibody. Polytene chromosomes from both wild-type and BR-C null (npr1³) animals were squashed and stained on the same slide. While both chromosome sets were visualized using the DNA stain Hoescht 33258 (Fig. 5B), only the wild-type chromosome set contained BR-C proteins (Fig. 5A, left). These proteins are localized to over 200 sites with varying degrees of intensity. BR-C protein is primarily associated with interbands, although not all interbands immunofluoresced. No protein could be detected in the centromere region or on the fourth chromosome. A similar distribution pattern was also observed using the anti-BRcore polyclonal serum. A cytogenetic listing of the BR-C-binding sites will be published elsewhere. Our result shows that BR-C proteins synthesized in vivo bind to chromosomes as expected from the in vitro studies of von Kalm et al. (1994).

To test whether different BR-C isoforms exhibit distinct temporal accumulation patterns, we used isoform-specific antibody reagents to probe western blots containing proteins from late third instar, prepupal and early pupal animals. Accumulation of the Z1 and Z3 isoforms could be measured directly with isoform-specific mAbs while accumulation of Z2 isoforms was deduced from the accumulation of the p57-Z2 and p64-Z2 isoforms detected on blots probed with anti-BRcore mAb. All BR-C isoform families are expressed during this early metamorphic period (Fig. 6).

In agreement with whole animal mRNA accumulation studies (Karim and Thummel, 1992; Karim et al., 1993; Andres et al., 1993; C. Bayer and J. Fristrom, personal communication), BR-C proteins are detected at low levels prior to the major pulse of ecdysone (−18 hour time point). Although little protein can be detected in these whole animal western blots, BR-C signal is clearly visible in similarly staged tissue whole mounts (Fig. 4B,F,J,M,P). Following the hormone pulse that occurs approximately 6 hours prior to pupariation, the Z3 and the Z2 isoforms are induced and reach maximum levels by the time of puparium formation (0 hour). Z3 isoform accumulation is brief and decreases to almost preinduction levels by 2 hours after pupariation while the Z2 isoforms persist much longer and disappear by 10 hours after pupariation. In contrast, Z1 isoforms are first induced at puparium formation (0 hour), attain maximum levels for a 10 hour period (4-14 hours after pupariation), then diminish to preinduction levels in the early pupal period. This Z1 profile closely approximates that of the corresponding BR-C 4.5 Kb mRNA size class (Chao and Guild, 1986) but with a 3-4 hour delay suggesting that additional post-transcriptional events are required prior to protein accumulation.

To test whether different BR-C isoforms are expressed in tissue-specific formats, we performed developmental western experiments using blots containing proteins from individual tissues. Protein extracts were prepared from dissected salivary glands, imaginal discs and CNS isolated from late third instar larvae and prepupae aged up to 8 hours. As before, individual isoforms were distinguished using a combination of mAb detection reagents and gel mobility criteria.

Salivary glands predominantly express Z1 isoforms (Fig. 7). In contrast to the whole animal developmental profile, late third instar salivary glands contain a substantial amount of Z1 protein (−18 hour) which is induced to higher levels after the major ecdysone pulse (−4 hour larvae and 0 hour prepupae) and then decreases in concentration. This protein developmental profile agrees well with previous salivary gland BR-C mRNA developmental profiles (Chao and Guild, 1986; Huet et al., 1993).

We detect little, if any, Z3 isoform in salivary glands during the late third instar period. This was unexpected because Z3 transcripts are easily detected during the mid to late third instar interval (Huet et al., 1993; von Kalm et al., 1994). It is possible that Z3 protein levels fall below our level of detection because Z3 transcript levels are declining during late third instar development (von Kalm et al., 1994) and are undetectable at puparium formation (Huet et al., 1993). Alternatively, production of the Z3 encoding isoform in late third instar salivary glands may be subject to additional post-transcriptional regulatory events.

Imaginal discs switch from the production of Z2 isoforms to Z1 isoforms after puparium formation (Fig. 7). The production of Z2 isoforms in discs reflects the profile seen in whole animals. The low levels of p64-Z2 and p57-Z2 detected in late third instar animals (−18 hour) are stimulated to maximum levels after the major ecdysone pulse (−4 hour), remain steady for several hours (0, 2, 4 hour prepupae) and then begin to decrease in concentration (8 hour prepupae). The p57-Z2 isoform is the dominant Z2 isoform at all times. Accumulation of the Z1 isoforms is also induced after the ecdysone pulse (−4 hour), remains steady for several hours (0 and 2 hour prepupae) and then increases in concentration (4 and 8 hour prepupae). In addition, there is a dramatic increase in the p91-Z1 accumulation in 8 hour prepupal discs. Thus, the Z1/Z2 isoform ratio shifts during prepupal development with the Z2 isoforms predominating at puparium formation and the Z1 isoforms predominating 8 hours later. A contribution of the Z3 isoform to the p91 band in the 8 hour disc preparation was ruled out by probing similar preparations with anti-Z1 and anti-Z3 monoclonal antibodies. These experiments showed virtually no Z3 expression in 0 hour discs (Fig. 7B) or in 8 hour discs (not shown).

Wing discs represent a large percentage of total disc protein. In order to test whether this Z1/Z2 ratio shift occurs in smaller discs, we repeated the developmental western with both wing discs and the set of non-wing discs. In both cases, the same ratio shift took place (not shown). We conclude that this developmental switch in isoform expression probably occurs in most discs.

Prepupal CNS development is accompanied by an increase in the concentration of the Z3 isoform (Fig. 7). Late third instar (−18 hour) CNS preparations contain relatively low levels of the Z1, Z2 and Z3 isoforms. All seem to be stimulated to higher levels after the major ecdysone pulse (−4 hour larvae and 0 hour prepupae). While the levels of Z1 isoforms and Z2 isoforms (p57-Z2 predominating) seem to remain steady during prepupal development (2-8 hour prepupae), the Z3 isoform shows substantial concentration increases during this period. Thus, the Z3 isoform becomes the predominant BR-C isoform during prepupal CNS development.

The BR-C coordinates the ecdysone response in many tissues. Most BR-C mutations result in developmental arrest during metamorphosis (e.g., Table 1). This failure to complete meta-morphosis is probably due to the summation of many tissue-specific defects (see Introduction). In agreement with this hypothesis, we find that BR-C proteins are induced during third instar development after the major pulse of ecdysone (Fig. 6) and accumulate in the nuclei of most, if not all, tissues early in metamorphosis (Fig. 4) suggesting that the BR-C temporally coordinates ecdysone-regulated gene expression in each of these tissues (Burtis et al., 1990; Thummel et al., 1990).

Several lines of evidence indicate that two BR-C functions play important roles in regulating salivary gland gene expression. First, rbp mutations delay intermolt gene activation in mid-third instar larvae and completely block late gene activation in prepupae (Guay and Guild, 1991; Karim et al., 1993). Second, 2Bc mutations can delay both the activation and repression of mid-third instar intermolt gene expression (Karim et al., 1993) and the wild-type activity is required for expression of the hsp23 and hsp28 genes (Dubrovsky and Zhimulev, 1988). In addition, 2Bc⁺ function coordinates the intermolt and late gene transcriptional cascades by providing the competence for other early gene promoters to be fully activated (Karim et al., 1993). Evidence that BR-C-dependent regulation is mediated directly by BR-C isoform-DNA binding was recently obtained for the intermolt target gene Sgs-4 (von Kalm et al., 1994).

Taken together, these results suggest that the BR-C regulates several different temporal classes of salivary gland genes and is consistent with the observation that a large number of BR-C protein-binding sites can be visualized on the polytene genome (Fig. 5). Interestingly, the number and distribution of BR-C protein-binding sites does not parallel that of two other early gene proteins, E74A and E75A, that bind to a moderate number of sites (70 in the case of E74A) and most of these correspond to ecdysone-induced loci (Urness and Thummel, 1990; Hill et al., 1993). BR-C proteins bind to approximately 200 sites suggesting that BR-C isoforms function as metamorphosis-specific general transcription factors required for the activation and/or repression of many genes.

The BR-C is known to regulate expression of specific genes in other tissues including the larval fat body (Lepesant et al., 1986; Nelson et al., 1991; K. Crossgrove and GMG, unpublished observations) and we would predict loci-specific chromosomal localization of BR-C proteins in this tissue as well. In agreement with this idea, we see localization of BR-C proteins to numerous loci on the fat body polytene genome (L. Wagner, IFE and GMG; unpublished observations).

The three groups of complementing BR-C mutations [rbp, br and 2Bc] appear to affect different BR-C functions. There is a strong genetic correlation between the br⁺ function and the Z2 isoforms based on mapping a P element insertion site in the br²⁸ mutation to the Z2 encoding exon (DiBello et al., 1991). In agreement with this data, we detect no wild-type Z2 isoforms in br²⁸ animals (Fig. 3). Instead, an apparently truncated protein product is made. The molecular weight of this protein (55.5 kDa) coupled with the map position of the P element insert is consistent with the idea that the novel br²⁸ product(s) is a BR-C/P element fusion protein lacking the Z2 zinc finger DNA-binding domain (C. Bayer and J. Fristrom, personal communication).

We show a strong biochemical correlation between the rbp⁺ function and Z1-containing isoforms. We detected no wildtype Z1 isoforms in rbp⁵ animals (Fig. 3). As before, a novel and apparently truncated protein product accumulates. The molecular weight of this protein (53.5 kDa) in combination with the absence of the epitope recognized by the anti-Z1 mAb suggests that the novel rbp⁵ product(s) is prematurely terminated approximately 50 amino acids after the BRcore domain and contains no DNA-binding domain. If the loss of the DNA-binding domain inactivates a BR-C isoform, we would predict that complete loss of all functional Z1 or Z2 isoforms in rbp⁵ and br²⁸ mutants, respectively, would lead to the complete loss of the associated genetic functions.

We have no strong biochemical evidence correlating the 2Bc⁺ function with any isoform. However, close examination of the mobility in polyacrylamide gels of the Z3 isoform synthesized in 2Bc¹ animals shows a more compact protein band, which appears to migrate slightly faster than the wild-type isoform (Fig. 3), perhaps reflecting alterations in post-translational modifications. Additional biochemical and genetic experiments are needed to decide whether the Z3 isoform can supply the 2Bc⁺ function. We find that the Z3 isoform is the predominant BR-C protein expressed in fat body early during metamorphosis (L. Wagner, I. F. E. and G. M. G., unpublished observations). Interestingly, 2Bc mutations have been shown to disrupt the expression of at least two fat-body-specific genes, Fbp1 (Lepesant, et al., 1986) and Fbp2 (Rat et al., 1991; Nelson et al., 1991; K. Crossgrove and GMG, unpublished observations), providing circumstantial evidence for the correlation between the 2Bc⁺ function and the Z3 isoform in fat body.

There is probably not a simple one-to-one correspondence between BR-C genetic functions and BR-C encoded protein isoforms. Saturation mutagenesis of the BR-C region revealed three complementing genetic functions (Belyaeva et al., 1980; Kiss et al., 1988) while molecular analysis indicates the production of at least four BR-C protein isoform classes generated by alternative splicing (DiBello et al., 1991; C. Bayer and J. Fristrom, personal communication). Either one isoform is responsible for an, as yet, undiscovered genetic function or multiple isoforms can provide at least one of the genetic functions. We favor the second (redundancy) hypothesis for two reasons. First, genetic complementation experiments indicate that loss of either the rbp⁺ or the br⁺ function can lead to overlapping phenotypes (Kiss et al., 1988) suggesting that some genes may be regulated by more than one BR-C isoform. This interpretation is complicated by BR-C autoregulation. For example, npr1 null mutants fail to show regression of the 2B5 early puff (site of the BR-C gene) during the prepupal period (Belyaeva et al., 1981) suggesting BR-C products are needed to turn off its own transcription. Moreover, loss of Z2 isoform expression in br²⁸ larvae results in lower expression of other BR-C isoforms (Fig. 3). Thus, it is easy to imagine situations where aberrant expression of one isoform might lead abnormal expression of other isoforms resulting in a complex phenotype. Second, in vitro DNA footprinting experiments using the Sgs-4 enhancer suggest that different BR-C isoforms can recognize a single binding site (von Kalm et al., 1994). Such shared sites could be present near many genes and function as regulatory elements that are responsive to more than one BR-C isoform.

In addition to coordinating ecdysone-induced gene expression in different tissues, the BR-C may also direct gene expression programs that specify tissue-specific metamorphic outcomes. For example, rbp mutations severely disrupt salivary gland gene expression, which ultimately alters the metamorphic outcome from tissue histolysis to tissue persistence (Restifo and White, 1992). In agreement with the correlation between the rbp⁺ function and the Z1 isoforms, we find that the Z1 proteins represent the predominant BR-C isoforms in salivary glands early during metamorphosis. It seems likely that the Z1 isoforms control the synthesis of salivary gland products whose activity results in the timely histolysis of this tissue during metamorphosis.

Synthesis of the Z2 isoforms may be associated with cellular differentiation and morphogenesis. For example, mutations in the br⁺ function result in evagination defects during imaginal disc development (Kiss et al., 1988). The Stubble-stubbloid gene encodes a transmembrane serine protease that is required for disc morphogenesis (Appel et al., 1993). Interestingly, the Stubble-stubbloid gene is known to interact genetically with br function (Beaton et al., 1988) suggesting that the BR-C and Stubble-stubbloid act in the same morphogenic pathway. In agreement with the correlation between the br⁺ function and the Z2 isoforms, we find that the Z2 proteins represent the predominant BR-C isoforms in discs during the developmental window when evagination occurs (the 9 hour time period after the major pulse of ecdysone; Condic et al., 1991). These observations suggest that Z2 isoforms coordinate disc evagination by interaction with disc-specific genes.

CNS remodeling during metamorphosis involves persistence and reorganization of larval neurons, new cell growth and differentiation, and some cell death. All three BR-C functions are required for this process (Restifo and White, 1991). In agreement with these data, we find Z1, Z2 and Z3 isoforms present in CNS preparations during the late third instar. However, during prepupal development, the Z3 isoform predominates. The larval fat body and malpighian tubules also persist through metamorphosis but without extensive cellular remodeling. We also find that the Z3 protein is the predominant BR-C isoform expressed by these two tissues during the late third instar (L. Wagner, I. F. E. and G. M. G.; unpublished observations) suggesting that the Z3 isoform is associated with larval tissues that survive metamorphosis to form organs in the adult.

The process of metamorphosis is set in motion by a single molecule, the steroid hormone ecdysone. However, the ensuing physiological responses are complex and include tissue histolysis, differentiation and reorganization. This diversity of responses may originate from the tissue-specific accumulation of different combinations of primary ecdysone-responsive gene products. In addition to the BR-C, three other primary response genes (EcR, E74 and E75) have been characterized at the molecular level and each encodes a family of transcription factors (Burtis et al., 1990; Thummel et al., 1990; Segraves and Hogness, 1990; Koelle et al., 1991). Developmental analysis of BR-C, EcR (Talbot et al., 1993; Truman et al., 1994) and E74A (Karim and Thummel, 1991) isoform expression shows that distinct isoform combinations accumulate in different tissues. It seems possible that selective usage of this exceedingly rich array of transcription factors could be used to dictate and coordinate the component processes of metamorphosis. It will be interesting to determine whether transgenic perturbation of isoform combinations or ratios will alter the tissue-specific expression of effector genes and, as a result, the metamorphic fate of particular tissue. Ultimately, the combination of BR-C-encoded primary response isoforms present in a tissue is a consequence of tissue-specific splicing. Perhaps this aspect of gene expression is also under the control of ecdysone.

We thank Cindy Bayer, Linda Restifo and Igor Zhimulev for fly stocks; Linda Restifo for her expert instruction in interpreting the CNS antibody staining experiments; Kirsten Crossgrove for providing several BR-C protein expression constructs; and Sabine Baxter and Linda Wagner for their help in establishing and characterizing the monoclonal reagents. We also thank Cindy Bayer, Jim Fristrom, Laurie von Kalm, Linda Restifo, Igor Zhimulev, Elena Belyaeva, Kirsten Crossgrove, John Emery and Larry Wright for helpful comments on the manuscript and for generously sharing unpublished results. I. F. E. was supported by a National Science Foundation pre-doctoral fellowship and by a Ford Foundation dissertation fellowship. This work was supported by research grants from the American Cancer Society (NP-796 and DB-89A) and the National Science Foundation (DCB-8905031).