Puffs and PCR: the in vivo dynamics of early gene expression during ecdysone responses in Drosophila

Hormones coordinate the regulation of networks of gene expression that characterise the development and differentiation of higher organisms. In insects, the steroid ecdysone regulates gene expression in many stages and tissues, however the most detailed description of its effects is the sequence of puffs induced in the giant polytene chromosomes of the larval and prepupal salivary glands of Drosophila melanogaster. The early in vivo studies of Becker (1959) were later extended by Ashburner (1967) and Zhimulev (1974) to include some 340 puff sites that change their activity in a strictly defined temporal pattern between 110 and 132 hours of development, a period that ends with the histolysis of the larval gland. Ashburner (1971) initiated a series of in vitro experiments with 20-OH ecdysone that reproduced all of the essential features of the late larval response in cultured salivary glands and gave rise to a formal model (Ashburner et al., 1974) which has stimulated molecular studies. Recent results confirm the similarities between the ecdysone receptor (EcRKoelle et al., 1991) and vertebrate steroid receptors and add a level of complexity to the model with the demonstration that heteromeric receptors mediate the ecdysone response in Drosophila (Yao et al., 1992; Thomas et al.,1993). If this response requires two (or more) members of the steroid receptor superfamily, the immediate targets, the early responding puffs of the salivary gland are equally complex. Three of these puffs (2B5, 74EF and 75B) each encode a family of proteins related to known vertebrate transcription factors and are expressed in many tissues (see Andres and Thummel, 1992 for review). Thus the salivary gland remains an interesting model for our understanding of the regulation of steroid coordinated networks of gene activity.

In vitro puffing studies (Richards, 1976a,b) revealed two further phases in the salivary gland ecdysone response following the late larval response: a first period from 3 to 8 hours after pupariation (which occurs at 120 hours) where ecdysone titres must fall dramatically for the progression of puffing, followed by a second increase in ecdysone titre which regulates the late prepupal ecdysone response from 8 to 14 hours after pupariation. Interestingly, although some of the early puffs of the late prepupal response are in common with those of the late larval response, others are stage specific as are the global responses of the tissue to hormone (glue secretion and gland histolysis respectively). Later puffing studies (Richards, 1982) compared the ecdysone response in chromosomes of the fat body to those of the salivary gland and showed once more that although many early puffs were common to the two tissues there were differences in their response which might find their counterparts at the molecular level. This idea has been developed in the tissue coordination model (Burtis et al., 1990; Thummel et al., 1990) which suggests that one aspect of the stage and tissue-specific response to hormone may consist of differences in the levels of the different isoforms of the transcripts of the early puffs induced by ecdysone in each response. The specificity may lie in the direct action of these factors on target genes or be further elaborated by subsequent cross-talk between the products of these regulatory genes.

The strength of the original cytological studies lay in the fact that they were based on the simultaneous analysis of gene activity at many loci in individual nuclei of a given salivary gland which led to the assignment of larvae to a puff stage (PS, Becker 1959) thus defining the developmental age of the individual. Because of the rapidity of the response to hormone there is necessarily a heterogeneity in the developmental age of late third instar larvae, which has obscured the fine detail of the in vivo response at the molecular level, although elegant in vitro studies with E74 (encoded by the 74EF puff), E75 (75B), BR-C (2B5) and EcR have revealed the precision of the response (Karim and Thummel, 1991, 1992). From the culture of mass isolated larval organs with different concentrations of 20-OH ecdysone, Karim and Thummel (1992) have proposed two general categories of early gene transcript based on their ecdysone dose-response characteristics: Class I transcripts (e.g. EcR and E74B) respond to low levels of hormone (first induced from 2 × 10-9 M, 50% activity at 2 × 10-8 M) while Class II transcripts (e.g. E74A, E75A and E75B) respond to higher levels of hormone (induced from 10-8 M, 50% activity between 1 and 3 × 10-7 M). E75A is the most sensitive of the Class II transcripts being some three-fold more sensitive to hormone than E74A and E75B. Their analysis of BR-C RNAs suggested transcripts with both Class I and Class II characteristics. Note that we use the term ecdysone for 20-OH ecdysone when referring to in vitro experiments but for all active ecdysteroids in vivo.

While extractions of RNA from 5–10 staged prepupae may be sufficient for the conventional northern analysis of early gene transcripts, our conviction was that for in vivo studies, an assay based on the analysis of tissues from individual larvae or prepupae was indispensable. We describe here a RT-PCR based assay and show its application to the salivary gland, where we have followed transcripts of EcR and those of the isoforms of E74, E75 and BR-C (grouped by their zinc-finger domains) together with those of the gland-specific target gene Sgs-3 from 86 to 132 hours of larval and prepupal development. We then used the assay to establish profiles of early gene transcripts in a number of tissues.

We confirm the precision of the biological system as shown by the studies of Karim and Thummel (1991, 1992) and reveal unsuspected detail in both the temporal and spatial response. We show that EcR transcript levels in the salivary gland decrease at least 50-fold during the classic late larval puffing response but increase again prior to the late prepupal response. In both responses E74 isoform switching occurs prior to morphological changes in the 74EF puff. Although we detect transcripts from all of the early genes in the tissues we have examined, E75 shows the most significant differences between tissues and the most dynamic patterns within tissues. In particular the E75B ‘one-fingered receptor’ isoform (Segraves and Hogness, 1990) is rapidly but transiently induced during the salivary gland responses to hormone. The late larval response includes a novel induction of E75B in early prepupae. As suggested by the analysis of cDNAs (DiBello et al., 1991), BR-C shows a complex pattern of expression throughout this period. The developmental profiles of BR-C transcripts reveal combinations of isoforms encoding Z1, Z2 or Z3 zinc-finger domains at all stages in the salivary gland and in all tissues during the late larval response to ecdysone. Our results suggest that for BR-C both promoter usage and alternative splicing are developmentally regulated. This in vivo analysis of early gene isoform transcripts confirms that the early response to hormone is stage and tissue specific and suggests new elements that may contribute to this specificity.

The Drosophila melanogaster wild-type Oregon-R line, maintained at 25°C on a standard cornmeal medium, was used throughout. In these conditions, the first embryos hatch around 22 hours after egg laying, the second to third larval moult is at 72 hours, pupariation occurs at 120 hours and adults emerge at 240 hours. For the mid to late third instar samples, females were placed in split bottles for a 1-2 hour laying period. Larvae were selected at the second to third moult by size and anterior spiracle morphology, placed in 1.5 cm diameter tubes (approx. 20 larvae per tube) and returned to 25°C until they reached the desired age. Larvae synchronised in this way leave the food at approx. 112 hours and pupariate in a 4–8 hour period around 120 hours. Wandering larvae (112–120 hours) were staged by puff analysis of at least five nuclei from one salivary gland. For later stages animals were selected as white prepupae, a stage which lasts approximately 15 minutes, and then aged at 25°C in fresh food tubes for 0–14 hours. These animals were also analysed for their puff stage (see text).

Larvae and prepupae were dissected in insect Ringer. One salivary gland lobe was transferred to cytological fixative (3:1, ethanol:propionic acid) while the sister lobe was placed in 25 μl of ice-cold extraction buffer (3 M LiCl, 6 M urea, 10 mM sodium acetate, pH 5.0, 0.2 mg/ml heparin, 0.1% SDS) and immediately frozen in liquid nitrogen. Where appropriate, further tissues of the same larva were treated in parallel in 25 μl aliquots of extraction buffer. All samples were stocked at −80°C until the first salivary gland lobe had been analysed cytologically following propioniccarmine-orcein staining (see Ashburner, 1967).

The RNA preparation is derived from the method of Richards et al. (1983). Tubes were allowed to thaw on ice and then vortexed vigorously, which was as effective for the extraction of RNA as conventional grinding, and reduced variability resulting from debris sticking to the piston (data not shown). The tubes were then stored at least 1 hour at 4°C before the addition of 25 μl of 4 M LiCl, 8 M urea. After vortexing followed by centrifugation for 15 minutes at 4°C, the pellet was resuspended in 25 μl of 0.1 M sodium acetate (pH 5.0), 0.1% SDS, by shaking 5 minutes at room temperature. Phenol (25 μl) and chloroform (25 μl) were added and the tube shaken for a further 20 minutes at room temperature. After 5 minutes in a microfuge, 25 μl of the supernatant were transferred to a fresh tube, adjusted to 0.2 M sodium acetate (pH 5.0) and precipitated with 2 volumes of ethanol for 1 hour at −80°C with two agitations at 20 minute intervals. The RNA was recovered by centrifugation 15 minutes in a microfuge at 4°C and resuspended in 10 μl of distilled water. Dilutions were made in distilled water and stocked at −80°C as necessary.

General considerations for designing RT-PCR assays can be found in Wang et al., (1989), Chelly et al., (1990), Makino et al., (1990) and Innis et al., (1990).

In order to standardise reaction conditions, we used 20-mer oligonucleotide primers (with a few exceptions, see Table 1), having a GC content as close to 50% as possible. In each case the RTase-PCR primer was positioned so that a second oligonucleotide could be placed between it and the splice site to serve as a hybridisation probe for confirming PCR products or in subcloning experiments (see below). The 5′ primers were placed so that products were in the size range of 150–700 base pairs (bp) to ensure efficient PCR amplification. Primer sequences are given in Table 1 and their position and the size of the RNA derived products are shown in Fig. 2. We used a series of dilutions of a bulk RNA preparation to determine sub-saturation conditions for the PCR products (see for example Fig. 8C). Then, in a first series of glands, two dilutions of template, approximately 10and 100-fold lower than that leading to saturation were used. In later series we used only the higher concentration as this was clearly sufficient for quantitation. This value is given in Table 1 for each set of oligonucleotides for the salivary gland experiments.

The combined reverse transcription PCR reaction was set up on ice in 50 mM KCl, 1.5 mM MgCl₂, 25 mM Tris-HCl pH 8.3, 100 μg/ml BSA, 100 μM TMAC (tetramethylammonium chloride), 50 μM of each dNTP with 12.5 pmol/primer plus the RNA or DNA template and distilled water to 49 μl. After the addition of two drops of paraffin oil, denaturing and annealing was performed in the PCR apparatus (DNA Thermal Cycler, Perkin Elmer Cetus) for 3 minutes at 94°C followed by a programmed step (10 minutes) to 55°C. 1 μl of a mix of AMV RTase (2.5 units, Pharmacia) and Taq polymerase (1.25 units, Cetus) was added and the tube left for 20 minutes at 55°C. The subsequent amplification programme was 3 minutes at 94°C followed by 30 cycles of 94°C (1 minute), 55°C (2 minutes) and 72°C (3 minutes) followed by a hold step at 4°C. For the analysis of rp49 transcripts we used MMLV RTase (50 units per tube, BRL). In this case annealing was programmed to 37°C (13 minutes) and we left the reaction 30 minutes at 37°C before starting the PCR cycles. Both a positive (known template) and a negative (no template) control was included in each set of reactions. For E75, the PCR primers were added after the RTase reaction as one of these primers initiated reverse transcription which gave rise to a non-specific PCR product. To confirm that the PCR products were those predicted from earlier studies, they were subcloned into pEMBL plasmids and sequenced on double stranded mini-prep DNA templates by the Sanger dideoxy method.

If the intron between the PCR primers is short (<1 kb), in addition to the product corresponding to the spliced transcript, we detect a second band that may derive from either pre-mRNA or contaminating DNA (see Fig. 2). To distinguish between these possibilities, for each gene, we compared four aliquots from a miniextract treated with or without DNAse I followed by a standard RT-PCR (with AMV RTase and Taq polymerase) or PCR alone (i.e. Taq polymerase alone). After DNase I treatment (40 units DNase I/RNase-free (Boehringer), 100 mM KCl, 3 mM MgCl₂, 50 mM Tris-HCl, pH 8.3, 200 μM TMAC with 30 units RNasin (Amersham), 30 minutes at 37°C followed by 15 minutes at 65°C to inactivate the enzymes) the reaction volume was doubled with the addition of the remaining components of the RT-PCR reaction as described above.

We have developed an RT-PCR assay for the simultaneous analysis of ecdysone regulated transcripts, in tissues dissected from individual staged larvae and prepupae (see Fig. 1). Our approach implies the processing of a large number of microextracts (approx. 50 ng of total RNA/sample). The general principle is that an oligonucleotide primer is placed 3′ to an intron and serves for first strand cDNA synthesis, thereafter DNA amplification procedes using this primer and a second primer 5′ to the intron. This distinguishes, by size, products from spliced transcripts from those resulting from pre-mRNA or DNA contamination of the RNA sample.

We have studied the following genes (see Fig. 2):

• (A) The ribosomal protein 49 (rp49). These transcripts serve as a standard for RNA extraction and the subsequent RT-PCR assay.

• (B) The salivary gland secretion protein (‘glue’) gene Sgs-3. This is a tissue-specific target gene of the ecdysone regulatory network which gives rise to abundant transcripts in the late larval period (110–120 hour) where our assay requires 0.01% of a salivary gland extract. Previous studies (Georgel et al., 1991) suggested its interest as a marker of a mid-third instar larval ecdysone response occurring around 96 hours.

• (C) The ecdysone receptor (EcR). As Koelle et al. (1991) report the possibility of alternate 5′ exons for this gene, our primers are positioned either side of the third intron (see Fig. 2C) so as to measure all isoforms.

• (D) The early ecdysone induced gene E74 which contains an ets-related DNA binding domain (see Burtis et al., 1990). This gene gives rise to two major forms of transcripts, E74A and E74B. E74B transcripts have two initiation sites separated by 300 bp (Burtis et al., 1990) which we have not sought to distinguish. A common oligonucleotide primer for reverse transcription is placed in the first shared exon and an E74Aand E74B-specific primer is positioned in the corresponding 5′ exon (Fig. 2D).

• (E) The early ecdysone induced gene E75 which is a member of the steroid receptor superfamily. Three transcripts have been reported, E75A, B and C, which encode distinct protein isoforms (Segraves and Hogness, 1990; Feigl et al., 1989, see Fig. 2E). As for E74, a primer for reverse transcription is located in the first shared exon with E75A-, E75B and E75C-specific primers in the corresponding 5′ exons.

• (F) The Broad-Complex (BR-C), localised in the early ecdysone puff 2B5. In this complex transcriptional unit, transcripts, initiated from one of at least three promoters, contain a common ‘core-region’ encoding segment that is differentially spliced to one of at least three zinc finger encoding motifs, Z1-Z3 (DiBello et al., 1991). As protein synthesis initiates in the common core region, the finger region is the principal source of functionally distinct protein isoforms and accordingly we have devised a fingerspecific strategy to detect Z1, Z2 or Z3 encoding transcripts. For Z1 transcripts, unless otherwise noted, the Z1 product corresponds to the Q¹ variant (Fig. 2F).

In preliminary experiments we determined an optimal distribution of the RNA microextract for the various assays (see Materials and methods and Table 1). In general, regulatory gene transcripts showed comparable maximum levels which were at least 100-fold lower than those of Sgs-

3. In cases where a short intron is involved, we observe a product that derives from contaminating DNA (rp49, EcR and E75) or pre-mRNA in the case of Sgs-3, where 100fold less template is required (data not shown, see Materials and Methods).

We established the reproducibility of the assay using late third instar wandering larvae (a period characterised by dramatic puffing changes) by puff staging one salivary gland lobe and analysing the second lobe by RT-PCR. Following protocols derived from puffing studies, we processed at least four glands for each puff stage (PS; see Ashburner, 1967) in each experimental series. An example is given in Fig. 3 for three glands of each of PS1, 2, 3 and 4, taken from a single series. The extraction is clearly reproducible as levels of the rp49 product are similar for all 12 glands. The products of the two early genes E74 and E75 show the precision of this approach. At PS1 the E74B product is predominant, by PS2 the ratio of E74A to E74B is close to 1:1, by PS3 this ratio has become approx. 5:1 and by PS4 E74B is visible in just one of the three glands.

In parallel, increased E75B activity is detected late in PS1 (see below), but while E75B persists in PS2 and 3 before diminishing at PS4, E75A increases from late PS2 to PS4. Note that RT-PCR is more sensitive than PS analysis. A PS is an approximation as adjacent nuclei in a gland may show slight differences in their puffing pattern, particularly during key phases of the hormonal response (Ashburner, 1967). With RT-PCR we can order glands within a puff stage, for example from the E75A profile, the gland PS2c is a late PS2. Hereafter we will present a single representative gland for each time point. In total for this study, we have analysed 550 microextracts in some 4000 RT-PCR reactions. All of the ethidium bromide stained gels presented were transferred and the membranes hybridised with both isoform-specific and/or common probes (data not shown, see Fig. 8 for examples). We reserve the terms ‘absence’ or ‘disappearance’ for cases where products are at least 100-fold lower than their maximum levels as estimated from serial exposures of the autoradiographs. For convenience, we will refer to these PCR products as transcripts, although of course our assay detects only a specific portion of the transcript. A comparison of our results for E74A and E74B (see below) with those of Karim and Thummel (1991), who used conventional northern analyses, suggests that we measure full-length transcripts.

We have analysed transcripts in the salivary gland from the mid third instar to the histolysis of the larval gland some 12 –15 hours after pupariation, the period of the classical chromosome puffing studies (see Ashburner and Berendes, 1978). This period is most conveniently divided into three phases, each of which includes a distinct response to ecdysone. For the first phase larvae are staged at the second to third larval instar moult (see Materials and methods) and then sampled at 2 hour intervals from 86 hours onwards. The second phase is the post-feeding, wandering phase that ends with pupariation at 120 hours. Here larvae were staged as for Fig. 3. From pupariation (PS11) until the histolysis of the larval salivary gland at PS21 animals were collected as white prepupae and staged both in terms of hours post pupariation (pp) and puff stage. Development is relatively synchronous up until 8 hours pp, thereafter hormone induced changes are rapid and puff staging is a more reliable index (see Richards, 1982).

While early puffing studies used PS1 to describe the period from the mid-third instar until the late larval ecdysone response in wandering larvae (Fig. 1), later puffing studies (Richards, 1982) and more recent molecular studies (Andres and Cherbas, 1992; Andres and Thummel, 1992 for review) have shown that there are further hormonally regulated events in this period. Studies in this period are difficult as mid instar larvae are about half the size of late wandering larvae and show considerable developmental variability between individuals (synchronised 24 hours earlier at the second to third instar moult). Fig. 4 shows results from a single series sampled at 4 hour intervals; although different series differ by up to 4 hours in development, the sequence of events is always maintained. Low basal levels of E75A are seen throughout this period (see also PS1 glands Fig. 3) but certain individuals showed exceptional levels in the early part of this period (e.g. the gland at 94 hours, Fig. 4). The analyses of the other transcripts revealed less variation between individuals. The Class I transcripts (Karim and Thummel, 1992) EcR and E74B increase from 86 to 94 hours and then remain relatively constant until the premetamorphic pulse of ecdysone (at approx. 114 hours) which induces the transition to PS2. A similar profile is seen following hybridisation for E75C which thus appears with the Class I transcripts. In the salivary gland E75C levels are highest during this period but remain low compared to those of E75A and E75B from PS2 to 10 (see below). For BR-C, Z1 containing transcripts increase gradually from 94 hours until 114 hours, while those containing Z3 appear in parallel with E74B, peak between 94 and 98 hours and then decrease. This is consistent with the observation that the salivary gland acquires competence to respond to ecdysone by GEBF-I synthesis and Sgs-3 transcription at approx. 96 hours and that both processes require BR-C activity (Georgel et al., 1991). Sgs-3 is induced some 100,000-fold between 98 and 106 hours (see below).

By 112 –114 hours there is heterogeneity in the developmental age of larvae (see Fig. 4, 114h for example), and we staged animals in this period by the PS of one salivary gland lobe. The PS1 and PS2 animals of both series provide the junction between the two periods. It is difficult to determine the duration of each PS (Ashburner, 1967); wandering lasts some 8 hours in our Oregon stock and from the frequency in random sampling, PS1 may continue for 2 –3 hours of wandering. The ecdysone induced progression PS2 to PS11 (pupariation) therefore takes 5 –6 hours, with an average of some 30 minutes per PS.

The first observation is that EcR transcripts, that appear to be induced by a small increase in ecdysone titre between 90 and 94 hours (Fig. 4), may increase at PS2 but diminish dramatically from PS7 onwards (Fig. 5). EcR is localised in the region of the 42A puff (Koelle et al., 1991) which is active at PS1 and regresses at PS10 (Ashburner, 1967). This suggests that the puff reflects EcR transcription although chromosome refolding may continue after transcription stops.

There is a striking similarity between our RT-PCR results for E74 and those of Karim and Thummel (1991, their Fig. 2) using northern analyses for in vivo studies of whole animals prior to pupariation. They showed that at −12 hours E74B was abundant, at −3 hours E74A and E74B were at similar levels and that at pupariation there was essentially only E74A. While, as expected, E74B is abundant at PS1, we were surprised by the rapidity of the isoform switch as E74A transcripts are at the same level as E74B by PS2 and predominate by PS3. E74A transcripts reach their maximum at the time of the 74EF puff maximum, remain abundant during 74EF regression (PS8-10) and then decrease abruptly at PS11.

E75 transcripts are closely correlated with those of E74 (see also Fig. 3), as would be predicted from puffing studies. The Class II E75B transcript is induced in parallel with the E74A transcript but is short lived in that it disappears from PS5 onwards while the E75A transcript increases from PS3 onwards. From PS5 to PS11 the E74A and E75A transcripts show a similar profile. As the E75A and E75B promoters have similar sensitivities to ecdysone (Karim and Thummel, 1992, but see Discussion) the delay in the appearance of the E75A transcript relative to E75B may be a consequence of primary transcript length (see Fig. 2). In contrast the fact that the E75B primary transcript is nearly 40 kb shorter than E74A suggests that it is induced somewhat later or that it is not spliced as rapidly as E74A. By comparison with E75A and E75B, the E75C transcript is much less abundant in the salivary gland and is rarely detected on the ethidium bromide stained gels. Hybridisation with an E75C-specific probe reveals two major periods of activity, the first initiated in mid instar larvae (Fig. 4) continues between PS1 and PS5, the second is at PS9 and 10.

The interpretation of BR-C transcripts is complex as the same RT-PCR product may derive from different promoters in the course of the response. Z1-containing transcripts are present at PS1, diminish at PS2 and thereafter are moderately abundant from PS3-PS10. In addition to the Q¹-Z1 transcripts visible on the stained gel, low levels of TNTQ¹-Z1 and Q²-Z1 transcripts, following a similar profile, are seen by hybridisation and are specific to this period in the salivary gland (data not shown, but see Fig. 7). This is in accord with the frequency of these forms amongst Z1containing cDNAs (Q¹-Z1, 20/25; TNT-Q¹-Z1, 3/25; Q²Z1, 2/25 DiBello et al., 1991). Z2-containing transcripts (barely detectable before 114 hours, see Fig. 4) show a weak activity at PS1 and 2 that increases between PS3 and10 although there is heterogeneity in these samples. The pattern of Z3-containing transcripts is close to the sum of those of E74A and B, that is an activity at PS1 and 2 that increases appreciably by PS3, remains abundant until PS10 and diminishes by PS11. If the BR-C promoters resemble those of the other early genes this may indicate that a Class I promoter active at PS1 is rapidly repressed by the major increase in ecdysone titre and that simultaneously Class II transcripts are initiated from more distal promoters.

The 68C puff, containing Sgs-3, regresses from PS2 onwards in vivo and is rapidly repressed by ecdysone in vitro. Similarly Sgs-3 transcription is reduced within 15 minutes of hormone treatment in vitro (Crowley and Meyerowitz,1984), suggesting that the gene should be transcriptionally silent by PS2 or 3. In contrast Sgs-3 transcripts remain constant and abundant until PS10, suggesting the presence of a large stable pool of transcripts following puff regresssion. The level of these transcripts then drops at least ten-fold by PS11 and is a further example of an event which, at pupariation, reduces the levels of most of the transcripts that we have analysed. It may be significant that this decrease appears specific to the salivary gland (see below) at the time of glue expulsion.

The late larval ecdysone response continues beyond pupariation (PS11) until PS14, some 4 hours into the prepupal period. There follows a period of low hormone titre that prepares the gland for its final response to the late prepupal peak of ecdysone at approx. 8 hour pp which ends with the histolysis of the larval gland at 13 –15 hour pp. These three phases are marked by changes in the transcripts we have followed (Fig. 6).

Following the global reduction of transcripts at PS11, E74A and E75B transcripts are detected from PS12 to PS14, a co-expression similar to that seen from PS2 to PS5 (Fig. 5). For E75 this corresponds to a further isoform switch. Between PS12 and PS14, there is a parallel but less dramatic reinduction of Z2 and Z3 transcripts while Z1-containing transcripts appear later between 4 and 6 hours pp.

EcR transcripts, which are minimal from PS7/8 onwards, increase at PS13/14 at the end of the late larval ecdysone response and thereafter remain relatively stable until PS21. This suggests that EcR transcripts are induced following the drop in ecdysone titre, that their induction is a prerequisite to the late prepupal response (see Discussion) and that the smaller late prepupal peak of ecdysone does not lead to the repression of EcR transcription. The reinduction of E74 and E75 from PS17 to 21 resembles a compression of their larval activity, i.e. Class I transcripts followed by Class II transcripts. However, there is at least one significant difference in the order of induction. E74B is induced by PS17, closely followed by E75B, an induction that clearly precedes that of E74A. The delay in E75A induction relative to E75B is very similar to that seen in PS1-4. E75C transcripts, detected by hybridisation, are even lower between PS11 and PS21 than in the larval period (data not shown). The duration of this response is much shorter than that of the late larval response, confirming results of both puffing studies and prior analyses of E74A products (see Boyd et al., 1991).

For BR-C, Z2 activity in the salivary gland is highest during the late larval response, although low levels of Z2 can be detected by hybridisation throughout the prepupal period (data not shown). Z1 increases between PS18/19 and PS20, while Z3 transcripts increase during ecdysone responses (PS11-PS14 and PS18-20 in Fig. 6) but have a relatively abundant basal level at most stages.

The above results show that there are differences in both the isoforms of the key regulatory genes that are expressed and the timing of their expression during the responses of the salivary gland to ecdysone at three developmental stages. The prediction from puffing, genetic and molecular analyses is that similar differences exist between tissues and serve to translate a single ecdysone peak into tissue-specific responses. To test this we have undertaken the simultaneous analysis of transcripts in a number of tissues from individual larvae. An overall picture of the late larval response can be derived from animals at PS1, PS3, PS7 and PS11 (Fig. 7). We present results from five tissues (salivary glands, SG; gut, GT; malpighian tubules, MT; fat body, FB and wing discs, WD) from animals staged by the PS of one salivary gland. As expected from the tissue coordination model we detected transcripts of EcR, BR-C, E74 and E75 in all tissues during this period.

As shown above (Fig. 5), EcR transcripts drop dramatically in the SG from PS7 onwards. This decrease is slightly delayed in the MT, is less dramatic in the FB and GT while in the case of the WD a relatively low level of EcR transcripts remains unchanged throughout this period. This may be a consequence of tissue differences in EcR isoform expression. The E74B to E74A switch is slightly faster in the SG than the MT and WD while both the GT and FB have appreciable levels of E74B at PS11. This analysis is in accord with Boyd et al., (1991) who described essentially minor tissue differences in E74A expression, notably a slower repression in certain tissues. More significant differences appear with the E75 analysis, where we distinguish four different profiles. At PS1 E75A is active in the GT (as is E75C), but not in the other tissues. By PS3 both E75B and E75A are active in the SG, GT, MT and FB, but only E75A is active in the WD. E75B transcripts disappear by PS7 in the SG but are still present in the GT, MT and FB. At PS11, both E75A and E75B are present in the GT and WD, E75B persists in both the MT and FB while E75 transcripts are absent from the SG. In these tissues E75C is a minor isoform which decreases in all tissues between PS1 and PS11. At PS1, Z1-containing transcripts from the BRC are abundant only in the GT where they appear to remain constant until PS11 when a second shorter product corresponding to the Q²-Z1 isoform (see above and Fig. 2 legend) is observed. In the SG and MT, Z1 transcripts are seen at PS3 and PS7, but not PS11 while in both the FB and WD low levels of transcripts increase between PS3 and PS11. Note the presence of low levels of Q²-Z1 in the PS7 SG. Results with Z2and Z3-containing transcripts show important quantitative differences between the tissues. In the SG there are traces of Z2 while Z3 is abundant until PS7, dropping dramatically at PS11. For the other tissues the profiles are constant but with significant differences between the tissues, i.e. MT, GT and FB have mainly Z3 with a trace of Z2 (seen on hybridisation for the MT and SG data not shown), while the WD have mainly Z2 with a trace of Z3. These results confirm the predictions of DiBello et al., (1991), based on genetic analyses, that Z1 and Z3 should be expressed in salivary glands and Z1 and Z2 in imaginal discs.

There is an important difference in the abundance of Sgs3 transcripts compared to those of the regulatory genes such as EcR, the former being at least 100-fold more abundant in the late larval salivary gland as determined by template dilutions (see Table 1). A second difference lies in the maximum and minimum levels of these transcripts, a problem that we have studied by hybridisation with radioactive probes (Fig. 8). For Sgs-3 we first compared transcript levels at key stages by serial dilutions of the input RNA, so as to obtain bands of similar intensity on ethidium bromide stained gels (data not shown). We then assembled composite gels, which after transfer and hybridisation gave comparable bands on autoradiographs (Fig. 8A). The combination of the dilution factor and the autoradiographic signal enables us to establish relative levels for this period of development (Fig. 8B). If we set the initial value for Sgs3 in 98 hour glands at 1 (although we did not detect a band even on extended exposure of the autoradiograph) we estimate that it reaches 10 by 100 hours, 100,000 in late wandering larvae (PS1 to 10), and then decreases to 5,000 at PS11, 2,000 in 2 hour pp, 1,000 in 8 hour pp and remains between 100 and 10 at the time of gland histolysis in 14 hour pp. For EcR (Fig. 8C), unlike Sgs-3, we detect transcripts at all stages. The use of serial dilutions of a standard RNA extraction (Fig. 8Ci) confirms both that the assay is quantitative when using sub-saturation levels of template and establishes a scale for estimating relative levels in the microextracts. From 86 hour, PS2 and PS11 larvae, we obtain values of 2, 100 and 1 respectively (Fig. 8Cii).

The study of relative transcript levels shows that regulatory genes may differ from their downstream target genes not only by the overall abundance of their transcripts (the sum of the EcR isoform transcripts is between 100and 1000fold less abundant than those of Sgs-3) but also by their activation/repression characteristics. EcR transcripts are always present in this period with levels that vary at most a hundred-fold. We assume that this is reflected in the EcR proteins (although post-transcriptional mechanisms may modulate their presence or activity) and that even the lowest level is physiologically important for the autoregulation of EcR. Thus, for EcR, hormonal regulation appears to be overlaid on a mechanism that ensures basal levels of activity. Switching of Sgs-3 is more dramatic. Prior to its activation at approx. 100 hours we do not detect transcripts, which demonstrates that although our RT-PCR assay is highly sensitive it does not detect transcripts at the level of ‘illegitimate’ or leaky transcription (Chelly et al., 1989). Once Sgs-3 is activated, there is at least a 100,000-fold increase in expression as would be expected from a strong enhancer. The profile of the decrease in transcripts following pupariation is surprising. The 50-fold drop from PS10 to 2 hour pp corresponds to the image previously obtained by northern analyses (e.g. Richards et al.,1983) but thereafter RT-PCR detects non-negligible levels of transcripts until the destruction of the gland in 13-15 hour pp. This suggests either that RNA that escapes the degradation seen in PS11 salivary glands is relatively stable or that the mechanism of Sgs-3 repression is not as effective as that which ensures its silence in early larval stages. Hybridisation analyses of E74, E75 and BR-C isoforms resemble those of EcR (data not shown), and it is possible that the differences between EcR and Sgs-3 in promoter regulation (modulation or on/off switching) are characteristic of regulatory and ‘working’ genes respectively.

It would be possible, using the relevant in vitro synthesised transcript as a standard for each reaction, to quantify the RT-PCR products in absolute terms (see Wang et al., 1989 for a related approach). An RNA standard is essential as the quantity of cDNA product obtained from the same transcript but using adjacent oligonucleotide primers may differ several-fold. Thus, within a gene, the use of a single RT primer for E74 (or E75) isoforms facilitates their comparison while the relative levels of Z1, Z2 and Z3 forms of BR-C may be distorted by the use of three distinct RT primers.

We have worked with ethidium bromide stained gels not only to develop a non-radioactive approach, but equally so as to detect non-specific PCR products that incorporate primers but are not recognised by an internal labelled probe. In initial experiments we eliminated a number of primers that gave such products. With hybridisation, the sensitivity of the method increases a minimum of ten-fold, so that RNA from a single gland could be analysed for at least one hundred transcripts, if suitable primer pairs have been established.

Ecdysone studies in the mid to late third larval instar pose problems due to the heterogeneity in the relatively large samples necessary for radioimmune assays (see Richards, 1981 for discussion). The general conclusion is that there are one (or several) small peaks from the mid instar onwards but that the massive increase in titre most probably coincides with the PS1-PS2 transition some 6 –8 hours before pupariation. Although the minor peak(s) has biological significance (e.g. inducing glue gene transcripts in the salivary gland, Georgel et al. 1991; see also Andres and Cherbas, 1992), its form remains undetermined. The characteristics of this first response are that it requires only a small titre increase, presumably with a sub-optimal occupation of receptors (so that only Class I promoters with high affinity for the hormone-receptor complex are activated) and that it is very slow in comparison to those of the late larval and prepupal period. Our observation of a few animals with significant levels of E75A transcripts suggests that there may be pulses of hormone release that occasionally result in the brief induction of the most sensitive Class II promoter. As mid instar larvae fed high levels of ecdysone (5 × 10⁻⁴ M) respond by glue gene transcription but not pupariation we presume that either there exists a mechanism that metabolises excess hormone in this period, and maintains levels sufficient for the sustained activity of Class I promoters only, or that tissues are not yet competent for the late larval response. Although only Class I early transcripts (i.e. E74B, EcR and presumably one of the BRC promoters) are induced, target genes such as Sgs-3 are activated as a late response to hormone.

The simplest interpretation of the PS1 to PS2 transition is that an increase in ecdysone titre activates Class II promoters (E74A, E75A, E75B and at least one BR-C promoter) and represses directly or indirectly Class I promoters (EcR, E74B and at least one BR-C promoter). The precision of the mechanism is apparent from the synchronous switching of E74 and E75 transcripts at this stage. One factor in the timing of isoform appearance is the elongation rate of RNA polymerase II determined for E74 at approx. 1.1 kb/minute (Thummel et al., 1990; Karim and Thummel, 1991) and estimated at 1.5 kb/minute for E75 (Karim and Thummel, 1992). E75B primary transcripts are 40 kb shorter than those of E74A (see Fig. 1) which our RT-PCR assay detects at the earliest 45 –60 minutes after activation of the E74A promoter (allowing for both transcription and splicing of exon 5 to 6; LeMaire and Thummel, 1990). As we detect these two transcripts simultaneously it suggests that the E75B promoter is induced slightly later than the E74A promoter despite the similarities in the dose-response curves determined in vitro (Karim and Thummel, 1992). Although E75A appears three-fold more sensitive to ecdysone than E75B in vitro, we detect E75A one PS later than E75B suggesting that in the salivary gland a rapid increase in ecdysone titre activates E75B shortly after E75A and that, as proposed by Karim and Thummel (1992), the timing of the appearance of transcripts is essentially a consequence of the 30 kb difference in primary transcript length.

Although EcR repression could be a direct consequence of the rising ecdysone titre, we favour the hypothesis that it involves the protein product of a gene induced following the PS1-PS2 transition, both because the timing coincides with late puff induction and because Karim and Thummel (1992) have shown that, in vitro, high levels of ecdysone lead to a decrease of EcR transcripts after some 3-4 hours of culture of larval organs.

While the rising titre and transcript length are sufficient to explain the beginning of the response, the reappearance of E75B from PS12 to PS 14 requires further promoter switching which may be EcR or ecdysone independent. Although E74A transcripts are equally present from PS12 to 14, E75A transcripts are not induced at this stage so that not all Class II transcripts are present in the early (0 to 4 hour) prepupae. There are two precedents for this activity from the extended culture of PS1 larval glands with ecdysone. There is a significant plateau in the regression profile of the 74EF and 75B puffs from 8 to 10 hours (Richards, 1976b) and E75B transcripts are reinduced at 8 hours of culture (Segraves and Hogness, 1990). Our in vivo analysis adds an unexpected twist in that this occurs in glands with minimal levels of EcR transcripts. However E74A transcripts are detected in late embryos and late pupae where their regulation may be ecdysone independent (Thummel et al., 1990). The early prepupal E74A transcripts are less abundant than those of late larvae and may simply reflect the final phase of the larval response. Similarly, the absence of the E74A protein in 2 and 4 hour prepupal salivary glands (Boyd et al., 1991) may indicate that these transcripts are not translated.

The E75B isoform is intriguing as it encodes a member of the superfamily of steroid receptors that has a single zinc finger from the DNA binding region and may not be able to bind DNA (Segraves, 1991). In the salivary gland its rapid but transient induction may indicate that it plays a role in heteromer formation between members of the steroid receptor family early in the ecdysone response (see Richards, 1992). Note that all three inductions in the late larval and prepupal period (from PS2 to PS5, PS12 to PS 14 and PS 17 to PS19/20) show a similar time course and level of expression, which may reflect a fundamental aspect of the regulation of E75B and its role in the salivary gland.

From 4 to 8 hours after pupariation the salivary gland must experience low ecdysone titres for the progression of puffing patterns and the acquisition of competence (between 5 and 6 hours) for the late prepupal response to ecdysone (Richards, 1976a,b). The simplest interpretation of the present results is that the increase in EcR transcripts from 5 hours onwards, is the molecular basis of competence. Although glands are competent by 6 hours, in vivo they respond at 8 hours as seen by the PS18/19 transition. During this shorter response, EcR transcript levels appear to remain stable, which contrasts with their induction in the mid larval period and their repression during the late larval response. This suggests that EcR is not a Class I transcript in this period and may reflect changes in EcR isoform expression between the larval and prepupal stages. From 8 hours pp the increasing ecdysone titre sequentially activates both Class I (E74B) and Class II (E74A, E75A and E75B) transcripts. This resembles a compression of the activity seen in the larval period, where each class is associated with a distinct biological response.

At PS1 in wandering larvae there is little sign of puffing activity at 74EF and 75B although E74B transcripts are abundant. This may be a consequence of the short E74B transcription unit, as puff formation is related to both promoter strength and transcript length (Simon et al., 1985). PS2 is defined by the first cytological appearance of 74EF and 75B induction, which can be detected within 5 –10 minutes of ecdysone treatment in vitro (Ashburner, 1973). However at PS2, E74A transcripts, which we detect at the earliest 45 –60 minutes after activation of the E74A promoter (see above) are abundant. In vitro, full length E74A transcripts are detected between 1 and 1.5 hours of culture (Karim and Thummel, 1991). Thus in vivo the E74A promoter must be activated late in PS1, prior to puff formation. At PS4/5 when the 74EF and 75B puffs have reached their maximum size, levels of the two shorter transcripts have dropped dramatically while the longer transcripts are at their maximum levels. Puff regression is followed by the disappearance of the long transcripts, E74A and E75A, although E74A persists in early prepupae. However, Boyd et al. (1991) detected nuclear E74A in salivary glands of late larvae but not in 0 hour or 2 hour prepupae suggesting either that the transcripts are entirely cytoplasmic at this point or below the levels of detection of their in situ analysis. In the prepupal response, E74B and E75B transcripts are detected at PS17 (i.e. just prior to the cytological sign of the reinduction of 74EF and 75B) and thereafter are replaced by the E74A and E75A forms respectively. As the puff 75B remains partly open throughout the prepupal period it may facilitate the activation of E75A which precedes E74A in this response. The duration of puffing at 74EF and 75B and expression of the E74A and E75A transcripts is shorter in the prepupal response than the larval response.

If large puff formation is related to the presence of long Class II transcripts, short pulses of ecdysone in the period 90 to 112 hours might lead to the occasional presence of small puffs at 74EF and/or 75B seen in some nuclei (data not shown). This is consistent with the observation that PS1 glands exposed briefly to high concentrations of hormone respond by a short-lived induction of 74EF and 75B without progression to the late response (Ashburner and Richards, 1976).

In general agreement with the analysis of DiBello et al. (1991), we found that for Z1-containing transcripts, the Q¹ isoform was the most frequent, Q² occasionally reached similar levels (notably at PS11 in the gut, Fig. 7) while the maximum levels of the TNT-Q¹ form (in late larval salivary glands) are about one hundred-fold lower as determined by hybridisation with an oligonucleotide common to all forms (data not shown). Similarly the Z2 isoform, found in both whole animal and imaginal disc cDNA libraries (DiBello et al., 1991), was most abundant in the wing disc. In contrast we observed consistently higher levels of Z3 products than would be predicted from that analysis.

The unravelling of the regulation of the BR-C promoters and the function of the protein isoforms will no doubt prove to be complex and an extension of our present assay to include promoter usage will be necessary. The developmental profiles of Z1, Z2 and Z3 in the salivary gland are not identical, which would be the case if the relative (but fixed) strength of the splice acceptor sites determined the splicing of the core-region to the Z1, Z2 and Z3 encoding exons. The changes we observe suggest that ecdysone may regulate trans-acting factors, which modify splice site usage. If the hormone induces both promoter switching, as for the other early genes, and changes in the splicing pattern, there may be preferential association of certain promoters with specific zinc finger encoding regions in the BRC transcripts. It is evident that in addition to this developmental regulation there are tissue-specific patterns of BR-C splicing (Fig. 7).

The link between BR-C transcripts and protein function may not be straightforward. A combination of molecular and genetic analyses suggest that Z1 transcripts correspond to rbp function and Z3 to l(1)2Bc (see DiBello et al., 1991). Our analysis in the period 88-98 hours reveals a peak of Z3 transcripts that precede Sgs-3 activation. However, the fact that glue gene transcripts are found in l(1)2Bc mutants but not rbp mutants (Guay and Guild, 1991) suggests that in this period it is the less abundant Z1 transcripts that are essential for inducing glue gene transcription.

These studies, taken with other recent results add considerably to the complexity of the molecular description of the ecdysone response in Drosophila. Currently, the basis of stage and tissue-specific early responses may involve differences in the ecdysone titre(s), the isoforms of EcR, other members of the steroid receptor superfamily (see Yao et al., 1992; Thomas et al. 1993) and the accessibility of the different promoters of the early genes.

The analysis of transcripts from different tissues of the same animal confirms the predictions of the ‘tissue coordination model’ (Burtis et al., 1990; Thummel et al., 1990) and excludes problems that derive from the developmental heterogeneity of groups of larvae. For the late larval response alone we can distinguish four major differences between tissues. The first is the timing of the initiation of the response, which derives from the observation of E75A activity at PS1 in the gut, i.e. prior to the hormone response in the other tissues. The second is the rapidity of the response, e.g. both E74B to E74A switching and the disappearance of EcR transcripts occurs more rapidly in the salivary gland than in the other tissues. The third is the difference in profiles of E75A and E75B transcripts, which show four distinct patterns in our PS1-11 study. Finally although BR-C profiles are relatively constant in this period they show clear tissue differences suggesting that tissuespecific factors modulate splice site selection.

The first two points suggest that effective hormone titres or receptor concentrations may differ between tissues. Effective hormone titres depend on the ecdysone-metabolising enzymes present in each tissue, which are not uniformly distributed. Although we have used ecdysone as a synonym for 20-OH ecdysone in vitro, at this stage we cannot exclude the possibility that in vivo the active ecdysone(s) may differ between tissues. This difference may be increased by differences in EcR levels (or isoforms) or the presence or absence of other members of the receptor family leading to tissue-specific heteromers. For EcR there are differences between tissues (Fig. 7), which require further quantitative and qualitative (isoforms) analyses. From the activity of E75 shown in Fig. 7 and the analysis of other animals (data not shown) it appears that the late larval ecdysone response is seen first in the gut followed by the wing disc and fat body and finally in the salivary gland and malpighian tubules. For E74, apart from minor differences in the rapidity of the response, the profiles are very similar in the different tissues even if the abundance of E74 transcripts relative to those of the other early genes differs between tissues (compare transcript levels in the gut and fat body, Fig. 7). Unlike the salivary gland, E74A transcripts remain abundant in the fat body at PS11, which as Boyd et al. (1991) have noted would explain the extended activity of the 74EF puff in this tissue (Richards, 1982). In vitro, Karim and Thummel (1991) detected differences from this pattern in the ecdysone induction of E74A and E74B in brains and anterior gut. In vivo, we did not observe such differences in brain (data not shown) and whole gut analyses (Fig. 7) and our results support the conclusion, based on the overall similarity of E74A expression in most tissues (Boyd et al., 1991), that although E74 plays a central role in the ecdysone response it is not directly responsible for tissue-specific responses. It remains to be seen if the tissue differences in E74 transcript profiles are reflected in the levels of the corresponding proteins and, more importantly, have consequences for network regulation.

Results with E75 suggest two sources of tissue specificity. In the salivary gland the transcript profiles can be explained by the activation of the E75A promoter closely followed by the E75B promoter with transcript length then determining the order of appearance (Karim and Thummel, 1992). This clearly does not apply to the other tissues, where, in addition, E75B repression is also different. One interpretation is that pre-existing tissue-specific factors alter the promoter sensitivities relative to those in the salivary gland. If this is the case the assignment of an early promoter to Class I or Class II may be tissue dependent. Indeed, based on the timing of its induction in the WD, E75B would be classed as a late transcript. A more interesting possibility is that the ecdysone titre may start to increase earlier but does so less rapidly in other tissues. In particular, the insect gut is important for ecdysone metabolism and excretion (see Isaac and Slinger, 1989) and in consequence much of the hormone delivered by the haemolymph may be inactivated rather than binding to nuclear receptors. If this is the case, a slower increase in ecdysone titre may trigger the E75A promoter significantly earlier than the E75B promoter so that in spite of the differences in transcript length, it is the E75A transcript that appears first in this tissue. Similar arguments will explain E75 profiles in other tissues. However, the two mechanisms are not exclusive and both may contribute to the tissue differences we observe. In the tissues we have followed, E75C appears to be a minor isoform, which may explain the absence of E75C in the report of E75 salivary gland cDNAs (Segraves and Hogness, 1990), the isoform, being first isolated from an adult head cDNA library (Feigl et al., 1989). It remains to be seen whether the corresponding protein plays a regulatory role during this period.

By RT-PCR we have obtained a molecular overview of the events underlying the regulation of gene networks first studied by chromosome puffing. The technique provides an in vivo characterisation of each ecdysone response in terms of early gene isoform expression and generalises the analysis to tissues and stages where puffing analysis is not possible. Its extension to studies of other genes and other developmental problems is evident. As most of these molecules have homologues in vertebrates and recent results show the high degree of conservation for heterodimerisation and DNA binding amongst invertebrate and vertebrate members of the steroid receptor superfamily (Yao et al., 1992; Thomas et al., 1993), there is an obvious application to the study of hormonal responses in vertebrates. The approach is complementary to studies that use isoform-specific antibodies to establish developmental profiles of such regulatory molecules. As we detect transcripts of a number of known or putative transcription factors in all of the tissues examined, it should prove possible to follow the regulatory molecules of other gene networks (even if they are ten-fold less abundant). This is particularly appropriate where the tissues are relatively homogenous. The ability to work in vivo (or in organ culture) with tissues from individual staged larvae will prove useful for the analysis of crosstalk in the ecdysone-induced networks where transcripts of key genes can be analysed either in existing mutants or in transgenic larvae carrying in vitro modified alleles.

We thank Greg Guild, Felix Karim and Carl Thummel for helpful discussions and sharing results prior to publication. We are grateful to Frank Ruffenach for the synthesis of oligonucleotides. We thank Angela Giangrande and Bohdan Wasylyk for a critical reading of the manuscript which was prepared with the help of Bernard Boulay. We thank Luc Maroteaux and Jean-Luc Dimarcq for suggestions concerning details of the RT-PCR assay.

This work was supported by funds from the Institut National de la Santé et de la Recherche Médicale, the Centre National de la Recherche Scientifique, the Centre Hospitalier Universitaire Régional, the Association pour la Recherche sur le Cancer and the Fondation pour la Recherche Médicale. F. H. was supported by a MRES studentship.