Glue protein genes in Drosophila virilis: their organization, developmental control of transcription and specific mRNA degradation

In third instar larvae of Drosophila, the major function of the salivary glands is the production of a mucoprotein glue. This glue is extruded shortly after puparium formation and serves to attach the pupal case to the substrate (Fraenkel and Brooks, 1953). In D. melanogaster, the genes coding for five glue proteins have been cloned and characterized (for review see Meyerowitz et al. 1987). Cis-acting sequences in flanking regions that are responsible for the concerted inactivation of these genes have been sought by sequence comparison, functional tests and spotting of DNasel-sensitivity (Garfinkel et al. 1983; Shore and Guild, 1986; Hofmann and Korge, 1987; Meyerowitz et al. 1987; Martin et al. 1988; Jongens et al. 1988; Ramain et al. 1988; Martin et al. 1989). But until now, no coherent picture as to their number, structure or position has emerged. The identification of functionally important motifs should be speeded up, if functional tests concentrate on conserved upstream sequences of distantly related Drosophila species. This strategy has been applied successfully for the recognition of the β 2-tubulin gene promoter (Michiels et al. 1987, 1989).

It was our intention to start a similar approach in the case of larval glue protein genes, using D. virilis for comparison. This species belongs to the subgenus Drosophila which is separated from the subgenus Sophophora (of which D. melanogaster is a member) by at least 52 million years of independent evolution (Maclntire and Collier, 1986). In D. virilis two glue proteins have been described (Kress, 1982). The gene of the major component lgp-1 was isolated from microcloned DNA (Kress et al. 1990). In the present report, we analyse this gene at the DNA level and characterize its transcripts which are subjected to specific degradation during development. We compare the deduced amino acid sequence with those of D. melanogaster glue proteins. Furthermore, we look for conserved sequence elements in the 5′ flanking region which might be involved in the regulation of glue protein gene expression. Finally, we briefly describe attempts to identify other glue protein genes in the proximity of the Lgp-1 gene.

All experiments were carried out with a Drosophila virilis wild-type strain, obtained from the University of Bochum collection.

Plasmid purification, DNA blotting and RNA blotting were done by standard procedures (Maniatis et al. 1982). Hybridization probes were prepared by random primed labeling or nick-translation. End-labeled probes were obtained by fill-in reactions with Klenow polymerase or by phosphorylation with T4 polynucleotide kinase. For determination of gene orientation a probe was prepared using ssDNA derived from mpl8/mpl9 subclones as template for Klenow polymerase synthesis in the presence of ³²P-dATP. The reaction was initiated by the anneal of primers complementary to the 5′ region of the multiple cloning site.

The clone λ DvA31 had been isolated from a mini library obtained by microcloning of EcoRI digested DNA excised from the 16A region on chromosome X of D. virilis (Kress et al. 1990). The 2.92 kb insert of genomic DNA (fA31) was subcloned into the EcoRI site of the vector pUC8. The resulting plasmid was called pDvA31.

DNA sequence was determined by the dideoxy chain-termination method (Sanger et al. 1977) either with Klenow polymerase or sequenase, using single-stranded mpl8/mpl9 subclones as templates.

Nuclease protection assays were carried out according to Calzone et al. (1987). Total RNA isolated from 10, 20, or 50 salivary gland pairs was denatured together with 150 – 200 ng of the respective end-labeled fragment in 10 μ 1 of a buffer containing 40mM-Pipes pH 6.4, ImM-EDTA, 400mM-NaCl and 80% (v/v) formamide for 15 min at 90 °C. The mixture was cooled rapidly to the calculated hybridization temperature (44% C+G content: 50°C; 48% C+G content: 51°C) and was incubated for at least 3.5 h. Where a short DNA-RNA duplex (60bp) was expected, the hybridization was carried out in 50 or 65% (v/v) formamide at 43 °C. Hybridization was followed by nuclease digestion for 30 min at 25 °C and 37 °C, respectively, with SI nuclease concentration of 250 i.u. ml^-1. The protected fragments were analyzed on 6% or 7% polyacrylamide/urea sequencing gels.

The DNA of adult animals was isolated by a modification of the method of Kirby (1962). About 10 g of flies were crushed in a chilled (N₂) mortar. After addition of 10 ml TKM buffer (10 mM-Tris-HCl pH 7.8, 5mM-KCl, 3mM-MgCl₂) and Img RNase A the fly powder was homogenized in an ice-cold glass homogenizer. The homogenate was centrifuged for 5 min at 3000 revs min^-1 (4°C). The pelleted nuclei were resuspended in a solution containing lOOmM-EDTA-NaOH pH 7.5, 400mM-NaCl, 0.2% SDS. After addition of 1 mg Proteinase K the nuclei were lysed by slow shaking at 37 °C. The viscous mixture was extracted three times with phenol/chloroform/ isoamylalcohol (25:24:1) and once with chloroform. After precipitating the DNA with an equal volume ethanol (4°C), the DNA was removed with a glass rod and dissolved in TE buffer.

Salivary gland RNA was prepared according to the method described by Kress et al. (1990).

For RNase H digestion, the RNA was denatured for 5 min at 65 °C in 29 μl of a buffer containing 10 mM-Tris-HCl pH 7.5, 5mM-MgC1₂, 50mM-KCl, 50 μ M-DTT, 2.5% (w/v) sucrose and 350 μ g ml^-1 oligo(dT) primers. The anneal was performed by cooling the RNA-primer mixture slowly to room temperature over a period of 30min. After the addition of 1 μl RNase H (2 i.u. μ ^-1), the mixture was incubated for 30 min at 37°C. The reaction was stopped by adding phenol and chloroform. The RNA was precipitated by the addition of 0.1 volumes of 3M-sodium acetate pH 5 and 3.5 volumes of ethanol.

Poly(A)⁺ RNA and poly(A)^- RNA were isolated by chromatography on oligo (dT)-cellulose. Total RNA was dissolved in 10mM-Tris-HCl pH 7.6, 500mM-LiCl, 10 HIM-LiEDTA and 0.2% (w/v) LiDS and was applied onto an equilibrated column (same buffer). After washing the column with this buffer, the poly(A)⁺RNA was eluted with 10 mM-Tris-HCl pH 7.6, 0.2% (w/v) LiDS.

The clone pDvA31 contains a 2.92 kb EcoRI genomic insert (fA31) originating from puff 16A DNA of D. virilis (Kress et al. 1990). This fragment was characterized by restriction mapping and subfragments were used for probing Northern blots of late third larval instar salivary gland RNA. DNA from the region between the restriction sites Ssil and EcoRV (Fig. 1A, black bars) hybridized with highly abundant transcripts approx. 970 – 1070nt in length (transcript population A). Fragments between the XbaI and Hindi sites (Fig. 1A, hatched bar) hybridized with transcripts of lower abundance, 1400 – 1470 nt in length (transcript population B; Kress et al. 1990).

The orientation of the transcription units was determined by hybridization of RNA with single-stranded M13 probes containing one of the complementary strands of the corresponding fragments. Both coding regions are localized on the same DNA strand (results not shown).

The complete 2.92 kb fA31 fragment was sequenced according to the strategy shown in Fig. 1B. An open reading frame of about 670 nt was found in the region corresponding to transcript A (Fig. 2). There was no ATG start codon in frame. The presence of possible intervening sequences in this gene was tested by nuclease protection experiments. An end-labeled 485 bp SstI – RsaI fragment (Fig. 1C, probe B) was hybridized with total RNA from salivary glands of late third instar larvae and subjected to S1 nuclease digestion. The analysis of the protected fragments showed a nuclease-resistant product of 380 nt (Fig. 3A). This indicated the presence of the 5′ boundary of an exon 380 nt upstream of the RsaI site. The nucleotides 5′ of that position correspond to the 3′ splice site consensus sequence PyAG. There was an ATG start codon about 87 nt upstream of this splice site. The sequence G′GTAAAA located 28 nt downstream of this ATG represents a 5′ splice site which is relatively uncommon in invertebrate genes (Shapiro and Senapathy, 1987). The ATG of this exon is in frame with a stop codon 288 nt downstream of the RsaI site in the second exon. It may, therefore, act as start codon.

The transcription start was determined by S1 nuclease protection assays. An end-labeled 141 bp Taql fragment (Fig. 1C, probe A) was used as probe. A group of protected fragments, differing in length by one nucleotide, respectively, was detected (Fig. 3B). We assume that they were caused by the low stringency of the assay conditions. The fragment corresponding to the most intensive band had a length of 61 nt. This indicates that the start point for transcription (position + 1) is 66 nt upstream of the start codon ATG (A being +66). Thus, the corresponding RNA begins with the sequence motif ATCAG ITT. This is identical with the motif at the D. melanogaster Sgs-3 transcription start (Garfinkel et al. 1983).

The TATA box of the gene is localized 30 nt upstream of the transcription start. It is part of the sequence TATATAAAGCG (– 30 to – 22) which is homologous to the TATA box region of the Sgs-4 gene of D. melanogaster (Hofmann and Korge, 1987).

The polyadenylation site of the A transcripts was mapped by hybridization of an end-labeled 524 bp NcoI-HpaII fragment (Fig. 1C, probe C) with RNA from late third instar salivary glands. The analysis revealed an SI nuclease-resistant product of 302 nt length (Fig. 3C). This indicates the addition of the poly(A) sequence to the dimer CA (position +975/976)which is 156nt downstream of the last codon. The dimers CA or UA are the most common polyadenylation sites (Bimstiel et al. 1985). The putative polyadenylation signal (AATATA) is then located 15 nt upstream of this poly(A) addition site (position +961 to +966). This is within the usual distance range described for eucaryotic genes (Bimstiel et al. 1985).

The amino acid sequence of the corresponding protein was deduced from the open reading frame of this gene, starting with the first ATG codon behind the transcription start site (Fig. 2). This region (CGAAATG) is in accordance with the consensus sequence (C/A A A C/A ATG) of the translation initiation site defined for Drosophila genes (Cavener, 1987). The open reading frame is interrupted by an intron which causes a frame shift. The predicted translation stop codon is localized at position +821 and is followed by three further in-frame stop codons within the next 90 nucleotides. The open reading frame codes for a primary translation product comprising 232 amino acids.

A general scheme of the protein primary structure is shown in Fig. 4A. The first 25 amino acids at the N-terminus contain a high proportion of hydrophobic residues, which is a characteristic feature of signal peptides of most secretory proteins (Watson, 1984). Accordingly, we found two possible signal cleavage sites, one behind position 23 (alanine) and one behind position 25 (cysteine). The former is more probable, since its weight matrix score calculated according to von Heijne (1986) is higher (5.6:3.9, respectively).

There is an extended central domain of tandemly repeated threonine-rich elements between the signal peptide and the carboxy-terminal 61 amino acids. All threonines of the sequence and most of the prolines are localized in this repetitive central domain. In total, the protein comprises about 45 % of threonines and 13 % of prolines. The repetitive elements show different unit types. Heterogenous blocks of differing length and of the general sequence T_4-8PCPT are followed by repeated elements with the sequence TTTTX (X=P, R, Q; Fig. 4A). The carboxy-terminal part of the protein shows a high portion of cysteines.

This primary structure identifies this protein as a typical Drosophila glue protein (see discussion). The biochemical data (Kress, 1982), genetic data (Kress, 1986), stage- and tissue-specificity of expression (Kress et al. 1990) and the sequence data presented above show that the gene, from which the A transcripts are released, represents the gene for glue protein lgp-1.

A second open reading frame spanning about 300 nt was found in the region hybridizing with B transcripts (Fig. 2, positions 1561 to 1857). The two other possible reading frames contain several stop codons. The open reading frame shows no ATG start codon but a putative 3′ splice site (TAG′) at position 1558 and at position 1858 a TAA stop codon. SI nuclease protection analysis of this DNA region showed that the sequences between positions 1561 and 1998 are present in B transcripts. The 3′ terminus TTGCA′ is identical to that of the A transcripts.

The amino acid sequence deduced from this open reading frame is very similar to the carboxy-terminus of the lgp-1 protein. The region of the 54 C-terminal amino acids shows 63 % identity (Fig. 4B). It is striking that all cysteines are at identical positions. Upstream of the C-terminus five identical threonine- and prolinerich repeats of the sequence PKTIPTTT are evident (Fig. 2). It is obvious that transcript B codes for another glue protein which is related to lgp-1. We term the corresponding gene Lgp-3 (see discussion).

The Lgp-1 mRNA accumulates only during the third larval instar in salivary glands. It could be detected at first in larvae of about 110 h after oviposition. During the following 20 h, transcript concentration increased rapidly to a maximum between 130 and 140 h and subsequently dropped dramatically until puparium formation, which takes place at about 150 h. We estimate that the amount of the specific RNAs at this stage was less than 1 % of the maximum level. The decrease of concentration is accompanied by a size reduction of the transcripts. The first detectable transcripts had a length of about 1100 nt (Fig. 5). After puparium formation their length was about 900 nt. This is close to the length of the non-adenylated transcript deduced from the sequence (917nt).

The gradual change in size of the Lgp-1 transcripts indicates that the decrease of transcript length during development is continuous. For further characterization of this process, we analyzed poly(A)⁺ RNA and poly(A)^- RNA from salivary glands of third instar larvae and from 2h prepupae. Lgp-1 mRNA of third instar larvae was found mainly in the poly(A)⁺ RNA fraction. In contrast, most of the prepupal Lgp-1 transcripts were found in the poly(A)^- fraction (Fig. 6A). This suggests that the decrease of transcript length is caused by a reduction of the poly(A) tail during development. To test this assumption, we used RNase H, an endoribonuclease specific for RNA-DNA duplices. Digestion of total RNA from salivary glands of third instar larvae and prepupae in the presence of excess oligo(dT) produced deadenylated transcripts of approximately the same size (Fig. 6B). The difference in length of the adenylated and deadenylated larval Lgp-1 transcripts is about 150 to 180 nt. Moreover, deadenylated transcripts produced sharper RNA bands after electrophoresis. This suggests that in late third instar larvae the Lgp-1 mRNA consists of a population of transcripts with heterogenous length of poly(A) tracts. No difference in length was seen between RNase H treated and untreated Lgp-1 transcripts from prepupae, indicating that the in vivo prepupal Lgp-1 RNA has no or only very short poly(A) tracts. Consequently, size variation of the Lgp-1 specific transcripts during the transition from larvae to prepupae is due to a trimming of the poly(A) tract.

The reduction of the poly(A) tract in Lgp-1 transcripts was correlated with loss of these transcripts in the total RNA population. Therefore, we analyzed RNA from early prepupal salivary glands for possible specific degradation products. The 3′ ends of the prepupal Lgp-1 transcripts were mapped by SI nuclease assays using an end-labeled 868 bp NcoI-XbaI fragment as hybridization probe (Fig. 1C, probe D). Two protected fragments with minor differences in length were detected: 298 and 293 nt (Fig. 7, lane 4 and 5). With larval RNA as a control, the protected fragment had a length of 302 nt (Fig. 7, lane 2 and 3). The 3′ terminus of the RNA complementary to this fragment is identical with that of the undegraded transcript (compare Fig. 3C). The 3′ ends of the prepupal RNAs corresponding to the shorter fragments were reduced by 4 and 9 nt, respectively. Both of these smaller transcripts appeared only in the prepupal RNA population and were not detectable in larval RNA, although poly(A) shortening had already commenced at this stage of development. The two distinct, main decay products appeared to be relatively stable. Whether the smaller discernible fragments (Fig. 7) represent further in vivo degradation products or whether they are only caused by nonspecific SI nuclease effects is unclear.

Two different larval glue proteins, lgp-1 and lgp-2, were previously identified in salivary glands of D. virilis (Kress, 1982). The present study shows that the gene encoding lgp-1 resides in the puff 16A on the X-chromosome. As yet the gene encoding the minor component lgp-2 could not be localized. However, the sequence data indicate that there is a second glue protein gene in this puff. Transcript B codes for a glue protein which is obviously related to the lgp-1 protein and was called Lgp-3. In fact, separation of isolated glue proteins in prolonged urea-PAGE reveals the presence of a third glue protein which can hardly be separated from lgp-1 (data not shown). We suppose that this is the lgp-3 protein. This is substantiated by the observation that this protein shows an increased electrophoretic mobility in a mutant strain that has shorter B transcripts (data not shown). Furthermore, it is improbable that the Lgp-3 gene encodes the lgp-2 protein for two reasons. (1) We estimate that the B transcripts possess the coding capacity for a protein of a relative molecular mass of about 30000 (see next paragraph), while the electrophoretic mobility of the lgp-2 protein suggests one of 10000 – 12000 (Kress, 1982). (2) The glue protein lgp-2 is only weakly glycosylated (Kress, 1982), while even the known fragmentary sequence of Lgp-3 indicates a translation product with a high number of glycosylation sites.

Although the ORF of the 300 bp fragment hybrid-izing with B transcripts starts with a putative 3′ splice site (TAG′), we are not sure whether this ORF represents an exon of a longer transcription unit. Preliminary analysis of overlapping genomic fragments upstream of the fA31 fragment revealed a 1.1 kb DNA region, about 2 kb 5′ of the Lgp-1 gene that hybridizes also with B-transcripts. Sequence data indicate that this coding region is localized on the same strand as the fA31 coding sequences. The deduced amino acid sequence existing so far reveals the presence of internal threonine and proline-rich repetitive units and a 3′portion which is almost identical to that deduced from the 300 bp ORF of the LA31 fragment (data not shown). It is probable, therefore, that this 1.1 kb region, which codes for a peptide of about 300 amino acids, represents a great portion of the genuine Lgp-3 gene.

By comparing the Lgp-1 gene with all known sequences of larval glue genes from D. melanogaster, we detected a striking similarity of exon 1 with the first exons of the D. melanogaster genes Sgs-3, Sgs-7 and Sgs-8 (Garfinkel et al. 1983). The amino acids encoded by the first ten codons of Lgp-1 display about 70 % identity to those of the D. melanogaster Sgs-8 gene. There are identical amino acids at positions [1], 2, 3, 5, 6, 9 and 10. The degree of correspondence with the Sgs-3 and Sgs-7 genes is lower (40% in either case). In all four genes, the intron is inserted between the first and second nucleotides of codon 10 (alanine). We conclude from this similarity that our proposal of two exons (exon 1: 28 nt and exon 2 : 668 nt) which are interrupted by a short intervening sequence of 59 nt in the Lgp-1 gene is correct.

The predicted amino-terminal signal sequence of lgp-1 contains 23 amino acids. This is also the length of the experimentally determined signal peptides of the three D. melanogaster proteins, sgs-3, sgs-7 and sgs-8 (Crowley et al. 1983). Fifteen amino acids of the lgp-1 signal peptide are at identical positions in sgs-3, sgs-7 and sgs-8 (Fig. 8A). Moreover, the predicted signal cleavage site of lgp-1 and the cleavage site of sgs-3 are followed by a homologous region which spans the five amino acids DCGCP. Provided that this predicted cleavage site is used in vivo, then the N-termini of the processed lgp-1 and sgs-3 proteins are identical in the first positions with the exception of an additional cysteine at the beginning of sgs-3.

Both lgp-1 and sgs-3 contain large portions of threonine- and proline-rich tandem repeats, the threonines being the sites of oligosaccharide linkage (Beckendorf and Kafatos, 1976; Korge, 1977; Kress, 1982). In sgs-3 there are 37 tandem repeats each of 5 amino acids (Garfinkel et al. 1983). Similarly, lgp-1 contains 14 repeats also of 5 amino acids. They follow additional threonine-rich elements of differing length. Similar repeats were also found in lgp-3. The occurrence of threonine- and proline-rich sequences as repetitive elements is apparently a common property of the glue proteins lgp-1, lgp-3, and sgs-3. Threonine- and prolinerich repeats were also found in sgs-4 of D. melanogaster (Muskavitch and Hogness, 1982). However, there is no significant resemblance in the primary structures of sgs-4 and lgp-1.

In the C-terminal regions of lgp-1 and lgp-3, other characteristic features are evident when compared with the three 68C glue proteins. Seven cysteines and a number of other amino acids are at identical positions in all five proteins (Fig. 8B). These striking similarities between the D. virilis glue protçins and the D. melanogaster 68C glue proteins indicate common ancestry. The function of the cysteine-rich carboxy-terminal domains of the glue proteins is unclear. Obviously, the cysteines are more conserved than the other amino acids. Their possible function may be the production of specific structures like disulfide bonds. It should be noted that in sgs-4 a cysteine-rich C-terminus is also present, although with a different composition (Muskavitch and Hogness, 1982).

The hormone ecdysone plays a substantial role in coordinating the control of glue protein gene expression (Meyerowitz and Hogness, 1982; Crowley and Meyerowitz, 1984). Ecdysone acts at two levels: (i) in early third instar larvae low titers of ecdysone are obviously required for initiation of transcription of glue protein genes (Hansson and Lambertsson, 1983); (ii) in late third instar larvae high titers of ecdysone are responsible for the stop of transcription of these genes and the degradation of their transcripts. Several DNA motifs (Hoffmann and Corees, 1986; Mestril et al. 1986; Riddihough and Pelham, 1987) have been postulated as possible binding sites for ecdysone receptors (ECR elements). However, there is no uniform view of the interaction between hormone-receptor complexes and ECR-elements.

We analyzed the 5′ untranscribed region of the Lgp-1 gene for homologies to the described ECR elements. At positions – 318 to – 281 relative to the transcription start, the sequence motif ATTTGTCCAT-18-ATAA-CAAAT was found, which displays a striking similarity to the hsp23 element ATTTTCCAT-19-ATGGCAGAT from D. melanogaster (Mestril et al. 1986; Fig. 9A). The two half sites of each element represent incomplete inverted repeats. In hsp23 both half sites appear to act as independent ECR binding sites, which are required for hormonal regulation of the gene (Mestril et al. 1986). A further motif (AAGGGTTCA), similar to a part of the hsp27 element (Riddihough and Pelham, 1987), was found at positions – 303 to – 294, i.e. in the center of the hsp23-like dyad element (Fig. 9A). A combination of hsp23 and hsp27 promoter sequences has also been reported for the Sgs-4 enhancer (Jongens et al. 1988).

In the proximal 5′ untranscribed region of Lgp-1, two further conserved motifs were identified which are shared with other glue protein genes. The Lgp-1 sequence TAITTGCTC (positions – 122 to – 113, Fig. 9B) shows an obvious homology to a similar element found in Sgs-3, Sgs-4, Sgs-5, Sgs-7 and Sgs-8. The distance of these elements to the transcription start site ranges between -80 to -140. The consensus sequence (T A/G TTTG C/T N C/T) corresponds in 7 positions to the OCTA consensus sequence (Al l 1G-CAT) and in 8 positions to the related U2 or IG octamer sequence (Bohmann et al. 1987). The OCTA element is a functional component of mammalian enhancers (for review see Falkner et al. 1986). The octamer sequence acts as binding site for transcription factors (Levine and Hoey, 1988; Müller et al. 1988; Robertson, 1988; Scheidereit et al. 1988). Functionally similar proteins were also found in D. melanogaster (Thali et al. 1988). Our assumption is that the conserved octamer-like sequence, described here, may act as a binding site for a transcription factor. We call this proximal box ^‘Prox 1 motif’.

In Lgp-1 and Sgs-3, an additional conserved sequence (ATTT-4-AATTG) was found at an identical distance from the Prox 1 boxes of both genes. This element (Prox 1′) is localized 19 nt downstream of the Prox 1 motif (Fig. 9B). It is also found at similar positions in the promoter regions of Sgs-4, Sgs-7 and Sgs-8 (Fig. 9B). In Sgs-3, deletion of short DNA fragments containing either the Prox 1 or Prox 1′ motif causes impairment of transcription, while their simultaneous deletion prevents transcription completely (Martin et al. 1989). Therefore, these motifs may be involved in the control of transcription of glue protein genes.

The analysis of the Lgp-1 transcripts during the transition from larvae to prepupae showed a size reduction correlated with the loss of the transcripts. The decrease of transcript length was shown to be due to a continuous shortening of the poly(A) tracts. It should be noted that Lgp-3 transcripts show a comparable shortening during the same developmental period (data not shown), suggesting a similar regulation.

Developmental changes in the length of the poly(A) tail of specific mRNAs were repeatedly observed (Colot and Rosbash, 1982; Dworkin and Dworkin-Rastl, 1985; Rosenthal and Ruderman, 1987; Rosenthal and Wilt, 1986). The transcripts of the six co-regulated 71E ‘late’ genes of D. melanogaster, which are expressed in the prepupal salivary glands, undergo progressive shortening of their poly(A) tracts (Restifo and Guild, 1986). In all of these cases, poly(A) shortening appears to be a specifically regulated process since other cytoplasmic mRNAs are not affected. This holds true also for the Lgp-1 transcript deadenylation as the poly(A) tracts of other mRNAs in D. virilis salivary glands remain stable during the same developmental period (Th ü roff, personal communication).

Many studies have shown that poly(A) reduction below a critical length is correlated with destabilization of mRNA (Wilson et al. 1978; Deshpande et al. 1979; Colot and Rosbash, 1982). Poly(A) shortening of the Lgp-1 transcripts starts in the late third larval stage and is followed by further degradation at the 3′ untranslated end. The succession of these processes led us to assume that deadenylation is the initial process in Lgp-1 mRNA degradation. Subsequent specific nucleolytic attacks begin with the removal of 4 or 9 nt, respectively.

The detection of such defined degradation products is in accordance with other reports about the initiation of nucleolytic cleavages near the 3′ terminus (for review see Brawerman, 1987; 1989). Similar correlations between poly(A) trimming and mRNA decay have been described for the human c-myc mRNA in vitro (Brewer and Ross, 1988) and for the mouse c-fos mRNA in vivo (Wilson and Treisman, 1988).

The Lgp-1 mRNA accumulates to substantial levels in salivary glands of third instar larvae. This mRNA seems to be very stable since its translation into the protein continues unimpaired for 5 h in the presence of actinomycin D. In contrast, application of ecdysone induces a complete stop of lgp-1 synthesis during the same period. This process is dependent on RNA synthesis for at least 2h after hormone application (Kress, 1979). Thus, the hormone may induce the synthesis of products which are responsible for the specific degradation of the Lgp-1 mRNA at the end of the third larval instar. Increased instability of 68C transcripts in isolated salivary glands after addition of ecdysone has also been described by Crowley and Meyerowitz (1984).

In summary, the cessation of glue protein synthesis at the end of the third larval instar seems to be controlled by ecdysone at two different levels: (i) by the repression of transcription and (ii) by specific inactivation of distinct mRNAs. While in the first case preexisting transcription factors and hormone receptors seem to be involved, the second process requires the synthesis of new gene products.

We thank Dr E. Thiiroff for discussion and reading of the manuscript. We are grateful to Ms S. Maletz for excellent technical assistance. This work was supported by the Deutsche Forschungsgemeinschaft (Kr 509/7).