Akh Biosynthesis: Transcriptional and Translational Control of Two Co-Localised Prohormones

Neurosecretory cells and peptidergic neurones typically produce more than one peptide by the processing of precursor polypeptides which contain the sequences of more than one peptide. In cells producing multiple peptides from a single precursor, the ratios of peptides can be regulated by mechanisms involving segregation of different parts of the precursor into separate subcellular compartments followed by differential degradation (Fisher et al. 1988; Sossin et al. 1990a,b). Such mechanisms can produce peptides in ratios very different from their representation in the precursor. The fact that peptidergic neurones can configure their peptide cocktails in different ways has led to the idea that peptide ratios may represent an important part of the signal from such cells.

We have studied peptide ratios in neurosecretory cells of the locust (Schistocerca gregaria) which produce a set of five peptides (two monomers and three dimers) by the co-expression of two prohormones. Peptide ratios are altered by regulating the relative in vivo synthesis rates of the two prohormones, segregation followed by degradation is not involved and peptide ratios can change while remaining in stoichiometric relationship to the precursor structure (Hekimi et al. 1989, 1991). Here we examine the mechanisms by which this differential regulation of prohormone biosynthesis is achieved.

In the locust Schistocerca gregaria, the paired glandular lobes of the corpora cardiaca (CC) are remarkable structures consisting of a homogeneous population of neurosecretory cells (about 6000 in the adult animal) that synthesize two major neuropeptides called adipokinetic hormone I and II (AKH I and II). AKH I and AKH II are ten and eight amino acids long, respectively (Stone et al. 1976; Siegert et al. 1985), and are members of an arthropod peptide family that has a wide range of important physiological functions (reviewed by Gäde, 1990). AKH I predominates in the adult CC, where it is co-localised in a ratio of about 4.5:1 with AKH II. They are released into the circulation during flight and stimulate lipid metabolism, enabling locusts to undertake prolonged migratory flights without feeding. The AKHs have also been implicated in the modulation of activity of neurones, skeletal muscle, gut and heart (for reviews, see Orchard, 1987; O’Shea and Rayne, 1992).

Studies on the biosynthesis of AKH I and II showed that they are derived from two very small and structurally simple prohormones called proAKH I (or A-chain) and proAKH II (or B-chain). ProAKH I contains one copy of AKH I, followed by a processing site (Gly-Lys-Arg) and a 28-residue C-terminal sequence called the β -chain (Hekimi et al. 1989). ProAKH II contains one copy of AKH II, a processing site (Gly-Arg-Arg) and another 28-residue C-terminal sequence called the β-chain (Hekimi et al. 1991). Further characterization of AKH biosynthesis revealed the surprising finding that, prior to processing, the prohormones are arranged into three dimeric precursors called P1, P2 and P3 by the formation of a single disulphide bond between the prohormone subunits. P1 is a proAKH I homodimer (A–A), P2 is a heterodimer (A–B), and P3 is a proAKH II homodimer (B–B). Dimerisation is a prerequisite for in vivo precursor processing that will yield five peptides: the two monomeric AKH peptides and three dimeric peptides formed from the α and β sequences of the two prohormones. The dimeric peptides are called the AKH-precursor-related peptides or APRP 1, 2 and 3 (Hekimi et al. 1991). As yet, the APRPs have no known functions but they are released with the AKHs (Hekimi and O’Shea, 1989).

In studies undertaken to investigate how the homodimeric precursor P1 is made, the first issue was to determine whether a messenger RNA encoded more than one copy of proAKH I (A-chain), or whether each subunit was independently translated from a small mRNA containing only one copy. Results described by Schulz-Aellen et al. (1989) show unambiguously that the CC contains two small mRNAs which are translated in vitro into two very small proteins. The larger of these proteins was identified as preproAKH I. This was further confirmed by DNA sequencing of a clone isolated from a cDNA library, made from CC poly(A)+ RNA, which contains only one copy of proAKH I preceded by a 22 amino acid signal sequence. These results demonstrated that P1 is not made from one larger protein containing two copies of the subunit but from independently translated small protein subunits.

In the present paper, we describe an mRNA encoding preproAKH II and show that it corresponds to the second major product of in vitro translation and that it codes for only one copy of an AKH prohormone (proAKH II). This result indicates that AKH biosynthesis involves the production of dimeric precursors from two independently translated and very small preprohormones. The synthesis of the heterodimer P2 (an AKH I and AKH II precursor) therefore requires the translation of two different mRNAs. Previously we showed that the relative synthesis rates of the dimeric precursors P1, 2 and 3, the AKHs and the APRPs change systematically during postembryonic development (Hekimi et al. 1991). All of the complex developmental changes in peptide ratios can be explained by the differential regulation of synthesis of the two prohormone subunits of the dimeric precursors. Moreover, our results suggest that the differential regulation of prohormone biosynthesis is driven by both transcriptional and translational mechanisms. A general model accounting for all the postembryonic changes in peptide ratios is presented.

Locusts Schistocerca gregaria (Forskål) used in this study were reared in our laboratory culture. Both males and females were used since no sex-dependent differences can be measured in the regulation of AKH biosynthesis (S. Hekimi, unpublished results).

One thousand adult CC and 200 fourth- and fifth-instar CC were dissected and immediately frozen in liquid nitrogen. They were stored at -80°C until further use. Total cellular RNA was extracted using the guanidinium isothiocyanate–hot phenol method described by Feramisco et al. (1982). Several adult CC total RNA preparations were used in this study, yielding approximately 0.56 μg/CC. Total RNA was also obtained from fourth- and fifth-instar CC yielding about 0.25 μg/CC. Poly(A)+ RNA from adult CC was obtained after chromatography on an oligo(dT)-cellulose column. Routinely, 15 μg of poly(A)+ RNA was isolated from 1000 adult CC.

The design of the first oligonucleotide probes used in some experiments was based on the amino acid sequence of the A (41 residues) and B (39 residues) subunits of the AKH precursors P1 (Hekimi et al. 1989) and P2 (Hekimi et al. 1991), which were determined by direct protein sequencing. These probes were used for the screening of the cDNA library and for the hybrid arrest experiments. Their sequences and corresponding protein residues are as follows. A-chain (proAKH I)-specific oligonucleotides: oligo α2, 5′-GCT GTA GGG GTC TCC GAA GTC CGC AGC GTC-3′ (A-chain 14–23, antisense probe); oligo α3, 5 ′-GAT GCC GCC GAT TTC GGC GAT CCC TAC TCC TTC CTG TAC CGC CTG ATC-3′ (A-chain 14–29, sense probe). B-chain (proAKH II)-specific oligo-nucleotides: oligo (β1, 5′-GGC CAT GGG GTC AGC GTT GGG GTC GGC GTA-3′ (B-chain 12–21, antisense probe); oligo (32, 5′-TAC GCC GAT CCC AAC GAT GAT CCC ATG GCC TTC CTG TAC AAG CTG ATC-3′ (B-chain 12–27, sense probe).

New oligonucleotides were designed later, on the basis of the cDNA sequences, and used for the Northern analyses presented in this paper. A-chain antisense oligonucleotide probes: oligo AKH I, 5′-GGT GCC CCA GTT GGG GGT GAA GTT GAG CTG-3′ (A-chain 1–10); oligo α3′, 5′-AGA GCA CCC AGA CAT CTT CCT GGC TTC AGC-3′ (A-chain 31–40). B-chain antisense oligonucleotide probes: oligo AKH II, 5′-ACC CCA ACC GGT TGA GAA GTT GAG CTG-3 ′ (B-chain 1–9); oligo β3, 5′-CGC CAT GGG GTC AGC GTT GGG ATC TGC GTA-3′ (B-chain 12–21).

The construction of the CC cDNA library in Xgt10 has previously been described in detail (Schulz-Aellen et al. 1989).

A library containing about 3×10⁵ recombinant phages was screened using standard methods (Benton and Davis, 1977). Oligo β1 was ³²P-end-labelled using polynucleotide kinase to a specific activity of about 18 μCi μg^-1. Hybridization was done at 50°C in a 6× SSPE, 5× Denhardt’s, 0.1% SDS, 100 μg ml^-1salmon sperm DNA for 16h and posthybridization washes were performed with 6× SSPE (60mmol l^-1 NaH₂PO₄, 90mmol l^-1 NaCl, 7.5mmol l^-1 EDTA, pH7.4) at 50°C followed by with 2X SSPE at 20°C. The positive recombinant phages were submitted to a second screening under identical hybridization conditions. We here report on the further analysis of one recombinant phage containing a 600 bp insert that hybridized strongly with oligo β1. Southern blot analysis (Southern, 1975) on restriction fragments derived from this insert showed that hybridization with the probe was restricted to a fragment of 270 bp which was transferred into plasmid pUC18 and sequenced.

Sequencing of DNA was carried out using both the chemical cleavage sequencing method (Maxam and Gilbert, 1980) and the dideoxy chain termination method (Sanger et al. 1977). Each nucleotide was read an average of seven times and at least twice in both directions.

Poly(A)+ RNA (0.25 μg) from adult CC was translated in the wheat germ cell-free translational system (Amersham) in the presence of [³⁵S]methionine (1455Cimmol^-1, Amersham). After a 2-h incubation at 25°C, the translated products were analyzed by SDS–PAGE on a 17.5% polyacrylamide gel. The gel was stained with Coomassie Blue, then treated for 30min with an Enhancer solution (Enlightning, New England Nuclear), dried and exposed under Kodak XAR-5 film. The proteins used on molecular weight standards were from BRL (low range 3–43kDa).

Hybrid arrest was performed as described by Minshull and Hunt (1986). Prior to the addition of mRNA to the translational mix, CC mRNA (0.25 μg) was incubated for 10min at room temperature in 5.5 μl of 0.18mol l⁻¹ KCl in the presence of 10 units of RNAsin (Genofit) containing either no oligonucleotide or 31pmol (0.31 μg) of oligo α2 or oligo (31 (antisense oligonucleotides), or 31pmol (0.5 μg) of oligo α3 or β2 (sense oligonucleotides). Each sample was then added to the wheat germ cell-free system and incubated in appropriate conditions. An aliquot of each sample was analyzed by SDS–PAGE (see above).

Oligonucleotides (0.5 μg) were 5′-end-labelled using 50 μCi of γ-[³²P]ATP (6000Cimmol^-1, Amersham) by polynucleotide kinase T4 (Biofinex), as described by Sambrook et al. (1989). Labelled probes were separated by chromatography on Sep-Pak C₁₈ (Waters). Specific activities were 30–50 μCi μg^-1 oligonucleotide. Northern blot analyses were performed by adding 1.5×10⁶–3×10⁶ ctsmin^-1 of labelled oligonucleotide to 10ml of hybridization buffer.

Poly(A)⁺ RNA (0.5 μg) from adult CC and total RNA of adult, fourth and fifth instars CC (5 μg) were denatured in a buffer containing 50% formamide and separated by electrophoresis on 1% agarose–formaldehyde gels according to the procedure described by Sambrook et al. (1989). The RNA was then transferred overnight to nitrocellulose membranes. Following the transfer, the damp membranes were irradiated with ultraviolet light to crosslink the RNA, as described by Church and Gilbert (1984). Membranes were then hybridized with ³²P-labelled oligonucleotides. Prehybridization (2–4h) and hybridization (16–24h) were done at 42°C in 6× SSPE, 5× Denhardt’s, 1% SDS, 100 μg ml^-1of salmon sperm DNA. The filters were washed with 2× SSPE and 0.1 % SDS at successively higher temperature starting at 42°C up to 60°C and exposed under Kodak XAR-5 film. The same filter was occasionally hybridized with different probes. In such cases, before hybridization with the next probe, previous probes were washed away by incubating the filter at 85–90°C for 20min with 1XTE and 0.1% SDS. The filter was exposed overnight to ensure that the label had been removed. Autoradiograms were scanned on a chromoscan 3 (Joyce Loebl) using a 100W tungsten–halogen lamp as light source (400–750nm). To compare mRNA levels, poly(A)+ RNA or total RNA was charged in duplicate on the gel and, after transfer, half of the blot was used for hybridization with the proAKH I mRNA-specific probe (α2) and the other half with the proAKH-II-specific probe (β3).

Translation of CC poly(A)+ RNA in a wheat germ cell-free system yields two major labelled proteins of small molecular weight estimated by SDS–PAGE to be 6.8 and 5.7kDa (Schulz-Aellen et al. 1989). Hybrid arrest experiments demonstrated that the heavier product corresponded to preproAKH I but the identity of the smaller protein remained unresolved because the sequence of the second AKH prohormone was not known.

A CC cDNA library constructed in λgt10 was screened for proAKH II using an oligonucleotide (oligo (31) based on the amino acid sequence of the β-chain of APRP 2. APRP 2 is the dimeric byproduct of processing of P2 (Hekimi et al. 1991), a heterodimer of proAKH I and proAKH II. The oligo β1 would therefore be expected to identify clones encoding proAKH II. A positive recombinant phage was characterized by DNA sequencing of the insert and our results show that the insert encodes a 61 amino acid protein, consisting of a 22 amino acid signal sequence, followed by one copy of AKH II, a processing site (Gly-Arg-Arg) and the 28 residue sequence of the β-chain(Fig. 1). This is the expected structure of preproAKH II.

Calculated molecular weights of the preproAKH I and preproAKH II proteins inferred from the cDNA clones are 6.84kDa and 6.64kDa, respectively. Proteins of such similar size cannot be separated under the experimental conditions used for the SDS–PAGE. The two major translation products in the wheat germ system were, however, clearly separated and estimated to be 6.8 and 5.7kDa. The larger is the correct size for the preproAKH I and its translation is specifically blocked by appropriate antisense oligonucleotides (Fig. 2). The smaller protein, however, appears to be too small to be preproAKH II as postulated from cDNA cloning. It was important, therefore, to show that the smaller protein did represent the product of translation of the preproAKH-II-encoding mRNA. Although the protein translated in vitro is smaller than the expected weight of preproAKH II inferred from the cDNA clone, a hybrid arrest experiment (Fig. 2) performed using a proAKH-II-specific oligonucleotide, designed to the least homologous portion between the two prohormones, does specifically suppress synthesis of the 5.7kDa protein. Fig. 2 shows a hybrid arrest experiment using both proAKH-I- and proAKH-II-specific oligonucleotides in their sense and antisense forms. Using the proAKH-I (α-chain specific) antisense oligonucleotide, the upper band (6.8kDa) could be specifically suppressed, as expected. The lower band of 5.7kDa is abolished only with the proAKH-II-specific antisense oligonucleotide, which does not block the translation of the upper 6.8kDa protein. Both of the sense oligonucleotides are without effect. The most likely explanation for this is that the smaller protein is indeed preproAKH II, but that the second methionine present in the signal sequence and located eight amino acids downstream from the presumed first methionine (see Fig. 1: second methionine underlined) is used preferentially as the starting codon in the wheat germ cell-free translation system. The calculated molecular weight of a protein starting at this second methionine is 5.63kDa, corresponding to the estimated size of the smaller translational product. We cannot rule out the possibility that the first starting codon is also used in vitro. In fact, whereas the proAKH II antisense oligonucleotide totally abolishes the 5.7kDa protein, prehybridization with the proAKH-I-specific antisense probe does not entirely suppress the 6.8kDa product (Fig. 2). There is usually about 8% of the radioactivity in the 6.8kDa band which cannot be suppressed with the proAKH I antisense oligonucleotide. It is therefore possible that the first methionine is used in vitro as the starting codon for preproAKH II but only about 8% of the time, resulting in the translation of some complete preproAKH II, which co-migrates with preproAKH I.

An estimation of the size of the full-length AKH II mRNA was obtained by Northern blotting. As shown in Fig. 3, the proAKH-II-specific oligonucleotide (oligo β3) hybridizes to a unique band corresponding to an mRNA of about 560 bases, which is very similar to that estimated for the AKH I mRNA (Schulz-Aellen et al. 1989).

The sequence identity between preproAKH I and preproAKH II was assessed at the nucleic acid and amino acid levels. Table 1 shows that the sequences encoding the A-and B-chains are more conserved (70% homology) than the signal sequences (59 % homology). There is a high number of single-base mutations leading to the amino acid differences and the prohormones therefore have a higher homology at the nucleic acid level (64.5%) than at the amino acid level (59%).

In summary, the results of the DNA sequencing and in vitro translation show that AKH I and II biosynthesis depends on the expression of two different small messenger RNAs. These encode the two subunits of the three dimeric AKH precursors (P1, P2 and P3) identified and characterised by HPLC and direct protein sequencing (Hekimi et al. 1989, 1991). How can this arrangement explain the changing ratios of the five peptides generated from these three precursors in the CC?

In the adult CC, there is approximately 4.5 times more AKH I than AKH II because the rate of AKH I synthesis is 4.5 times that of AKH II in vivo (Hekimi et al. 1991). In accordance with this and if transcription were the only mechanism regulating relative AKH expression, one might expect 4.5 times more AKH I mRNA than AKH II mRNA. In our Northern analysis of adult CC poly(A)+ RNA (Fig. 3), using probes that can discriminate between the two mRNAs and labelled to the same specific activity, we have compared the relative amounts of these messages. Densitometrical measures performed on this Northern blot and on several independent poly(A)+ RNA preparations showed that the mean ratio of proAKH I mRNA to proAKH II mRNA is about 1.7:1 (mean 1.65±0.2, S.D.). Thus, in the adult the mRNA ratio is far from the expected 4.5:1 indicated by in vivo protein synthesis.

The discrepancy between the 4.5:1 ratio of AKH I:AKH II synthesis and the 1.7:1 ratio of the mRNAs could be explained if the AKH I mRNA were more efficiently translated than the AKH II mRNA by a factor of about 2.6. Evaluation of mRNA translatability was made in the in vitro wheat germ translation system. Densitometrical analyses of the two proteins synthesized in vitro are consistent with the results of Northern analyses, giving a mean ratio of 1.8:1 (preproAKH I to preproAKH II), i.e. similar to the ratio of the mRNAs. This indicates that both mRNAs are translated in vitro with equal efficiency, producing an amount of protein directly related to the relative mRNA levels. We know, however, that during translation in vivo there is relatively more proAKH I produced. This indicates that the relative translatability of the two mRNAs in vivo is not equal, the AKH I mRNA being more efficiently translated by a factor calculated from the mean relative levels of the two mRNAs and the relative levels of AKH I and AKH II synthesis. These calculations suggest that in vivo AKH I mRNA is translated about 2.6 times more efficiently than AKH II mRNA and indicate the presence in the CC, but not in the wheat germ system, of a factor or factors which differentially regulates translation of the two mRNAs. We cannot, of course, distinguish between the relative upregulation of AKH I mRNA or the downregulation of AKH II mRNA. Nevertheless, the bias towards proAKH I translation is a significant factor which determines ultimate peptide ratios and this view is further supported by developmental results described below, which show that levels of the two transcripts change while the translational bias remains constant.

The adult AKH I:AKH II ratio is achieved following systematic changes in the relative rates of precursor biosynthesis during postembryonic development, progressively to favour AKH I synthesis (Hekimi et al. 1991). There are in the locust five larval stages or instars prior to adulthood. In the first stage (first instar), the ratio of AKH I to AKH II is about 1:1, it reaches 2.5:1 in the fourth instar, 3:1 in the fifth instar and thereafter increases to the adult ratio of 4.5:1, sometimes reaching 5:1 in older adult animals. In order to assess the contribution of transcript levels to these changes we have determined the relative amounts of AKH I and AKH II mRNAs present at different stages of development. Owing to the minute amount of tissue available from the CC of larval animals, experiments were performed on total RNA preparations using the specific antisense oligonucleotides described above. Furthermore, it proved impractical to perform Northern analysis on RNA from stages of development earlier than the fourth instar. As the most pronounced changes in relative AKH I and AKH II synthesis occur between the fourth and fifth instars and the adult, these stages provide the best opportunity to detect correlated changes in relative mRNA levels.

Comparisons were made, using Northern analysis and densitometry, between amounts of AKH I mRNA and AKH II mRNA present in the same quantity of total RNA extracted from fourth and fifth instars and adult (Fig. 4). Levels of both mRNAs increase but they do not increase at the same rate. In the fourth instar, the two messengers are present in approximately equal amounts, with slightly more AKH II mRNA. By the fifth instar AKH I mRNA predominates, giving a ratio of 1.14:1 (AKH I mRNA:AKH II mRNA), and this difference increases further to the adult ratio of 1.7:1 (Fig. 5). Note that the adult ratio obtained from a total RNA preparation is almost precisely that obtained by analysis of polyA+ RNA (see above and Fig. 3). Fig. 5 also shows that changes in relative transcript levels and suggests that differential transcriptional parallel the changes in the AKH ratios over this period of development and suggests that differential transcriptional regulation can account for the trend of the changing peptide ratios. Notice that the abrupt change in the transcript ratios between the fifth instar and the adult is accounted for by a reduction in the rate of increase in AKH II mRNA rather than by an upregulation of AKH I transcript expression. This feature was also noted for developmental changes in AKH I and II biosynthesis (Hekimi et al. 1991).

While differential control of transcript levels clearly operates during postembryonic development, it alone cannot explain the differences in the levels of AKH I and AKH II biosynthesis. For example, as described above for the adult, the AKH I:AKH II translational ratio is 4.5:1 in vivo, whereas the corresponding transcript ratio is only 1.7:1. It was argued from the translation ratio in vitro that this difference indicated an in vivo translational bias of about 2.6 favouring proAKH I production. This factor appears to be preserved in fourth and fifth instars also, as indicated in Fig. 6, in which the mRNA and peptide ratios are compared. At the fourth instar, the peptide ratio is 2.77 times higher than the corresponding mRNA ratio, at the fifth instar 2.63 times and in the adult 2.64 times. It appears, therefore, that the developmental changes in peptide ratio can be explained by a combination of the differential regulation of transcript levels played against a background of a consistent difference in the relative translatability of the two mRNAs in vivo.

In all the experiments performed on several different total RNA preparations, a second band was observed when α-chain oligonucleotides were used for hybridization. This small RNA of about 250 bases is present in instar and adult total RNA (Fig. 4) but not in poly(A)+ RNA preparations. It hybridizes neither with the β-chain-specific probe nor with probes corresponding to the AKH I and AKH II sequences. A probe complementary to the proAKH I mRNA, encoding the C-terminal of the α-chain (oligo α3′), hybridizes with the same stringency as the N-terminal α-chain-specific oligonucleotide (α2) (result not shown). This suggests that this small RNA may encode the complete α-chain of proAKH I but not AKH itself. The amount of this RNA compared with that of the proAKH I mRNA was estimated by scanning the autoradiograms. In five different preparations, the radioactivity associated with the smaller RNA was 60–70% of that of the proAKH I mRNA. Although there is no clue at the moment as to the primary structure and possible role of such RNA, it seems to be related to the α-chain only.

Our experiments show that two different mRNA species encode two prohormones that subsequently dimerise to form the three precursors of AKH I and II. These mRNAs are homologous in arrangement and coding sequence and are of similar size, around 560 bases. They both have one open-reading frame encoding equal-length but different hydrophobic signal peptides followed by the hormone sequence (AKH I or AKH II), a glycine residue for carboxy-terminal amidation, a dibasic processing site (Lys-Arg or Arg-Arg) and a C-terminal chain of 28 residues (α or β). By blocking their translation in vitro with specific oligonucleotide probes we have shown that the AKH I mRNA yields a preproAKH I protein of 6.8kDA and the AKH II mRNA yields a protein of 5.7kDA. The smaller in vitro protein corresponds to preproAKH II but appears to have a signal peptide nine amino acids shorter than inferred from cDNA sequencing. This may be due to the preferential use of a second starting codon located within the leader sequence, although the first AUG seems to lie in a more optimal context for initiation when the sequences flanking both AUGs are compared with the optimal nine-nucleotide consensus sequence, CCA/GCCAUGG, determined by Kozak (1986) for initiation in eukaryotic mRNAs. The underlined positions -3 and -4 are particularly important to ensure a high efficacy of translation initiation. Inspection of the sequence surrounding the first starting codon (GCAGGATGA) shows a better match of five nucleotides out of nine with position -3 conserved, whereas the sequence around the second codon (CTCTGATGC) shows a poor match of only four nucleotides out of nine, with those in positions -3 and -4 not conserved. Secondary structures of mRNA are involved in facilitating recognition of initiator codons (Kozak, 1990, 1991) and several small inverted repeats have been noticed within the cDNA sequenced, but the possible ‘hairpin’ structures that may result from these would seem to favour the first starting codon even more. Whether this second starting codon is preferentially used in vivo is unknown, but if it were, one may argue that the 13-residue signal peptide that would result from translation initiated at the second AUG may not be long enough for a protein to be translated into the endoplasmic reticulum for subsequent secretion. However, the analysis of von Heijne on the minimal requirements for a signal peptide (von Heijne, 1985) indicates that 13 residues would be the minimum length. Furthermore, the (-3,-1) role for the determination of the cleavage site of the signal peptide (von Heijne, 1984a) is followed here with suitable residues in these positions (serine and alanine respectively) or preproAKH II. The net charge of the signal peptide has been shown to influence both the level of translation and the efficiency of transport: a positively charged N-terminal sequence of the peptide appears to be advantageous (Vlasuk et al. 1983; von Heijne, 1984b). In this respect, a signal peptide of 22 residues would be favoured because it presents a net charge, according to principles established by von Heijne (1984b), of +2, whereas the shorter form of 13 residues bears no charge. A net charge of zero has been shown to affect synthesis of a prokaryotic proprotein, but not its secretion (Vlasuk et al. 1983). Therefore, if the second AUG codon is indeed used in vivo, then this could be the factor, or one of the factors, limiting translation of the proAKH II mRNA.

Homologies between the prohormones and mRNAs were determined in S. gregaria and compared with the sequences obtained by Noyes and Schaffer (1990) in another locust species, Schistocerca nitans. The overall identities at the nucleic acid level are higher between preproAKHs of the two locust species (97.4% for preproAKH I and 96.7% for preproAKH II) than between preproAKH I and preproAKH II in the same animal (64.5%, Table 1). The same is observed at the amino acid level, where preproAKH I and II share an overall identity of 96.8% and 93.4%, respectively, between the two species, whereas in S. gregaria the two preproAKHs only have 59% homology. The ancestral proAKH gene must therefore have duplicated and diverged before phylogenetic divergence of the two locust species.

Cells that express the AKH peptides contain all three forms of precursors P1, P2 and P3 generated by the dimerisation of two co-localized prohormones. Prohormone dimerisation appears to be a random process because it results in the production of binomial distributions of the dimeric precursors (Hekimi et al. 1991). The relative amounts of proAKH I and proAKH II determine the shape of the binomial distribution of precursors and, following enzymatic processing, the ratios of the peptides. For example, a 5:1 ratio of proAKH I:proAKH II produces a 25:10:1 binomial for P1:P2:P3 and, following processing, a 5:1 AKH 1:AKH II ratio (because there is one AKH copy per prohormone) and a 25:10:1 distribution of the dimeric APRPs. An explanation of the shifting peptide ratios therefore resides in the mechanisms that regulate the differential production of proAKH I and proAKH II.

Fig. 7 represents the ideal situation where the proAKH I and proAKH II are produced in a ratio of 5:1, resulting in a 5:1 ratio of AKH I and II (approximately the expected adult ratio). To achieve this ratio, we have assigned a 2:1 ratio of transcripts and a bias of 2.5 towards the transcription of proAKH I over proAKH II. These figures are based on our experimental findings, but the adult transcript ratio has been rounded up and the translational bias rounded down to 2.5. These changes simplify the binomial calculations, producing whole numbers, and result in peptide ratios (AKH and APRP) close to those normally achieved in the adult CC.

Although Fig. 7 approximates the adult situation, it can be simply adapted to generate correct peptide ratios at any stage of development. For example, in the fourth instar the transcript ratio is (simplified) 1:1. Application of the transcriptional bias of 2.5 produces a 2.5:1 ratio of the prohormones, which generates the binomial distribution 6.25:5:1 for the dimeric precursors (P1:P2:P3). The binomial distribution is obtained by expanding (2.5x+y)², where x is proAKH I and y is proAKH II. Following processing, the expected 2.5:1 AKH I:AKH II ratio and the above binomial distribution of the APRPs characteristic of the fourth instar are produced.

During postembryonic development, the absolute levels of both mRNAs increase as a proportion of total RNA in the CC (Fig. 4). Throughout postembryonic development there is also an increase in the number of AKH-producing neurosecretory cells (S. Kirshenbaum and M. O’Shea, in preparation) and this increase in cell number probably accounts for the absolute increase in mRNA levels. The change in the relative levels of the two mRNAs, however, indicates transcriptional differential regulation of AKH I and AKH II genes as development progresses.

Why AKH I and II peptide ratios change systematically during postembryonic development is unknown, but the most rapid changes occur late in development when the animal’s wings start to differentiate. The change may, therefore, be related to the transition between peptide function in the larval stages and in the adult, when flight behaviour first appears, suggesting that the peptide ratios must be configured appropriately for their adult functions.

Dimerisation between different subunits encoded by distinct mRNAs is clearly a process that considerably extends the diversity of the final proteins and can lead to mature products with totally different activities, or even opposite activities, as in the case of the mammalian activin/inhibin gonadal dimeric peptides (Vale et al. 1990). The AKH precursors are formed from two identical or different subunits that each carry a neurohormone with similar biological activity, i.e. AKH I and II, and a 28-residue chain, α or β. Physiological functions for the dimeric APRPs, generated by the enzymatic processing of these precursors, remain to be uncovered. One may be tempted to draw an analogy with the dimeric activin/inhibin family. These peptides have opposite effects on synthesis and secretion of FSH (follicle-stimulating hormone) from the anterior pituitary (Vale et al. 1990). In a similar manner, enzymatic processing of the AKH precursors would furnish the flying insect with two hormones (AKHs) to control fuel metabolism. The APRPs might exert different functions on identical targets or similar functions on distinct targets, possibly in relation to the complex physiology of flight behaviour. Interestingly, a potential dibasic processing site is observed in the sequence of the α- and β-chains (Arg-Lys at positions 21–22 of both 28-residue chains). Cleavage at these sites does not occur within the neurosecretory cells of the CC (Hekimi et al. 1991). It is possible that these sites are not accessible for endoproteolysis because they are in the vicinity of the cysteine residue that forms the disulphide bridge or they might be cleaved at target sites. We are now looking forward to elucidating the physiological functions of the dimeric APRPs.

We thank M. Guidon Ayala of the Glaxo Institute for Molecular Biology for synthesizing the oligonucleotide probes, M. Marc Ackermann for nursing the animals and M. Slim Chraiti for photographic work. We also thank Glaxo IMB for their hospitality to J.F.-L. during part of this work and Ms Magali Leemann-Husler and Annie Bacon for secretarial assistance. M.-F.S.-A. was the recipient of a grant from Ciba-Geigy. This work was supported in part by the Swiss National Fund (Grant no. 3.181.085), by the Science and Engineering Research Council (Grant no. GR/G52524) and by the Universities of Geneva and London.