Transcriptional regulation of neuropeptide and peptide hormone expression by the Drosophila dimmed and cryptocephal genes

Neuropeptides and peptide hormones are small chemical transmitters that carry physiological messages to the attention of specific target cells. They are present in vertebrates and invertebrates and control diverse processes,including growth, stress responses, reproduction, homeostasis and memory(Nässel, 2002; Strand, 1999). A central feature of neuroendocrine signaling is the regulation of the synthesis and secretion of neuropeptides by inputs to neurosecretory cell afferents(Burbach, 2002). In many systems, the mechanisms regulating neuropeptide expression and secretion, and the genes and cell signaling pathways underlying these processes, are largely unknown.

The Drosophila melanogaster dimm gene encodes a bHLH protein(DIMM) in the Atonal family of transcription factors(Hewes et al., 2003). This family includes NeuroD, Neurogenin, Mist1 and Olig, which play essential roles in the determination and execution of cell fate decisions in many tissues(Hassan and Bellen, 2000). Likewise, DIMM determines secretory protein levels in diverse neuropeptidergic cells. The dimm gene is highly expressed in neuroendocrine cells, and dimm mutant animals display strikingly reduced cellular levels of various neuropeptides, neuropeptide biosynthetic enzymes(Hewes et al., 2003) and a dopamine receptor (Park et al.,2004). In contrast, dimm mutations do not disrupt cell survival or the differentiation of neuropeptidergic cell types, and the functions of dimm are largely restricted to development of the neuropeptide secretory pathway (Hewes et al., 2003). Does dimm regulate the expression of other transcription factors or structural proteins required for secretory granule biosynthesis, or does dimm directly regulate the expression of many secretory proteins?

In the present study we examined whether dimm is required for normal expression of neuroendocrine genes. We monitored 16 genes encoding neuropeptides, peptide hormones, neuropeptide biosynthetic enzymes, secretory granule proteins, and enzymes involved in synthesis of biogenic amines. Levels of these transcripts in dimm mutants and in control genotypes were measured by quantitative real-time polymerase chain reaction (qRTPCR) and in situ hybridization. To disrupt dimm expression, we used several genetic aberrations that differentially disrupt dimm and/or a neighboring gene, crc. crc encodes a basic-leucine zipper (bZIP)transcription factor that is orthologous to activating transcription factor-4,ATF-4 (Hewes et al., 2000), an important mediator of the unfolded protein response to endoplasmic reticulum stress (Blais et al., 2004). We have previously tested several crc alleles (crc¹, crc^E98, crc⁹²⁹ and Df(2L)TW1)by immunostaining with anti-neuropeptide antisera (myomodulin and–RFamide) and anti-neuropeptide biosynthetic enzyme antisera (Furin 1 and Amontillado), and in each case, the levels of these markers were unaffected by disruption of crc(Hewes et al., 2003; R.S.H.,unpublished). Therefore, we predicted that secretory protein mRNAs would be found at normal levels in crc¹/crc¹larvae, and we included this genotype as a control for the qRTPCR experiments.

Levels of three neuropeptide mRNAs, Dromyosuppressin(Dms), FMRFamide-related (Fmrf) and Leucokinin (Lk), were all reduced by disruption of dimm and not crc. However, crc was required for expression of the ETH gene in the endocrine Inka cells. Comparative genome sequence analysis revealed putative recognition elements in the ETH promoter for factors in the ecdysteroid response pathway and CRC. Our results suggest that DIMM controls the transcription of multiple neuroendocrine genes. Additionally, the molting defects in animals bearing the crc¹ mutation, a classical allele first discovered in 1942(Hadorn and Gloor, 1943),result from loss of a key endocrine regulator of ecdysis behavior.

Drosophila melanogaster Meigen stocks were cultured on standard cornmeal–yeast–agar media at 22–25°C. The following alleles were used to disrupt genes in the 39D1 region of chromosome 2L(Fig. 1, Table 1, and Table S2 in the supplementary material): Df(2L)Rev8 (Rev8), Df(2L)Rev4 (Rev4) and crc¹(Hewes et al., 2000); P{SUPor-P}dimm^KG02598 (dimm^KG02598)(Hewes et al., 2003); and P{EPgy2}dimm^EY14636(Bellen et al., 2004). The ETH-EGFP reporter line (Park et al., 2002) was kindly provided by Michael Adams (University of California, Riverside). Mutations were balanced over CyO-y⁺ or CyO, P{Ubi-GFP.S65T}PAD1(CyO, Ubi-GFP). Oregon-R was used as the wild-type strain.

RNA extractions were performed on 24.5°C collections of 50 hatchling larvae on apple juice–agar plates supplemented with yeast paste. At this stage, the central nervous system (CNS) fills approximately 20–30% of the total body volume. The CNS volume:body volume ratio decreases with larval growth, and early collections maximized the relative yield of CNS mRNAs that could be obtained from whole animals. In addition, several of the neuropeptide transcripts measured in this study are expressed exclusively or primarily in the CNS (e.g. Fmrf and Dms)(Nichols, 2003; Schneider et al., 1991). Larval genotypes were distinguished by scoring for yellow(y) or UBI-GFP.

Tissues were disrupted by Polytron homogenization (Brinkmann, Westbury, NY,USA) on speed 5 for 5 min on ice. Total RNA was extracted in Trizol(Invitrogen, Carlsbad, CA, USA), in two extractions separated by a DNase treatment (RQ1 DNase kit; Promega, Madison, WI, USA). We synthesized cDNA from total RNA using random hexamer primers with the ISCRIPT kit (Bio-Rad,Hercules, CA, USA). One complete reaction and one `No Enzyme' (NoE) reaction was performed for each RNA sample, with 50 ng (by spectrophotometry) of total RNA per reaction (reverse transcriptase was omitted from the NoE reactions).

Three sets of PCR primers were designed using Primer3(Rozen and Skaletsky, 2000)for each gene in our analysis (Table 2). Based on product quality and purity using genomic DNA templates (judged by the presence of a single band of the correct size in 2%agarose electrophoresis gels and by the homogeneity of amplicon T_m values in qRTPCR dissociation curves), the best pair of primers was then selected (see Table S1 in the supplementary material). Primer concentrations were picked according to the nearest neighbor thermodynamic parameters method with salt corrections(SantaLucia, Jr, 1998) to match the conditions of the ABI qRTPCR cycle protocol (50 cycles: 15 s at 95°C followed by 1 min at 60°C on an ABI 7000; Applied Biosystems,Foster City, CA, USA).

Gene-specific qRTPCR reactions were performed with 1 μl of the reverse transcriptase mix, a pair of gene-specific primers, and SYBR green dye (ABI SYBR green PCR master mix). Each qRTPCR run was performed on a 96-well plate,providing transcript level information for 11 genes and the Ribosomal protein L32 (RpL32) control (see below) for two experimentally paired genotypes. For each gene on the plate, we performed three technical qRTPCR replicates per genotype and one `No Template' (NoT) reaction. NoT reactions lacked cDNA and were used to detect potential template-independent PCR amplification. For each genotype, we included two technical replicates with the RpL32 primer set and the NoE control to check for potential genomic DNA contamination. In all cases, PCR products in NoE and NoT reactions were at least 50-fold less concentrated than the gene-specific qRTPCR products. Thus, contamination with genomic DNA and primer-related templates was negligible. In addition, melting temperatures of the gene-specific amplicons were always consistent across the technical and biological replicates and across all genotypes (data not shown).

We performed relative quantitation analysis on qRTPCR data using the housekeeping gene, RpL32 (rp49), as a control. Levels of RpL32 mRNA were not significantly different between paired genotypes(data not shown) and were therefore not affected by mutations in the dimm region. For each PCR reaction, we obtained a Ct value representing the number of PCR cycles necessary to reach a threshold amplicon concentration. Ct values were normalized to RpL32 to obtain ΔCt values (ΔCt=Ct_{test gene}–Ct_RpL32), which were then averaged across the three technical replicates. By comparing levels of each transcript to RpL32, we confirmed consistency of the mRNA extraction, cDNA synthesis, and loading for the two paired genotypes within each experiment. In addition, normalization of test gene Ct values to those of RpL32 allowed us to compare transcript levels across experiments.

Anti-Manduca pre-ecdysis triggering hormone (anti-PETH)immunostaining (Park et al.,2002; Zitnan et al.,1999) and ETH-EGFP imaging, preparation of digoxigenin-labeled DNA probes (from genomic templates), and whole-mount larval or CNS in situ hybridization were performed as described(Hewes et al., 2003). Control and experimental genotypes were always processed in parallel within a given experiment, using the same reagents, to minimize variability. In addition, for the in situ hybridization analysis, all reactions were stopped at the same time (when the most intense signals first became dark to prevent overstaining). We then measured the intensity of each cellular signal(intensity index) as described (Hewes at al., 2003). Briefly, confocal (fluorescence) and CCD (brightfield)images were obtained after adjusting exposure settings to optimize detection without saturating the signal. For a given neuron, identical settings were used for all preparations and genotypes, and the mean pixel intensity for the area covering each soma (S), and the neighboring background (B), was measured using Adobe Photoshop (San Jose, CA, USA). The intensity index=(S–B)/B. Images shown in the figures were chosen to best represent the mean intensity index values.

Statistical analyses were performed using NCSS 2001 (Kaysville, UT, USA). Sequential Bonferroni corrections were performed to minimize type I errors in multiple pair-wise comparisons (Rice,1989). We used parametric statistics, because the data generally followed a normal distribution. All values are means ± s.e.m.

Drosophila genome sequences were visualized with VISTA (VGB2.0)(Frazer et al., 2004), using AVID and SLAGAN alignments, on the UCSC Genome Browser at http://genome.ucsc.edu/(Karolchik et al., 2003) and with the MAVID multiple alignment server at http://baboon.math.berkeley.edu/mavid/(Bray and Pachter, 2004). The alignments included sequences from eight Drosophila genomes: D. melanogaster (January 2003 assembly)(Celniker et al., 2002); D. pseudoobscura (July 2003)(Richards et al., 2005); D. yakuba (April 2004) and D. simulans (December 2004;Genome Sequencing Center, Washington University School of Medicine); D. ananassae (July 2004; The Institute for Genomic Research); D. mojavensis (August 2004), D. erecta (October 2004) and D. virilis (July 2004; Agencourt Bioscience Corporation). Consensus sequences (IUPAC code) were obtained using the TRANSFAC (see below)adaptations of the Cavener rules (Cavener,1987). The code was capitalized when the nucleotide was present in at least seven sequences in the eight-species alignment.

The conservation track (phastCons) in the UCSC Genome Browser was based on a MULTIZ alignment of the D. melanogaster, D. yakuba and D. pseudoobscura genomes. These scores present an estimate of evolutionary conservation based on phylogeny, nucleotide substitution rates and autocorrelation of conservation levels along the genome(Siepel and Haussler, 2005). Putative transcription factor binding sites were identified using rVISTA(Loots et al., 2002), using the TRANSFAC Professional 7.4 library of binding site sequences (BIOBASE Biological Databases, Wolfenbüttel, Germany).

We used qRTPCR to analyze neuropeptide gene expression because of the sensitivity of this method. This was true even with whole-animal RNA samples,because a large majority of the cells that express dimm-dependent neuroendocrine peptides in our analysis (e.g. the populations of cells that express LK and ETH; Table 2)are affected in dimm mutant animals(Hewes et al., 2003). Based on our earlier immunocytochemical studies, we chose the neuropeptide and peptide biosynthetic enzyme genes for this analysis based on whether they were expressed in patterns largely or completely overlapping with dimm,and whether they showed reductions in protein levels in dimm mutant larvae (Hewes et al., 2003). However, we also included neuropeptide genes (e.g. Pdf, Eh) encoding proteins that are known not to be affected in dimm mutants as internal controls.

Because all of the loss-of-function alleles of dimm were also loss-of-function alleles of the crc gene(Fig. 1), we used three different paired genotype comparisons, in order to reveal the effects of dimm specifically on levels of secretory protein mRNAs(Table 1). First, we performed qRTPCR to monitor transcript levels in Rev8/Rev8 larvae and Rev8/+ controls. The Rev8 deletion is a null allele of crc and a strong loss-of-function allele of dimm(Hewes et al., 2003; Hewes et al., 2000). Second,we compared dimm^KG02598/Rev4 mutants to dimm^KG02598/+ controls. The Rev4deletion is a null mutation of both crc and dimm. In contrast, the dimm^KG02598 mutation is a strong dimm loss-of-function allele, but a weak loss-of-function allele of crc (Hewes et al.,2003). Therefore, in both of the first two experiments we tested the effects of double-mutant combinations of dimm and crc,but in the second experiment, the crc defects were much less severe. In the third experiment, we compared crc¹/crc¹ larvae to crc¹/+ controls. crc¹ is a strong crc loss-of-function allele, but it does not disrupt dimm (Hewes et al.,2003; Hewes et al.,2000).

We first examined the effects of the above genotypes on genes in the 39D1 region: dimm is flanked by crc and a second gene, Tetraspanin 39D (Tsp39D). As expected, dimm and crc transcript levels were reduced in Rev8/Rev8 larvae(Fig. 2A). Rev8deletes the crc gene (Fig. 1), resulting in a dramatic decrease in crc mRNA levels(although some crc mRNA is maternally loaded)(Hewes et al., 2000). Rev8/Rev8 mutants also display markedly reduced dimm mRNA levels (Hewes et al., 2003),presumably due to the deletion of dimm gene regulatory regions. In dimm^KG02598/Rev4 mutants, levels of crc,dimm and Tsp39D transcripts were all lower than in the heterozygous controls (Fig. 2B). This result is consistent with our earlier observation that KG02598 is not only a strong hypomorphic allele of dimm but also a weak hypomorphic allele of crc(Hewes et al., 2003). We suspect that the broad effects of this insertion on genes in 39D1 are due to chromosomal insulator elements contained within the KG02598 element(Roseman et al., 1995). Finally, dimm, crc and Tsp39D transcript levels were not significantly different between crc¹/crc¹ and crc¹/+ (Fig. 2C). This result was expected because crc¹ is a missense mutation specific to crc, and because our earlier crc in situ hybridization analysis of crc¹/crc¹ animals showed no change in crc mRNA levels (Hewes et al.,2000). Thus, the levels of dimm, crc and Tsp39Dtranscripts behaved as predicted in the three qRTPCR experiments.

Only three neuropeptide mRNAs, Dms, Fmrf and ETH, varied significantly between paired genotypes in at least one of the three qRTPCR experiments. Dms transcript levels were reduced by 46% in dimm^KG02598/Rev4 animals(Fig. 2B). Levels of Dms were also down by 44% in Rev8/Rev8 mutants(Fig. 2A), although this difference was not statistically significant after the Bonferroni correction(P=0.047). In contrast, Dms transcript levels were normal in crc¹/crc¹ animals(Fig. 2C). We obtained similar results for Fmrf. Fmrf mRNA levels dropped 83% in Rev8/Rev8(Fig. 2A), and they were down 51% in dimm^KG02598/Rev4(Fig. 2B), although the latter difference was not statistically significant (P=0.13). Fmrftranscript levels were normal in crc¹/crc¹ animals(Fig. 2C). Based on our in situ hybridization data (see below), the relatively low Pvalues, and the conservative nature of the Bonferroni correction, it appears likely that the reductions of Dms in Rev8/Rev8 and of Fmrf in dimm^KG02598/Rev4 were incorrectly judged as not significantly different due to type II error (false negatives). Notably, we previously observed reduced in situhybridization with an Fmrf probe in dimm^KG02598/Rev4 larval CNS(Hewes et al., 2003). Therefore, the combined qRTPCR results suggested an effect of dimm,but not crc, on levels of Dms and Fmrf mRNA. These findings are consistent with the cellular reductions in immunocytochemical staining for the neuropeptide products of these two genes(Table 2).

The last of the three affected neuropeptide/peptide hormone mRNAs was ETH, which was reduced by 90% in the Rev8/Rev8 mutants(Fig. 2A) and by 60% in the crc¹/crc¹ mutants(Fig. 2C). While the reduction in ETH levels caused by the Rev8 chromosome was consistent with our previous studies (Table 2), the reduction in crc¹/crc¹ animals was novel, and we explored this relationship further (see below).

In the qRTPCR experiment comparing Rev8/+ and Rev8/Rev8,we did not observe significant differences in transcript levels for three neuropeptide genes, Pdf, Ccap and EH, two genes that encode known or putative components of secretory granules in neuropeptidergic cells(ia2 and Caps), and two genes, Ddc and ple, encoding enzymes involved in synthesis of biogenic amines(Fig. 2A). For Pdf and Ddc, these results are consistent with previous immunostaining data(Table 2). Thus, these seven transcripts were not affected by disruption of either dimm or crc, and we excluded them from the subsequent qRTPCR analysis of dimm^KG02598/Rev4 and crc¹/crc¹(Fig. 2B,C).

Finally, there were five genes, amon, Dsk, Fur1, Lk and Phm, for which we observed no change in mRNA levels(Fig. 2) despite marked reductions in levels of their protein products(Table 2). In some cases, these differences may be due to indirect regulation of protein levels by dimm, such as through transcriptional regulation of other elements of the regulated secretory pathway (see Discussion).

The pattern of in situ hybridization with a Dms probe was similar to the reported immunostaining pattern(Nichols, 2003). Dmswas expressed in ∼14–16 cells, with one pair in the subesophageal region (SE) and at least three pairs in each brain lobe (LB, MP2 and SP)(Fig. 3A). Additional, faintly labeled cells were sometimes visible. In dimm^KG02598/Rev4 larval CNS, we observed significantly less signal in two cell types, SP and SE, than in the dimm^KG02598/+ controls(Fig. 3B). There was also a reduction in Dms levels in the MP2 cells, although this trend was not statistically significant. In contrast, we found no significant variation in the intensity of Dms hybridization between Rev8/Rev8 and Rev8/+ (Fig. 3C) or between crc¹/crc¹ and crc¹/+ (Fig. 3D). The reason for the effect of dimm^KG02598/Rev4 but not Rev8/Rev8 on Dms transcript levels is unclear, although dimm^KG02598 may simply be a stronger dimm allele than Rev8. However, these results are in general agreement with the qRTPCR data, and we conclude that dimm, and not crc, likely upregulates Dms gene expression and/or increases the stability of the Dms mRNA.

Previously, we found a marked reduction in levels of anti-LK immunostaining in Rev8/Rev8 mutants (Hewes et al., 2003). The qRTPCR results here, however, showed no change in Lk transcript levels in Rev8/Rev8 mutants, indicating that the regulation of LK protein levels in this genotype may be post-transcriptional. Therefore, we performed Lk in situhybridization on Rev8/Rev8 larvae to further test this hypothesis. In first instar larval CNS, we detected hybridization with an Lkantisense DNA probe in a pair of cells (Br1) in the brain lobes, in two pairs of cells in the subesophageal region (SE), and seven pairs of more weakly Lk-expressing cells (A1–A7) in the ventral nerve cord (VNC)(Fig. 4A). This pattern of expression appears to be identical to the immunostaining pattern(Hewes et al., 2003). In the A1–A7 cells of Rev8/Rev8 mutant larvae, the strength of Lk hybridization was strongly reduced relative to Rev8/+controls (Fig. 4B). In contrast, levels of Lk in the SE and Br1 cells appeared to be increased in Rev8/Rev8 animals, although the increase observed in the Br1 cells was not statistically significant. These results are consistent with our qRTPCR data, since increased Lk mRNA levels in the six Br1 and SE cells likely masked a decrease in Lk levels in the 14 more weakly Lk-expressing A1–A7 cells.

These cell type-specific changes in Lk mRNA levels also mirror our anti-LK immunostaining results. In multiple different dimm^–/– genotypes, the A1–A7 cells display a greater reduction in anti-LK immunostaining than SE or Br1, although all three cell types are affected (data not shown). While we cannot exclude the possibility that crc regulates LK levels, the loss of dimm alone can account for these findings, since LK protein levels are also reduced by RNA interference directed against dimm(Hewes et al., 2003). These results show that in A1–A7, dimm likely upregulates Lkgene expression and/or increases the stability of the Lk mRNA. In SE and Br1, the responses are more complex, since dimm may downregulate Lk gene expression in these cells while increasing LK protein levels. Thus, in some cases, dimm may regulate neuropeptide synthesis at the transcriptional level as well as at a later step in the regulated secretory pathway (see Discussion).

To further test the dependence of ETH levels on crc, we performed in situ hybridization with an ETH probe in crc mutant larvae. To facilitate preparation of larval fillets, we used third instar larvae, and we observed strong ETH hybridization in seven pairs of Inka cells (O'Brien and Taghert, 1998; Park et al.,2002). ETH-positive cells were located on the dorsal-longitudinal tracheal trunks in tracheal metameres 1 and 4–9(TM1, TM4–TM9) (Manning and Krasnow,1993). Compared to heterozygous controls, we found reduced ETH hybridization in dimm^KG02598/Rev4(Fig. 5A). The cause of the difference in the results for dimm^KG02598/Rev4 in the qRTPCR versus the in situ hybridization analysis was not determined, but these experiments were performed on different larval stages,and the cumulative effects of dimm^KG02598/Rev4 on crc-dependent processes may be more pronounced in older animals. Notably, ETH in situ hybridization was markedly reduced in crc¹/crc¹ larvae(Fig. 5B), consistent with the qRTPCR results (Fig. 2C). In addition, we observed a severe reduction in anti-PETH immunostaining(Park et al., 2002) in crc¹/crc¹ Inka cells (data not shown). This antiserum interacts with ETH-like peptides from diverse insect species(Zitnan et al., 2003), and it labels peptides in the Drosophila Inka cells that are presumably ETH1 and/or ETH2 (Park et al.,2002). These results provide strong additional evidence for an important role of crc in regulating ETH expression.

Does DIMM contribute to the regulation of ETH levels in vivo in addition to CRC? We have previously shown that dimm is expressed in the Inka cells (Hewes et al.,2003), but without a specific dimm mutant allele, this question could not be addressed directly. However, shortly before our completion of these experiments, the Drosophila Gene Disruption Project (Bellen et al., 2004)reported a P element insertion, P{EPgy2}dimm^EY14636 (dimm^EY14636),inserted in the dimm open reading frame in exon 2. To determine whether dimm^EY14636 disrupts crc, we performed lethal complementation analysis with other dimm and crcalleles (see Table S2 in the supplementary material). The dimm^EY14636 allele was semi-lethal (6–50% survival)in combination with Rev4 and Rev8, and it was subvital(51–85% survival) over dimm^KG02598. In contrast, dimm^EY14636 complemented crc¹. Therefore, dimm^EY14636 selectively disrupts dimmand not crc.

In dimm^EY14636/dimm^EY14636 larvae,we observed a marked reduction in CNS levels of anti-LK immunostaining relative to dimm^EY14636/+ controls (data not shown). We also observed a small decrease in the intensity of anti-PETH immunostaining in the Inka cells in dimm^EY14636/dimm^EY14636 larvae,although the strength of ETH in situ hybridization was unaffected(see Fig. S1 in the supplementary material). Thus, crc and dimm regulate ETH through distinct mechanisms. crccontrols ETH transcription, whereas dimm can regulate ETH levels without altering ETH mRNA expression.

Park et al. defined a 382 bp ETH enhancer region that is sufficient to direct expression of an ETH-Enhanced green fluorescent protein (ETH-EGFP) transgene specifically to the 14 Inka cells(Park et al., 2002). To determine whether this regulatory region is sensitive to regulation by crc, we monitored EGFP fluorescence in crc¹/Rev4, ETH-EGFP and dimm^KG02598/Rev4, ETH-EGFP third instar larvae. In dimm^KG02598/Rev4, ETH-EGFP CNS, we observed slightly reduced levels of EGFP relative to +/Rev4, ETH-EGFP controls(Fig. 6A), but this difference was not statistically significant (P=0.056, Inka^TM5; P=0.35, Inka^TM8). We observed a much stronger reduction in EGFP fluorescence in crc¹/Rev4, ETH-EGFP larvae(Fig. 6B). These findings,together with the qRTPCR and in situ hybridization results,demonstrate crc-dependent control of ETH gene expression.

We predict that CRC controls ETH transcription by binding to regulatory sequences directly upstream of the ETH promoter. To identify potential CRC recognition elements, we obtained a comparative genome alignment of a 404 bp sequence extending from immediately 3′ of the stop codon in the Origin recognition complex subunit 4 (Orc4)gene through the first 10 bp of the ETH coding sequence(Fig. 7). This region contains the 382 bp ETH promoter region used to create the ETH-EGFPline (Park et al., 2002). In pairwise VISTA alignments of the sequence from D. melanogaster with the corresponding sequences from five other Drosophila species(pseudoobscura, yakuba, ananassae, mojavensis, and virilis),we detected three highly conserved regions(Fig. 7A). One was centered on the translational start site, and the other two conserved regions (CR1 and CR2) were located 91–171 bp upstream of the ETH translational start site [77–157 bp upstream of the predicted transcriptional start site (Park et al., 1999)].

Using MAVID, we added the corresponding sequences from two additional species, D. erecta and D. simulans, to the alignment. Based on these eight genomes, we obtained a Drosophila genus consensus sequence for CR1 and CR2 (Fig. 7B). An rVISTA analysis to detect putative ATF-4 binding sites resulted in three matches, at the same position in each species, in the aligned D. yakuba, D. pseudoobscura, D. ananassae, D. mojavensis and D. virilis sequences. One was located within the most conserved portion of CR1, a second was found in CR2, and the third was located 21 nucleotides upstream of the ETH translational start site. In D. yakuba, all three hits were conserved [at least 80% identical over a 24 bp window (Loots et al.,2002)], and the CR2 hit was conserved in D. pseudoobscura. The other hits in D. pseudoobscura and in the other three species did not meet the 80% conservation threshold. In addition to these matches, we also obtained one conserved hit in CR2 in D. yakuba and D. pseudoobscura for the TRANSFAC consensus binding sequence for Drosophila transcription factors encoded by the E74 early ecdysone-inducible gene (Eip74EF)(Fig. 7B). This sequence was an imperfect match to a consensus binding site for E74A determined by random oligonucleotide selection (E74A cons)(Urness and Thummel, 1990). Finally, the conserved portion of CR1 also contains a putative ecdysteroid response element (DR4), which consists of an imperfect direct repeat of AGGTCA separated by 4 nucleotides (Park et al.,1999). This sequence was also a conserved hit (with the DR4 consensus) in our rVISTA analysis of the D. yakuba and D. pseudoobscura sequences (data not shown).

We compared the Drosophila genus consensus sequence for the predicted ATF-4 sites in CR1 and CR2 to the ATF-4 binding site in the rat Grp78 promoter (ATF4-Grp78) (Luo et al., 2003), the CAATT-enhancer binding protein(C/EBP)-activating transcription factor (ATF) composite site in the hamster chop promoter (ATF4-chop)(Fawcett et al., 1999; Ma et al., 2002), and the cAMP response element (CRE) in the rat phoshpenolpyruvate carboxykinase(PEPCK) gene (PEPCK CRE) (Vallejo et al., 1993) (Fig. 7B). All of these confirmed ATF-4-binding sites were imperfect matches to the ATF4 rVISTA hits in the Drosophila sequences. The best match (7 of 8 nucleotides) was between the CR1 hit and the PEPCK CRE. The latter has been shown to bind ATF-4-C/EBPβ heterodimers(Vallejo et al., 1993). Thus,there is strong conservation of two sequences in the ETH promoter that are close, but imperfect matches to known binding sites for ATF-4, the mammalian ortholog of CRC. We predict that one or both of these putative CRC binding sites is required for CRC-dependent expression of ETH.

DIMM has been proposed as a direct regulator of neuroendocrine gene expression in most neuropeptidergic cells(Allan et al., 2005; Hewes et al., 2003). Here we present qRTPCR results, supplemented by in situ hybridization,showing that DIMM upregulates the levels of mRNAs derived from at least three neuropeptide genes, Fmrf, Lk and Dms. These findings provide strong support for DIMM as a key regulator of multiple neuroendocrine genes.

The LIM-homeodomain gene apterous (ap) also controls Fmrf and Lk gene expression(Allan et al., 2005; Allan et al., 2003; Benveniste et al., 1998; Herrero et al., 2003; Park et al., 2004). ap acts cell-autonomously to stimulate dimm expression, but the AP and DIMM proteins can also physically interact, and they may function together in regulating Fmrf (Allan et al., 2005). Several other factors, including the transcriptional co-factors encoded by dachshund and eyes absent (Miguel-Aliaga et al.,2004), the zinc-finger gene squeeze, and the retrograde bone morphogenetic protein (BMP) pathway, act in combinatorial fashion with dimm and ap to control Fmrf expression(Allan et al., 2005; Allan et al., 2003). However,other neuropeptidergic cells appear to use only portions of this code. For example, ap and dimm appear to contribute to the expression of Lk in Fmrf-negative cells (A1–A7 and possibly Br1). Even within the population of Lk cells, loss of dimm results in very different effects in different neurons, with a reduction in Lk transcript levels in cells A1–A7, and an increase (or no change) in Lk levels in the Br1 and SE neurons(Fig. 4). How do these relatively widely expressed factors interact with other regulatory proteins to produce cell type-specific patterns of neuropeptide gene expression? It will be of interest to determine which other elements of the combinatorial pro-Fmrf code are used to control Lk and Dmsexpression, and to identify additional factors that interact with dimm to control expression of these neuropeptide genes.

We did not detect changes in levels of three neuropeptide biosynthetic enzyme mRNAs, Phm, Fur1 and amon, in the qRTPCR analysis. This is in contrast to our earlier immunocytochemical studies, in which we observed a marked reduction in the protein products of these genes in dimm mutant CNS (Hewes et al.,2003). In some cases, these differences may reflect the spatial insensitivity of the qRTPCR methods, as was confirmed by our in situhybridization analysis of Lk expression(Fig. 4). Phm, in particular, may belong in this category. Although levels of PHM and DIMM expression are strongly correlated (Allan et al., 2005; Hewes et al.,2003), PHM is also highly expressed in many other tissues(Jiang et al., 2000) that do not express dimm. Any dimm-dependent change in Phmexpression may have been obscured by the much larger pool of dimm-independent Phm mRNA in our whole-animal qRTPCR analysis.

DIMM may regulate levels of other neuroendocrine proteins through a route that does not involve interactions between DIMM and cis-regulatory elements in the respective genes. We obtained the first evidence in support of this hypothesis in our earlier analysis of an ectopically expressed neuropeptide in dimm mutant cells; levels of ectopic PDF protein were strongly reduced while dimm had no effect on levels of the cognate Pdf mRNA (Hewes et al.,2003). Here, we show that larvae homozygous for a specific loss-of-function mutation in dimm displayed reduced levels of endogenous ETH-like protein(s), but not ETH mRNA, in the endocrine Inka cells (see Fig. S1 in the supplementary material), a site of dimm gene expression (Hewes et al., 2003). This may occur simply through a dimm-dependent change in levels of one secreted protein, such as PHM,that may disrupt the formation of multi-protein aggregates required for neuropeptide sorting into secretory granules(Arvan and Castle, 1998; Brakch et al., 2002). Alternatively, recent studies on the mouse ortholog of dimm, Mist1,suggest that dimm may play a more direct role in the management of secretory granule budding from the trans-Golgi network. In Mist1knockout mice (Mist1^KO), pancreatic exocrine cells display reduced intracellular organization (Pin et al., 2001). Moreover, the Mist1^KO phenotype is partially phenocopied in animals mutant for the Rab3D gene, a small GTPase involved in secretory granule trafficking(Johnson et al., 2004). Further studies on the regulation of ETH, PHM and Rab3-like proteins, and on the biochemical interactions among them, may shed light on the cellular mechanisms underlying the indirect actions of DIMM.

Mutations in the crc gene result in pleiotropic defects in ecdysone-regulated events during molting and metamorphosis(Hewes et al., 2000). Many of the morphological defects are associated with a failure of the insect to properly complete ecdysis, a stereotyped set of behaviors required for shedding of the old cuticle at the culmination of each molt. Several neuropeptides and peptide hormones, including ETH, play critical roles in organizing and triggering ecdysis behavior(Ewer and Reynolds, 2002).

Here we provide four independent lines of evidence that demonstrate a crucial role for crc in the upregulation of ETH mRNA levels. First, we observed a marked reduction by qRTPCR in levels of ETHtranscripts [but not in mRNAs encoding CCAP or EH, two other neuropeptides involved in the neuropeptide hierarchy controlling ecdysis(Ewer and Reynolds, 2002)] in crc mutant larvae (Fig. 2). Second, in situ hybridization revealed a strong reduction in ETH mRNA levels in the endocrine Inka cells in crc mutant larvae (Fig. 5). Third, the intensity of anti-PETH immunoreactivity was markedly reduced in crc¹/crc¹homozygotes. Fourth, EGFP fluorescence driven by an ETH-EGFP reporter gene was reduced in crc mutant larvae(Fig. 6). Therefore, CRC is a strong activator of ETH gene expression, and loss of CRC results in a corresponding reduction in levels of the ETH protein.

Despite the molecular identification of the crc locus(Hewes et al., 2000), almost six decades after the original description of the first crc allele(Hadorn and Gloor, 1943), the causes of the molting and metamorphosis defects in crc mutants remained unclear. Our current results suggest a simple model to explain the crc mutant phenotype. Strong hypomorphic or null mutations in crc and ETH both severely disrupt ecdysis. These defects include weak, irregular and slower ecdysis contractions and a failure to shed old cuticular structures, leading to retention of two and sometimes three sets of mouthparts into the next larval stage(Chadfield and Sparrow, 1985; Park et al., 2002). These similarities in molting defects, taken together with our observation that crc is required for normal expression of ETH mRNA and ETH protein, point to the loss of ETH signaling as the likely proximate cause of the ecdysis defects observed in crc mutants.

Despite the specific effects of crc on ETH transcription in the Inka cells, crc is widely expressed(Hewes et al., 2000),suggesting a cellular housekeeping function. The vertebrate ATF-4 protein is also ubiquitously expressed (Hai and Hartman, 2001). In addition, the upregulation of ATF-4constitutes a milestone of many cellular stress response pathways including oxidative stress, amino acid deprivation(Rutkowski and Kaufman, 2003),and hypoxia (Blais et al.,2004). In the tobacco hornworm, Manduca sexta, levels of ETH fluctuate during the molts and are regulated by circulating ecdysteroids(Zitnan et al., 1999). We hypothesize that CRC contributes to the regulation of ETH gene expression during this period, perhaps in response to signals from the tracheae.

Peaks in circulating levels of the ecdysteroid hormone, 20-hydroxyecdysone(20HE), initiate and coordinate each molt. A subsequent decline in 20HE levels is required for ecdysis, and the activation of these behaviors involves a hierarchical cascade of peptide hormone and neuropeptide signals that is triggered by ETH (Ewer and Reynolds,2002). Is CRC required in order to maintain ETH expression, or is CRC involved in regulating transcription of the ETH gene during the molts? While it is not known whether ETH mRNA levels fluctuate during Drosophila post-embryonic development, the regulation of ETH levels by ecdysteroids in molting Manduca sexta, and our analysis of the CR1 and CR2 sequences, provides tantalizing clues to possible coordinate regulation of ETH gene expression by CRC and ecdysone response genes. There is substantial overlap between the predicted CRC binding site in CR1 and a putative ecdysteroid response element (EcRE) (cf. Park et al., 1999). In addition, we found a potential binding site in CR2 for products of the E74 early ecdysone-inducible gene. E74 expression is induced directly by 20HE, and it encodes transcription factors that regulate other ecdysone response genes (Fletcher and Thummel, 1995). Mutations that specifically disrupt E74B, which likely binds the same consensus as E74A(Urness and Thummel, 1990),display defects associated with pupal ecdysis that closely phenocopy crc. In future studies we hope to determine if ETHexpression is regulated by elements in both CR1 and CR2 in an ecdysteroid-dependent manner, and whether CRC, E74B and other factors in the ecdysone-response pathway interact competitively or cooperatively at these sites.

 
• 20HE

  20-hydroxyecdysone

 
• amon

  amontillado

 
• ap

  apterous

 
• ATF-4

  activating transcription factor-4

 
• B

  background

 
• bHLH

  basic helix-loop-helix

 
• BMP

  bone morphogenetic protein

 
• bZIP

  basic-leucine zipper

 
• C/EBP

  CAATT-enhancer binding protein

 
• Caps

  Calcium activated protein for secretion

 
• Ccap

  Cardioacceleratory peptide

 
• CNS

  central nervous system

 
• CR

  ETH promoter conserved region

 
• crc

  cryptocephal

 
• CRE

  cAMP response element

 
• Ct

  cycles to threshold amplicon concentration

 
• Ddc

  Dopa decarboxylase

 
• dimm

  dimmed

 
• Dms

  Dromyosuppressin

 
• Dsk

  Drososulfakinin

 
• EcRE

  ecdysteroid response element

 
• EGFP

  enhanced green-fluorescent protein

 
• Eh

  Eclosion hormone

 
• Eip74EF (E74)

  Ecdysone-induced protein 74EF

 
• ETH

  Ecdysis triggering hormone

 
• Fmrf

  FMRFamide-related

 
• Fur1

  Furin 1

 
• Lk

  Leucokinin

 
• NoE

  No Enzyme

 
• NoT

  No Template

 
• Orc4

  Origin recognition complex subunit 4

 
• Pdf

  Pigment–dispersing factor

 
• PEPCK

  phospenolpyruvate carboxykinase

 
• PETH

  pre-ecdysis triggering hormone

 
• Phm

  Peptidylglycine-α-hydroxylating monooxygenase

 
• ple

  pale

 
• qRTPCR

  quantitative real time polymerase chain reaction

 
• RpL32 (rp49)

  Ribosomal protein L32

 
• S

  soma

 
• TM

  tracheal metamere

 
• T_m

  melting temperature

 
• Tsp39D

  Tetraspanin 39D

 
• VNC

  ventral nerve cord

 
• y

  yellow

We thank Chad Hargrave for statistical guidance, Mike Adams for fly stocks and antisera, David Durica for comments on the manuscript, and Audrey Kennedy,Kendal Milam, Matthew Mote, Jeremiah Smith and Elizabeth Pearsall for technical assistance. This work was supported by grants from the National Science Foundation (NSF IBN0344018), NSF EPSCoR and the Oklahoma State Regents for Higher Education (NSF-0132534), and the Oklahoma Center for the Advancement of Science and Technology (HR03-048S) to R.S.H.