The bHLH protein Dimmed controls neuroendocrine cell differentiation inDrosophila

The differentiation of a neuroendocrine phenotype involves the selection and production of a specific peptide hormone. For both endocrine and neuroendocrine cells, recent genetic analyses in mice have revealed transcriptional hierarchies controlling such developmental events. The process typically involves multiple stages of gene activation and extinction, and it represents the actions of multiple regulatory cascades. For example, the proliferation and specification of hypothalamic neurosecretory neurons that produce vasopressin, oxytocin and corticotrophin-releasing factor are controlled by early expression of Orthopedia(Acampora et al., 1999) andSim1 (Michaud et al.,1998). These effects are mediated largely by induction or maintenance of a secondary factor, Brn2, which directly regulates neuropeptide gene expression (Schonemann et al., 1995; Nakai et al.,1995). Likewise, in pituitary and in pancreatic endocrine cells,tissue-specific (e.g. Ptx1/2) and secondary cell type-specific transcription factors (e.g. neurogenin3) promote patterned expression of peptide hormones(Sheng and Westphal,1999).

Neuroendocrine cell differentiation also involves the integrated assembly of cellular machinery needed to produce large amounts of secretory peptides. Such mechanisms coordinate several events associated with the regulated secretory pathway (Arvan and Castle,1998): the ability to synthesize, process, sort, traffic and accumulate dense-core secretory granules and their contents. Neurons differ greatly and reproducibly in the amount of secretory peptides that they produce, and in their elaboration of the secretory pathway. For example,mammalian motoneurons produce low levels of neuropeptides such CGRP or galanin(Streit et al., 1989), and at the ultrastructural level, their terminals contain many small, clear vesicles,but very few large, dense-core (peptide-containing) granules(Hall and Sanes, 1993). By contrast, hypothalamic neurosecretory neurons produce large amounts of vasopressin, oxytocin, or corticotrophin-releasing factor, and they contain correspondingly large numbers of dense-core secretory granules(Burbach et al., 2001). Cells also transiently modify their levels of secretory activity following injury(e.g. Blake-Bruzzini et al.,1997) or stimulation (e.g.Herman et al., 1991). The mechanisms underlying these differences in levels of secretory activity are unknown.

Because the amplified expression of the secretory pathway is a stable and cell-specific feature of neuroendocrine cells, we hypothesize the existence of genetic factors that control this phenotype. Identifying such factors will facilitate a detailed, mechanistic analysis of neuroendocrine cell organization and physiology. Such an analysis will be crucial to a general understanding of neuroendocrine cell biology and will support efforts to produce a program of neuroendocrine differentiation from stem cells in vitro. We describe a Drosophila bHLH gene, dimmed (dimm),with an expression pattern that corresponds precisely to the neuronal and endocrine cells that accumulate large amounts of secretory peptides. We present both loss-of-function and gain-of-function analyses to argue thatdimm confers a pro-secretory phenotype within these diverse cells,and that its actions appear confined to that aspect of cellular differentiation. Thus, we propose a novel and general mechanism, of whichdimm is an essential component, for the amplification of the regulated secretory pathway by dedicated secretory cells.

Flies were cultured at 22-25°C on a standard cornmeal-yeast-agar medium. The molecular and genetic characterization of c929, R6 andRev8 has been described previously(Hewes et al., 2000).Rev4 and Rev18 are X-ray revertants ofP{PZ}l(2)k05106⁰⁶³¹¹ (M. Horner and C. Thummel, personal communication). y^-w^- revertants ofKG02598 (e.g. dimm^S2a) were obtained by transposase-mediated excision, and each line was characterized by PCR with primers flanking the original insertion site. Except as noted, all other strains are described elsewhere (Lindsley and Zimm, 1992; FlyBase,1999) and were obtained from the Bloomington stock center, the BDGP Gene Disruption Project and other sources.

Eggs were collected on apple juice-agar plates supplemented with yeast paste. Larvae were collected from the plates, and heterozygotes (y*w* and balanced over CyO-y⁺) were distinguished by mouthpart color. After scoring, size-matched pairs ofy^- and y⁺ larvae were dissected and stained in parallel.

Immunostaining was performed as described previously(Benveniste et al., 1998). Briefly, tissues were fixed in 4% paraformaldehyde (PFA), Bouin's, or 4%paraformaldehyde/7% picric acid (PFA-PA). Polyclonal or monoclonal primary antisera were used (overnight at 4°C) to detect the following proteins:β-gal (1:1000, PFA-PA; Promega); PHM (1:750 pre-absorbed to PHM^-/- larvae, Bouin's) (Jiang et al., 2000); -RFa (`PT2')(1:2000, PFA-PA)(Taghert, 1999); FMRF (1:2000,PFA-PA) (Chin et al., 1990);corazonin (1:500, Texas Red-conjugated; PFA-PA)(Veenstra, 1994); LK (1:500,PFA-PA) (Nässel and Lundquist,1991); CCAP (1:400, PFA-PA)(Ewer and Truman, 1996); PDH and PAP (each 1:2000, PFA-PA) (Renn et al., 1999); MM (1:800, PFA-PA)(O'Brien and Taghert, 1998);dopa decarboxylase (affinity purified 1:100, PFA)(Scholnick et al., 1991);Furin-1 (1:1000, Bouin's) (Jiang et al.,2000); and Myc (1:500, PFA-PA; a gift from Y.-N. Jan; Sigma). Secondaries used were goat Cy3, FITC, Texas Red or ALEXA 488 conjugates(Jackson ImmunoResearch) at a 1:500 dilution. Confocal z-series projections were obtained using an Olympus Fluoview microscope.

A cDNA library was made from RNA of y w adult heads using commercial reagents (Clontech). 5′ RACE was performed according to manufacturer's recommendations using CG8667-specific primers.

RNAi was performed based on the methods of Kennerdell and Carthew(Kennerdell and Carthew, 1998)and Clemens et al. (Clemens et al.,2000). The template for RNA synthesis was generated by PCR, using primers containing a T7 promoter sequence(5′-GAATTAATACGACTCACTATAGGGAGA-3′) at the 5′ ends and P1 DNA (DS00532) as the PCR template. The gene specific primers 5′-CAGATTCCAGTTCGCAAAGCGAT-3′ and 5′-GGGCTCGTCGAAATTATCATTGATA-3′ amplified a 951 bp segment of open reading frame in exon 3, including the entire bHLH domain. Transcription and analysis of the double-stranded RNA (dsRNA) were performed as described(Clemens et al., 2000). dsRNA(3 μM) was injected into syncytial blastoderm embryos (Canton-S) ∼75%along the anteroposterior axis. For the mock controls, all steps were performed in parallel, except that the P1 DNA was omitted from the PCR reaction. Mock- and RNA-injected larvae were dissected during or within 6 hours after hatching.

The predicted coding region of CG8667 was amplified by PCR using cDNA generated from y w adult head RNA (Clontech). The primers 5′-CAGATCTCGACGATTTTTGTTCAGCCAT-3′ (5′ UTR) and 5′-TGCGGCCGCAGAAACTCTCGAAAGGGCT-3′ (end of the ORF) were used to construct a 1236 bp fragment that was cloned into pBSK+ and then transferred to pP{UAST-Myc} at the BglII and NotI sites. Transgenic lines containing P{UAS-dimm::Myc} insertions were created using standard techniques(Benveniste and Taghert, 1999).UAS-dimm::Myc^2-A-3, Rev8/CyO, Act-GFP flies were crossed to c127-Gal4, UAS-GFP; Rev4/CyO, Act-GFP, and first instar larvae were scored for Act-GFP- and c127-specific GFP labeling patterns.

Whole-mount in situ hybridization(Tautz and Pfeifle, 1989) was performed using single-stranded, digoxigenin-labeled RNA or DNA probes(Patel, 1996) prepared from P1 or cDNA templates.

Cells were imaged on a Zeiss Axioplan fitted with a SPOT CCD camera and software (Diagnostic Instruments) or in confocal z-series scans. Exposure settings were adjusted to optimize detection without saturating the signal. For a given neuron, identical settings were used for all preparations and genotypes. Mean pixel luminosity for the area covering the soma (S) was measured for each neuron using Adobe Photoshop. An adjacent area was sampled to measure the background signal (B). The intensity index=(S-B)/B. Cells not visible were scored 0. Cells that were obscured or lost due to tissue damage were not analyzed. Brightfield images were inverted before quantification. CNS size was measured as an additional control — in each case, mean brain lobe diameters were not significantly different between genotypes (data not shown). Statistics were performed using the NCSS-2000 Statistical Analysis System or StatView (MANOVAs; Games-Howell). Variances are reported as±s.e.m.

The c929 P-element insertion was isolated in a P{Gal4}enhancer detection screen for genes expressed in the Tv neuroendocrine neurons(O'Brien and Taghert, 1998). In addition to the Tv neurons, c929 drove reporter gene expression(GFP or β-galactosidase) in ∼200 neurons scattered throughout the larval CNS and in neuroendocrine projections to the ring gland, the dorsal neurohemal organs and the transverse nerves(Fig. 1A). Outside the CNS,this pattern included at least three classes of endocrine cells: intrinsic cells of the corpora cardiaca (Fig. 1A), 10-20 midgut cells (data not shown) and the peritracheal myomodulin-immunoreactive cells. The latter appear homologous to the endocrine Inka cells (O'Brien and Taghert,1998). c929 reporter expression also appeared in several other tissues, including peptidergic PNS neurons (LBD neurons; D. Allan and S. Thor, personal communication), fat body, epithelial cells and salivary glands(data not shown).

To determine whether c929-positive neurons express neuropeptides,we performed double-label experiments for the c929 reporter and for the peptide biosynthetic enzyme, peptidylglycine-α-hydroxylating mono-oxygenase (PHM). In Drosophila, PHM is a marker for most peptidergic cells. It is required for neuropeptide amidation(Jiang et al., 2000), which is a highly specific modification of secretory peptides(Eipper et al., 1993); greater than 90% of all known or predicted Drosophila peptide transmitters are amidated (Hewes and Taghert,2001). Most if not all c929-positive CNS neurons were immunostained very strongly by PHM antibodies (n=8 specimens;Fig. 1B,E).

Conversely, most neurons displaying strong PHM immunostaining were alsoc929 positive, while most weakly PHM-positive neurons were notc929 positive (data not shown). In addition, PHM was expressed in all three c929-positive endocrine cell types and in the LBD peripheral neurons (O'Brien and Taghert,1998) (data not shown). Thus, in the larval CNS and in several peripheral tissues, c929 primarily labels neuroendocrine cells as its expression was highly correlated with the production of large amounts of amidating enzyme, amidated neuropeptides and peptide hormones.

To assess the degree of heterogeneity among c929-positive cells,we compared the expression pattern of c929 with a variety of other peptidergic cell markers. This population of cells was chemically diverse. For example, seven bilateral pairs of c929-positive neurons were double-labeled with the PT2 antiserum (Fig. 1C,E). PT2 is a marker for -RFamide containing neuropeptides,which include the products of at least three Drosophila genes(Taghert, 1999). Additional subsets of c929-positive neurons were immunostained with antisera directed against a variety of neuropeptides. These included theDrosophila FMRF propeptide (n=8 specimens), cockroach corazonin (n=5), cricket leucokinin-1 (LK), crustacean cardioactive peptide (CCAP; n=4), crustacean beta-PDH (n=4) andAplysia myomodulin (MM; Fig. 1D,E). Finally, a distinct subset of 34 c929-positive neurons (see below) was immunopositive for an additional, putativeDrosophila peptide biosynthetic enzyme (n=10; P.H.T. and M. Han, unpublished) Furin 1 (De Bie et al.,1995). Based on their positions, cellular morphologies, and immunostaining with the above markers, the cells within the c929pattern represent more than 26 distinct classes of peptidergic neurons and endocrine cells. No c929-positive neurons were stained with an antiserum to dopa decarboxylase (n=8), an enzyme required for synthesis of the biogenic amines, serotonin and dopamine(Hirsh, 1989).

No single transmitter system we tested was entirely c929 positive. For example, among the 17 Fmrf cell types(Benveniste and Taghert, 1999),only the Tv neurons were c929-positive. However, in third instar larvae there were some c929-negative neurons, such as the peptidergic MP1s and VAs, which displayed weak and/or transient c929 reporter expression during other stages of development (e.g.Fig. 5G). Thus, our identification of c929-positive peptidergic neurons is likely to be an underestimate of the total population of peptidergic cells that express the reporter gene.

To test for roles of a putative `c929' gene in the development and/or function of peptidergic cells, we generated deletions flanking thec929 insertion site (Fig. 2A). These deletions caused recessive lethality, owing to disruption of at least one essential gene, cryptocephal(crc) (Hewes et al.,2000). However, many homozygous mutant animals survived into the larval stages, when we could examine the fates of CNS peptidergic neurons.

By immunostaining the mutant animals for PHM, we detected a novel phenotype: R6/Rev8 trans-heterozygous animals contain small deficiencies around the c929 insertion site that are ∼ 12 and∼ 35 kb respectively (Fig. 2A). Transheterozygous larvae displayed marked reductions in PHM protein levels in all strongly c929-positive CNS neurons(Fig. 2B).c929-negative neurons were unaffected in R6/Rev8 larvae, and weakly or transiently c929-positive neurons, such as the VAs, showed smaller reductions in PHM immunostaining(Fig. 2B). The mutant phenotype was detectable at the time of larval hatching and throughout all larval stages. By contrast, heterozygous R6 or Rev8/+ larvae were essentially wild type, although these alleles displayed mild haploinsufficiency (n=24; data not shown) with other markers (see below). These results demonstrate a requirement for ∼ 10 kb of DNA flanking the c929 insertion site for the normal expression and/or maintenance of PHM in c929-positive CNS neurons. We named the affected gene dimmed to reflect the diminished staining.

We used six additional neurosecretory markers in dimm mutant larvae, and found that all six displayed moderate to severe reductions in immunostaining in spatial patterns corresponding to the c929 reporter pattern. The affected proteins included several known or presumed neuropeptides — MM (Fig. 2C), LK (n>25), the FMRF propeptide (n>12)and several PT2 positive neuropeptides (n>50) — and the putative neuropeptide biosynthetic enzyme Furin 1 (see below). Allc929-positive neurons displayed the mutant phenotype for at least one marker, PHM (Fig. 2B); many showed reduced immunostaining with multiple markers. For example, the Tv neurons had reduced levels of PHM, the FMRF propeptide, —RFamide peptides and Furin 1 (Fig. 2B,see Fig. 3B, seeFig. 6B). Thus, in a large and diverse population of CNS peptidergic neurons, dimm regulates levels of a broad array of secretory proteins.

As the three classes of c929-positive endocrine cells also likely secrete peptide hormones, we also tested them for effects of the dimmmutations. The ring gland (n=15) and tracheal endocrine cells(n>50) displayed severe reductions in peptide immunostaining for PHM and/or MM in dimm^-/- mutants(Fig. 2E; data not shown); the gut endocrine cells were not tested. Taken together, these results suggest a crucial role for dimm in controlling bioactive peptide levels in diverse neuronal and endocrine secretory cells.

Using chromosomal deletions, we genetically mapped the dimm gene. We performed peptide immunostaining on Rev8 homozygotes(n=15) and on hemizygotes (n>50) bearing one copy ofR6 (or Rev8) over one of several independently derived deficiencies of the entire 39C4-D1 region of chromosome 2L (e.g.Rev4). In each case, the effects on peptide immunostaining were comparable, although the reduction in MM staining in larvae homozygous forRev4, a null allele (Fig. 2A), was more pronounced than in R6/Rev8trans-heterozygotes (n=12; data not shown). Thus, R6 andRev8 are hypomorphic alleles. Normal peptide immunostaining was restored in male Rev8 homozygotes (n=6) bearing a duplication of chromosome bands 35A-40 [Tp(2;Y)J54], consistent with the location of dimm in 39C4-D1.

In contrast to R6/Rev8 mutants, larvae with disruptions in thecrc gene (c929 homozygotes, n=6;crc¹/R6 trans-heterozygotes, n>40; R2homozygotes, n=5), or deletions of DNA extending up to 200-300 kb towards the telomere (TW1/Rev18 trans-heterozygotes, n=7)displayed wild-type neuropeptide levels (see Table S1). Thus, dimm is not crc, nor is it any other gene located distal to the site of the c929 insertion.

The closest gene proximal to c929 is CG8667(Mistr), found within 25 kb (Fig. 2A). It encodes a basic helix-loop-helix (bHLH) protein that is a member of the Atonal subfamily of transcription factors(Moore et al., 2000). Its bHLH domain displays 79% identity with the mouse Mist1 protein(Pin et al., 1999). InRev8 homozygous embryos, CG8667 mRNA expression was markedly reduced, but not eliminated (n>50; data not shown), consistent with the identification of Rev8 as a hypomorphic dimmallele. After 5′ RACE identification of the 5′ end ofCG8667, we identified a P-element insertion(dimm^KG02598) located 111 bp upstream(Fig. 2A).dimm^KG02598 displays homozygous lethality (see Table S2),and represents a severe hypomorphic dimm allele, becauseCG8667 mRNA expression appeared low or undetectable indimm^KG02598 homozygous mutant embryos(Fig. 3A). Hatchlingdimm^KG02598/Rev4 larvae displayed reduced immunostaining for PT2-positive neuropeptides (n>15;Fig. 3B). Normal PT2 immunostaining was restored (n>15;Fig. 3B) after precise excision of the dimm^KG02598 P element. Consistent with the conclusion that dimm and crc are separate genes,KG02598 was lethal when trans-heterozygous with Rev4, but not with crc¹ (see Table S2). The dimm^KG02598 mutation also reduces levels of secretory peptide mRNAs in the Tv neuroendocrine cells, which display high levels ofFmrf mRNA expression (Schneider et al., 1993): when assessed using in situ hybridization, the mean number of Fmrf-positive Tv neurons per CNS was 5.57 in dimmheterozygotes (n=7; Fig. 3C) and 2.33 in dimm hemizygotes (n=9;Fig. 3C; P<0.01). These combined data indicate that in the absence of dimm, there is a reduction in levels of both secretory peptide mRNAs and secretory peptides.

To examine further the effect of disruptions in CG8667 expression,we performed RNAi analysis and observed reduced levels of MM immunostaining in hatchling stage larvae (Fig. 2D). The reduction in MM immunostaining was comparable with the phenotype in null dimm^-/- mutants(Fig. 2C). We obtained the same results using two additional antisera, PT2 and anti-LK (n=5 andn=6; data not shown). We also tested the ability of aUAS-dimm::Myc transgene to restore neuropeptide levels indimm^-/- animals. We used the c127-Gal4 line to drive dimm::Myc expression in a small set of ventral CNS neurons,which included the 14 LK-positive cells in abdominal neuromeres(Fig. 4A). Expression ofdimm::Myc selectively restored normal levels of LK immunostaining inRev8/Rev4 animals (n=19;Fig. 4C), but not in the absence of the Gal4 driver (n=17;Fig. 4B). The rescue displayed cell specificity: the FMRF-positive MP2 neurons did not expressUAS-GFP by c127-Gal4, and they were not rescued(n=10; data not shown). Together, these results support the hypothesis that dimm is the Drosophila Mist1 ortholog,CG8667.

We performed a gain-of-function analysis by driving UAS-dimm::mycin an otherwise wild-type background. When misexpressed using a pan-neuronalelav-GAL4 driver, most embryos died (data not shown). This suggested that the effects of dimm on shaping neuronal properties can be widespread. To permit a more restricted analysis, we used ap-Gal4(Fig. 4E-G), a P{Gal4}reporter inserted in the apterous (ap) gene(O'Keefe et al., 1998). When overexpressed in a subset of brain neurons, dimm increased the brightness of LK immunostaining in the cell body and processes of the LK-positive Br1 neuron (Fig. 3F,G). dimm overexpression did not produce widespread LK misexpression, but it reproducibly increased the number of LK-positive neurons from one (in animals lacking the ap-Gal4 element, n=18 hemispheres) to two (n=22 hemispheres). The additional LK-positive neuron was always adjacent to the normal one. Thus, dimm can alter the properties of normal neuroendocrine cells, and it can affect the number of cells displaying a neuroendocrine phenotype.

CG8667 mRNAs were ubiquitous in pre-cellular blastoderm embryos(Moore et al., 2000; data not shown) and later were expressed in the developing nervous system(Moore et al., 2000). Presumed zygotic CG8667 expression was first visible as nascent transcripts scattered throughout the CNS in stage 12 embryos. Cytoplasmic CG8667hybridization was visible in many of these cells beginning around stage 14(Fig. 5A), was strong by stage 16 (Fig. 5B) and persisted in stage 17 embryos (Fig. 5C) and in hatchling larvae less than 24 hour old(Fig. 5G).

The pattern of CNS CG8667 in situ hybridization resembled thec929 reporter pattern (Fig. 5A-C,G). Based on their positions and morphologies, more than 12 separate types of CG8667-expressing neurons were putatively identified as c929 positive. These included dorsal chain neurons(e.g. d3-d5), T1-3v, LP1, MP1, MP2, SP1, T1-3vb and VA, as well as several bilateral clusters of neurons: large, midline protocerebral brain cells (MC),lateral protocerebral brain cells (LC), ventral subesophageal neurons (SE) and lateral abdominal neurons (neuromeres N1, N4 and N5).

We also observed expression of the c929 reporter andCG8667 in strikingly similar patterns in peripheral tissues(Fig. 5). These sites included the LBD neurons and several endocrine tissues: intrinsic cells of the corpora cardiaca, Inka cells and a few midgut cells. Numerically, all peripheral cell types were equally represented, except that there were fewerCG8667-expressing gut cells in embryos than c929-positive gut cells of larvae. CG8667 was not expressed in any other location,except for a few unidentified non-CNS cells scattered throughout the anterior and lateral regions (stages 12-15). Thus, in CNS, PNS and endocrine tissues,expression of the c929 reporter closely mirrored CG8667expression. These expression analyses support the genetic mapping, genetic identification and RNAi data. Thus, from this point onwards we refer toCG8667 as dimmed.

We next determined whether secretory cells survived and differentiated indimm^-/- mutant animals. In larvae homozygous for the null allele, Rev4, some of the affected cells displayed low residual immunostaining for secretory proteins (e.g.Fig. 2B,C). Thus, somedimm-expressing cells survived in mutant larvae and were at least partially differentiated. Others displayed a complete loss of peptide immunostaining, and their status was unclear.

In order to determine the fates of the latter cells, we usedGal4/UAS mosaics to express ectopic, non-secretory proteins indimm mutant neurons. We studied 34 CNS neurons that co-expressed three different markers: the c929 reporter, the putative peptide biosynthetic enzyme Furin 1, and ap-Gal4(Fig. 6A; P.H.T. and M. Han,unpublished). In dimm^-/- larval CNS, all 34 neurons displayed strongly reduced, and often undetectable, immunostaining for Furin 1(Fig. 6B). Usingap-Gal4 to drive heterologous expression of a tau::Myc fusion protein, we found that all 34 of these neurons were present and displayed normal morphology in the dimm^-/- larvae(Fig. 6C). In addition, the intensity of anti-Myc immunostaining was not affected(Fig. 6D). We obtained identical results using green fluorescent protein (GFP) to mark the cells(n=6; data not shown). Thus, dimm mutant neurons displayed multiple differentiated features and synthesized non-secretory proteins at normal levels throughout larval development.

We also examined the effects of dimm on the terminal arbor of the LK-positive neurons. These cells displayed reduced soma LK immunostaining indimm^-/- CNS (Fig. 4B). Each neuron had a single efferent axon that projected across the posterior muscle 8 surface and terminated dorsally near a tracheal branch. In third instar dimm^-/- larvae, these axons also displayed reduced LK immunostaining. However, a sufficient number of immunoreactive boutons remained to indicate a normal axonal expanse (see Fig. S1). Thus, the effects of dimm on this LK neuron appear limited to expression of the transmitter phenotype.

Our earlier measures of the dimmed phenotype were restricted to analysis of proteins abundant in the regulated secretory pathway. We also tested for an effect of dimm on constitutively secreted proteins. With ap-Gal4, we directed expression of a CD8::GFP fusion protein(UAS-CD8::GFP) to a subset of dimm-expressing neurons. CD8 is an integral membrane protein that is targeted to the plasma membrane inDrosophila cells (Zito et al.,1997). In dimm^-/- mutant larvae, all 34ap-Gal4 (Furin-1) neurons expressed CD8::GFP and displayed normal neuritic projections. However, CD8::GFP levels were significantly lower inc929-positive neurons in the dimm^-/- background(see Fig. S2). This effect was more subtle than the effects on levels of regulated secretory proteins. However, it suggests that dimm influences both regulated and constitutive secretory activity in neuroendocrine cells.

Because ap-dependent expression of transgenes was unaffected bydimm (Fig. 6C), we were able to uncouple neuropeptide transcription from potential effects ofdimm on secretory activity. Thus, when ap-Gal4 drove ectopic expression of the pdf neuropeptide gene, ectopic pdf mRNA levels were unaffected in dimm^-/- larvae(Fig. 7A,B). By contrast,ectopic PDF protein levels were severely reduced. We performed immunostaining for two peptide epitopes of the proPDF precursor(Renn et al., 1999): PAP(Fig. 7C,F) and PDF(n=20; data not shown). All 34 (c929-positive) neurons displayed severely reduced immunostaining for both PDF-related epitopes. Additional ventral abdominal neurons served as internal controls. These included 44 neurons that also displayed ectopic pdf expression driven by ap-Gal4, and a set of approximately eight native pdfneurons (not ap-positive). All of the internal control cells werec929 negative, and PAP/PDF immunostaining in these neurons was unaffected in dimm^-/- larvae(Fig. 7C-F). Thus,dimm was required within c929-positive neurons for the maintenance of ectopic PDF neuropeptide levels, but not of ectopicpdf mRNA.

Dimm is the first example of a dedicated pro-secretory factor. Dimm is necessary to confer neuroendocrine features onto peptidergic neurons that, in its absence, survive with normal neuronal properties. In addition, Dimm overexpression produces supra-normal levels of neuropeptide expression in peptidergic neurons and the appearance of additional cells with neuroendocrine features. From this genetic analysis, we suggest that neuroendocrine cell differentiation includes two interrelated, but separate sets of instructions. The first specifies the identity of the neuropeptide(s) or peptide hormone(s)to be expressed, while the second, which involves Dimm, specifies the level of regulated secretory activity.

The bHLH domain of the predicted Dimm protein showed the highest degree of sequence identity with the mouse, rat and human Mist1 proteins. These proteins may be orthologs (Moore et al.,2000). Interestingly, mouse Mist1 is present in many adult peripheral tissues, but within these tissues it is found only in serous exocrine cells (Pin et al.,2000). The restriction of mouse Mist1 expression to dedicated secretory cells suggests that dimm and mouse Mist1 may both control levels of secretory activity, and so may perform evolutionarily conserved functions. Other members of the Atonal family are expressed in both differentiating and terminally differentiated cells (e.g. NeuroD)(Morrow et al., 1999). Several mammalian Atonal family bHLH proteins have previously been implicated in earlier stages of endocrine cell development, including cell lineage commitment (e.g. Yang et al.,2001) and endocrine cell differentiation(Sheng and Westphal,1999).

In Drosophila, dimm performs a novel, pro-secretory function in a diverse population of peptidergic CNS and PNS neurons and endocrine cells. In its absence, peptidergic cells complete many aspects of their differentiation— some express low levels of appropriate peptide transmitters. However,they uniformly fail to display normal amplified levels of secretory activity,which is a characteristic and fundamental property of peptidergic secretory cells (Arvan and Castle, 1998). How such cells acquire and maintain this capacity is largely unknown. We have shown that it is under the control of specific genetic mechanisms, as revealed by animals deficient in expression of the dimm gene. These experiments indicate that dimm plays a fundamental role in the differentiation of neuroendocrine lineages.

We propose a working model in which Dimm directly regulates transcription of genes required for production of a neuroendocrine phenotype — genes encoding neuropeptides, peptide hormones and peptide biosynthetic enzymes. Consistent with this model, we found that dimm reduces the normally high levels of Fmrf neuropeptide mRNA in specific neuroendocrine cells. In addition, Dimm also may regulate expression of proteins (e.g. transcription factors, or structural or regulative proteins of dense core granules) that are important for the function and amplification of the secretory pathway [e.g., as suggested by Kim et al.(Kim et al., 2001)]. Dimm functions after cell fate determination and during the early differentiation of these neurons — in dimm mutants, affected peptidergic neurons are present, arborize normally and often express low levels of appropriate neuropeptides.

Some secretory proteins form dense aggregations (`progranules') in the trans-Golgi network prior to their uptake into immature secretory granules. Similarly, condensation of secretory proteins during subsequent granule maturation may be required for their retention in maturing granules(Arvan and Castle, 1998). Therefore, direct reductions in the levels of a small number of target secretory proteins in dimm mutant cells may lead to a secondary disruption in aggregation or condensation of other proteins. In turn, these effects could lead to loss of most secretory proteins by mis-routing and degradation. This may account for our observation that secretory peptide levels could be reduced in a dimm mutant background, despite the artificial elevation of the cognate secretory peptide mRNA(Fig. 7).

Does dimm also regulate the constitutive secretory pathway?Although constitutive secretion was quantitatively affected by loss ofdimm function, mutant neurons maintained their normal cellular morphology. These observations suggest that Dimm has only moderate effects on the constitutive secretory pathway. Given the physical interactions between cargoes destined for the regulated and constitutive pathways(Arvan and Castle, 1998), the reduction in constitutive secretion may reflect an indirect effect of disruptions in the regulated pathway.

We favor the view that during development and maturity, dimmexpression is a crucial determinant of high secretory protein expression in neuroendocrine cells. This hypothesis was supported by the gain-of-function analysis. Overexpression of dimm in a wild-type background produced higher levels of LK expression in the normally LK-positive Br1 neuroendocrine neuron. It also increased the number of cells that display the specific LK neuroendocrine phenotype, but only within the immediate proximity of Br1. In this case, dimm overexpression was driven by a promoter(ap-GAL4) that is only expressed in postmitotic neurons. Therefore,it appears likely that the additional LK immunoreactive neurons represent cells that normally express LK but at levels that are too low to be detected. In addition, the limited number of ectopic leukokinin cells is likely a function of the specific GAL4 driver used (ap is only expressed in a subset of cells), and the marker assayed (LK is only expressed in ∼20 out of 10,000 neurons). Although the complete extent of the effects of dimm, when overexpressed, is not yet known it is likely to be large, as UAS-dimm produces large-scale embryonic lethality when driven by the pan-neuronal elav-GAL4 (D. P., unpublished).

Accordingly, we propose that dimm promotes diverse neuroendocrine cell fates in different cellular locales, depending on local cellular context and identity. We observed dimm expression soon after cells cease dividing, and in its absence, most of these cells were deficient in`transmitter expression'. Thus, Dimm appears to function like NeuroD proteins,which are also members of the Atonal family and which act as cell differentiation factors (Hassan and Bellen, 2000).

Analysis within the identified, neuroendocrine Tv neurons may be especially informative to reveal further details of the mechanisms of dimmaction. Four regulatory factors have now been defined that affect FMRF neuropeptide levels in Tv neurons. Loss-of-function ap(Benveniste et al., 1998),Chip (Van Meyel et al.,2000) and dimm (this report) alleles all decrease Tv-specific FMRF expression, but do not influence Tv survival or morphology. Likewise, the squeeze (sqz) gene helps regulate Tv-specific FMRF levels (S. Thor, personal communication). Within Tv neurons, ap,Chip, dimm and sqz may function in a linear pathway to regulateFmrf gene expression, akin to the sequential actions of the bHLH protein MASH1 and the Phox2 homeoproteins in neurons of the locus coeruleus(Pattyn et al., 2000). Alternatively, they may work in parallel fashion, akin to the synergistic interactions between the bHLH NeuroD1 and the LIM homeoproteins Lmx1.1 and Lmx1.2 to control insulin expression(Ohneda et al., 2000). As a first step, we have shown that ap promoter function is independent ofdimm. Further work will permit description of the molecular pathways controlling qualitative and quantitative aspects of neuroendocrine cell differentiation in vivo.

We thank Weihua Li and Aloka Amarakone for technical assistance, and Hans Agricola, Doug Allan, Hugo Bellen, Gabrielle Boulianne, Adelaide Carpenter,Heinrich Dircksen, Chris Doe, Dan Eberl, John Ewer, Jeff Hall, Mike Horner,Yuh-Nung Jan, Iris Lindberg, Dick Nässel, Jae Park, Anton Roebroek, Steve Scholnick, Amy Sheehan, John Thomas, Stephan Thor, Carl Thummel, Jan Veenstra,Klaude Weiss and Andrew Zelhof for information, DNA, antibodies or fly stocks. We thank Lou Muglia, Jim Skeath and Stefan Thor for comments on the manuscript, the Bloomington Stock Center for fly stocks, and the BDGP for DNA sequence. This work was supported by an American Cancer Society Postdoctoral Fellowship PF4212 (R.S.H.) and by a grant NS21749 from the NIH (P.H.T.).