Abstract
The cryptocephal (crc) mutation causes pleiotropic defects in ecdysone-regulated events during Drosophila molting and metamorphosis. Here we report that crc encodes a Drosophila homolog of vertebrate ATF4, a member of the CREB/ATF family of basic-leucine zipper (bZIP) transcription factors. We identified three putative protein isoforms. CRC-A and CRC-B contain the bZIP domain, and CRC-D is a C-terminally truncated form. We have generated seven new crc alleles. Consistent with the molecular diversity of crc, these alleles show that crc is a complex genetic locus with two overlapping lethal complementation groups. Alleles representing both groups were rescued by a cDNA encoding CRC-B. One lethal group (crc1, crcR6, and crcRev8) consists of strong hypomorphic or null alleles that are associated with mutations of both CRC-A and CRC-B. These mutants display defects associated with larval molting and pupariation. In addition, they fail to evert the head and fail to elongate the imaginal discs during pupation, and they display variable defects in the subsequent differentiation of the adult abdomen. The other group (crcR1, crcR2, crcE85, crcE98, and crc929) is associated with disruptions of CRC-A and CRC-D; except for a failure to properly elongate the leg discs, these mutants initiate metamorphosis normally. Subsequently, they display a novel metamorphic phenotype, involving collapse of the head and abdomen toward the thorax. The crc gene is expressed throughout development and in many tissues. In third instar larvae, crc expression is high in targets of ecdysone signaling, such as the leg and wing imaginal discs, and in the ring gland, the source of ecdysone. Together, these findings implicate CREB/ATF proteins in essential functions during molting and metamorphosis. In addition, the similarities between the mutant phenotypes of crc and the ecdysone-responsive genes indicate that these genes are likely to be involved in common signaling pathways.
THE development of Drosophila and other insects is punctuated by several molts, during which the animal produces a new external cuticle and sheds the old one (Riddiford 1993). The larval molts are initiated and coordinated by steroid hormones, the ecdysteroids (hereafter called ecdysone). At the onset of metamorphosis, a high titer pulse of ecdysone triggers pupariation, which is followed ~12 hr later by a brief ecdysone pulse that causes head eversion and the prepupal-pupal transition. Subsequently, a large, prolonged surge of ecdysone directs adult development. These remarkable developmental changes involve the programmed cell death of most larval tissues, extensive remodeling of other larval tissues, and the generation of adult tissues from islands of undifferentiated imaginal cells (Riddiford 1993; Trumanet al. 1994; Jianget al. 1997).
Detailed studies of the responses to ecdysone in the larval salivary glands provided key insights into the gene regulatory pathways controlling molting and metamorphosis (Thummel 1996). Expression of a small set of about six early genes is triggered rapidly and directly by ecdysone. Together with ecdysone, these genes regulate numerous late genes. Four early genes have been molecularly characterized. E74 encodes a family of ETS proteins (Burtiset al. 1990), E75 encodes a family of orphan nuclear receptors (Segraves and Hogness 1990), the Broad-Complex (BR-C) encodes a family of zinc-finger proteins (DiBelloet al. 1991), and E63-1 encodes a novel Ca2+-binding protein (Andres and Thummel 1995). E74 and the BR-C control developmental responses to ecdysone in diverse larval and imaginal tissues (Kisset al. 1988; Restifo and White 1991; Fletcheret al. 1995) at least in part through direct transcriptional regulation of the late genes (Urness and Thummel 1995; Crossgroveet al. 1996). Thus, the early genes are near the top of a complex gene regulatory hierarchy.
Mutations in several ecdysone-responsive genes reveal distinctive phenotypes that reflect important developmental functions. For example, an E74B mutant displays incomplete differentiation of the adult abdomen and failed gas bubble translocation at pupation (Fletcheret al. 1995). A mutation of βFTZ-F1, which functions as a competence factor for pupal stage-specific responses to ecdysone, displays similar translocation defects (Broaduset al. 1999). Mutations in E74, βFTZ-F1, and two other ecdysone-response genes, crol and DHR3, each display several additional common features. These include defective eversion of the adult head and (with the exception of DHR3) incomplete leg disc elongation (Fletcheret al. 1995; D'Avino and Thummel 1998; Broaduset al. 1999; Lamet al. 1999). The head eversion defect is called the “cryptocephal” phenotype, named after cryptocephal (crc1), a mutation that displays all of the above-mentioned defects (Hadorn and Gloor 1943). These phenotypic parallels indicate that crc and the ecdysone-response genes are likely to be involved in common regulatory pathways.
Fristrom (1965) examined chitin biosynthesis in the crc1 mutant and concluded that the head eversion defect is due to excess chitin deposition in (and increased stiffness of) the cuticle. Sparrow and Chadfield (1982) tested this hypothesis with a different crc1 strain and found normal chitin deposition. At pupation, crc1 mutants display contractions of the abdomen that are slower, more irregular, and weaker than in wild-type animals, indicating that behavioral abnormalities may cause at least some of the phenotypic defects observed in these mutants (Chadfield and Sparrow 1985). However, behavioral abnormalities likely do not explain other aspects of the crc1 mutant phenotype, such as incomplete abdominal differentiation. In a discussion of the similar phenotypic defects displayed by E74B mutants, Fletcher et al. (1995) hypothesized that premature muscle death accounts for the full range of defects observed in these cryptocephalic mutants. To distinguish among these and other competing models, it will be important to characterize crc gene function at the molecular and cellular level.
Here we show that crc encodes multiple proteins in the activating transcription factor 4 (ATF4) subfamily of CREB/ATF basic-leucine zipper (bZIP) transcription factors. ATF4 proteins have been implicated in several important developmental and disease processes, including wound healing (Esteset al. 1995), long-term synaptic facilitation (Bartschet al. 1995), stress responses (Fawcettet al. 1999), apoptosis (Kawaiet al. 1998), and cancer (Mielnickiet al. 1996). We add to this list by showing that the Drosophila ATF4 homologs play critical roles in molting and metamorphosis. We have isolated seven new crc alleles, which reveal multiple functions of the gene in larval molting, pupariation, pupation, and adult differentiation. These tissues include several targets of ecdysone signaling as well as the endocrine source of ecdysone, the ring gland. Our findings implicate CREB/ATF transcription factors for the first time in the hormonal regulation of molting and metamorphosis. Moreover, these results indicate that there are likely to be important interactions between signaling by crc and the ecdysone-response genes.
MATERIALS AND METHODS
Fly strains and culture: Flies were grown at 22°–24° on a standard cornmeal-yeast-agar medium. All crosses were performed at 25°. The P{P-GawB}c929 insertion was generated in the laboratory of Dr. Kim Kaiser (University of Glasgow). The w; Sp/CyO; P[629-1N] stock contains an insertion of the GAL4 coding region fused to the hsp70 promoter (Halfonet al. 1997). All other mutations are described in Lindsley and Zimm (1992) or FlyBase (Gelbartet al. 1997) and were obtained from the Bloomington Drosophila stock center and other sources.
Molecular biology: Standard molecular biology techniques were performed as described in Sambrook et al. (1989). Single-fly PCR (Glooret al. 1993) was conducted on larvae, adult males, or virgin females. Plasmid rescue of c929 (Pirrotta 1986) was used to generate probes for screening phage λEMBL3 and cosmid CoSpeR genomic libraries. The phylogenetic analysis was performed using FITCH software (Dayhoff PAM matrix; Felsenstein 1993). For cDNA isolation, probes were prepared from the plasmid rescue fragment and two SacI fragments (3.2 and 5.7 kb) flanking the c929 insertion site. These probes were used to screen cDNA libraries made from adult heads (>750,000 plaques) and 0–24-hr embryos (>250,000 plaques).
Sequencing of the crc1 allele was performed on the products of at least two independent PCR reactions from single y w; crc1/crc1 larvae. The genomic region containing exons 5 and 6 was amplified using the oligonucleotides 5′-AGCGCTCGAATTGTATCTCGGCTTT-3′ and 5′-AAGCTGGCAGACTTGATTAGCTAACT-3′. The genomic region containing exons 7 and 8 was amplified using 5′-GACCAAAAGGCAACGTTGAACTGCAT-3′ and 5′-GGACAGGCTGCGTACTTTGGTGATT-3′. Except for short regions near the template ends, the sequencing was performed on both strands. In addition to the Q171R mutation described in results, we identified the following six polymorphisms and two conservative substitutions (the second base of each pair listed in parentheses refers to the sequence of the crc1 chromosome): confirmed polymorphisms, TCGAATTG(T/A)ATCTCGGC (5′untranslated region [UTR]), GTGTATCA(A/T)TTAAAATA (5′ UTR), TTTACCGT(T/C)TTAATTTT (5′ UTR), GGAGCGGC(G/A)CAACGAAA (open reading frame; H354R), CAGCTGAT(C/T)CGAGAGTT (open reading frame; I364I), GGCTTTAT(G/T)TTCTTTGA (3′ UTR); potential mutations, ATCTTCAT(T/A)ATATAAGG (5′ UTR), TTTCAGCC(T/G)AACATTAA (open reading frame; P107P).
Microscopy and photography: Preparations were imaged on a Zeiss Axioplan fitted with a SPOT CCD camera and software (Diagnostic Instruments, Sterling Heights, MI). Live pupae were imaged with a Sony CCD-IRIS video camera.
Germ-line transformation: Germ-line transformation was performed as described (Benveniste and Taghert 1999) using a pP{UAS-crc} expression plasmid containing a 2.5-kb insert ligated into the EcoRI to NotI sites of the pP{UAST} expression vector (Brand and Perrimon 1993). The insert was made through fusion of two cDNAs at an internal EcoRI site in exon 5a. We obtained two independent insertions, and additional independent insertions were created by transposition of these onto a chromosome bearing the crcRev8 mutation.
Generation of crc deletions: We generated white-eyed revertants of c929 as described (O'Brienet al. 1994) and molecularly screened 37 for deletions of crc. One deletion, Rev8, was identified. Rev8 was mapped by Southern blot and by PCR on y w; Rev8/Df(2L)Rev4 larvae. We also generated deletions by imprecise excision and P-induced male recombination (Prestonet al. 1996) using an al dp c929 px sp chromosome. We screened 104 hemizygous revertant F2 males by single-fly PCR, and we characterized 15 independent recombinants (Svedet al. 1990) by Southern blot analysis and PCR. Five deletions of crc were identified in this screen (R1, R2, R6, E85, and E98).
Genomic organization of 39C4. (A) The organization of crc and gene Y is shown in simplified schematic organization with the direction of transcription indicated (arrows). Deletions are indicated as wavy lines; the shaded bars and arrows indicate breakpoint uncertainties. Other deficiencies of the entire region were used (e.g., Rev4 and TW161) but are not shown. P elements are indicated with triangles. Phage (λ), cosmid (cos), and plasmid rescue (4A) clones used to construct the walk are shown below. EcoRI sites determined by restriction mapping are shown as tick marks on the genomic clones and as half-tick marks below the top line. The halftick marks above the top line represent EcoRI sites predicted from the DS01560 sequence. On the proximal side of crc, there were no ESTs or clear gene homologies (other than transposons) within 20 kb. B, pBluescript side of c929; black triangles, known insertion sites (Spradlinget al. 1999); gray triangle, tentative location mapped by Southern blot and in situ hybridization (Spradlinget al. 1995); white triangle, location inferred from complementation analysis. (B) In situ hybridization to polytene salivary chromosomes using cos 2-1A as a probe. Other genomic clones from the chromosomal walk also hybridized to overlapping segments of 39C-D (not shown).
R1, R2, and c929 were backcrossed to y w67c23 for 7 generations, and R6 was backcrossed to y w67c23 for 14 generations. Unless noted otherwise, these recombined chromosomes were used for all subsequent experiments.
mRNA in situ hybridization: Whole-mount in situ hybridization was performed on stage 1–16 embryos and on larval tissues (Tautz and Pfeiffle 1989). We prepared the probes from a cDNA subclone containing sequence between the ClaI site in exon 7 and the EcoRI site in exon 8. This sequence is common to the crc-a, crc-b, and crc-c mRNAs and encodes amino acids 98–373; it includes the bZIP domain and the first 218 bp of the 3′ UTR, but no sequence corresponding to crc-d.
RESULTS
Molecular cloning of crc: The cytological location of crc1 is 39C2-4 (Wrightet al. 1976). We initiated the cloning of crc1 using a P-element insertion, c929, which we mapped to 39C4 by in situ hybridization. Subsequent to plasmid rescue, we obtained several overlapping genomic clones covering a region of ~50 kb (Figure 1A). The DS01560 Berkeley Drosophila Genome Project (BDGP) clone covers this region (GenBank accession no. AC005130).
We isolated seven cDNAs from diverse libraries (see materials and methods). All were products of the same gene, which we named crc (GenBank accession nos. AF201914–AF201924). These clones represent six distinct mRNAs (Figure 2A). Southern analysis showed that the crc locus is represented only once in the genome (data not shown). The BDGP has generated >40 crc expressed sequence tags (ESTs) derived from several stages and tissues, indicating crc expression throughout development. Among clones from adult head and embryo cDNA libraries, the isoforms crc-a and crc-b are the most abundant, representing ~85% and ~10% of the total, respectively. Each of the other crc mRNA isoforms is represented by a single cDNA.
The predicted crc open reading frames are shown in Figure 2B. CRC-A is encoded by crc-a, whereas crc-b and crc-c encode an identical isoform, CRC-B. CRC-A and CRC-B differ only at the N terminus. The crc-d transcript encodes CRC-D, a truncated isoform of CRC-A. The C terminus of the 288-amino acid region common to CRC-A and CRC-B contains basic DNA-binding and leucine zipper protein dimerization motifs (Figure 2B). The basic DNA-binding region is immediately preceded by a PEST-like sequence (PEST score 8.21, Rodgerset al. 1986). Thus, CRC may display PEST-mediated instability.
The bZIP domain of CRC displays the strongest homology with other members of the CREB/ATF superfamily of transcription factors. CRC belongs to the ATF4 subfamily, on the basis of phylogenetic analysis (Figure 3A) and the conservation of several characteristic residues in the bZIP domain (Figure 3B). CRC is most closely related to mouse and human ATF-4 (>40% sequence identity within the bZIP domain); CRC is much more distantly related to Drosophila CREB-A and CREB-B (Figure 3A).
crc encodes at least six mRNA and three protein isoforms. (A) Exons and introns of the crc transcripts are shown as boxes and horizontal lines, respectively. The putative open reading frames (black) and untranslated regions (white) are indicated. Exons were numbered as indicated below (gray boxes). The predicted translational starts of crc-a/d and crc-b contain Cavener consensus sequences (Cavener and Ray 1991) and are preceded by stop codons in each reading frame. (B) Organization of the three predicted protein isoforms. Hatched boxes, isoform-specific sequences; S, SacI; R, EcoRI; closed circles, consensus casein kinase II phosphorylation sites; open circles, consensus tyrosine kinase phosphorylation site; open squares, consensus cAMP/cGMP-dependent kinase phosphorylation site; PEST, putative protein degradation signal; BASIC, basic DNA-binding region; ZIP, leucine zipper.
Generation of crc alleles: We generated additional crc alleles using imprecise P-element excision and male recombination. We found six partial or complete deletions of crc (Figure 1A). crcRev8 (Rev8) is a complete null; it removed all of the crc exons and several exons from gene Y. The remaining five deletions (R1, R2, R6, E85, and E98) are all partial disruptions of crc. At least two of these alleles, R2 and E98, also disrupt gene Y. E85 appears to be a specific CRC-D mutant, since it affects only exon 4b. By contrast, R1, R2, and E98 disrupt both CRC-A and CRC-D. R6, which deleted the exons encoding the bZIP domain, disrupts CRC-A and CRC-B. Because exons 1–4 remain intact in R6, this allele may not disrupt CRC-D and two small 5′ RNAs (crc-a and crc-f).
R1, R2, R6, and E98 each retained some or all of c929, the P element used for the mutant screens. In E85, c929 appears to have been excised completely. An additional recombinant line, R20, contained a precise excision of c929.
Complementation analysis of the crc alleles: Complementation analysis revealed at least three lethal groups in 39C2-4, two of which (the “5′ group” and “3′ group”) were associated with deletions of crc exons (Tables 1 and 2). The 5′ group includes R1, R2, E85, and E98, all deletions of 5′ crc exons, as well as crc929 (929). The 3′ group includes crc1 and R6, a deletion of the 3′ crc exons. A third lethal complementation group was associated with disruptions of gene Y (Table 3).
With the exception of crc1, all of the crc mutant alleles share the same parental chromosome, 929. Precise excision of the c929 P element (R20) fully restored the viability of animals bearing this chromosome in trans over Rev8, a lethal deletion of the entire crc locus (Figure 1A), and over Rev4, a larger deletion of 39C (Table 1). Thus, the parental 929 chromosome displayed no lethality in 39C2-4 independent of the P-element insertion.
The crc 5′ complementation group was associated with isoform-specific disruptions of the crc gene. For example, 929 was semilethal in trans over TW161, Rev4, and Rev8 (all of which completely delete crc) but not over TW1, which leaves intact the entire crc gene as well as ~15 kb of DNA upstream of the putative crc-a transcriptional start site (Table 1). Since the 929 P element is inserted in an intron of crc upstream of the putative crc-b/c transcriptional start site (Figure 2A), the lethality caused by 929 may reflect a specific disruption of the crc-a mRNA isoform. We obtained similar results with R1, which deletes all of the 5′ exons of crc (leaving the exons encoding crc-b and crc-c intact). Both R1 and 929 displayed similar lethality (with variable penetrance) in crosses to the deficiencies TW161, Rev4, and Rev8. R1 was semilethal in homozygotes, whereas 929 homozygotes were fully viable. Thus, R1 appears to be a slightly more severe allele. This difference may stem from the fact that R1 disrupts the crc-d-f mRNAs in addition to crc-a (Figure 2A). Consistent with this hypothesis, E85, a smaller deletion that disrupts an exon specific to crc-d, displayed significant lethality in trans with TW161 and Rev4 (Table 1). The E85 chromosome also appears to bear a lethal mutation at a second, distant site: E85 homozygotes displayed greater lethality than E85 hemizygotes, and in contrast with the larger R1 deletion, E85 displayed some lethality in trans with TW1. Finally, there were two stronger lethal alleles, E98 and R2, and the degree of lethality associated with these alleles (Table 1) was correlated with the distal extent of these deletions (Figure 1A).
CRC is the Drosophila homolog of ATF4, a member of the CREB/ATF family of transcription factors. (A) An unrooted phylogenetic tree for CREB/ATF bZIP domains. Proteins separated by shorter horizontal branches are more closely related. Percentage identities between CRC-A/B and other bZIP domains are shown in parentheses. (B) Alignment of selected bZIP domains. Positions marked by boldfaced letters distinguish members of the ATF4 subfamily from other bZIP proteins. The numbers and lowercase letters to the upper right indicate the heptad repeats of the leucine zipper (see Vinsonet al. 1993). Dots, residues strongly conserved in bZIP proteins; N and R, strictly conserved residues; asterisk, strictly conserved C or S residue (except in the ATF4 subfamily); L, positions of zipper leucines; -, acidic residue available for interhelical salt bridge in heterodimers; inverted carets, residues available for interhelical salt bridges in CRC homodimers. Sequence accession nos.: dCREB-B, Drosophila CREB-B (AH004367); hTREB5, human TREB5 (NM 005171); hATF-a, human ATFa (NM 006856); rATF-3, rat ATF3 (P29596); cATF-4, Caenorhabditis elegans ATF-4 (CAA93757); hATF-4, human ATF-4 (P18848); mATF-4, mouse ATF-4 (CAA43723); mATFx, mouse ATFx (AAB21705); ApCREB-2, Aplysia CREB-2 (U40851); dCREB-A, Drosophila CREB-A (M87038); hCREM, human CREM (NM 001881); mCREB, mouse CREB (Q01147); mATF-1, mouse ATF-1 (P81269); mC/EBP, mouse C/EBP (M62362); cFOS, chicken cFOS (M37000).
The crc 3′ complementation group was associated with disruptions of both of the major crc mRNA isoforms, crc-a and crc-b. R6 deletes all of the 3′ crc exons shared by these two mRNAs. crc1 and R6 both were lethal in trans with deletions of the crc locus (Table 1), and crc1 and R6 failed to complement each other. By contrast, crc1 was fully viable over deletions that extend distally from the c929 P-element insertion site (Tables 1 and 2). Thus, the wild-type function(s) of the crc gene must include contributions by transcription units located proximal to the c929 insertion, such as crc-b and crc-c. We observed up to 2% adult escapers among hemizygous crc1 progeny (Table 2; see Figure 5A). Hemizygous R6 adult escapers were never observed. Thus, although both crc1 and R6 are very strong hypomorphs, we conclude that R6 is a more severe allele. R6 is not a complete crc amorph, since it complements E85.
The 5106/8036 complementation group maps to a separate gene: BDGP ESTs identify a novel gene (gene Y) located ~820 bp distal to crc, and we identified three independent P-element insertions in or near this gene (Figure 1A). The 06311 insertion was fully viable in trans with deletions of 39C (Table 3). The other two insertions, l(2)k05106 and l(2)k08036, form a lethal complementation group (“5106/8036”; Spradlinget al. 1995) that was independent of the 5′ and 3′ groups of crc alleles (Table 3). TW1, with a breakpoint ~15 kb distal to crc (Figure 1A), uncovers the 5106/8036 group (Table 3) but does not uncover crc1 (Table 1). The 5106/8036 group includes R2 and E98, but it does not include R1. Finally, l(2)k05106, l(2)k08036, R2, and E98 all fully complement crc1, and l(2)k05106 and l(2)k08036 complement R6, a deletion of the crc bZIP domain. Therefore, l(2)k05106 and l(2)k08036 are not crc alleles, and the 5106/8036 group corresponds either to disruptions of gene Y, or another, more distally located gene.
Complementation of crc alleles with deficiencies of 39C4
Complementation among crc alleles
Phenotypic analysis of the 3′ group of crc alleles: The crosses shown in Tables 1 and 2 revealed two largely distinct phenotypes, each generally associated with only one of the crc complementation groups. This result further indicates that the 5′ and 3′ groups represent distinct genetic functions. For the 3′ group (crc1 and R6), there were several lethal phases during larval, pupal, and adult development. Both crc1 and R6 hemizygotes displayed 15–50% of their lethality after pupariation. At each stage, the R6 allele displayed a more severe phenotype than crc1.
The molts between successive larval stages were disrupted in crc1 mutants, and this phenotype was accompanied by significant lethality (Chadfield and Sparrow 1985). We observed a comparable phenotype in the R6 mutants (Figure 4G), and the presence of supernumerary mouthparts was strongly correlated with larval lethality (data not shown). In addition to the larval molting defects, R6 hemizygotes showed delayed and defective pupariation. By contrast, crc1 hemizygotes pupariated normally (cf., Hadorn and Gloor 1943), consistent with the weaker hypomorphic phenotype of crc1 (Table 2). Although ~5% of the hemizygous R6 puparia were indistinguishable from wild type, the rest were aberrant to varying degrees. These defects included a failure to evert the anterior spiracles and a retention of a larval shape, which was thinned, elongated, and sometimes curved to one side (Figure 4H). In the most severe cases, the abdominal gas bubble, which normally forms ~6 hr after pupariation (Bainbridge and Bownes 1981), did not appear, although the larval mouthparts were later expelled.
Complementation analysis of gene Y alleles
crc1 and R6 mutant pupae displayed a range of defects associated with pupation and subsequent development (Figure 4, A–D; Table 4), as previously described for the crc1 allele (Hadorn and Gloor 1943; Fristrom 1965; Chadfield and Sparrow 1985). The pupal phenotypes of these two alleles were similar. The mutants often failed to expel or translocate the abdominal gas bubble. Head eversion failed or was incomplete, and the leg and wing discs did not completely elongate. In addition, segmentation and differentiation of the abdomen usually failed, although in some cases the anterior abdominal segments differentiated (Figure 4B). Other aspects of adult development proceeded normally, resulting in the appearance of mature eye pigments and darkened macrochaetes and differentiation of the wings and legs (Figure 4, A–D).
Adult, hemizygous crc1 females displayed markedly decreased fecundity. In addition, 5–50% (depending upon the genetic background) of the hemizygous crc1 adults of both sexes failed to expand their wings and fully tan the adult cuticle. Other hemizygous crc1 adults displayed more subtle defects involving the wings, legs, scutellum, scutellar bristles, halteres, and dorsal thorax. Many of these defects could be explained by incomplete tanning of the adult cuticle after eclosion.
Morphological defects associated with mutations in the 39C4 region and crc-c misexpression. In A–D, the puparium was removed (the pupal cuticle was left intact). In E–H, animals were photographed through the puparium. (A) Wild-type, dorsal view. (B) crc1/Rev4, dorsal view. (C) Wild type, ventral view. (D) crc1/Rev4, ventral view. (E) Ventral view of a wild-type pharate adult. (F) Ventral view of a R2/R6 pupa displaying the head/abdomen-collapsed phenotype. (G) Multiple mouthparts in a third instar y w; R6/Rev8 larva. (H) Dorsal views of a Rev8 or R6/CyO, y+ pupa (left) and a R6/Rev8 sibling (right). p, larval mouthparts; t, reduced thorax; a, naked, unsegmented abdomen; h, cryptocephal head; 3, third instar mouthparts; 2, second instar mouthparts; s, partially everted anterior spiracle; arrowheads, incompletely elongated legs. Bar, 1 mm (in G, 0.19 mm).
Pupal phenotypes associated with crc alleles
Phenotypic analysis of the 5′ group of crc alleles: The lethal phase for the 5′ group of crc alleles (R1, R2, E85, E98, and 929) was primarily after pupariation, since the number of dead pupae observed on the sides of the vials was approximately equal to the total amount of lethality. We did not observe larvae with multiple mouthparts, and the puparia were normal in size and shape (Figure 4F). In addition, gas bubble translocation, expulsion of the larval tracheae and mouthparts, and head eversion were all completed successfully. The 5′ group of alleles displayed defects in leg and wing disc elongation that were similar to those observed for the 3′ group, but they also caused novel defects in adult development (“head/abdomen-collapsed” phenotype; Table 4). In contrast to crc1 (Figure 4, A–D), the distal portions of the everted leg discs often darkened abnormally and did not differentiate further. After pupation, the abdomen shrank markedly and withdrew to a dorsal position. Subsequently, the head collapsed partially or completely into the thoracic cavity (Figure 4F). Despite these events, many pupae developed eye pigmentation and other signs of adult differentiation.
Phenotypic overlap between the 5′ and 3′ crc complementation groups: Although most of the mutations within the 5′ group fully complemented the 3′ group, R6 was an exception. R6 uncovered crc1 (3′ group) as well as R1, R2, E98, and 929 (5′ group; Table 2). In addition, R6 mutants displayed a cryptocephal phenotype when crossed to crc1 and the head/abdomen-collapsed phenotype when crossed to alleles from the 5′ group (Table 4). When placed in trans with R6, the 929, R1, R2, and E98 alleles each displayed a similar degree of lethality (independent of deletion size), indicating that each of these crc 5′ alleles displayed comparable defects in the function of the crc gene.
Interestingly, the 5′ group alleles and R6 (but not crc1) also displayed the head/abdomen-collapsed phenotype when heterozygous over either CyO, y+ or a second balancer, In(2LR)SLM. This dominant effect was associated with variable, but significant pupal lethality. Because the CyO, y+ and In(2LR)SLM chromosomes were created independently (I. Duncan, personal communication), it appears unlikely that these chromosomes share dominant enhancers of the head/abdomen-collapsed phenotype. Rather, this result suggests that the 5′ group alleles and R6 are crc haploinsufficient in some genetic backgrounds.
Sequencing of the crc1 allele: Because deletions distal to c929 complemented crc1 (Table 2), we sequenced the genomic regions containing exons 5 and 6 and exons 7 and 8 from crc1/crc1 larvae. We identified nine differences between the crc1 and wild-type sequences. Of these, six corresponded to wild-type polymorphisms, and two were conservative substitutions (see materials and methods). The remaining substitution (GATGCACAGCCAAAA; the underlined residue is G in crc1) results in a nonconservative change from glutamine to arginine (Q171R; Figure 2). Because Q171R was the only nonconservative substitution in the crc1 coding sequence, we speculate that it is the cause of the associated phenotypic defects.
Rescue of crc lethality by germ-line transformation: To confirm the molecular identification of the crc gene, we rescued the lethality of mutant crc alleles using germline transformants expressing a crc cDNA. On the basis of the complementation analysis, we predicted that ectopic expression of the CRC-B protein isoform (encoded by the crc-c mRNA, Figure 2) would rescue crc1 lethality. To test this hypothesis, we made multiple independent germ-line transformants with crc-c under the control of a GAL4 upstream activating sequence (UAS-crc). In six of seven lines, UAS-crc rescued 8–21% of the lethality in crc1/Rev8 heterozygotes (Figure 5A). The rescue was constitutive (without heat shock), presumably reflecting basal expression of UAS-crc. Heat-shock-induced expression of UAS-crc under the control of an hs-GAL4 driver caused substantial lethality in an otherwise wild-type background; thus the hs-GAL4 driver only lowered the degree of rescue seen (data not shown). The degree of rescue also was influenced by the parental genotype; for insertion 9-2, ~40% rescue was obtained when both parental stocks were balanced with CyO, y+ (data not shown).
We also attempted rescue of 5′ group functions. We chose 929 as a representative 5′ group allele for two reasons. First, the complementation (Tables 1 and 2) and molecular (Figure 2) analyses indicated that 929 displayed lethality due to disruption of crc-a without any confounding disruption of gene Y. Second, c929 is a GAL4 enhancer trap P element, which allowed heterogeneous expression of the UAS-crc transgene in 929 mutants. The c929 GAL4 reporter gene is expressed in larvae in peptidergic central nervous system (CNS) neurons, intrinsic cells of the ring gland, salivary gland, fat body, patches of the epidermis, the PM peritracheal cells, and a few other scattered locations (Schaefer and Taghert, Society for Neuroscience Abstract 1995; O'Brien and Taghert 1998; R. S. Hewes and P. H. Taghert, unpublished results). By contrast, there is very restricted c929 reporter gene expression in the imaginal discs and no detectable expression in the skeletal muscles and abdominal histoblasts. Several independent UAS-crc insertions partially or completely rescued the lethality observed in 929 hemizygotes (Figure 5B). Thus, there were two separable crc functions, and both were rescued by transgenic expression of CRC-B. Moreover, given the inclusion of numerous neurosecretory neurons in the pattern of c929 reporter gene expression, we speculate that crc may function in close association with ecdysone biosynthesis/secretion. This issue will be examined in detail in future studies.
Rescue of crc1 and 929 lethality by multiple independent UAS-crc lines. (A) Rescue of crc1/Rev8 by UAS-crc. The parental female genotype was y w; crc1/SM6. (B) Rescue of 929/Rev8 by UAS-crc. The female parents were y w; crc929. In A and B, the expected number of Cy+ adults in each cross was calculated from the number of Cy− siblings. χ2-tests were performed assuming that 3% (A) or 46% (B) of the hemizygous mutants would survive to the adult stage without rescue. All crosses were maintained at 25°. The parental male genotypes were y w; Rev8, UAS-crcx/CyO, y+, where x indicates one of nine independent UAS-crc lines (none = no insertion). For line 9-2, the parental males were y w; Rev8/CyO, y+; UAS-crc9-2/+. The total progeny in each cross was as follows (listed in order: A, B): none (695, 198), q121 (381, 180), 9-2 (203, 382), d48 (423, 339), c10 (103, 328), a6 (110, 112), x5 (73, 149), b67 (332), z124 (380, 338). * P < 0.05; ** P < 0.01; *** P < 0.001.
Whole-mount crc in situ hybridization in stage 12 embryos. All of the embryos carried a single copy of the en-GAL4 insertion, and approximately half of the embryos carried a copy of the UAS-crc9-2 insertion (the other half carried a wild-type third chromosome). (A and B) Antisense crc RNA probe. (C) Sense probe.
Expression pattern of crc transcripts: We performed in situ hybridization to determine the expression pattern of crc mRNAs. We first tested the specificity of the probes on embryos expressing UAS-crc under the control of an engrailed-GAL4 (en-GAL4) driver. Consistent with the pattern of en expression, ~50% of the stage 12 embryos showed a 14-stripe pattern (Figure 6A). This result confirmed the functionality of the UAS-crc construct and confirmed hybridization of the antisense probe to crc mRNAs. The remaining embryos showed lower, ubiquitous hybridization (Figure 6B), and this signal was detectable prior to the onset of zygotic transcription. Comparable ubiquitous staining was observed in wild-type embryos (data not shown). Thus, crc transcripts appear to be maternally loaded. No signal was detected with the sense crc probe (Figure 6C).
In wild-type, wandering stage third instar larvae, we observed specific hybridization in several tissues (Figure 7). The imaginal discs and CNS displayed the strongest signals. There was strong, relatively uniform staining in the T1-T3 leg discs and detectable, though weaker, staining in the wing discs, labial discs, and in large cells associated with the anterior spiracles. Within the CNS, the strongest expression was observed in or near the optic lobe proliferation zones (Figure 8A). The rest of the brain and ventral nerve cord showed strong, uniform hybridization, although less hybridization was observed in the posterior abdominal neuromeres. We also observed specific hybridization in patches of small cells located throughout the midgut (data not shown).
Whole-mount crc in situ hybridization in wild-type third instar larvae. (A, C, E, G, and I) Antisense crc mRNA probe. (B, D, F, H, and J) Sense probe. (A and B) Eye (e) and antennal (a) discs. (C and D) T1 leg discs. (E and F) Cells associated with the anterior spiracles (h). (G and H) CNS and ring gland (r). (I and J) Salivary gland. Arrows, small patches of strong staining within the large salivary gland nuclei.
We examined the effects of several crc alleles on the pattern of crc in situ hybridization in the CNS of feeding third instar larvae (Figure 8). There was strong hybridization in the CNS of +/Rev8 larvae (Figure 8A). By contrast, no signal was detected in larvae bearing a complete deletion of the crc locus (Rev8/Rev4, data not shown), nor was there signal in R6/Rev8 (Figure 8B) and R1/Rev8 (Figure 8C) larvae. Both R1 and R6 delete portions of crc-a, but R1 and R6 may have differential effects on crc-b/c (Figure 1A). Thus, under the hybridization and detection conditions used for this experiment, it appears that the crc-a isoform accounts for most if not all of the visible signal, while crc-b and crc-c were below detection. Finally, the pattern of hybridization in crc1/Rev8 larvae (Figure 8D) was the same as the pattern observed in the control, +/Rev8 larvae (Figure 8A). Thus, crc expression in the CNS appeared to be normal in crc1 mutants, consistent with our interpretation that a defect at the protein level (Q171R) likely accounts for the crc1 mutant phenotype.
DISCUSSION
The crc gene encodes multiple ATF4-like protein isoforms: crc is a complex locus encoding multiple mRNA and protein isoforms. The two most abundant forms are CRC-A and CRC-B; on the basis of their representation among ESTs, the transcripts encoding CRC-A outnumber those encoding CRC-B by approximately nine to one. Consistent with this observation, most of the in situ hybridization signal observed with a probe for both isoforms was attributable to the transcript encoding CRC-A (Figure 8). CRC-A and CRC-B differ at the N terminus, while a common C-terminal region contains identical bZIP protein dimerization and DNA-binding domains. Therefore, CRC-A and CRC-B likely share dimerization partners and show identical DNA-binding properties.
crc in situ hybridization in crc1 mutants. In this experiment, the color reaction was not allowed to proceed as far as in Figure 7. (A) y w; Rev8/CyO, y+ with antisense (top) and sense (bottom) crc RNA probes. (B) y w; al dp R6/Rev8. (C) y w; R1/Rev8. (D) y w; crc1/Rev8. Inset in A, ring gland displaying weak but detectable hybridization; horizontal arrows, optic proliferation zones; vertical arrows, border of stronger, uniform hybridization in the anterior ventral nervous system.
In addition to the two major mRNA isoforms, there were three uncommon transcripts, crc-d–f, which may serve regulatory functions. crc-d encodes CRC-D, a C-terminally truncated form of CRC-A. Therefore, CRC-D lacks the bZIP domain and could function as a dominant negative regulator by competing with CRC-A (or other factors) for protein-binding sites. The expression of CRC-D may be essential for viability; the crcE85 mutation, which partially deletes a CRC-D-specific exon, displays significant lethality. The crc-e and crc-f transcripts have very small open reading frames that are preceded by suboptimal translational start sites, indicating that they may not be efficiently translated. Rather, these transcripts may participate in the regulation of the crc gene (e.g., Smithet al. 1989). For mammalian CREB, the expression of truncated forms has been proposed to interrupt a positive feedback loop involving autoactivation of the gene (Waeberet al. 1991). Therefore, similar mechanisms may be involved in the regulation of crc expression.
Isoform-specific mutations reveal multiple crc gene functions: Genetic analysis demonstrated that crc is a complex locus consisting of at least two overlapping lethal complementation groups (Table 1). These complementation groups correlate with the molecular structure of the crc gene, indicating that the different CRC protein isoforms have overlapping, but distinct functions. We propose the following hypothesis to explain the correlation between the molecular and genetic results. The 3′ complementation group phenotypes reflect the functions of both CRC-A and CRC-B. Consistent with this prediction, the one observed sequence alteration in crc1 (Q171R) was found in a region common to CRC-A and CRC-B. The 5′ group phenotypes reflect CRC-A- and/or CRC-D-specific functions that do not require CRC-B.
Nevertheless, we anticipate some overlap in the functions of the different CRC proteins. The lethal phenotypes of both crc complementation groups were rescued by ectopic expression of a single RNA isoform encoding CRC-B. Furthermore, the C-terminal 288 amino acids of CRC-A and CRC-B are identical, and both lethal complementation groups displayed similar defects in leg disc elongation. R6, which deleted this common region, failed to complement both 5′ and 3′ group alleles.
crc performs critical functions during molting and metamorphosis: crc mutant alleles displayed several defects associated with molting and metamorphosis. The mutant phenotypes associated with the two lethal complementation groups were distinct, although there was some overlap. Therefore, these mutations define multiple roles for crc during development.
In insects, the molts between successive larval stages are initiated and coordinated by pulses of ecdysone (Riddiford 1993). This process appears to require crc. As previously described for the crc1 allele (Chadfield and Sparrow 1985), both crc1 and R6 displayed larval lethality associated with failure to shed the old larval mouthparts (Figure 4G). These alleles comprise the 3′ complementation group and involve disruptions common to the crc-a, crc-b, and crc-c transcripts. Therefore, CRC-A and/or CRC-B perform an important role(s) in the regulation of larval molting. Similar larval phenotypes have been described for mutations in the dare gene, which encodes an adrenodoxin reductase likely to be involved in ecdysone biosynthesis (Freemanet al. 1999). Likewise, mutants in EcR-B, which is a component of heterodimeric ecdysone receptors, and PHM, an enzyme involved in neuropeptide biosynthesis, both displayed this larval molting phenotype (Schubigeret al. 1998; Jianget al. 2000). These similarities indicate that CRC-A and CRC-B may perform necessary functions in the peptidergic neurons that stimulate ecdysone biosynthesis, in the ecdysone-producing prothoracic gland cells, and/or in the tissues that respond to the ecdysone signal.
During the third larval instar, pulses of ecdysone trigger the onset of metamorphosis (Thummel 1996). A late high titer pulse of ecdysone triggers puparium formation. Approximately 12 hr later, a subsequent brief pulse of ecdysone directs pupation. crc mutants displayed defects in pupariation and pupation, indicating that both of these developmental transitions require crc. The pupariation defects seen in R6 hemizygotes (Figure 4H)—retention of the larval shape, failure to form the abdominal gas bubble, and incomplete eversion of the anterior spiracles—are reminiscent of similar defects described for late-arrested EcR-B mutants (Schubigeret al. 1998) and for mutations in E74B (Fletcheret al. 1995) and DHR3 (Lamet al. 1999).
At pupation, crc1 and R6 both displayed the cryptocephal phenotype as well as defects in imaginal disc elongation (e.g., Figure 4, A–D). Similar pupation defects are associated with mutations in several ecdysone-response genes, including E74B (Fletcheret al. 1995), crol (D'Avino and Thummel 1998), βFTZ-F1 (Broaduset al. 1999), DHR3 (Lamet al. 1999), and the BR-C (Kisset al. 1988). Unlike crc1 and R6, the leg and wing discs in the 5′ group mutants remained bulbous and undifferentiated, and often discolored. A phenotype similar to that of 5′ group mutants has been reported for βFTZ-F1 (Broaduset al. 1999). Therefore, lesions in crc and the ecdysone-response genes generate common defects in the larval, prepupal, and pupal responses to ecdysone signaling. These similarities indicate that crc has a central role in the regulation of ecdysone biosynthesis/secretion or in determining the responses of target tisin our analysis of crc function, we plan to examine whether crc is also an ecdysone-response gene.
After comparing aspects of the E74B pupal phenotype and the phenotypes of mutations affecting larval muscle development, Fletcher et al. (1995) proposed that premature death of the larval muscles might account for the defects observed at pupariation and pupation in those mutants. Due to similarities in phenotype between E74B and crc, this model could also account for the pupariation and pupation defects observed in crc mutants, but it probably does not explain the crc larval molting and adult fecundity defects. Moreover, as is true for E74B (Fletcheret al. 1995), most crc1 and R6 mutants display normal larval locomotion (R. S. Hewes, unpublished observations), indicating that the larval muscles develop and function normally prior to metamorphosis.
Rescue of crc by germ-line transformation: Transgenic UAS-crc lines rescued the lethal phenotype of both the 5′ and 3′ lethal complementation groups (929 and crc1), confirming the identification of crc. The rescue was partial, and some aspects of the mutant phenotype, such as the reduction in female fecundity and the defects in adult wing expansion and tanning, showed no improvement. Several factors may have contributed to the incomplete rescue. These include requirements for expression of the CRC-A and CRC-D isoforms, or for more precise temporal and/or spatial regulation of CRC expression.
One aspect of the rescue experiments did not fit simple predictions but may be explained by technical details of the transgene expression. The rank order of potency for the rescue of crc1 by the different UAS-crc lines was reversed for the rescue of 929. The variation in the degree of crc1 rescue was likely due to position effects that led to constitutive, low level expression of the transgene. By contrast, c929 is an enhancer trap P element that drives heterogeneous GAL4 reporter gene expression in several tissues. Thus, to explain the second observation, c929 may rescue the wild-type pattern of crc expression to a significant degree, while minimizing the negative effects of crc misexpression in other tissues.
Potential interactions between crc and ecdysone signaling pathways: By analogy to other CREB/ATF proteins, the roles of crc during molting and metamorphosis are likely to involve heterodimerization with other bZIP proteins and competition with them for DNA-binding sites (Hai and Curran 1991). Similarly, we hypothesize that ecdysone-responsive signaling pathways include crc. For example, by convergence on the transcriptional coactivator, CREB-binding protein (CBP), CREB/ATF proteins can antagonize the activity of members of the nuclear receptor superfamily (Kameiet al. 1996). This family includes several ecdysone-response genes. Therefore, further analysis of crc may elucidate several points of interaction between crc and these hormonal signaling pathways.
Acknowledgments
We thank Rod Murphey, David Shepherd, Carl Thummel, Ian Duncan, Haig Keshishian, and the Berkeley Drosophila Genome Project for fly stocks. We also thank the Berkeley Drosophila Genome Project/HHMI EST Project for ESTs and genomic sequences. We are grateful to Mike Horner, Carl Thummel, Martin Burg, and William Pak for the sharing of information, clones, and fly stocks from 39C4, John Tamkun for the genomic libraries, and Paul Salvaterra, Gerald Rubin, and Bruce Hamilton for cDNA libraries. We thank Sonalee Jilhewar, Marie Roberts, and Dimitri Reznikov for technical assistance and Dianne Duncan for advice with in situ hybridization. This work was supported by National Institutes of Health grant NS21749 (P.H.T.) and American Cancer Society Postdoctoral Fellowship PF4212 (R.S.H.).
Footnotes
-
Communicating editor: R. S. Hawley
- Received January 13, 2000.
- Accepted April 10, 2000.
- Copyright © 2000 by the Genetics Society of America
