A Genetic Screen of the Drosophila X Chromosome for Mutations That Modify Deformed Function

Corresponding author: William McGinnis, Department of Biology 0349, 4305 Bonner Hall, 9500 Gilman Dr., University of California, San Diego, CA 92093-0349., wmcginnis{at}ucsd.edu (E-mail).

Communicating editor: T. C. KAUFMAN

	ABSTRACT

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

We have screened the Drosophila X chromosome for genes whose dosage affects the function of the homeotic gene Deformed. One of these genes, extradenticle, encodes a homeodomain transcription factor that heterodimerizes with Deformed and other homeotic Hox proteins. Mutations in the nejire gene, which encodes a transcriptional adaptor protein belonging to the CBP/p300 family, also interact with Deformed. The other previously characterized gene identified as a Deformed interactor is Notch, which encodes a transmembrane receptor. These three genes underscore the importance of transcriptional regulation and cell-cell signaling in Hox function. Four novel genes were also identified in the screen. One of these, rancor, is required for appropriate embryonic expression of Deformed and another homeotic gene, labial. Both Notch and nejire affect the function of another Hox gene, Ultrabithorax, indicating they may be required for homeotic activity in general.

HOMEOTIC (Hox) genes of the Antennapedia and Bithorax Complexes of Drosophila encode transcriptional regulators that bind DNA through a 60-amino-acid homeodomain (SCOTT et al. 1989 ). They provide positional cues along the anterior-posterior axis of developing embryos, as do their homologs in other animals (MCGINNIS and KRUMLAUF 1992 ; MANAK and SCOTT 1994 ). Hox mutations often result in homeotic transformations where one body part takes on the appearance of another (GARCIA-BELLIDO 1977 ; LEWIS 1978 ). This is usually caused by the inappropriate activity of an underlying or ectopically expressed homeotic gene (STRUHL 1982 ; STRUHL and WHITE 1985 ). For example, the absence of the Ultrabithorax (Ubx) gene causes the third thoracic leg to have the morphological characteristics of a second thoracic leg because of inappropriate activity of the Hox gene Antennapedia (MORATA and KERRIDGE 1981 ; STRUHL 1982 ). The Hox proteins specify structures in concert with dorsal-ventral and anterior-posterior signals that position those structures within a segment. For example, in the Drosophila embryo, the signaling proteins Hedgehog, Wingless, and Decapentaplegic (Dpp) determine the position of cuticular features, such as sensory organs, tracheae, and ventral denticles (IRISH and GELBART 1987 ; BEJSOVEC and WIESCHAUS 1993 ; AFFOLTER et al. 1994 ; HEEMSKERK and DINARDO 1994 ; WAPPNER et al. 1997 ). The segmentally unique patterns of these structures are controlled by the Hox proteins (LEWIS 1978 ; KAUFMAN and ABBOTT 1984 ; CASTELLI-GAIR et al. 1994 ). The involvement of Hox specification in the many disparate developmental pathways required to pattern a metameric animal suggests the Hox proteins integrate their segmental identity functions with many different cofactors.

Understanding how the Hox proteins differentially specify regulatory targets and, thus, determine unique segmental identity has been problematic. The Hox homeodomains are highly related, causing different Hox proteins to have similar or identical monomeric DNA-binding specificity (EKKER et al. 1991 , EKKER et al. 1992 , EKKER et al. 1994 ; LAUGHON 1991 ). These proteins can also bind a wide range of sequences in vitro and in vivo, allowing them to interact with the regulatory regions of a great many potential downstream genes (DESPLAN et al. 1988 ; WALTER and BIGGIN 1994 , WALTER and BIGGIN 1996 ). Auto- and cross-regulation of Hox genes further complicate the issue, as factors affecting Hox expression may also affect the ability of Hox proteins to differentially regulate other targets. An alteration in Hox DNA-binding specificity is accomplished by dimerization of Hox proteins with the Extradenticle (Exd) protein (VAN DIJK and MURRE 1994 ; KNOEPFLER and KAMPS 1995 ; VAN DIJK et al. 1995 ; CHAN et al. 1997 ). However, because most Hox proteins can dimerize with Exd on similar sites (VAN DIJK et al. 1995 ; CHAN et al. 1997 ; NEUTEBOOM and MURRE 1997 ), this is unlikely to explain entirely the distinct morphologies controlled by the homeotics. The appropriate regulation of target genes by Hox proteins is likely to require interactions with many other regulatory factors.

Screens for genetic modifiers of homeotic functions have constituted one method for identifying regulatory factors that interact with the Hox system. Previous Hox modifier screens have identifed many members of the repressive Polycomb Group (PcG) or activating trithorax Group (trxG) genes (reviewed in KENNISON 1995 ; PIRROTTA 1998 ). PcG genes are required to prevent the spread of Hox expression after the establishment of their initial expression patterns, and they encode proteins involved in the formation of suppressive chromatin structure. Genes of the trxG are required for the maintenance of Hox expression and have been proposed to be a more diversified group than PcG (KENNISON 1995 ). As might be expected, some of the TrxG factors, such as Brahma, a likely chromatin-remodeling factor (TAMKUN et al. 1992 ; PAZIN and KADONAGA 1997 ; ELFRING et al. 1998 ), may function by directly reversing the effects of PcG proteins. Others, such as Zeste (reviewed in KENNISON 1995 ) and Trl (CROSTON et al. 1991 ; FARKAS et al. 1994 ), are transcription factors that function in part by interacting with chromatin components. At least one TrxG protein, vacuolar H⁺-ATPase 55-kD B subunit (encoded by Vha55, DAVIES et al. 1996 ), does not have an obvious role in modifying chromatin structure nor in transcription, emphasizing the diversity of the factors that influence Hox function.

Another class of Hox-interacting genes (extradenticle, teashirt, homothorax, and cap'n'collar) fall into a class that we call the Hox modulators. These genes can mutate to give homeotic transformations, but they appear to act in parallel to Hox proteins, in contrast to the upstream regulatory functions of the PcG and trxG (PEIFER and WIESCHAUS 1990 ; RODER and KERRIDGE 1992 ; MOHLER et al. 1995 ; RIECKHOF et al. 1997 ). Because of the diversity of the trxG, there may be some overlap between the genes of the Hox modulator and trxG classes. The modulator proteins that have been characterized perform a variety of biochemical roles, which include enhancing the DNA-binding affinity of Hox proteins (MANN and CHAN 1996 ), regulating their transcriptional activities (PINSONNEAULT et al. 1997 ), and regulating the nuclear entry of Hox cofactors (RIECKHOF et al. 1997 ).

We have previously screened the second and third chromosomes of Drosophila for genes that show dose-sensitive interactions with the Hox gene Deformed (Dfd; HARDING et al. 1995 ; GELLON et al. 1997 ). The Dfd gene is required for the identity of two head segments (maxillary and mandibular), and it is required for larval viability (MERRILL et al. 1987 ). Dfd is distinct from the trunk homeotics in that its embryonic expression pattern does not overlap that of other Hox genes (with the exception of proboscipedia, which has no obvious morphogenetic function in embyryos, PULTZ et al. 1988 ). Without underlying homeotic expression, loss-of-function Dfd mutations result in deletions of embryonic structures rather than the homeotic transformations caused by mutations in trunk Hox genes (GARCIA-BELLIDO 1977 ). Therefore, underlying homeotic expression cannot compensate for mutations in genes that limit Dfd function. In general, the genes isolated as Dfd interactors that have been also molecularly characterized fall into two categories: those involved in cell-cell signaling (Collagen type IV, viking, devenir, Laminin A, Vha55, hedgehog, and Serrate) and those involved in transcriptional regulation (Ecdysone Receptor, apontic, cap'n'collar, Pc, and trx). Mutations in many of the Dfd-interacting genes suppress the dominant Pc phenotype (kismet, Ecdysone Receptor, sallimus, devenir, moira, Vha55, and hedgehog), indicating that they are likely to have general roles as regulators of homeotic expression patterns (KENNISON and TAMKUN 1988 ; HARDING et al. 1995 ; GELLON et al. 1997 ). However, some Dfd-interacting genes do not play a role in regulating homeotic expression (GELLON et al. 1997 ; K. HARDING, N. MCGINNIS, E. WIELLETTE and W. MCGINNIS, unpublished data) and, therefore, are likely to act in parallel or downstream of Hox proteins.

The X chromosome presents practical difficulties in modifier screens, and it has been relatively neglected when screening for mutations that interact with homeotic genes. We present here a screen of the X chromosome of Drosophila for genes involved in Dfd function. We have screened >2000 lethal chromosomes, isolating 14 alleles that show dose-sensitive interactions with Dfd hypomorphic alleles. These alleles map in seven different complementation groups. Three of these groups correspond to previously characterized X-chromosome genes involved in either cell-cell signaling (Notch) or transcriptional regulation (nejire and extradenticle). Another gene identified in the X-chromosome screen corresponds to a previously isolated but uncharacterized locus, l(1)6Ee (NICKLAS and CLINE 1983 ), which we have renamed strung out (stout) based on its embryonic phenotype (see text). Mutations in stout and three other interacting genes cause larval head defects in Dfd-dependent structures, indicating that they are required for Dfd embryonic function. However, only two of these mutants exhibit changes in Dfd expression. One of these, rancor (rnc), is a novel gene that is required for the proper regulation of both Dfd and labial. In addition, mutations in Notch and nejire modify a dominant Ubx phenotype, indicating that they may be generally involved in homeotic function.

	MATERIALS AND METHODS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Drosophila strains:
P{cosP479BE}7-20, N⁺ was obtained from S. Artavanis-Tsakonas (RAMOS et al. 1989 ). l(1)6Ee⁴, Df(1)Sxl^RA, and Dp(1;Y)ct⁺y⁺ were obtained from T. Cline (JOHNSON and JUDD 1979 ; NICKLAS and CLINE 1983 ). Df(1)Desi^S3 was obtained from A. Christensen (DORER et al. 1993 ). Df(1)C52 and Dp(1;Y)lz⁺ were obtained from A. SCHALET, and Dp(1;Y)FF1 was obtained from P. Santamaria (SANTAMARIA and RANDSHOLT 1995 ). Tp(1;Y)1 and l(1)11Ed were obtained from B. Ganetsky (COYLE-THOMPSON and BANERJEE 1993 ). Dp(1;f)y⁺ and mew were obtained from D. Brower (BROWER et al. 1995). nej³ was obtained from S. Smolik (AKIMARU et al. 1997 ). Ubx^M1 was obtained from G. Morata. All other strains were obtained from the Bloomington Stock Center. y²·YL is C(1;Y)1 marked with y² and YS is C(YS)1.

Mutagenesis:
EMS mutagenesis was performed according to standard protocols (GRIGLIATTI 1986 ). The screen is outlined in Figure 1. Briefly, isogenic y²·YL/YS males were fed mutagen and individually crossed to virgin vinscy (inscy, v^Of) females for 2 days. The females were then transferred to fresh vials, and the males were discarded. Single F₁ females carrying a uniquely mutagenized X chromosome were mated to vinscy; Ki Dfd¹³/TM6B, Tb Hu males (Dfd¹³ was originally called Dfd^rV8). Viable lines were identified by the presence of v⁺ F₂ males and discarded. F₂ females of lethal lines carrying TM6B were used for outcrossing and creating a stock over FM7c. The remaining *, y²·YL/vinscy; Ki Dfd¹³/+ F₂ females were mated to v/Y; Dfd³/TM6B males at 29° (Dfd³ is a temperature-sensitive allele originally called Dfd^rC11). On days 3 and 5, the flies were transferred to fresh vials. All Tb⁺ Hu⁺ F₃ female progeny were scored. Half of these females carry the mutagenized X chromosome (marked with v⁺), and an independently assorting half are Dfd³/Dfd¹³ (marked by the closely linked dominant Ki mutation). The ratio of Dfd mutant to Dfd⁺ animals was determined for flies carrying the mutagenized X chromosome (experimental: v⁺; Ki to v⁺; Ki⁺) and normalized to a similar ratio for flies carrying vinscy (control: v; Ki to v; Ki⁺):

where N is the number in each class, and I_S is the interaction strength shown in Table 2. For potentially positive lines, this interaction test was repeated at least twice with three initial vials.

View larger version (18K): In this window In a new window Download PPT slide   Figure 1. Screen to isolate dominant enhancers of Dfd function on the X chromosome. For the P, F1, and F2 generations, males are on the left, and females are on the right. Lethal X chromosomes were screened for enhanced lethality of Dfd13/Dfd3 hypomorphs at 29° in an F3 screen. Dfd females were identified by the presence of a closely linked Ki mutation (HARDING et al. 1995 ) and as being Tb+. The presence of the mutagenized X chromosome was determined by v+ eye color. F2 from progeny were also used to establish lines (see MATERIALS AND METHODS). Asterisks indicate mutagenized chromosomes; asterisk in parentheses indicates a possibly mutagenized chromosome. y2·YL/YS males were used to facilitate the collection of virgins in the F1 generation.

Determining map positions:
Recombinant map positions of the Dfd interactors were determined by mapping against isogenic, viable y⁺ cv v g² f or y⁺ sn³ wy car chromosomes. Appropriately marked chromosomes from these crosses were kept to provide "clean" chromosomes. Cytological positions were determined by crossing lines to flies carrying appropriate deficiencies and duplications (see Table 1). Complementation tests were performed between appropriate lines and available genes having the same map position. No duplication that covered the lethality of UC119 was found.

Lethal phase and lethal cuticular phenotype:
Embryos were collected on standard apple juice agar plates for 12 hr. These were washed and evenly distributed on clean agar plates. Yeast was added around the edges of the plate to attract larvae. After 36–48 hr, egg cases and dead embryos were counted. A line was considered embryonic lethal (E) if >75% of the expected number died before hatching and embryonic/larval lethal (E/L) if 25–75% died. Hemizygous males of the remaining lines, which did not reach pupal stages, were considered larval lethal (L).

Cuticles of unhatched larva were fixed for 3 hr in 1:4 glycerol:acetic acid and then mounted in 1:2 Hoyer's mountant:lactic acid. Aberrant cuticular structures were scored in ~20 cuticles from the embryonic lethal lines. Defects observable in 80% of the animals are listed as part of the phenotype. Other defects that were not as penetrant are also listed but qualified as such, e.g., often or less often observed.

Ubx interaction:
Ubx^M1/TM6B males carrying an isogenic Ubx^M1 chromosome were crossed to females of FM7c balanced lines. Flies were allowed to lay eggs for 3 days at 25° and then discarded. All B⁺ Tb⁺ female flies were collected from each vial and preserved in ethanol. After collections, 20 females were randomly selected. Their halteres were dissected, mounted in Hoyer's, and scored for ectopic bristles. The experiment was repeated at least once, and the total number of ectopic bristles from each experiment was used to determine an average. When standard errors of the mean >25% were obtained, a third experiment was performed. Pc⁴, trx^5C4, and the parental y²·YL chromosomes were used as controls.

In situ staining:
Lines to be analyzed molecularly were placed over FM7c carrying an actin:lacZ transgene (FM7c, P{w^+mC = act-lacZ.B}GD-1 P{w^+mC = act-lacZ.B}GD-2; MITCHELL et al. 1996 ). Embryos were stained using a standard RNA in situ protocol with an RNA digoxigenin-labeled lacZ probe to detect animals carrying the balancer.

	RESULTS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

We have screened the X chromosome of Drosophila for mutations that show dominant interactions with the homeotic gene Dfd. This was done using a slightly modified version of earlier screens (HARDING et al. 1995 ; GELLON et al. 1997 ; see MATERIALS AND METHODS). Briefly, males carrying an isogenic X chromosome were treated with EMS by standard methods. The resulting lethal X chromosomes were placed over a wild-type chromosome in a semilethal Dfd¹³/Dfd³ hypomorphic background and tested for a dominant reduction in viability (see Figure 1; MATERIALS AND METHODS). An interaction was considered positive if 65% of the expected Dfd adult female hypomorphs survived. Potentially interacting chromosomes were rescreened at least twice to confirm the interaction. In all, 2048 lethal chromosomes were screened, resulting in 17 chromosomes that consistently showed an interaction with Dfd. Lethals on these chromosomes were mapped by meiotic recombination. Four chromosomes had multiple lethals that individually failed to show an interaction with Dfd, and those will not be discussed. The remaining 14 chromosomes carried lethals that interacted with Dfd. These mutations were mapped against appropriate deficiencies and duplications to obtain cytological map positions (Table 1). Lethal phases were also determined (see MATERIALS AND METHODS; Table 2). If suitable duplications were available, appropriate lines were tested for complementation. The screen led to the identification of three known and four novel genes that genetically influence the function of Dfd (Table 2; Figure 2). Given that a single allele was isolated for four genes, it is likely that the screen was not saturating.

View larger version (30K): In this window In a new window Download PPT slide   Figure 2. Summary of Dfd-interacting genes in the Drosophila genome. The X chromosome genes were identified in the current work (see Table 2 and text for details). Genes on the second chromosome were identified by GELLON et al. 1997 , and genes on the third chromosome were identified by HARDING et al. 1995 . E(raf)2A is allelic to poc (E. WIELLETTE and W. MCGINNIS, unpublished data), EcR is allelic to snt (E. WIELLETTE, B. FLORENCE and W. MCGINNIS, unpublished data), and btl is allelic to dev (Szeged stock center, FlyBase). The fourth chromosome was not screened systematically, but ciD also showed an interaction with Dfd (M. MARTIN and W. MCGINNIS, unpublished data; see text).

The three known complementation groups are Notch (N), nejire (nej), and extradenticle (exd). Each group displayed a range of Dfd interaction strengths (I_S, Table 2) that did not notably change when alleles were placed on unmutagenized chromosomes (data not shown). Notch was originally identified as a neurogenic gene and was found to encode an epidermal growth factor repeat transmembrane receptor (reviewed in ARTAVANIS-TSAKONAS et al. 1995 ). The three Notch alleles isolated in the screen showed interaction strengths ranging from weak (N^UB53) to strong (N^S177). Many other Notch alleles were tested as well, but these failed to show an interaction with Dfd.

The largest number of interacting alleles were isolated in nej, which encodes a member of the CBP/p300 family of transcriptional adaptor molecules (AKIMARU et al. 1997 ). Two of these alleles (nej^S103 and nej^TA57) showed intermediate interactions with Dfd, and three other alleles (nej^WA69, nej^S342, and nej^Q7) had strong interactions with Dfd. There are other alleles of nej, e.g., nej^TC41 and the extant null nej³, which did not interact with Dfd. Only one allele of exd was isolated (exd^S136; preliminary results were reported in PINSONNEAULT et al. 1997 ). The variation of interaction strength within the Notch, stout, nej, and exd complementation groups suggests some alleles do not carry null mutations; other indications of the nature of these mutations will be pointed out below.

The four novel or uncharacterized loci are, proximally to distally, WC1, stout, rnc, and UC119. Only stout has multiple alleles, and these showed a strong interaction with Dfd. The three remaining loci, WC1, rnc, and UC119, showed intermediate interactions with Dfd.

Genetic interactions with other homeotics:
Although these loci were identified by their ability to interact with the gnathal homeotic Dfd, they might also have a general role in Hox-mediated patterning. We and other researchers have investigated such relationships by observing the genetic modification of the Pc phenotype, particularly the number of ectopic sex combs in adult males (KENNISON and TAMKUN 1988 ; LANDECKER et al. 1994 ; HARDING et al. 1995 ; GELLON et al. 1997 ). Because the mutations isolated in this work are male lethal, this test was not practical. Other dominant Pc phenotypes may be scored in females, but their variability in expression and penetrance would make the results difficult to interpret. A screen based on the modification of female Pc phenotypes failed to uncover homeotic interactors on the X chromosome (LANDECKER et al. 1994 ). Therefore, we chose to test the ability of the Dfd-interacting mutations to interact with mutations in another homeotic gene, Ubx.

Heterozygotes for Ubx display ectopic bristles on the haltere, an indication of a partial transformation toward wing. Heterozygosity at a locus required for Ubx function could potentially enhance this phenotype. Conversely, heterozygosity for a locus that normally represses Ubx haltere function would be expected to suppress the Ubx phenotype. The Dfd-interacting alleles were placed in a Ubx^M1 heterozygous background, and the number of bristles on ~80 halteres was scored for each line. Only nej mutations showed an enhancement of the Ubx phenotype, while Notch alleles strongly suppressed ectopic bristle formation (Figure 3). The remaining Dfd-interacting alleles showed no effect, nor did the Pc⁴ and trx^5C4 mutations (data not shown).

View larger version (55K): In this window In a new window Download PPT slide   Figure 3. Effect of the Dfd-interacting alleles on the dominant Ultrabithorax haltere phenotype. The number of ectopic bristles for each allele in a Ubx background was normalized to the parental y2·YL control (where 1 = 0.32 bristles/haltere). The arrowheads mark the expected location of genetic enhancers and repressors of Ubx function. Error bars indicate standard error of the mean for at least two separate experiments.

Phenotypes:
Dfd is required for the formation of the mouth hook, ectostomal sclerite, H-piece bar, anterior lateralgräten, cirri, ventral organ, and maxillary sense organ (MERRILL et al. 1987 ; Figure 4A). Ablation experiments have provided a rough fate map of the embryonic head for the larval cuticular structures (JURGENS et al. 1986 ), and they show that these Dfd-dependent structures are derived from the maxillary and mandibular segments where Dfd is expressed (JACK et al. 1988 ; JACK and MCGINNIS 1990 ). The Dfd targets Distalless (Dll; O'HARA et al. 1993 ), paired (prd; VANARIO-ALONSO et al. 1995 ), and Serrate (Ser; FLEMING et al. 1990 ; SPEICHER et al. 1994 ; E. WIELLETTE and W. MCGINNIS, unpublished data) are all expressed in subsets of Dfd-expressing cells. The Dll, prd, and Ser mutant phenotypes in the maxillary segment are strongly correlated with the maxillary subregions in which they are expressed. The effect of the Dfd interactors on Dfd-dependent structures and target genes is discussed below.

View larger version (96K): In this window In a new window Download PPT slide   Figure 4. Embryonic cuticular phenotypes. (A) Head phenotypes for wild type, Dfd16 nulls, and selected mutants. For rnc mutant cuticles, the arrow points to a dorsal papillus that is inappropriately located next to the antennal sense organ, and the arrowhead points to a multiply serrated mouth hook. In the stout mutant cuticle, the white arrow points to the strung-out dorsal bridge material. DBr, dorsal bridge; VP, vertical plate; eps, epistomal sclerite; MT, median tooth; ASO, antennal sense organ; MSO, maxillary sense organ; Ci, cirri; MH, mouth hook; hys, hypostomal sclerite; Ha, H-piece arm; Hc, H-piece crossbar; LG, lateralgräten. (B)Trunk phenotypes of wild type, stoutTA181, exdS136 (at 29°), and exd1 showing second and third thoracic and first abdominal segments. The white arrowhead in stoutTA181 points to the reduced width of the third thoracic denticle belt. The black arrowheads point to the first abdominal denticle belt that is transformed (exd1) or partially transformed (exdS136) in exd. White arrowheads in the exd mutants point to enlarged third thoracic denticles that represent a partial transformation toward an abdominal denticle pattern.

Notch: The classical dominant Notch phenotype includes notched wing margins. Although all three Dfd-interacting alleles showed notched wings in adults, their phenotypes were less severe than a Notch null, suggesting that all three alleles are hypomorphic. Mutant embryos lacking Notch have a neurogenic phenotype (reviewed in ARTAVANIS-TSAKONAS et al. 1995 ) where the ventral epidermis is transformed into nervous tissue and, therefore, only the dorsal cuticle is produced. However, approximately half the N^S177 cuticles had laterally derived maxillary cirri, whereas N^UB53 animals did not (data not shown). This suggests that N^S177 is a weaker mutation than N^UB53 even though N^S177 showed the strongest interaction with Dfd.

nejire: The extant nej³ allele has been molecularly characterized as a null (AKIMARU et al. 1997 ). However, in the Dfd interaction test, the viability of Dfd hypomorphs was not affected by heterozygous nej³ (114% of the expected progeny). The failure of a nej null to interact with Dfd suggests that the Dfd-interacting nej alleles are not amorphic. Consistent with this interpretation, the lethal phases of our nej alleles differ from that of the nej³. We found that only ~15% of nej³ males died as embryos, demonstrating that the nej maternal component (AKIMARU et al. 1997 ) is sufficient for embryogenesis. However, 100% of both nej^Q7 males and nej^TA57j/nej^Q7 females, as well as ~35% of nej^S342/nej^Q7 females, died as embryos (Table 2). The premature lethality of the Dfd-interacting alleles indicates they provided less functional maternal component, perhaps because of an antimorphic Nej protein. In nej^TA57/nej^Q7 or nej^Q7/Y cuticles, the maxillary and antennal sense organs often showed a slight disruption in patterning, the mouth hooks and median tooth were reduced, and the proventriculus was sclerotized (data not shown). No other phenotypes were consistently observed in cuticles of any nej genotype.

A genetic interaction has been previously noted between nej and a dominant mutation of cubitus interruptus (ci^D; AKIMARU et al. 1997 ). We have also found ci^D to genetically interact with Dfd (M. MARTIN and W. MCGINNIS, unpublished data); therefore, the interactions between ci^D and nej may have relevance to Dfd function. The ci^D allele causes Ci to be ectopically expressed in the posterior of the wing, causing defects in vein, bristle, and margin formation (SLUSARSKI et al. 1995 ). In ci^D animals, the L4 wing vein is truncated, and small amounts of ectopic wing vein are occasionally present. The posterior wing margin is fused to the alula, has larger bristles, and is often nicked (Figure 5, arrows). This latter defect can be partially suppressed by the nej³ null allele (AKIMARU et al. 1997 ). The nej^Q7 allele isolated in our screen also suppressed this phenotype (compare black arrows in Figure 5). Paradoxically, nej^Q7 also strongly enhanced the formation of ectopic wing veins in the ci^D background (Figure 5, white arrows). The lack of a similar enhancement in ci^D animals that are heterozygous for the nej³ null mutation supports the idea that nej^Q7 is antimorphic.

View larger version (164K): In this window In a new window Download PPT slide   Figure 5. Interactions of nej3 and nejQ7 with ciD. Black arrows point to the wing margin, which was nicked in ciD heterozygotes and normal in nej ciD double heterozygotes. White arrows show ectopic wing vein material. Arrowheads in insets point to the larger bristles seen in ciD and nej ciD animals.

rancor: Cuticles of rnc^TA54, rnc^TA54/Df(1)Desi^S3, and Df(1)Desi^S3 embryos showed defects in derivatives of the antennal and ventral gnathal lobes, including Dfd-dependent structures. Specifically, the hypostomal sclerite and the H-piece arm were present but malformed, the base of the mouth hook was reduced, the mouth hooks were multiply serrated (Figure 4A, arrowhead), and head involution failed. The involution defect caused the maxillary and antennal sense organs to develop on the lateral aspect of the head instead of at the normal anterior-dorsal location. Less often, the H-piece cross-bar was reduced or failed to fuse medially, the dorsal-lateral or medial papillus was missing or misplaced (Figure 4A, arrow, DLP, or DMP), the ventral organ was disrupted, and the median tooth was translocated dorsally. The head involution defect is similar to that caused by mutations in another homeotic gene, labial (lab; MERRILL et al. 1989 ). In addition, both lab and rnc mutations affect many of the same head structures, such as mouth hooks, hypostomal sclerite, and H-piece.

strung out: Cuticles of stout^TA181 and stout^UA104 male and stout^TA181/Df(1)HA32 and stout^TA181/l(1)6Ee⁴ female embryos had very similar phenotypes, although the stout^TA181/l(1)6Ee⁴ phenotype was somewhat milder. Most striking in these mutants were the defects surrounding the dorsal sac. The dorsal bridge was diffuse and occasionally unfused. Strands of sclerotic material were "strung out" from the anterior vertical plate and ventral side of the dorsal bridge in the area between the dorsal sac and lateralgräten (white arrow in Figure 4A). Also, the H-piece lateral arm was bifurcated or widened, and the H-piece crossbar was widened or broken. The mouth hooks were of normal length, but thin, and the mouth hook bases were often reduced. The trunk of stout animals also showed narrow third thoracic denticle belts (Figure 4B, white and black arrowheads). A similar denticle phenotype was seen in mutants of another Dfd interactor, EcR (BENDER et al. 1997 ; called snt in GELLON et al. 1997 ). The stout gene appears to correspond to a previously identified lethal complementation group, l(1)6Ee (NICKLAS and CLINE 1983 ), as our stout alleles were lethal when placed over the l(1)6Ee⁴ chromosome.

Heterozygous stout^TA181 and stout^UA104 adults have thick aristae (Figure 6, arrows) and arced wings, and many have one or more kinked macrochaete, phenotypes that are not seen in deficiency heterozygotes for the stout locus. In addition, males carrying these alleles over two duplications that cover the stout locus [Dp(1;2)sn^13a1 or Dp(1;Y)ct⁺y⁺] were sterile. The presence of these dominant phenotypes indicates that the stout^TA181 and stout^UA104 alleles are either neomorphic or antimorphic. Another extant allele of stout, l(1)6Ee⁴, does not exhibit these dominant phenotypes, nor does this mutant allele exhibit an interaction with Dfd in our assay.

View larger version (122K): In this window In a new window Download PPT slide   Figure 6. Dominant adult phenotype of stoutTA181. Arrows point to the aristae, which are enlarged in stout heterozygotes.

extradenticle: The exd^S136 allele is embryonic-larval lethal (~40% die as embryos), while exd nulls are embryonic lethal (Table 2; WIESCHAUS et al. 1984 ; PEIFER and WIESCHAUS 1990 ). In addition, homeotic transformations of abdominal and thoracic cuticle (WIESCHAUS et al. 1984 ; PEIFER and WIESCHAUS 1990 ) were absent in exd^S136 animals at 25° and only slightly apparent at 29° (Figure 4B, white and black arrowheads). The different lethal phase and milder cuticular phenotype indicates that exd^S136 is hypomorphic and slightly temperature sensitive. The exd cuticular phenotype in the trunk has been described previously (PEIFER and WIESCHAUS 1990 ), but the zygotic mutant head phenotype has been only noted and not described in detail. Hemizygous exd¹ cuticles showed perturbations in structures derived from all head segments, except the antennal and hypopharyngeal lobes. All cuticular structures derived from the maxillary segment were malformed, with the exception of the cirri. The lateralgräten were short and often thick, and the hypostomal sclerite was elongated, perhaps fused to the ectostomal sclerite. Also, the structures surrounding the dorsal sac, such as the median tooth, epistomal sclerite, and dorsal bridge, were disrupted or fused (Figure 4A). These dorsal sac region phenotypes may be caused by defects in dorsal head involution (JURGENS et al. 1986 ).

Other loci:
The novel WC1 and UC119 alleles displayed larval and embryonic-larval lethality, respectively. Animals that died as embryos showed inconsistent phenotypes in all segments. These variable phenotypes included occasional random holes, head defects, and poorly differentiated cuticle (data not shown). Because of the variable and poorly penetrant mutant defects, we did not further analyze these mutant alleles in the present study.

Molecular analysis:
Wild-type Dfd expression is initiated in the blastoderm embryo in the primordia of the mandibular and maxillary segments (JACK et al. 1988 ; JACK and MCGINNIS 1990 ). In early germ band-extended embryos, the Dfd expression domain includes the hypopharyngeal, mandibular, and maxillary lobes. During stages 11 and 12, Dfd expression is repressed in the hypopharyngeal and anterior mandibular epidermis (Figure 7A). This refinement in the Dfd pattern requires the repressive function of an isoform from the cap'n'collar (cnc) gene (N. MCGINNIS, E. RAGNHILDSTVEIT, A. VERAKSA and W. MCGINNIS, unpublished data). Soon after the initiation of Dfd transcription, Dfd protein is required in nearly all maxillary epidermal cells for persistent activation of Dfd transcription (ZENG et al. 1994 ). Therefore, mutations that affect either Dfd transcription or protein levels would be expected to have an effect on Dfd transcript pattern in the maxillary epidermis. To determine if the X-linked, Dfd-interacting genes were required for the normal Dfd transcript pattern, hemizygous mutants were assayed for Dfd expression by RNA in situ hybridization. Only Notch (data not shown) and rnc mutant animals showed perturbations in Dfd transcription during any stage of embryogenesis.

View larger version (251K): In this window In a new window Download PPT slide   Figure 7. Gene expression in rnc mutants. (A) Stage 12/13 expression of Dfd, cnc, prd, and Ser in wild-type and rnc mutants detected by RNA in situ analysis. Arrows denote regions where differences in gene expression were detected between wild-type and rnc mutants. (B) Transcription pattern of lab in wild-type and rnc mutants at stages 10, 13/14, and stage 17. White arrows denote the procephalic intercalary region, and black arrows denote the procephalic tritocerebral primordia. Arrowheads indicate the lab midgut domain of expression.

In rnc mutants, ectopic Dfd expression was often seen in the anterior mandibular lobe (Figure 7A). Because cnc is required for the repression of anterior mandibular Dfd expression (N. MCGINNIS, E. RAGNHILDSTVEIT, A. VERAKSA and W. MCGINNIS, unpublished data), one possiblity is that the ectopic Dfd seen in rnc mutants is indirectly caused by a loss of cnc expression. Figure 7A shows that cnc expression is severely reduced in the hypopharyngeal and mandibular lobes of rnc mutant embryos, suggesting that the aberrant mandibular Dfd expression is caused by reduced cnc levels. Expression of cnc in the clypeolabrum is unaffected in these mutants.

Although it was not obvious from the cuticular phenotype, the altered expression patterns of Dfd and cnc suggested that the loss of rnc function may result in a partial homeotic transformation of cells in the mandibular segment to maxillary identity. If this hypothesis is correct, maxillary-specific target genes of Dfd should be inappropriately regulated in the mandibular lobe of rnc mutants. We examined the expression of the Dfd targets prd, Ser, and Dll in rnc mutants. In most rnc mutant embryos, two of these genes were expressed in maxillary-like patterns in the mandibular segment: prd being ectopically activated, and Ser inappropriately absent (Figure 7A); Dll expression was unaffected in rnc mutants (data not shown). These data are consistent with a partial transformation of the mandibular lobe toward maxillary identity.

As the rnc cuticular phenotype also had similarities to the lab phenotype, we examined the lab transcription pattern in rnc mutants. Early phases of lab expression were normal in the rnc mutant embryos (stage 10, Figure 7B). However, by stage 13/14, lab transcript abundance in the intercalary and tritocerebral primordia of the procephalic region was weaker or absent in rnc mutants than in wild-type embryos (Figure 7B). The pattern and amounts of lab transcript in the midgut of rnc mutants was comparable to wild type (DIEDERICH et al. 1989 ; Figure 7B). However, after stage 15, the second midgut chamber, as marked by lab expression, was smaller and of variable diameter along its length, whereas the other chambers appeared relatively normal (Figure 7B). Therefore, even though both procephalic and midgut lab-expressing tissues are disrupted in rnc mutants, lab expression is affected only in the procephalic region.

	DISCUSSION

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

This is the final paper in a series uncovering genetic requirements for the function of the homeotic gene Dfd (see Figure 2 for a summary). We have isolated mutations in seven loci on the X chromosome that significantly reduced the viability of Dfd hypomorphs. The Notch, exd, and nej genes have been characterized previously, while stout and rnc are either novel or uncharacterized. Many Dfd-interacting alleles we have isolated are either hypomorphic or antimorphic and have stronger interactions with Dfd than nulls. This indicates that the interactions between Dfd and other genes may not always be caused by reduced gene dosage, but by an antagonistic function of the mutant protein on wild-type activity, perhaps by interfering with particular subfunctions of interacting factors. Therefore, molecular characterization of antimorphic or hypomorphic proteins may provide insight as to how the respective wild-type proteins modify Dfd (or Hox) function.

Embryonic lethal Deformed interactors:
Mutations in a gene crucial for embryonic Dfd function would be expected to cause defects in or loss of Dfd-dependent cuticular structures, as is seen in Notch, exd, stout, and rnc mutants. As mutations for the other genes identified in our screen do not appreciably affect the formation of the gnathal cuticle, the functions of these genes might be supplied maternally or affect the development of Dfd-dependent tissues not visible in cuticular preparations (e.g., tentorium and subesophageal ganglion). Alternatively, some may interact with Dfd only during postembryonic development. These results are similar to those from the second and third chromosome screens, where zygotic mutants for one-third of the loci did not exhibit defects in Dfd-dependent embryonic cuticular structures.

Transcription:
Only 18 of the 27 genes identified in all three Dfd interactor screens have been sequenced, and their encoded products have been published. Of these 18, 16 may be grouped according to their roles in either cell-cell signaling (Notch, Collagen type IV, viking, devenir, Vacuolar H⁺-ATPase 55-kD B subunit, Laminin A, Serrate, and hedgehog) or transcription (exd, nej, Ecdysone Receptor, cap'n'collar, apontic, Polycomb, trithorax, and cubitus interruptus). Because the Dfd protein is a transcription factor, it is not unexpected to have isolated a large number of genes involved in transcription. One of these, exd, is required for many homeotic functions (PEIFER and WIESCHAUS 1990 ; GONZALEZ-CRESPO and MORATA 1995 ; RAUSKOLB et al. 1995 ; AZPIAZU and MORATA 1998 ). Physical interactions between Dfd and most of the genetically interacting transcription factors have not yet been investigated. However, direct interactions have been discovered between Exd and Hox proteins, including Dfd and Dfd homologs (LU et al. 1995 ; PHELAN et al. 1995 ; CHAN et al. 1997 ). In vitro, the formation of Hox/Exd heterodimers generates a DNA-binding surface that interacts with a greater number of bases than Hox monomers, resulting in an increase in DNA-binding specificity (CHAN et al. 1994 ; VAN DIJK and MURRE 1994 ; CHAN et al. 1995 ; JOHNSON et al. 1995 ; KNOEPFLER and KAMPS 1995 ; VAN DIJK et al. 1995 ; CHAN et al. 1997 ). From the ~2000 lethal X chromosomes screened, we isolated only one exd allele—the hypomorphic exd^S136. The interaction strengths of neither exd^S136 nor exd¹ vary substantially when placed on different X chromosomes (data not shown), indicating that the interaction with Dfd is not dependent on genetic background. It is not clear why exd^S136 shows a stronger Dfd interaction than exd nulls. However, as exd^S136 zygotic mutants often have normal head structures, it is possible the exd^S136 protein preferentially disrupts Dfd postembryonic activity or functions not apparent in cuticle preparations.

The other X-chromosome Dfd interactor known to participate in transcription in nej (AKIMARU et al. 1997 ), which encodes a Drosophila member of the mammalian CBP/p300 family (LUNDBLAD et al. 1995 ). These proteins contain an inherent acetyltransferase activity and act as transcriptional adaptor molecules (BANNISTER and KOUZARIDES 1996 ; OGRYZKO et al. 1996 ; GU and ROEDER 1997 ). A plethora of transcription factors physically interact with CBP/p300 proteins, which may act as a bridge to the basal transcriptional machinery (KWOK et al. 1994 ; NAKAJIMA et al. 1997 ) or possibly alter chromatin structure (BANNISTER and KOUZARIDES 1996 ; OGRYZKO et al. 1996 ) or transcription factor activity (GU and ROEDER 1997 ) by acetylation. Preliminary experiments have failed to demonstrate an in vitro interaction between either murine CBP and Exd, or CBP and Dfd (B. FLORENCE and W. MCGINNIS, unpublished data). However, two Dfd-interacting genes encode homologs of mammalian factors that physically contact CBP or p300. The cnc gene encodes a homolog of the p45 subunit of NF-E2 (CHENG et al. 1997 ), and Ecdysone Receptor (EcR) encodes a nuclear hormone receptor (e.g., KAMEI et al. 1996 ). In addition, nej interacts genetically and functionally with ci (AKIMARU et al. 1997 ), which is downstream of another Dfd interactor, hedgehog. The ci gene is on the fourth chromosome and, therefore, was not a target of our screens. However, the ci^D allele shows a moderate interaction with Dfd in our assay (M. MARTIN and W. MCGINNIS, unpublished data). The likely physical interaction of EcR, Cnc, and Ci proteins with the Nej CBP-like protein raises the possibility that the genetic interaction observed between Dfd and nej is indirect, which is consistent with our preliminary in vitro binding results; i.e., insufficient levels of Nej might simultaneously lower the activity of Ci, Cnc, and EcR, resulting in a synergism that further lowers Dfd hypomorphic function.

Cell-cell signaling:
Nearly half of the molecularly characterized loci from all three Dfd interactor screens (see Figure 2) have functions in cell-cell signaling. Four of the Dfd-interacting loci encode transmembrane receptors (devenir and Notch) or extracellular ligands (hedgehog and Serrate). Three other loci—Laminin A, Collagen type IV, and viking—encode components of the basal lamina that also provide signals that modulate cell differentiation and gene transcription (reviewed in LAFRENIE and YAMADA 1996 ). The Vha55 gene codes for the vacuolar H⁺-ATPase 55-kD B subunit (DAVIES et al. 1996 ), which is likely to be important for the recycling of receptor molecules to the cell surface (JOHNSON et al. 1993 ). It is also possible that the calcium channel encoded by Ca-1D could have a direct affect on Dfd function, as Ca²⁺ flux has been shown recently to have a potent affect on the function of some transcriptional activators (DOMETSCH et al. 1998 ; LI et al. 1998 ). Taken together, the isolation of these loci as Dfd interactors indicates that events at the cell membrane can have critical, dosage-sensitive effects on Dfd (and Hox) function in the nucleus.

The interaction of Hox proteins with the transcriptional effectors of signaling cascades has been noted previously. For example, Ubx modulates the ability of Wingless and Dpp signals, allowing appropriate activation of Dll expression in the leg primordia (COHEN et al. 1993 ; CASTELLI-GAIR and AKAM 1995 ). In other examples, Dpp signaling in the midgut requires lab protein function to refine and amplify the lab autoactivation circuit (GRIEDER et al. 1997 ), and signaling by Wingless in the midgut requires abd-A function to activate transcription of the pointed gene (BILDER et al. 1998 ). In a similar vein, it is possible that the functions of Dfd and of its signaling interactors converge on a subset of targets so that both Dfd and the transcriptional effectors of the signals are required for proper gene expression.

The Notch gene is expressed in every cell (FEHON et al. 1991 ; KOOH et al. 1993 ) and has a role in many developmental processes: it is required to determine the fate of neurons (reviewed in ARTAVANIS-TSAKONAS et al. 1995 ), muscle cells (BAKER and SCHUBIGER 1996 ; RUIZ-GOMEZ and BATE 1997 ), oocytes (RUOHOLA et al. 1991 ; CUMMINGS and CRONMILLER 1994 ), and probably many other cell types (reviewed in MUSKAVITCH 1994 ). Therefore, a priori, the functions and expression of Notch give no clear indication of what particular role it may play in Dfd function. However, one clue can be found in the mutant phenotype for another Dfd interactor, Serrate (E. WIELLETTE and W. MCGINNIS, unpublished data), which encodes a Notch ligand. Serrate null mutants have reduced mouth hooks (SPEICHER et al. 1994 ), which are Dfd-dependent structures, suggesting that Notch signaling is required for at least one epidermal Dfd function.

The Notch alleles that interact with Dfd are hypomorphic, with the strongest effect on viability (I_S) shown by the allele with the least severe cuticular phenotype. The reason for this paradox is unclear, as is the reason why 9 of the 12 Notch alleles we tested failed to interact with Dfd. One explanation could be that the Dfd-interacting hypomorphs have levels of active Notch protein within a critical range required to observe an effect. Alternatively, the aberrations in Dfd-interacting Notch proteins could preferentially affect a Dfd- (or Hox-) specific subfunction.

Notably, no obvious intermediates (e.g., kinases) between transmembrane receptors and the nucleus were isolated in our screens although others have been successful in identifying such factors (e.g., DICKSON et al. 1996 ; KARIM et al. 1996 ). These factors often participate in multiple signaling pathways. For example, protein kinase A transmits signals from Gurken (LANE and KALDERON 1994 ) and Hedgehog (JIANG and STRUHL 1995 ; OHLMEYER and KALDERON 1997 ), while Ras functions in the EGF, JAK, IL/IFN, FGF, and integrin pathways (reviewed in CLARK and BRUGGE 1995 ; WASSARMAN et al. 1995 ; MOULE and DENTON 1997 ). Mutations in these genes would affect many pathways, some of which may have opposing effects on Dfd function, and so a genetic interaction would be difficult to detect. In addition, some individual cell-cell signaling receptors use multiple intracellular signaling cascades, as has been described for integrins (PARSON 1996 ), the FGF receptor (GISSELBRECHT et al. 1996 ; KANAI et al. 1997 ), and Notch (AXELROD et al. 1996 ). A defect in any single branch of signal transmission might be insufficient for an observable affect on Dfd function.

Hierarchical relationships of unclassified loci:
The products encoded by stout and rnc, and therefore their molecular functions, are not known. However, the morphological and molecular nature of their phenotypes gives some indication of their roles in specification by Dfd. Mutants for rnc show alterations in Dfd expression (see below), while stout mutants show no detectable alteration in Dfd expression. In addition, extant alleles of stout do not have a maternal effect (PERRIMON et al. 1989 ), indicating that stout functions in parallel with or downstream of Dfd. Placement of rnc function upstream of Dfd and stout downstream from or in parallel to Dfd is evidence for their distinct relationships with Dfd.

strung out:
Although the major defects in stout mutants are in Dfd-independent dorsal structures, these cuticles also have reduced mouth hooks and a malformed H-piece lateral bar, which are Dfd dependent. In addition, stout mutants have reduced third thoracic denticle belts, indicating a possible role in thoracic Hox function as well. A similar phenotype is also seen in EcR mutants (BENDER et al. 1997 ), suggesting a relationship between Hox-, stout-, and EcR-dependent morphogenesis in the thorax. The dominant stout phenotypes (thick aristae, kinked bristles, arced wings, and male sterility) are induced only by the two Dfd-interacting alleles, suggesting a relationship between the cause of these phenotypes and the manner in which stout interacts with Dfd. Only a few genes have been characterized whose mutations cause phenotypes similar to the dominant stout bristle and wing defects. Three encode proteins intimately involved with the cytoskeleton. Mutations affecting the actin-binding proteins Fascin (singed; CANT and COOLEY 1996 ) and Profilin (chickadee; COOLEY et al. 1992 ) cause kinked bristles, while mutations in expanded, which encodes a cytoskeletal organizing protein (BOEDIGHEIMER and LAUGHON 1993 ; BOEDIGHEIMER et al. 1993 ), cause arced wings. The phenotypic similarity of mutations in these genes to those in stout suggests a possible role for stout in the organization of the cytoskeletal architecture.

rancor:
The similar phenotypes of rnc^TA54 and Df(1)Desi^S3 hemizygotes suggest that rnc^TA54 is a strong allele. The distinctive cuticular and molecular phenotypes of rnc mutants indicate an important role for rnc in cephalic homeotic regulation and head development. Mutants for rnc have defects mainly in derivatives of the mandibular, maxillary, and labial lobes, whose structures are present but disrupted. Some of these defects are apparently caused by a partial transformation of mandibular to maxillary identity, as some genes that are regulated in a Dfd-dependent manner are regulated in a maxillary-like pattern in the mandibular segment of rnc mutants [i.e., Dfd (KUZIORA and MCGINNIS 1988 ; BERGSON and MCGINNIS 1990 ; ZENG et al. 1994 ), Ser (E. WIELLETTE and W. MCGINNIS, unpublished data), and prd (VANARIO-ALONSO et al. 1995 )].

Loss of rnc results in reduced levels of cnc in mandibular cells. As cnc antagonizes Dfd expression and function in the mandibular lobe (MOHLER et al. 1995 ; N. MCGINNIS, E. RAGNHILDSTVEIT, A. VERAKSA, W. MCGINNIS, unpublished data), the misexpression of Dfd target genes in mandibular cells of rnc mutants may be explained most simply by the reduction in cnc expression (see Figure 8). The low levels of cnc expression still evident in rnc mutants may prevent high levels of mandibular Dfd activity, resulting in only a partial transformation of mandibular to maxillary identity. Alternatively, rnc function could independently repress Dfd transcription and activate cnc transcription in mandibular cells.

View larger version (21K): In this window In a new window Download PPT slide   Figure 8. A model for regulation of intercalary and mandibular identities by rnc function. Black arrows represent positive regulation, and barred lines represent repression. (A) Wild type. Intercalary autoregulation of lab and anterior mandibular cnc expression are both dependent on rnc function. Mandibular cnc activity represses Dfd transcription. (B) rnc mutant. Both cnc and intercalary lab transcription are reduced severely, and their ability to regulate target genes (indicated by superscript arrows) is limited. Reduced cnc expression allows persistent Dfd expression in the anterior mandibular segment, which, in turn, leads to the ectopic activation of some Dfd target genes, e.g., prd.

Another rnc phenotype is a partial failure of head involution, resulting in the maxillary and antennal lobes being more posterior and lateral. The head involution defect is distinct from and more severe than that seen in Dfd mutants, but it is similar to that observed in mutants for another homeotic gene, lab (MERRILL et al. 1989 ). In addition, rnc and lab mutations disrupt some of the same cuticular structures (the hypostomal sclerite and H-piece), suggesting related biological functions for rnc and lab. Mutants for rnc have lower levels of lab transcripts in the intercalary region of the head. The timing of this requirement for rnc coincides with embryonic stages during which Lab autoactivates its own transcription unit (CHOUINARD and KAUFMAN 1991 ), suggesting that rnc activity is required for lab autoregulation (Figure 8). In the second midgut chamber, lab expression is unaffected by the absence of rnc, but the tissue itself has an altered morphology. Unlike autoregulation of lab in the head, midgut autoactivation of lab is dependent on Dpp signaling (CHOUINARD and KAUFMAN 1991 ; TREMML and BIENZ 1992 ; GRIEDER et al. 1997 ), indicating that Dpp signaling may substitute for rnc activity in lab midgut regulation.

Interactions with other homeotics:
The molecular and phenotypic analyses of our mutants suggest that they play a role in the function of homeotic genes other than Dfd. We tested this possibility by looking at their genetic interactions with Ubx. Unlike the interaction with Dfd, the Ubx test will detect interactions in only a small subset of Ubx-expressing cells (the haltere margin) during a very narrow period of time [early third instar (KAUFMAN et al. 1973 ; SANCHEZ-HERRERO and M