Needs and Targets for the multi sex combs Gene Product in Drosophila melanogaster

Corresponding author: Neel B. Randsholt, Centre de Génétique Moléculaire, C.N.R.S., F-91198 Gif sur Yvette Cedex, France., randsholt{at}cgm.cnrs-gif.fr (E-mail).

Communicating editor: T. SCHÜPBACH

	ABSTRACT

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

We have analyzed the requirements for the multi sex combs (mxc) gene during development to gain further insight into the mechanisms and developmental processes that depend on the important trans-regulators forming the Polycomb group (PcG) in Drosophila melanogaster. mxc is allelic with the tumor suppressor locus lethal (1) malignant blood neoplasm (l(1)mbn). We show that the mxc product is dramatically needed in most tissues because its loss leads to cell death after a few divisions. mxc has also a strong maternal effect. We find that hypomorphic mxc mutations enhance other PcG gene mutant phenotypes and cause ectopic expression of homeotic genes, confirming that PcG products are cooperatively involved in repression of selector genes outside their normal expression domains. We also demonstrate that the mxc product is needed for imaginal head specification, through regulation of the ANT-C gene Deformed. Our analysis reveals that mxc is involved in the maternal control of early zygotic gap gene expression previously reported for some PcG genes and suggests that the mechanism of this early PcG function could be different from the PcG-mediated regulation of homeotic selector genes later in development. We discuss these data in view of the numerous functions of PcG genes during development.

THE strict control of the expression of the Hox and HOM genes, which specify identities along the antero-posterior body axis (KRUMLAUF 1994 ; LAWRENCE and MORATA 1994 ), involves a mechanism that appears conserved between flies and vertebrates (ALKEMA et al. 1997 ; CORE et al. 1997 ; reviews by PIRROTTA 1997 ; SCHUMACHER and MAGNUSON 1997 ). Silence maintenance requires the Polycomb group (PcG) genes that are negative trans-regulators of the homeotic genes, but they are also needed for control of other selector genes. In Drosophila, the group counts at least a dozen members that regulate HOM gene expression in embryos and larvae (JURGENS 1985 ; reviewed by PARO 1995 ; PIRROTTA 1997 ). Expression of these targets is initially correct in PcG mutants, but later becomes ectopic, demonstrating that PcG products fix the repressed state of the genes that they regulate (BUSTURIA and MORATA 1988 ; MCKEON and BROCK 1991 ; SIMON et al. 1992 ). The products of polyhomeotic (ph) and Polycomb (Pc) associate with other proteins in a multimeric complex (FRANKE et al. 1992 ) that binds to about one hundred sites on polytene chromosomes (ZINK and PARO 1989 ; DE CAMILLIS et al. 1992 ; FRANKE et al. 1992 ). Proteins encoded by Posterior sex combs (Psc), Suppressor 2 of zeste (Su(z)2), Polycomb-like (Pcl), and Enhancer of zeste [E(z)] share many binding sites with PH and PC, including the two HOM gene clusters (MARTIN and ADLER 1993 ; RASTELLI et al. 1993 ; LONIE et al. 1994 ; CARRINGTON and JONES 1996 ). These binding sites define a large set of potential targets for PcG-mediated regulation. extra sex comb (esc) products, which are only transiently present in the embryo, might regulate HOM gene expression by promoting the initial binding of the multimeric complexes to chromatin (SIMON et al. 1995 ). PcG products are thought to organize the structure of their targets in such a way that the targets remain repressed throughout development, possibly by limiting the access of different transcription factors to these genes (PARO 1990 ; SCHLOSSHERR et al. 1994 ; MCCALL and BENDER 1996 ).

HOM gene regulation is not the only process that depends on PcG control because several amorphic PcG mutations have dramatic effects on cell viability or chromosome morphology. Embryonic development cannot proceed without maternal polyhomeotic product, and loss of maternal pleiohomeotic (pho) or zygotic ph leads to incomplete head involution associated with holes in the cuticle, likely due to massive epidermal apoptotic cell death (BREEN and DUNCAN 1986 ; DURA et al. 1987 , DURA et al. 1988 ; ABRAMS et al. 1993 ; GIRTON and JEON 1994 ). The E(z) product participates in some basic cellular process governing chromosome structure because loss of E(z) causes abnormal chromosome morphology and condensation (GATTI and BAKER 1989 ; PHILLIPS and SHEARN 1990 ).

The Sex combs on midleg (Scm) protein shows sequence similarity with the product of lethal(3)malignant brain tumor (l(3)mbt), a Drosophila tumor suppressor gene that controls brain cell division and differentiation, suggesting a link between PcG proteins and the products controlling cell proliferation (WISMAR et al. 1995 ; BORNEMANN et al. 1996 ). Furthermore, preliminary studies of the pleiotropic PcG gene multi sex combs (mxc) have demonstrated that mxc is allelic with lethal(1)malignant blood neoplasm (l(1)mbn), another tumor suppressor gene that controls hemocyte proliferation in larvae (GATEFF 1978 ; SANTAMARIA and RANDSHOLT 1995 ). mxc is also needed in both males and females at all stages of germline development, where the most extreme defects caused by loss of mxc are similar to proliferation defects (DOCQUIER et al. 1996 ). In so far as the wild-type mxc product controls HOM gene expression and hemocyte as well as germline proliferation in Drosophila, it provides a possible link between maintenance of cell identities and tumor suppression. A key question concerning mxc is how control of proliferation and of selector gene expression could be connected. To understand this better, we have extended our genetic studies of mxc to determine in which cells mxc product is crucially needed for cell identities, cell division, and cell survival, and which target genes might be responsible for (some of) the pleiotropic mutant mxc phenotypes.

	MATERIALS AND METHODS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Fly stocks and culture:
Fly cultures were maintained on standard culture medium at 23° or 25°, unless otherwise stated in the text. Df(1)mxc¹ is a small deletion obtained by excision of a P element inserted close to mxc (F. FORQUIGNON, unpublished results). All other mxc deficiencies, duplications, and alleles used in the present study have been described by SANTAMARIA and RANDSHOLT 1995 or by DOCQUIER et al. 1996 . Their major characteristics are summarized in Table 1 (see also RESULTS section). All mxc strains are y¹, and mxc larvae were recognized by their yellow mouth-hooks. PcG gene alleles tested in mxc^M1/Y;PcG/+ transheterozygous males were the following: Additional sex combs, Asx^XF23; extra sex combs, esc^r4; Enhancer of zeste, E(z)¹, E(z)⁶⁰ E(z)⁶², E(z)⁶³, and E(z)⁷⁰ (provided by R. JONES); Polycomb, Pc^K; Polycomb-like, Pcl^XM3; pleiohomeotic, pho^b; Posterior sex combs, Psc¹ and Psc^O; Sex combs extra, Sce¹; Sex combs on midleg, Scm^D1; super sex combs, sxc¹. The Gull allele of fat (MAHONEY et al. 1991 ) was provided by P. BRYANT, the Deformed null allele Dfd¹⁶ by T. KAUFMAN, and the Dfd^D mutation (CHADWICK et al. 1990 ) by W. MCGINNIS. lacZ enhancer traps or reporter genes were the following: cubitus interruptus-lacZ, ci-lacZ (ORENIC et al. 1990 ); decapentaplegic-lacZ, dpp-lacZ (BLACKMAN et al. 1991 ); engrailed-lacZ, en-lacZ (HAMA et al. 1990 ); ftz-lacZ, FM7c p[(ry⁺)ftz:lacZ] where lacZ is expressed in the striped ftz pattern as early as the blastoderm; hedgehog-lacZ, hh-lacZ (LEE et al. 1992 ); patched-lacZ, ptc-lacZ (LEPAGE et al. 1995 ); spalt-lacZ provided by K. BASLER (NELLEN et al. 1996 ); Ultrabithorax, Ubx-lacZ (line 35UZ-3; IRVINE et al. 1991 ); and wingless-lacZ CyO, wg-lacZ (KASSIS et al. 1992 ). Line pbG1-lacZ, a gift from D. CRIBBS, is expressed in the head primordia and in several other imaginal territories. The decapentaplegic alleles dpp^d6 and dpp^shv (from W. GELBART and the Bloomington Stock Center) were described by ST. JOHNSON et al. (1990). For description of PcG gene mutant phenotypes of all other stains and variants, see LINDSLEY and ZIMM 1992 .

Cuticle preparations and phenotype ratings:
Embryos were dechorionated with bleach and mounted in Hoyer's medium. Larval cuticles were also mounted in Hoyer's medium. Tissues with melanotic pseudotumors were dissected from second or third instar larvae. Adult phenotypes were rated in at least 20 adults (40 halves) for each genetic combination. Flies were collected 24 to 48 hr after emergence or dissected from pupal cases for combinations lethal as pharate adults and incubated in acetic acid:glycerol (4:1) at 65° overnight. Flies were then dissected, mounted in Hoyer's medium, and examined at x40 magnification. Extra sex-comb teeth were counted on all three legs.

Embryos mutant for mxc^22a-1 and mxc^G48 represent one-fourth of progeny from y¹ mxc^22a-1/FM7c and y¹ mxc^G48/FM7c females mated to mxc⁺ males. For antibody staining, y¹ mxc^G48 was balanced with FM7c p[(ry⁺)ftz:lacZ] and nonmutant embryos were recognized by the striped ftz-lacZ pattern. Rescue of nos phenotypes by mxc was tested with mxc^G48, mxc^22a-1, mxc^mbn1, Df(1)lz^10-70d, and Df(1)mxc¹ at 25°. mxc/FM7hb^7Mnos^L7/TM3 females were crossed to +/Y;TM3/nos^L7 males. Batches of 2- to 3-day-old mxc/+;hb^7Mnos^L7/nos^L7 females were then crossed to mxc⁺ males. Segmentation was evaluated on resulting embryos. As a control in each experiment, sibling +/FM7;hb^7Mnos^L7/nos^L7 females were crossed to wild-type males and segmentation of progeny estimated. Similar crosses, using balanced stocks to recognize the different genotypes, served to construct mxc^G48/+;Df(2)vg^B/+;hb^7Mnos^L7/nos^L7 and mxc^G48/+;hb^7Mnos^L7/nos^L7;pho^b/+ flies. Possible effects of Gull were tested in the progeny of crosses between y ac z mxc/FM7 females and +/Y;al Gull b pr/CyO males. Phenotypic interactions between dpp and mxc were evaluated among the progenies of dpp^d6/CyO females crossed with +/Dp(1Y)FF1''p^d6/CyO males, or in y ac z mxc^G43/FM7''p^shv/dpp^shv and y ac z mxc^G43/FM7''p^d6/dpp^d6 flies. To evaluate phenotype enhancement by other PcG alleles, batches of mxc^M1/FM7 females were crossed to balanced strains carrying autosomal PcG mutations. mxc^M1/Y;PcG/+ males were identified by the absence of dominant balancer chromosome markers. Interactions between Dfd^D and mxc were observed in mxc^M1/Y;Dfd^D/+, mxc^G46/Y;Dfd^D/+, and mxc^G43/Y;Dfd^D/+ individuals from crosses between mxc/FM7 females and Dfd^D/Dfd^D males, and in mxc^G43/Y;Dfd¹⁶/+ compared to mxc^G43/Y;TM3/+ males.

Recombinant mitotic clones:
Mitotic recombinant clones homozygous for mxc^22a-1 were induced by irradiating batches of y¹ mxc^22a-1 f^36a/+;mwh jv/+ larvae 24, 48, 72, and 96 hr before puparium formation (BPF) with 1000 rad (45 kV and 25 mA for 3 min 10 sec) and observing the resulting adult flies under a dissection microscope. Experimental clones were marked with y¹ f^36a and control clones with mwh jv.

X-gal staining of embryos and imaginal discs:
Embryos were treated as described by DOCQUIER et al. 1996 . Imaginal discs were dissected in 1x PBS from staged larvae, fixed in 1x PBS, 3.7% formaldehyde for 5 min at room temperature, washed in 1x PBS, and stained as for embryos.

Antibody staining:
Embryos were treated as described by MACDONALD and STRUHL 1986 . Imaginal tissues were stained according to LAJEUNESSE and SHEARN 1996 . Primary antibodies were directed against proteins from Sex combs reduced (Scr) (provided by W. GEHRING), from Deformed and labial (lab) (a gift from M. SCOTT), from Ultrabithorax (Ubx) (antibody FP3.38, provided by T. BRAVERMAN), and from Abdominal-B (Abd-B) (a gift from S. CELNIKER).

Acridine orange staining of imaginal discs:
mxc^mbn1/Y and mxc^16a-1/Y third instar imaginal discs were dissected in Ringer's solution, incubated 5 min in 5 µg/ml acridine orange (AO), rinsed in Ringer's, and immediately observed under a fluorescence microscope.

Brain squashes:
Larval brains were dissected in 0.7% NaCl, fixed for 30 sec each in 45 and 60% acetic acid. Tissues were stained for 1 min in a drop of orcein solution (2% orcein in 45% acetic acid) on a coverslip and then squashed between the coverslip and a clean slide and observed under the microscope. Brains were dissected from second instar y¹ mxc^G48/Y and from third instar y¹ mxc^16a-1/Y larvae raised at 25°.

	RESULTS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Characteristics of mxc alleles:
Some features of all mxc alleles used in this study (Table 1) have been described previously (GATEFF and MECHLER 1989 ; SANTAMARIA and RANDSHOLT 1995 ; DOCQUIER et al. 1996 ). mxc^G48 and mxc^22a-1 are lethal during the second larval instar as hemizygotes or homozygotes but also in trans with each other or in trans with the deficiency Df(1)lz^10-70d that uncovers the mxc locus. Furthermore, females that carry a given hypomorphic mxc allele in trans with either Df(1)lz^10-70d, mxc^G48, or mxc^22a-1 exhibit similar phenotypes. mxc^G48 and mxc^22a-1 are thus both potential amorphic alleles. All mxc alleles are strictly recessive and can be totally rescued by Dp(1Y)FF1. Classification of hypomorphic alleles from mxc^M1 to mxc^16a-1 is based on their increasingly severe phenotypes, including fertility and germline development (DOCQUIER et al. 1996 ), homeotic transformations (SANTAMARIA and RANDSHOLT 1995 ), abnormal immune responses (GATEFF and MECHLER 1989 ; SANTAMARIA and RANDSHOLT 1995 ; and see below), and lethality stage.

mxc^- is lethal in larvae and in clones:
The potential amorphic alleles, mxc^G48 and mxc^22a-1, were used to determine the mxc null phenotypes and lethality stage. Hemizygous y¹ mxc^G48 and y¹ mxc^22a-1 embryos hatch with normal cuticles. These animals exhibit no transformation of their anterior segments toward abdominal segment 8 (A8). As mxc is important for head specification (see below), we paid special attention to head structures but no difference from wild type was observed. y¹ mxc^G48/Y and y¹ mxc^22a-1/Y larvae are smaller than wild type and attain the second instar with about 24 hr delay. Most of them die several days later without going through a second molt. They exhibit gut-specific melanizations or pseudotumors (SPARROW 1978 ) that first affect the imaginal hindgut ring and later cover the posterior gut (Figure 6). All their major imaginal discs are smaller than wild type, but they contain anterior and posterior compartments, as revealed by lacZ expression in the posterior compartment imaginal disc cells of second instar y¹ mxc^G48/Y;en-lacZ/+ larvae (data not shown).

View larger version (178K): In this window In a new window Download PPT slide   Figure 1. Maternal effect of mxc. All embryos are anterior to the left. Cuticle preparations (A to D) are shown in a ventral view. X-gal stained embryos (E to H) are presented in a latero-ventral angle. (A) Cuticle of wild-type embryo. (B and C) Embryos from mxcM1/mxcM1 females crossed to wild-type males. (D) Embryo of mxcM1/mxcM1 mother crossed to a mxcM1/Y male. (E) hh-lacZ expression in a wild-type embryo at germband retraction. (F) Same in embryo from mxcM1/mxcM1 mother crossed to an hh-lacZ/hh-lacZ male. (G) Expression of dpp-lacZ in a wild-type embryo at germband retraction. (H) Same in embryo from a mxcM1/mxcM1 female crossed with a dpp-lacZ/dpp-lacZ male.

View larger version (144K): In this window In a new window Download PPT slide   Figure 2. Rescue of hb nos/nos progeny by mxc. (A) Embryo from hb7MnosL7/nosL7 female. (B–C) Rescue of embryos from mxcG48/+; hb7MnosL7/nosL7 females can be partial as in B or total, as shown in C (compare C to wild-type embryo in Figure 1A).

View larger version (107K): In this window In a new window Download PPT slide   Figure 3. Imaginal disc phenotypes of strong mxc hypomorphs. Leg and wing discs are represented posterior to the right and dorsal up. Eye-antennal discs are shown ventral to the right. All mxc16a-1/Y larvae were raised at 19°. (A) Prothoracic leg discs from a pbG1-lacZ late third instar larva where lacZ is expressed in the imaginal leg discs. (B) Fused prothoracic leg discs from a mature mxc16a-1/Y larva. (C to H) wg-lacZ expression in imaginal discs. (C) Wild-type wing disc from a mature larva. (D) From a mid-third instar larva. (E) wg-lacZ expression in wing disc from a mature mxc16a-1/Y larva. (F) Expression pattern in eye-antennal discs from wild-type late third instar. (G) From wild-type mid-third instar. (H) From mature mxc16a-1/Y larvae.

View larger version (115K): In this window In a new window Download PPT slide   Figure 4. Interactions between mxc and dpp. The wild-type expression domain of dpp-lacZ in the (A) wing disc, (C) leg disc, and (E) genital disc is broader in (B) wing, (D) leg, and (F) genital discs from mature mxc16a-1/Y;dpp-lacZ/+ larva raised at 19°. The expression domain of (G) spalt-lacZ is also wider in (H) mature mxc16a-1/Y;spalt-lacZ/+ larva. (I) Wild-type prothoracic male leg showing sex comb on basitarsus, segmented tarsus, and claws. (J) dppd6/dppd6 males have duplicated sex combs and lack distal-most structures. (K) Prothoracic leg from mxcG43/Y;dppd6/dppd6 male. (L) dppshv/dppshv males lack claws (arrow). (M) mxcG43/Y;dppshv/dppshv males have modified sex combs and fusions of intermediate tarsal segments (arrow). (N) mxcG43/Y;dppshv/dppshv males exhibit reduced eyes and duplications that can affect the first (S1) and third (S3) antennal segments. *, prothoracic ectopic sex-comb teeth.

View larger version (109K): In this window In a new window Download PPT slide   Figure 5. Cell death in mxc16a-1/Y imaginal discs. (A) Wild-type wing disc. (B) mxc16a-1/Y wing disc from mature larvae grown at 25° and stained with acridine orange, revealing cell death scattered throughout the disc.

View larger version (115K): In this window In a new window Download PPT slide   Figure 6. mxc mutants exhibit AIR phenotypes. (A) Gut from a second instar mxc22a-1/Y larva, showing melanization of the imaginal hindgut ring. (B) Third instar mxcmbn1/Y larvae. (C) mxcmbnSO/mxc22a-1 pharate adults, showing melanotic spots due to an abnormal immune response.

Loss of zygotic mxc at later stages of development was achieved by irradiating y¹ mxc^22a-1 f^36a/+;mwh jv/+ larvae BPF. No y¹ f^36a clones were found on individuals irradiated earlier than 48 hr BPF; 0.2 clone per hemi-notum was observed between 24 and 48 hr BPF, and only 0.5 clone on the hemi-notums of flies irradiated between 0 and 24 hr BPF. All clones were tiny, formed at most by a few cells, because they never included more than a single y¹ f^36a bristle. Such clones would be too small to be seen in the wing blade. mxc^22a-1 clones were never observed on the abdomen either. Similar results were observed with y¹ mxc^G48 f^36a/+ adults that had been irradiated in larvae. Control mwh jv clones appeared on the wings of the irradiated y¹ mxc^22a-1 f³⁶/+;mwh jv/+ flies with expected frequency and size (about 4.6 mwh jv clones per wing and 22 cells per clone for flies irradiated 24 to 48 hr BPF). These data show that loss of zygotic mxc causes cell-autonomous lethality and that mxc^o cells divide only a few times before dying.

Maternal effects of mxc:
The lethality stage and mitotic clone data for the strongest mxc alleles suggested that mxc product is either first required in larvae or, alternatively, that maternal mxc product allows normal development until the second larval instar. To test this, we examined the maternal effect on embryonic development of a viable hypomorphic mxc allele. Germline clones that are homozygous for amorphic or strong hypomorphic mxc alleles do not develop and most mutant mxc females lay no eggs (DOCQUIER et al. 1996 ). Maternal mxc requirement was therefore evaluated in mxc^M1 progeny because a few embryos from mxc^M1/mxc^M1 mothers show some development. mxc^M1/mxc^M1 females were crossed to wild type (Cross 1) or to mxc^M1/Y males (Cross 2), to see whether the paternal mxc⁺ copy has an effect. The most-developed embryos from Cross 1 have posterior "filzkörper," spiracles, and denticle belts, but head involution is seldom observed, dorsal closure is incomplete, and their segmentation is abnormal, especially on more anterior segments (Figure 1B). The segmentation defects include fusions and deletions of segments or of parts of segments. Occasional lateralized patches of denticle belts are also found, but we detected no homeotic transformations on the cuticles of these animals. The remaining Cross 1 embryos contain tiny pieces of necrotic tissue and have either no recognizable structures, or exhibit patches of differentiated cuticle with a few denticles and one posterior spiracle or "filz-körper" (Figure 1C). Statistical treatment of data from Cross 1 is difficult because <10% of the embryos develop. We assumed the first category to be mxc^M1/+ embryos, because they made up 51% (35/68) of the largest batch that we examined. The others would thus be mxc^M1/Y individuals. All progeny from Cross 2 are small with very severe phenotypes. These embryos were similar to the putative mxc^M1/Y individuals from Cross 1 (Figure 1D). We conclude that maternal mxc is critically required for embryogenesis. A partial rescue of the maternal effect by paternal mxc⁺ is seen but cannot restore normal development.

The structural defects of these embryos could be correlated with deregulation of genes controlling embryonic patterning. To see if this were the case, we examined mxc^M1/mxc^M1 progenies that carried lacZ reporter genes for the segment polarity genes en or hedgehog (hh), or for the dorso-ventral specifying gene decapentaplegic (dpp). In wild type, en-lacZ and hh-lacZ are expressed in a series of stripes corresponding to the posterior compartments of each segment (Figure 1E). en-lacZ/+ or hh-lacZ/+ embryos from mxc^M1/mxc^M1 mothers exhibit unevenly distributed stripes or parts of stripes (Figure 1F, and data not shown). The wild-type dpp-lacZ reporter gene marks head structures and, in each segment, a cluster of cells that are part of the imaginal disc or the histoblast primordium (Figure 1G). In mxc^M1/mxc^M1 progeny, head-specific dpp-lacZ staining is seldom seen, and the number of latero-ventral cell clusters is reduced (Figure 1H). In the embryos from mxc^M1/mxc^M1 mothers, the en-lacZ or hh-lacZ stripes do not appear larger than in wild type, and the dpp-lacZ reporter gene shows no sign of ectopic expression either. These results indicate that the drastic maternal effects of mxc^M1 on embryonic development are likely not caused by deregulation of en, hh, or dpp.

Although maternal mxc⁺ product is sufficient to specify wild-type larval cuticle in amorphs, we tested whether homeotic gene expression was normal in mxc^G48/Y embryos. Indeed, elimination of zygotic PcG products can sometimes have an effect on homeotic gene expression before the physiological modifications are detected on the embryos, or without causing adult structures to be homeotically transformed (DURA and INGHAM 1988 ; N. B. RANDSHOLT, unpublished result). Embryos from mxc^G48/+ females were stained with antibodies directed against the proteins coded by Deformed, labial, Sex combs reduced, Ultrabithorax, and AbdominalB. All these products are normally expressed in mxc^G48/Y embryos (not shown). Hence the maternal mxc⁺ component covers all needs for homeotic gene regulation during embryogenesis.

A mutant mxc product can partially rescue the maternal nos phenotype:
The regionalization of the Drosophila embryo depends on the maternally supplied products of bicoid (bcd), hunchback (hb), and nanos (nos). NOS represses the translation of the maternal hb mRNA in the posterior embryonic region. This permits the expression of the zygotic gap genes knirps (kni) and giant (gt) that specify posterior identities. These genes would otherwise be repressed by HB (reviewed by ST. JOHNSTON 1993 ). Embryos from nos/nos mothers form no abdominal segments, but this phenotype can be rescued by a total lack of hb in the maternal germline. It can also be dominantly rescued by the mutation of maternally supplied regulator molecules that normally repress kni and gt in the zygote. PELEGRI and LEHMANN 1994 have shown that certain mutant products of the PcG genes E(z), Psc, and pho can partially rescue nos by such a maternal effect. To determine if mutation of mxc also affects this regulation, we examined the cuticles of embryos from mxc/+;hb^7Mnos^L7/nos^L7 mothers that were heterozygous for different mxc mutations. We used this genetic background because a decrease in the amount of maternal hb product can partially rescue the nos phenotype in F₁ embryos (IRISH et al. 1989 ). Such embryos can differentiate a few abdominal denticle belts (Figure 2) and form an adequate background to evaluate increased rescue of nos.

We compared the progenies of young hb^7Mnos^L7/nos^L7 mothers heterozygous for five mxc variants to those of their FM7/+;hb^7Mnos^L7/nos^L7 sisters (Table 2). An average of 2.0% of the embryos from the FM7/+;hb^7Mnos^L7/nos^L7 females differentiate three or more abdominal segments, although as many as 5.8% of such embryos were seen in one control (Table 2). No rescue of the maternal nos phenotype was seen with deficiencies for mxc or with the mutant alleles mxc^22a-1 and mxc^mbn1. The progeny of mxc^G48/+;hb^7Mnos^L7/nos^L7 females show a strongly increased rescue, because 21.7% of the embryos differentiated three or more abdominal segments (Table 2 and Figure 2). It is interesting that mxc^G48 and mxc^22a-1 behave differently in this genetic context, showing that only mxc^22a-1 is like the deficiency. By following the development of embryos from mxc^G48/+;hb^7Mnos^L7/nos^L7 mothers, we showed that this combination also permits a more than 10-fold increase in the number of hatching first instar larvae, and in the rescue of adult progeny (Table 3). The adult escapers have abnormal tergites and are mostly fertile. Together these data show that a mutant maternal mxc product can affect gap gene regulation during early embryogenesis, whereas a decrease in maternal mxc⁺ product apparently has no such effect.

Several PcG genes appear to participate with mxc in gap gene regulation, but the mechanism involved is not entirely clear. The effects of PcG mutations on homeotic gene regulation are synergistically enhanced in transheterozygotes that carry hypomorphic and especially amorphic mutations of two PcG genes (JURGENS 1985 ; MOAZED and O'FARRELL 1992 ), indicating that all these products are involved in a common regulatory process. If early gap gene regulation by PcG products functions in a similar manner, then greater rescue of nos phenotypes should be obtained from hb^7Mnos^L7/nos^L7 mothers that also carry two PcG mutations. To test this, we examined progenies of mxc^G48/+;Df(2)vg^B/+;hb^7Mnos^L7/nos^L7 mothers and mxc^G48/+;hb^7Mnos^L7/nos^L7;pho^b/+ flies. Df(2)vg^B is a deletion of the Posterior sex combs/Suppressor of zeste 2 complex (Psc/Su(z)2-C) and pho^b was described as an amorphic EMS-induced pleiohomeotic mutation by GIRTON and JEON 1994 . One copy of Df(2)vg^B in hb^7Mnos^L7/nos^L7 mothers partially rescues the maternal nos phenotype (data not shown; PELEGRI and LEHMANN 1994 ). We found that at least 50% of the embryos from hb^7Mnos^L7/nos^L7;pho^b/+ mothers differentiate more than three abdominal segments. The rescue due to Df(2)vg^B or pho^b was not significantly increased by adding one copy of mxc^G48 in the mother (data not shown). These data confirm that early gap gene regulation is somehow dependent on certain maternal PcG products. But the lack of enhanced effect in a PcG double mutant background suggests that the nature, or the parameters, of this regulation differ from those pertaining to PcG-mediated repression of homeotic genes later in development.

mxc product is critically required in imaginal tissues:
The needs for mxc during larval development were determined in the hypomorphic mutants mxc^16a-1, mxc^mbn1, and mxc^G43 (see Table 1). mxc^mbn1/Y and mxc^16a-1/Y larvae were raised at 25° and 19°, because most mxc phenotypes are thermosensitive and more penetrant at 25°. At 19°, mxc^mbn1/Y larvae reach normal size with a slight delay. They are smaller and develop slowly at 25°. All mxc^16a-1/Y larvae develop in this way. mxc^mbn1/Y or mxc^16a-1/Y larvae remain at the wandering stage for up to 7 days without forming pupae. They have reduced imaginal discs and brains. The eye-antennal, leg, and wing discs appear more deformed than the haltere and genital discs. At 25°, the prothoracic leg discs of most mxc^16a-1/Y larvae develop as a fused mass of tissue (Figure 3). Coalescence of the peripodial membranes of the prothoracic leg discs takes place in wild-type individuals during early pupariation (DEMEREC 1950 ), whereas the disc fusions in mxc^16a-1/Y larvae also affect the primordial leg tissue. The mxc leg disc phenotype thus reveals a developmental abnormality specific to the larval stages because the prothoracic leg disc primordia are specified on either side of the embryo (review by COHEN 1993 ) and are only brought in close contact in the larva (AUERBACH 1936 ).

The finer structure of the reduced mxc imaginal discs in late third instar larvae was determined using a wingless-lacZ (wg) reporter gene that labels specific territories in all major discs (COUSO et al. 1993 ) (Figure 3). wg-lacZ expression in discs from wandering mxc/Y larvae resembles wg-lacZ expression in wild-type discs from younger larvae (Figure 3). Null mutations of the PcG gene E(z), which also induce a small disc and brain phenotype, exhibit abnormal mitotic figures in the brain tissue and proliferation defects (SHEARN et al. 1978 ; GATTI and BAKER 1989 ; PHILLIPS and SHEARN 1990 ). To see if the mxc mutant phenotype was associated with similar defects, we examined mitotic chromosomes from mxc larval brains. Brain squashes from second instar mxc^G48/Y or from third instar mxc^16a-1/Y larvae exhibit mitotic figures and metaphase chromosome morphology without striking anomalies (not shown). Gull, an allele of the tumor suppressor gene fat (MAHONEY et al. 1991 ), can dominantly rescue the leg phenotype of four jointed (fj), which is associated with proliferation defects in the leg discs (VILLANO and KATZ 1995 ). As defective proliferation might be responsible for the mxc phenotypes, we tried to rescue mxc mutants with Gull. Most mxc^G48/Y;Gull/+ larvae die before or during the second instar and mxc^G43/Y;Gull/+ pharate adults are no less transformed than their mxc^G43/Y siblings (data not shown), indicating that Gull has no effect on mxc phenotypes. Therefore, the disc phenotypes of mxc larvae are likely not associated with abnormal chromosome morphology or with proliferation defects. Because the fused leg discs of mxc^16a-1/Y larvae suggest an abnormal development during several cell divisions, we concluded that more than a simple arrest of imaginal development during the third instar must take place in these animals.

Strong mxc hypomorphs induce imaginal cell death:
The mxc small disc phenotype could reflect abnormal patterning. This would be possible because (1) engrailed is an embryonic and imaginal target of several PcG genes (BUSTURIA and MORATA 1988 ; MOAZED and O'FARRELL 1992 ; N. B. RANDSHOLT, unpublished results), and (2) correct pattern specification in the discs depends on en expression in posterior compartment cells, which activates the hedgehog signaling pathway and leads to organization of the future limb (TABATA et al. 1995 ; ZECCA et al. 1995 ). We examined the expression patterns of en, hh, and of the hh pathway genes patched (ptc), cubitus interruptus (ci), and dpp in mxc^mbn1/Y and mxc^16a-1/Y discs, using lacZ reporter genes or enhancer traps. en, hh, ptc, and ci expressions are similar to those of early third instar wild-type discs (data not shown), indicating that the hh pathway is not affected in the reduced mxc mutant discs.

We found, on the other hand, that expression of decapentaplegic is modified in mxc mutants. dpp-lacZ expression expands into both compartments in the genital discs, in the antennal, haltere, leg, and wing discs (Figure 4). The transforming growth factor ß (TGFß) homologue DPP acts as an appendage organizer in the imaginal discs where it diffuses from the antero-posterior compartment border and creates a gradient that governs patterning (LECUIT et al. 1996 ; NELLEN et al. 1996 ; reviewed by LAWRENCE and STRUHL 1996 ). During normal disc patterning, DPP induces a stripe of spalt expression in the imaginal discs, in a concentration-dependent manner. Using a spalt-lacZ reporter gene in mxc^16a-1/Y larvae, we showed that DPP produced by mxc mutants induces a wider spalt-lacZ domain than in wild type, indicating that the downstream targets of dpp can still react to the concentration of DPP in an mxc mutant (Figure 4G and Figure H).

Because of the ectopic expression of dpp-lacZ in mxc mutant discs, we analyzed genetic interactions between the two genes by looking at the phenotypes of double mutant flies for mxc and for the dpp mutations dpp^d6 and dpp^shv. dpp^d6 is a recessive semiviable disk-specific allele that induces pattern abnormalities of the head, thorax, and appendages, and dpp^shv is a recessive viable short-vein allele that only causes venation defects (ST. JOHNSTON et al. 1990 ). The Dp(1:Y)FF1 chromosome that carries an extra mxc⁺ copy has no effect on the homozygous phenotypes of dpp^d6 and dpp^shv (data not shown). Phenotypes induced by mutation either of dpp or of mxc were rated in mxc^G43 dpp^d6 and mxc^G43 dpp^shv double mutants raised at 19° (Table 4). Loss of the distal-most leg structures, the claws, is seen in both mxc and dpp mutants and becomes fully penetrant in mxc^G43/Y;dpp^d6/+ flies (Table 4). Another common mxc and dpp phenotype, small eyes, is also increased in mxc^G43/Y;dpp^d6/dpp^d6 adults. dpp^d6-specific phenotypes are all enhanced in these animals, including small wings, loss of distal-most tarsi, and duplication of the remaining ones (Figure 4, I–K, Table 4). mxc^G43-specific phenotypes, on the other hand, appear unaffected by dpp^d6. More ectopic sex-comb teeth are seen on some legs of the double mutant flies, but they are likely due to the dpp^d6-induced duplications of ventral structures (HELD et al. 1994 ). The mxc^G43/Y;dpp^shv/dpp^shv adults show mxc^G43/Y leg phenotypes with similar penetrance and expressivity as mxc^G43/Y single mutants. These animals also exhibit increased venation defects, reduced eyes, and maxillary palps. Both of the latter phenotypes are observed in stronger dpp^shv mutants (ST. JOHNSTON et al. 1990 ) and indicate again that mutation of mxc increases dpp mutant phenotypes. Some tarsi and antennae of these flies show fusions and duplications that are never seen in dpp^shv flies (SEGAL and GELBART 1985 ; ST. JOHNSTON et al. 1990 ). Such phenotypes recall those induced by dpp^d6 in the antennae and legs, but whereas dpp^d6 always affects the distal-most appendage structures in the strongest way, the mxc^G43/Y;dpp^shv/dpp^shv abnormalities sometimes first affect other structures along the proximo-distal axis of the antennae or the leg (Figure 4, L–N). Our data show that mutation of mxc increases dpp expression in the discs, but it also enhances the loss of dpp in discs, according to the adult phenotypes of mxc dpp double mutants. This apparent contradiction suggests that both genes are implicated in a common phenomenon.

DPP plays several roles in Drosophila. One such role was reported by BROOKS et al. 1993 , who established that dpp is ectopically expressed in discs during regeneration following extensive cell death, in a broad domain similar to the pattern we observed in the mxc mutants. To see whether cell death was occurring in mxc mutants we stained mxc^mbn1/Y and mxc^16a-1/Y imaginal discs with AO. This revealed that apoptosis occurs in all territories of mxc imaginal primordia, demonstrating that mxc is required by most imaginal cells and that alteration of mxc in strong hypomorphs causes imaginal cell death (Figure 5). Cell death, followed by regeneration in hypomorphic mxc discs, could thus explain the broadening of the dpp-lacZ expression domain. Furthermore, prothoracic leg discs have been reported to fuse during regeneration after massive cell death induced during the first larval instars (POSTLETHWAIT and SCHNEIDERMAN 1973 ). Together, these data indicate that cells are dying in the discs of strong hypomorphic mxc mutants, possibly as early as during the first larval instar. As dpp mutants have also been reported to induce cell death in imaginal discs (BRYANT 1988 ), we conclude that the enhancement of dpp phenotypes by mxc could be attributed to the joint effect of two different patterns either of reduced viability or of cell death in the discs.

Alteration of mxc induces abnormal immune responses:
All mxc mutant larvae and some adults exhibit melanotic pseudotumors with variable penetrance and expressivity, from rare mxc^M1/Y larvae raised at 25° under crowded conditions to almost all L3 mxc^mbn1 or mxc^16a-1 and L2 mxc^22a-1 or mxc^G48 larvae. Melanotic spots in mxc^22a-1/Y and mxc^G48/Y larvae touch the hindgut imaginal ring and the hindgut, whereas gut and fat body tissue can be melanized in other mutants (Figure 6). Imaginal discs or central nervous system structures are not melanized in mxc mutants. Because melanized pseudotumors result from an aberrant immune response (AIR) (WATSON et al. 1991 ) of the individual toward parts of its own body where lamellocytes—specialized hemocytes—encapsulate tissue perceived as "alien" and are melanized afterwards (SPARROW 1978 ), the pseudotumors of mxc mutants indicate that recognition capacities or membrane properties have been changed either in the hemocytes in charge of the defense response, in the cells that are encapsulated by the lamellocytes, or in both, and this in viable as well as in lethal mxc mutants.

Standard homeotic transformations of mxc mutants are enhanced by PcG mutations:
A number of the homeotic phenotypes of mxc mutant flies are induced by many PcG mutations, whereas other modifications are specific to mxc (SANTAMARIA and RANDSHOLT 1995 ). Among the latter phenotypes are the ectopic sex-comb teeth that develop on tarsal segment 2 of the prothoracic legs of mxc hypomorphs. This transformation is also induced by particular alleles of some PcG genes (JONES and GELBART 1990 ) and by all viable alleles of the cramped gene (crm). It could reveal a change in proximo-distal identities along the prothoracic leg (SANTAMARIA 1993 ; YAMAMOTO et al. 1997 ). The first abdominal segment (A1) of mxc mutant males is consistently smaller than wild type (Figure 7). This modification is similar to a weak bithoraxoid (bxd) phenotype (CASTELLI-GAIR et al. 1992 ). Furthermore, a tuft of bristles sometimes replaces ommatidia in the antero-ventral eye region of mxc mutants (SANTAMARIA and RANDSHOLT 1995 ), mimicking the phenotype of the gain-of-function Deformed allele Dfd^D (CHADWICK et al. 1990 ). We wondered to which extent all these phenotypes depended on joint regulation of the corresponding targets by PcG products. Because phenotypes that are due to deregulation of common PcG targets, such as the homeotic genes, are synergistically enhanced in transheterozygous mutant flies, we scored homeotic transformations in adult males that were hemizygous for a viable mxc allele and carried one copy of a PcG mutation on the autosomes (mxc/Y;PcG/+ flies). Because mxc^G46 is lethal in trans with several autosomal PcG mutations, we used the mxc^M1 allele that on its own induces weak homeotic phenotypes in males (SANTAMARIA and RANDSHOLT 1995 ).

View larger version (155K): In this window In a new window Download PPT slide   Figure 7. Small tergites and Ubx deregulation in mxc mutants. Cuticle from (A), wild-type male; and (B), mxcM1/Y;ScmD1/+ male, showing a bxd-like phenotype. (C) Expression of the Ubx-lacZ reporter gene from line 35UZ-3 in wild-type wing discs and (D) haltere discs from late third instar larvae. (E) Same in wing disc from mature mxcmbn1/Y larva raised at 19°.

All mxc^M1/Y;PcG/+ animals are viable or die as pharate adults. No significant interaction is detected in mxc^M1/Y in trans with one mutant copy of esc^r4, Sce¹, or with E(z)⁶³ (Table 5). Some enhanced transformations are observed for sxc¹, Asx^XF2, pho^b, and Pcl^XM3. Strong synergistic interactions were observed in flies carrying one copy of the antimorphic allele E(z)⁶⁰, of Scm^D1, Psc¹, or Pc^k. We found that standard PcG phenotypes of antennae, legs, wings, or abdomen, due to gain of function of Antp, Scr, Ubx, abdA, and AbdB, are synergistically enhanced in mxc^M1/Y;PcG/+ flies (Table 5). The enhanced leg phenotype involves an increased number of sex-comb teeth on the basitarsus of posterior legs, and the differentiation of prothoracic-specific bristles on the tibias of the second and third legs of mxc^M1/Y;PcG/+ flies (data not shown; HANNAH-ALAVA 1958 ). Antennae into leg transformations in mxc^M1/Y;Psc¹/+ flies feature reduced aristae and swollen second and third antennal segment with coxa or tibia-specific bristles (not shown) (POSTLETHWAIT and SCHNEIDERMAN 1971 ). Maxillary palps are also reduced in these flies (not shown).

These data indicate that several homeotic genes require the mxc product together with other PcG gene proteins to maintain their wild-type expression pattern during development. We checked whether mxc⁺ is needed for regulation of Ubx expression in imaginal discs, using the P[35UZ] transposon that functions as an Ubx-lacZ reporter gene (IRVINE et al. 1991 ). In line 35UZ-3, lacZ is expressed following a wild-type Ubx-like epidermal pattern in the posterior compartments of the metathoracic imaginal discs (IRVINE et al. 1991 ). ß-Galactosidase staining of discs from mxc^mbn1/Y;35UZ-3/+ larvae reveals ectopic expression in the wing discs and in many other imaginal structures (Figure 7 and data not shown). This result confirms the need for mxc⁺ product for negative control of the Ubx promoter.

Among the phenotypes that appeared specific to mxc, only the bxd-like A1 modification was increased in trans by the PcG mutations that we tested (Table 5). In extreme cases, the tergites form tiny plates that hardly join along the midline, as in mxc^M1/Y;Scm^D1/+ males (Figure 7). The proximo-distal leg transformations were not significantly increased in any of these mxc^M1/Y; PcG/+ flies. Three E(z) alleles, E(z)¹, E(z)⁶², and E(z)⁷⁰, that on their own can induce ectopic sex-comb teeth on the second tarsal segment of prothoracic male legs (JONES and GELBART 1990 ) were also tested in combination with mxc^M1/Y. None of them enhanced this transformation (data not shown). The Dfd^D-like eye transformation was not enhanced by PcG mutations either. These data suggest that mxc has a particular role in proximo-distal identity determination and in head specification.

mxc product is needed for adult head specification and regulates Deformed:
The need for wild-type mxc in head specification is illustrated by the eye phenotype of viable mxc mutants and by the head phenotype of mxc^22a-6/mxc^mbnSO females. These animals die as pharate adults and exhibit extremely reduced eyes and head structures (Figure 8). Differentiated eye structures, such as ommatidia and bristles, are present on the rudimentary heads. Antennal segments 1 and 2 appear normal, whereas antennal segment 3, the arista, and maxillary palps are reduced. Head structures derived from the labial and clypeo-labial discs, such as the proboscis, labrum, and clypeus appear normal (not shown), indicating that the need for mxc product is particularly strong in eye-antennal disc derivatives.

View larger version (125K): In this window In a new window Download PPT slide   Figure 8. Head specification and Deformed regulation require mxc+. (A) Head of mxcmbnSO/mxc22a-1 pharate adult. (B) Head of DfdD/+ male, showing ectopic antero-ventral bristles (arrow). (C) Similar phenotype in mxcG43/Y fly. (D) Head of mxcG43/Y;DFdD/+ male. (E to I) Eye-antennal discs from mature third instar larvae stained with anti-DFD antibody. All discs are shown ventral to the right. (E) DFD expression in wild type and (F) in DfdD/+ discs. (G) Dfd expression in mxcmbn1/Y at 19°. Arrowheads indicate ectopic DFD expression. (H and I) mxcG43/Y DfdD/+ larvae have reduced eye discs.

We looked for genetic interactions between mxc and Dfd because the identities of the maxillary palps and the head capsule that are affected in mxc mutants are specified by Dfd. The Dfd^D-like transformation of mxc^G43/Y flies depends on Dfd⁺ dosage because mxc^G43/Y;Dfd¹⁶/+ flies do not show this phenotype, whereas their mxc^G43/Y;TM2/+ siblings do. Furthermore, the gain-of-function allele Dfd^D enhances the phenotype of mxc mutant males. A single Dfd^D copy in trans with mxc^M1/Y, mxc^G46/Y, or mxc^G43/Y induces head phenotypes that are more severe than those of Dfd^D homozygotes (Figure 8). These flies have no aristae, and the eye-disc-derived maxillary palps and eyes are extremely reduced even in mxc^M1/Y;Dfd^D/+ males, despite the fact that mxc^M1/Y flies have normal eyes. The defects observed in mxc/Y;Dfd^D/+ flies are similar to those of hsp70-Dfd-58 flies that have been exposed to ectopic DFD in all disc territories during larval development (CHADWICK et al. 1990 ), suggesting that ectopic DFD could be responsible for the mxc eye phenotypes.

Wild-type DFD is present in antennal disc cells, in the peripodial membrane and in the lateral anterior disc territories that specify maxillary palps and rostral membrane, whereas ectopic DFD expression and disc reductions are observed in Dfd^D eye discs (CHADWICK et al. 1990 ). Both mxc^16a-1/Y and mxc^mbn1/Y third instar imaginal discs stained with anti-DFD antibody exhibit weak ectopic staining. In the small mxc^mbn1 eye discs ectopic DFD expression is seen in a number of cells, including the cells which will form the eye territory where tufts of bristles replace ommatidia in mxc mutants (Figure 8). A high level of ectopic DFD is detected in the eye discs of mxc^G43/Y;Dfd^D/+ larvae, associated with a dramatic reduction in the eye primordia that give rise to the compound eye (Figure 8), explaining the almost total absence of ommatidia in mxc^G43/Y;Dfd^D/+ adults. From these genetic and expression data, we conclude that mxc product is required to maintain expression of Dfd in the eye antennal disc, identifying Dfd as one of the mxc targets which is responsible for the effect of mxc mutations on head specification.

	DISCUSSION

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

mxc functions as a PcG gene during development:
mxc functions as a Polycomb group gene according to three basic characteristics: (1) mxc phenotypes affect many segments of the adult body and mimic those of gain-of-function mutations in the ANT-C and BX-C genes; (2) these phenotypes are synergistically enhanced by mutations of other PcG genes; and (3) they are associated with ectopic expression of homeotic target genes in imaginal discs. The functions of trxG and PcG genes are often described as antagonistic, although LAJEUNESSE and SHEARN 1996 have shown that E(z) mutants have features of either group. In this context, it is noteworthy that the bxd-like loss-of-function phenotype of mxc hypomorphic males is enhanced by several PcG mutations, suggesting that more PcG genes than E(z) may be needed for HOM gene activation.

mxc has, like many PcG genes, a strong maternal effect that cannot be entirely rescued by the father (BREEN and DUNCAN 1986 ; DURA et al. 1988 ). The defects of the cuticle of the few escapers from mxc^M1/mxc^M1 mothers are likely due to the strong need for mxc product during oocyte formation (DOCQUIER et al. 1996 ), because eggs from mxc^M1/mxc^M1 females are abnormal, most of them are small, and they collapse rapidly after oviposition. Misdistributed maternal determinants in oocytes from the mxc^M1/mxc^M1 mothers could explain the large cuticle holes and abnormal segmentation of their progeny. These embryos do not show the posterior transformations of other lethal PcG mutants (BREEN and DUNCAN 1986 ; DURA et al. 1987 ; PHILLIPS and SHEARN 1990 ), but mxc product may still regulate HOM genes in embryos if the paternal allele provides sufficient mxc⁺ product, like the wild-type paternal allele of Sce (BREEN and DUNCAN 1986 ). Alternatively, mxc might regulate HOM gene expression only during the larval stages, which would place mxc in a category of its own.

Deformed is a target of mxc regulation:
Dfd is an mxc target since the Dfd^D-like eye phenotype of mxc mutants depends on Dfd⁺ dosage, because the Dfd^D/+ phenotype is enhanced by mxc and because DFD is ectopically expressed in mxc eye discs. The ANT-C gene labial also specifies posterior and lateral head capsule (DIEDERICH et al. 1991 ), but we saw no ectopic label in the eye primordia of mxc mutant discs stained with anti-LABIAL antibody (data not shown), indicating that labial is not responsible for this mxc phenotype. The regulation of Dfd is likely specific to mxc. Indeed, no other PcG mutation exhibits Dfd^D-like phenotypes, hypomorphic polyhomeotic alleles have no effect on the eyes of Dfd^D/+ flies (data not shown), and this phenotype is not enhanced in mxc/Y;PcG/+ mutants. Homeotic gene expression is maintained within its normal domain during development through negative trans-regulation by the PcG and positive regulation by trxG genes. GELLON et al. 1997 recently showed that trxG products control Dfd expression, and mxc product might play the opposite role in control of Dfd.

An interesting point concerning the head phenotype of mxc^G43/Y;Dfd^D/+ males is the replacement of the compound eye by naked cuticle, as well as the disappearance of the corresponding eye disc territories. Other PcG gene mutants exhibit phenotypes where structures are apparently lost, such as the loss of the humerus in ph mutants (DURA et al. 1985 ), and the partial loss of the sixth segment in PcG mutant males. The latter phenotype is likely a partial transformation of A6 toward A7, which bears no tergites in D. melanogaster males (SANTAMARIA and GARCIA-BELLIDO 1972 ). Such phenotypes could be interpreted as homeotic transformations toward structures that are intrinsically defined by lesser proliferation rates.

Gap gene regulation by mxc and Polycomb group genes:
PELEGRI and LEHMANN 1994 proposed that certain PcG genes are required for the maintenance of the expression domains of knirps and giant, through a mechanism similar to the regulation of HOM genes. If PcG regulation of the gap genes and of HOM genes later in development involved a similar mechanism, then loss-of-function mutations should have a strong effect on rescue, and the embryos from hb nos/nos mothers that have two PcG mutations in their genetic background should permit increased rescue of the nos phenotype.

Three E(z)^son (suppressor of nanos) alleles or a hypomorphic pho allele partially rescue the phenotypes of hb nos/nos progeny by a maternal effect; deficiencies covering E(z) or the Psc/Su(z)2 complex also allow some maternal rescue of hb nos/nos progeny, yet the strongest effect is observed with the gain-of-function E(z)^son alleles (PELEGRI and LEHMANN 1994 ). We found that the EMS-induced allele mxc^G48 rescues the hb nos/nos progeny phenotype, whereas a deficiency of mxc does not. We also observed, like PELEGRI and LEHMANN 1994 , some rescue with the Psc/Su(z)2 complex deletion Df(2)vgB, but in contrast to their data for a deletion of pho, we found strong rescue (consistently >50%) with an EMS-induced pleiohomeotic allele pho^b described as amorphic (GIRTON and JEON 1994 ). This suggests that pho^b and mxc^G48 are probably not amorphic alleles, and that maternal rescue of hb nos/nos progeny by a PcG gene is most efficient with a non-null mutation.

We looked at the segmentation of embryos from transheterozygous mothers. Although quantification of the nos rescue is difficult because it depends on many parameters (temperature, medium, age of the females), we found that rescue is not increased in progeny from such females. This was obvious for embryos from mxc^G48/+;hb nos/nos;pho^b/+ flies, compared to those from hb nos/nos;pho^b/+ females. Yet the mxc and pho genes interact in HOM gene regulation because the transformations of mxc^M1/Y;pho^b/+ males are much stronger than those of mxc^M1/Y males.

Because neither a reduction of wild-type PcG product nor two PcG mutations in trans in the hb nos/nos mothers increases nos rescue, our data strongly suggest that, whatever the mechanism of gap gene regulation by these PcG mutations may be, it does not function like the PcG-mediated maintenance of HOM gene expression in embryos and in imaginal discs. The strong rescue provided by several non-null EMS-induced mutations, which may produce mutant proteins, leads us to propose that modified PcG proteins are poisoning a normal process. How this process depends on wild-type regulation by PcG products has yet to be established.

Loss of mxc is cell lethal:
Cell death is a common feature of many mxc mutations: in the embryos from mxc^M1/mxc^M1 mothers, in the small imaginal discs of strong hypomorphs, and in clones of mxc null cells. Cell death is also observed in association with loss of other PcG products, in amorphic ph embryos (DURA et al. 1987 ; ABRAMS et al. 1993 ), in embryos from pho^- germline clones (BREEN and DUNCAN 1986 ), and in pho^- pharate adults (GIRTON and JEON 1994 ). In E(z) null mutants, the small imaginal discs are associated with a low mitotic index, mitotic chromosome breakage, and polyploidy (GATTI and BAKER 1989 ; PHILLIPS and SHEARN 1990 ), whereas Anti-E(z) antibody binds strongly to 40 discrete sites on polytene chromosomes and weakly all along the chromosome arms, showing that E(z) is required for maintenance of the overall chromosome structure (CARRINGTON and JONES 1996 ). By contrast, none of our data suggested that the overall chromosome structure depends on wild-type mxc product.

The small discs, the fused discs, and the AO staining of mxc larvae all show that mxc cells start to die when the maternally supplied mxc⁺ is used up. The random AO staining throughout the disc indicates that all cells are affected (Figure 6). This could provide an explanation for the increased dpp-like phenotypes of mxc^G43 dpp double mutants. Indeed, the phenotypes of flies that are homozygous for dpp disk-specific alleles are due to massive apoptotic cell death in the central disc regions during the third larval instar (BRYANT 1988 ). If an mxc mutation adds a second pattern of subviability or of cell death to this, then many more cells will likely die, especially around the center of the discs where cell death is already occurring, causing increased dpp-like mutant phenotypes. T