E(z): a polycomb group gene or a trithorax group gene?

Polycomb group (Pc-G) genes in Drosophila encode transacting factors that are responsible for preventing the transcription of homeotic selector (HOM-C) genes outside of their appropriate expression domains (Duncan and Lewis, 1982; Struhl and Akam, 1985; Paro, 1990). There are several criteria according to which genes are classified as members of the Polycomb group. Loss-of-function mutations in Pc-G genes produce phenotypes similar to those caused by loss of Polycomb function. In embryos, this results in the disruption of the anterior/posterior expression boundaries of HOM-C genes of the ANTP-C and BX-C leading to ectopic expression of abdA and Abd-B (Simon et al., 1992) and transformation of the thoracic and abdominal segments to the identity of the eighth abdominal segment. Mutations in Polycomb group genes enhance each other’s phenotype (Kennison and Tamkun, 1988). For example, adult male flies doubly or triply heterozygous for mutant alleles of Pc-G genes have enhanced penetrance and/or expressivity of the multiple sex combs phenotype. It has been estimated, based on this property, that there are as many as 40 members of the Pc-G (Jurgens, 1985). Loss-of-function mutations for Pc-G genes result in phenotypes reminiscent of gain-of-function mutations of certain homeotic selector genes of the ANTP-C and BX-C (Denell, 1978; Duncan and Lewis, 1982; Duncan, 1982; Struhl and Akam, 1985; Wedeen et al., 1986; Busturia and Morata, 1988; Simon et al., 1992). The products of two Pc-G genes, Polycomb and polyhomeotic colocalize to 100 bands on polytene chromosomes and are components of a multimeric complex containing at least twelve polypeptides (Franke et al., 1992; DeCamillis et al., 1992). In Schneider Line 2 cells, Polycomb protein is a component of the chromatin of the transcriptionally inactive loci of the bithorax complex, Ultrabithorax and abdominal A, but not the Abdominal B locus that is transcriptionally active (Orlando and Paro, 1993). These data have contributed to the idea that Pc-G genes encode proteins that regulate the expression of the HOMC genes through the manipulation and arrangement of local chromatin boundaries (Paro, 1993).

Enhancer of zeste (E(z)) has been classified as a Pc-G gene (Jones and Gelbart, 1990; Phillips and Shearn, 1990). A lossof-function allele of the E(z) gene was recovered in a screen for late larval/pupal recessive lethal mutations that cause imaginal disc abnormalities (Shearn et al., 1971). Larvae homozygous for that allele (originally called l(3)1902 and now called E(z)⁵) have a small disc phenotype (Shearn et al., 1978). In a screen for dominant modifiers of the zeste-white interaction two gain-of-function alleles were recovered (Kalisch and Rasmuson, 1974): E(z)¹, a neomorphic mutation that enhances the phenotype of z¹w^is and E(z)⁶⁰, formerly called either E(z)^S2 or Su(z)301, an antimorphic mutation that suppresses the z¹w^is phenotype (Kalisch and Rasmuson, 1974; Jones and Gelbart, 1990; Phillips and Shearn, 1990). Both E(z)¹ and E(z)⁶⁰ are homozygous lethal (Jones and Gelbart, 1990) and do not complement the lethality of E(z)⁵ (Phillips and Shearn, 1990). This served as the initial evidence that all three mutations are within the same gene. Evidence that initially classified E(z) as a member of the Pc-G was that male flies homozygous for hypomorphic E(z) alleles express a multiple sex combs phenotype (Shearn et al., 1978). The Pc-G classification was further supported by four additional observations. (1) Embryos from mutant mothers have anterior to posterior segment transformations (Jones and Gelbart, 1990; Phillips and Shearn, 1990) which are also characteristic of embryos with loss of Pc-G gene function. Such embryos have ectopic expression of abdA and AbdB (Simon et al., 1992). (2) There is ectopic expression of the homeotic selector genes Ultrabithorax and Sex combs reduced in the imaginal discs from larvae hemizygous for a temperature-sensitive allele, E(z)⁶¹ (Jones and Gelbart, 1990). (3) Chromosome binding of Pc-G proteins is reduced in E(z)⁶¹ larvae raised at a restrictive temperature (Rastelli et al., 1993) and (4) E(z) alleles and mutations in other Pc-G genes are suppressors of nanos (Pelegri and Lehmann, 1994).

In this paper we present data indicating that E(z) can also be classified as a member of the trithorax group (trx-G), a group of genes whose products antagonize the activity of the products of the Polycomb group. By contrast with Pc-G genes, trx-G genes are required to maintain the active transcription of HOM-C genes within their appropriate expression domains. Loss of function mutations in trx-G genes produce phenotypes similar to those caused by loss of trithorax function (Ingham and Whittle, 1980; Ingham, 1985). These phenotypes include homeotic transformations identical to those caused by loss-offunction mutations in segment-specific HOM-C genes (Duncan and Lewis, 1982; Ingham, 1985; Shearn et al., 1987; Kennison and Tamkun, 1988; Shearn, 1989; Mazo et al. 1990; Tamkun et al., 1992; Breen and Harte, 1993; Tripoulas et al., 1994). Although not required for the initiation of the expression of HOM-C genes, trx-G genes appear to be responsible for the maintenance of HOM-C gene expression (Breen and Harte, 1993, LaJeunesse and Shearn, 1995). Mutations in trx-G genes enhance each other’s phenotype. For example, adult flies which are doubly or triply heterozygous for mutant alleles of trx-G genes have enhanced penetrance and/or expressivity of characteristic HOM-C loss-of-function homeotic transformation phenotypes (Shearn, 1989). Recently the product of brahma, a trx-G gene, has been shown to be a component of a large multimeric complex (Dingwall et al., 1995) and as with Pc-G gene products, it has been speculated that trx-G gene products are required for the maintenance of local chromatin structure (Breen and Harte, 1993; Dingwall et al., 1995; LaJeunesse and Shearn, 1995).

An intriguing implication of the results presented in this paper is that at different developmental stages and in different tissues the product of E(z) can be involved in either the activation or repression of a given homeotic selector gene.

Flies were reared at 17°C, 20°C or 27°C on a basic Drosophila culture medium. The lethal mutant stocks were kept as heterozygotes over the balancer chromosomes TM1 or TM3. For a full description of these balancer chromosomes see Lindsley and Zimm (1992). All of the experiments were performed with mutant alleles of E(z) as hemizygotes with the deficiency, Df(3L)lxd^6. The mutant alleles used were: E(z)⁵ (AKA pco¹⁹⁰²), an amorphic allele; E(z)¹⁵ (AKA pco^NU808), a hypomorphic allele; E(z)¹² (AKA pco^MY939), a heat sensitive allele; E(z)⁶⁰ (AKA Su(z)301), an antimorphic allele; and E(z)¹, a neomorphic allele.

The hypomorphic mutant ash1 alleles used were ash1²⁸ and ash1²⁹ (Tripoulas et al., 1994). The ash1²⁹ allele is an inversion that causes replacement of the carboxy terminal 350 amino acids (out of 2144 total amino acids) with a novel sequence of 21 amino acids (Tripoulas et al., 1996). The molecular nature of the ash1²⁸ allele is not known. It is presumed to be a point mutation because its size on genomic Southern blots is like those of wild-type (unpublished observation). Phenotypically ash1²⁸ behaves like ash1²⁹. Adults were mounted in Faure’s medium; cuticular phenotypes were scored under a dissecting microscope.

For the temperature shift experiments, ten females of the genotype: yw⁶⁷; Df(3L)lxd⁶/y⁺TM3 were crossed to ten males of the genotype yw⁶⁷; E(z)¹²/y+ TM3. These parental flies were kept at 17°C (the permissive temperature) and transferred to a new vial once every 24 hours. After 10 days of transfer, all of these vials were incubated at 27°C (the restrictive temperature). Mutant third instar larvae, recognized by their yellow mutant mouthparts, were collected and placed into fresh vials at 27°C to determine their lethal period or examined for localization of HOM-C gene expression as described below.

Wandering third instar larvae were inverted in PBS (pH 7.0), fixed in 4% formaldehyde for 30 minutes, preincubated in a 0.3% hydrogen peroxide solution in methanol for 30 minutes, and incubated in a solution of 0.1% Triton X-100, 0.3% BSA, and 0.5% filtered horse serum in PBS for 30 minutes. Primary antibody was diluted in incubation mix: 1:50 anti-UBX (mouse monoclonal, F38.81; White and Wilcox, 1984; a gift from Rob White), 1:50 anti-ANTP (mouse monoclonal, ANTP8c11; a gift from Debbie Andrew), 1:500 anti-SCR (mouse monoclonal, 6H4; Glicksman and Brower, 1987; a gift from Danny Brower), or 1:500 anti-EN (rabbit polyclonal; a gift from Charles Girdham and Patrick O’Farrell). The samples were incubated in the primary antibody overnight at room temperature. The samples were washed six times, 30 minutes each time in PBT (0.3% BSA and 0.1% Triton X-100 in PBS) and then incubated in 1:200 diluted secondary antibody (Amersham anti-mouse or anti-rabbit/HRP for immunohistochemistry or Amersham anti-mouse or anti-rabbit/FITC or Texas Red for immunofluorescence). The secondary antibodies had been preabsorbed with several fixed third instar larvae overnight to reduce background. Samples were again washed six times for 30 minutes each in PBT. Samples prepared for immunofluorescence were counter stained using Hoechst DNA stain. Samples prepared for immunohistochemistry were treated with a horseradish peroxidase reaction (HRP) for 5-15 minutes: 0.5 μg/ml solution of 3,3′ diaminobenzidine (DAB) in PBT and 0.02% hydrogen peroxide. Samples were then washed several times in PBS and either mounted in Immunomount aqueous mounting medium or stored for up to 9 months at 4°C in PBS. Samples were photographed using a Zeiss Axioplan microscope. For immunofluorescence, the data was obtained as stacks of 15-20 images using a Photometrics CCD camera and ONCOR imaging software. The raw data stacks were deconvoluted using the FOCUS program. The deconvoluted images were assembled into composite figures in Adobe Photoshop 3.0.

Females of the genotype: y w⁶⁷; 35UZ; E(z)¹/y+ TM3 were mated to males; y w⁶⁷; 35UZ; Df(3L)lxd6/y+ TM3 or males: y w⁶⁷; 35UZ; +/+. 35UZ is a reporter transgene that has 35.4 kb of Ultrabithorax upstream regulatory region fused to the bacterial lacZ gene (Irvine et al., 1991). Hemizygous mutant third instar larvae were collected, inverted in PBS (pH 7.0), fixed for 10 minutes in a 4% formaldehyde solution in PBS, rinsed three times, washed once for 5 minutes in PBS, and placed for 1 hour in a 0.02% X-Gal solution that contained 31 mM ferrocyanide, 31 mM ferricyanide, 100 mM sodium phosphate (pH 7.2), 150 mM sodium chloride and 10 mM MgCl². Samples were refixed in 4% formaldehyde in PBS for 20 minutes to stop the reaction.

Double heterozygous combinations of recessive mutant alleles of different trx-G genes result in homeotic transformation mutant phenotypes (Shearn, 1989). Double heterozygous combinations of an E(z) amorphic allele, E(z)⁵, or a deletion that uncovers the E(z) locus, Df(3L)lxd⁶, with strong hypomorphic alleles of ash1, ash1²⁸ or ash1²⁹, result in homeotic transformations of the third leg to second leg and first leg to second leg (Table 1). Similar transformations are expressed in flies doubly heterozygous for the ash1²⁹ mutant allele and the original E(z) gain-of-function allele, E(z)¹.

Larvae hemizygous for an amorphic allele, E(z)⁵, express a small imaginal disc phenotype, with dorsal discs being more severely affected than ventral discs (Phillips and Shearn, 1990). The viability of the cells in the small imaginal discs from larvae hemizygous for E(z)⁵, was examined by uptake of acridine orange vital dye. The lack of dye uptake supports the argument that the small size of these discs is not due to cell death. Earlier transplantation experiments lead to the same conclusion. Although these small discs lack the ability to differentiate when transplanted into metamorphosing third instar larval hosts, they remain viable even after several rounds of transplantation into adult females (Shearn et al., 1978).

Localization of the products of the ANTP-C genes, Sex combs reduced (SCR) and Antennapedia (ANTP), of the BXC gene, Ultrabithorax (UBX), and the segmentation gene engrailed (EN) was examined in imaginal discs from larvae hemizygous for loss-of-function E(z) mutations. Imaginal discs from larvae hemizygous for the amorphic allele E(z)⁵ have no detectable SCR, ANTP, UBX, or EN (Fig. 1). Similar results were obtained for imaginal discs from larvae hemizygous for the hypomorphic allele E(z)¹⁵ (data not shown). The absence or reduced accumulation of these gene products is similar to what is observed in imaginal discs from larvae homozygous for mutations in the ash1 gene, a member of the trx-G (LaJeunesse and Shearn, 1995). One possible explanation for the observed lack of SCR, ANTP and UBX is that expression of these genes is repressed by ectopic Abdominal-B gene product (Lamka et al., 1992). This was shown not to be the case by the lack of any ectopic Abdominal-B gene product in the imaginal discs of E(z)⁵ hemizygous larvae (data not shown).

The complete loss of accumulation of HOM-C gene products in mutant imaginal discs conflicts with the results of Jones and Gelbart (1990) that larvae hemizygous for a heat sensitive allele, E(z)61, show wild-type endogenous localization as well as ectopic localization of UBX and SCR. The E(z)⁵ allele used in this study is an amorphic allele that causes a small disc phenotype (Phillips and Shearn, 1990); the E(z)⁶¹ allele used in the study of Jones and Gelbart (1990) is a relatively weak hypomorphic allele that does not cause a small disc phenotype. We hypothesized that the conflicting results were due to differences in the E(z) activity of these mutant alleles. To test this hypothesis we performed a temperature shift experiment using, E(z)¹², a different heat sensitive allele. The idea of these experiments was to remove E(z) activity at specific points during development and test for stage of lethality, growth of imaginal discs and CNS, and expression of ANTP, UBX, and EN in imaginal discs and in the CNS.

The heat sensitive allele, E(z)¹², proved to be ideal for such experiments. At a permissive temperature of 17°C, flies hemizygous for E(z)¹² survive until the pharate adult/adult stage, possess wild-type sized imaginal discs, and have a very weak E(z) phenotype. However, at a restrictive temperature of 27°C, the phenotype of larvae hemizygous for E(z)¹² is very similar to the amorphic phenotype of larvae hemizygous for E(z)⁵. Larvae hemizygous for E(z)¹² shifted to restrictive temperature prior to the beginning of the third larval instar have small discs and die just after puparium formation while larvae shifted to restrictive temperature after this stage have wild-type sized imaginal discs and die later in pupal development (Fig. 2).

EN is expressed in the posterior compartment of each imaginal disc and is required throughout development for proper differentiation of that compartment (Brower, 1986). The wing imaginal discs from larvae hemizygous for E(z)¹² grown at the permissive temperature or shifted to the restrictive temperature late in the third instar (3-4 days after the third instar molt) have a wild-type pattern of EN accumulation (Fig. 3A). Ectopic localization of EN in the anterior compartment of the wing imaginal can be observed in imaginal discs from E(z)¹² hemizygous larvae shifted to the restrictive temperature during the third larval instar (Fig. 3B-D). Larger regions of ectopic EN are observed with earlier temperature shifts, i.e., those closer to the third instar molt (Fig. 3B versus 3D). However, imaginal discs from larvae hemizygous for E(z)¹² shifted prior to the third instar molt have greatly reduced levels of EN accumulation (Fig. 3E-H); the remaining EN accumulation does not appear to be completely confined within the posterior compartment.

In wild-type wing imaginal discs ANTP is localized throughout the disc with the greatest concentration along the anterior margin and in the presumptive notal and pouch regions (Figs 1C, 4A); there is no accumulation of UBX in the wing imaginal disc proper (Fig 4E). A wild-type pattern of ANTP and UBX in wing imaginal discs is observed in larvae hemizygous for E(z)¹² reared at permissive temperature. Temperature shifts to the restrictive temperature during the third larval instar results in the accumulation of ectopic UBX in the posterior compartment of the wing imaginal disc that is strikingly similar to the results of Jones and Gelbart (1990). Furthermore, there is a concomitant loss of ANTP in these same regions. The earlier the shift to the restrictive temperature occurs during the third larval instar, the greater is the amount of ectopic UBX accumulation (Fig. 4F) and the greater the reduction of ANTP (Fig. 4B). The greatest accumulation of ectopic UBX was observed with a temperature shift the same day as the third instar molt (Fig. 4G). Since UBX can repress the transcription of Antennapedia (Hafen et al., 1984; Beachy et al., 1988), the loss of ANTP is probably a result of the ectopic accumulation of UBX. However, temperature shifts to the restrictive temperature prior to the third instar molt results in the complete loss of both ANTP and UBX in wing imaginal discs (Fig. 4D and 4H respectively).

In the CNS of wild-type larvae, ANTP is normally found in a set of three parallel bands of cells in the ventral ganglion that correspond to the three thoracic segments (Fig. 5A) and wildtype UBX accumulation is restricted to two bands of cells that correspond to parasegments 5 and 6 (Fig. 5E). Larvae hemizygous for E(z)¹² raised at the permissive temperature display wild-type distributions of ANTP and UBX in the CNS. Amorphic alleles of E(z) cause reduced proliferation of the CNS (Phillips and Shearn, 1990). The results of the temperature shifts described here suggest that different cells of the CNS require E(z) activity at discrete times during development.

Ectopic accumulation of ANTP in E(z)¹² hemizygous larvae shifted to the restrictive temperature during larval development can be found within the central region of the brain lobes and in regions more anterior and more posterior than the endogenous wild-type accumulation pattern in the ventral ganglion (Fig. 5B,C). The amount of ectopic expression decreases with earlier temperature shifts as does the size of the entire CNS (compare Fig. 5B and 5C with 5D). The accumulation of UBX in the CNS of larvae hemizygous for E(z)¹² shifted to restrictive temperature during the third larval instar, is essentially wild type (Fig. 5E). There is no ectopic accumulation of UBX in the brain lobe region or in the thoracic ventral ganglion anterior to the wild-type expression pattern (Fig. 5F). However, in mutant larvae shifted during or prior to the third instar molt, ectopic localization of UBX in the brain lobe region is observed (Fig. 5G,H).

Closer examination of the ectopic expression within the central nervous system revealed discrete patterns of ectopic ANTP and UBX accumulation (Fig. 6). The localization of ectopic ANTP within the brain lobes is within neuroblasts of the central brain (Fig. 6A, large arrow) and in those of the optic lobe (Fig. 6A, small arrow; Truman and Bate, 1988; Ito and Hotta, 1992). Larger numbers of cells within these distinct regions accumulate ectopic ANTP when larvae are shifted to restrictive temperature later in development (compare Fig. 6A with 6B and 6C). Since the ectopic ANTP appears to be restricted to some regions or types of cells, E(z) activity may be required in these cells in a non-redundant fashion for repressing Antp during the third larval instar and that some other activity or activities may be required in other cells within the central brain for repressing Antp.

Ectopic UBX is also expressed in a very restricted pattern during CNS development. In the CNS of E(z)¹² hemizygous larvae shifted to restrictive temperature during embryonic or early larval development, ectopic UBX can be observed in neuroblasts of the thoracic ganglion and central brain region in either single cells (Fig. 6F, small arrow) or pairs of cells. In the CNS of mutant larvae shifted to restrictive temperature a short time later in development (the late second instar, just prior to the third instar molt, Fig. 6E) the proportion of single neuroblasts expressing ectopic UBX is decreased and the proportion of pairs of cells expressing ectopic UBX is increased (Fig. 6E, large arrow). We presume that the pairs of cells are the daughters of the single cells that accumulated ectopic UBX in earlier shifts to restrictive temperature. Ectopic expression of UBX can also be seen in the presumptive optic lobe region of the central brain (Fig. 6E). In the CNS of mutant larvae shifted to restrictive temperature still later in development (just after the third instar molt; Fig. 6D) little or no ectopic UBX can be found within the thoracic region or in the brain lobes. Ectopic UBX is confined to a ring of cells defining the presumptive optic lobe region. In the brains from E(z)¹² hemizygous larvae shifted to restrictive temperature near the end of the third larval instar no ectopic localization of UBX can be found within the central brain or the ventral ganglion anterior to the endogenous expression (Fig. 5F).

The 35UZ reporter transgene contains 34.5kb of the upstream control region of the Ultrabithorax locus, also known as the BXD/PBX regulatory element, fused to the bacterial lacZ gene. In a wild-type genetic background, the 35UZ transgene leads to the accumulation of β-galactosidase in the posterior compartment of haltere and third leg imaginal discs (Irvine et al., 1993). However, expression of the 35UZ reporter transgene in E(z)¹ heterozygous larvae is derepressed in the first and second thoracic segments: 16% of E(z)¹ heterozygous larvae accumulate β-galactosidase ectopically in the posterior compartment of wing discs (Fig. 7A) and 43% of E(z)¹ heterozygous larvae accumulate β-galactosidase ectopically in the first and second leg discs in discrete spots (Fig. 7B,C). Examination of endogenous UBX localization in E(z)¹ hemizygotes revealed that in third leg and haltere imaginal discs there is no wild-type accumulation of UBX (data not shown). Nevertheless, 90% of the larvae hemizygous for E(z)¹ accumulate β-galactosidase ectopically in wing, first leg and second leg imaginal discs and in the lobes of the central brain (data not shown).

Although the E(z) gene has been classified as a member of the Pc-G (Jones and Gelbart, 1990; Phillips and Shearn, 1990), we have presented two lines of evidence that classify E(z) as a member of the trx-G. The first line of evidence is that double heterozygous combinations of recessive loss-of-function E(z) and ash1 alleles express homeotic transformation phenotypes similar to those expressed by double heterozygous combinations of recessive loss-of-function trithorax and ash1 alleles (Shearn, 1989). ash1 is a trx-G gene (Shearn et al., 1987; Shearn, 1989; Tripoulas et al., 1994) and is required for the proper differential expression of homeotic selector genes of the ANTP-C and BX-C (LaJeunesse and Shearn, 1995). The phenotypes observed in E(z)+/+ash1 double heterozygotes are those expected from partial loss of function of the homeotic selector genes Sex combs reduced and Ultrabithorax. A second line of evidence is that within thoracic imaginal discs of larvae hemizygous for amorphic mutant alleles of E(z) there is complete loss of accumulation of the homeotic selector gene products SCR, ANTP, and UBX and the segmentation gene product EN. Loss of accumulation of these gene products is also observed in imaginal discs of larvae mutant for loss-offunction alleles in the ash1 gene (LaJeunesse and Shearn, 1995). Since amorphic E(z) alleles cause phenotypes like those of amorphic trx-G mutations, our data might be interpreted to mean that E(z) is primarily a trx-G gene. To account for the Pc-G phenotype caused by some E(z) alleles it is necessary to postulate either that those alleles are gain-of-function mutations or that E(z) is also an activator of one or more PcG genes. Genetic analysis rules out the former postulate (Jones and Gelbart, 1990; Phillips and Shearn, 1990) but does not rule out the latter. So, the ectopic expression of HOM-C genes caused by those hypomorphic alleles, according to this interpretation, is not a direct consequence of the failure of the E(z) product to repress HOM-C genes but an indirect consequence of the failure to activate one or more Pc-G genes.

This interpretation of E(z) as primarily a member of the trx-G is both supported and contradicted by other evidence. E(z) encodes a nuclear protein with a SET domain (Tschiersch et al., 1994) that is also present in the products of two trx-G genes, trithorax (Jones and Gelbart, 1993; Stassen et al., 1995) and absent, small, or homeotic discs 1 (Tripoulas et al. 1996). This might lead one to argue that the SET domain is a signature of trx-G proteins, were it not for the fact that the SET domain is also present in the product of Suppressor of variegation-9 (Tschiersch et al., 1994). Where there is overlap, suppressors of position effect variegation are in Pc-G genes not in trx-G genes (Reuter and Spierer, 1992). Kuzin et al. (1994) showed that temperature-sensitive alleles of E(z) such as E(z)³² and E(z)⁶¹ cause decreased binding of TRX to polytene chromosomes of mutant larvae raised at a restrictive temperature. This is the result that might be expected if E(Z) and TRX were components of a trx-G multimeric complex and E(z) mutations disrupted the complex. However, one of those alleles, E(z)⁶¹, also causes reduced binding of the Pc-G proteins PSC and SU(Z)2 to polytene chromosome of mutant larvae raised at a restrictive temperature (Rastelli et al., 1993). Finally, Campbell et al. (1995) reported that increased E(z) dosage enhances the extra sex combs phenotype caused by heterozygosity for Pc¹⁶. While this is the result expected for a trx-G gene, they also observed that decreased E(z) dosage i.e. heterozygosity for a null allele, E(z)⁵, enhances the posterior transformation phenotype of embryos derived from esc⁻ mothers as expected for a mutation in a Pc-G gene.

We interpret these observations of others together with our evidence derived from temperature shift experiments with E(z)¹² to mean that the E(z) gene should be classified as a member of both the Pc-G and the trx-G. The E(z) gene product may function as a component of both a trx-G and a Pc-G complex or in the formation of these complexes or there may be multiple E(z) gene products some of which act as Pc-G proteins and some as trx-G proteins. Some precedents exists for the product of a single gene to act either as a transcriptional repressor or a transcriptional activator. As an example, the Krüppel gene product, a zinc finger protein, as a dimer acts as a transcriptional repressor, but as a monomer acts as a transcriptional activator (Sauer et al., 1995). As another example, low levels of Hunchback protein activate Krüppel expression while higher levels repress Krüppel expression (Schulz and Tautz, 1994). It has yet to be determined whether E(z) binds DNA or acts directly as a transcriptional repressor or activator molecule as do Krüppel and Hunchback, yet these examples show that many different factors can play a role in determining the activity of a protein. Perhaps, molecular interactions through the SET domain determines whether the product of E(z) acts as either a trx-G or Pc-G gene.

The requirement for E(z) activity as a Pc-G or a trx-G gene appears to change during development. For instance, E(z) activity is required to prevent ectopic UBX in the larval CNS only during the initial symmetric thoracic and optic neuroblast divisions. Loss of E(z) activity before these divisions results in ectopic UBX accumulation within the nascent larval neuroblasts and cessation of their proliferation. Loss of E(z) activity after these divisions results in no ectopic UBX localization. Similar phenomena are observed with ectopic EN localization within the anterior compartment of the wing imaginal disc and ectopic UBX within the wing imaginal disc. Loss of E(z) activity early during the third instar (just after a burst of cell proliferation) results in the largest areas of ectopic expression. In each case the E(z) mediated suppression of HOM-C gene expression appears to be linked to events within a developing tissue when patterns of gene expression and fate are perhaps becoming stabilized. The requirement for E(z) activity as a PcG or a trx-G gene also appears to be tissue specific. In regards to Ubx regulation it is especially clear that mutations in E(z) can cause ectopic expression in one tissue and reduced expression in another. We observed ectopic 35UZ expression in wing discs of E(z)¹ heterozygotes and hemizygotes and loss of UBX in haltere discs of E(z)¹ hemizygotes. This latter result could be explained as the loss of positive regulation of Ultrabithorax itself (Irvine, et al., 1993) or the gain-of-function of a negative regulatory activity.

Molecular studies in the future should clarify how E(z) can be a member of two groups of genes that have antagonistic functions.

We would like to thank Evelyn Hersperger for technical advice and comments about E(z) stocks and Amy Adamson for critical comments about the manuscript. We would also like to thank Nicholas Tripoulas and Richard Jones for thoughtful conversations about E(z). We also thank Deborah Andrew, Charles Girdham, Pat O’Farrell, Danny Brower and Rob White for providing antibodies and Sanjaya Jha for the 35UZ reporter construct. We are especially thankful to the Staff at ONCOR imaging for technical advice. This work was supported by a grant from NIH (GM53058) to A. S.