Ten different Polycomb group genes are required for spatial control of the abdA and AbdB homeotic products

Differentiation of segments along the anterior-posterior (A-P) axis in Drosophila is controlled by genes of the Antennapedia and bithorax complexes (Lewis, 1978; Kaufman et al., 1980). These gene complexes encode homeotic proteins that are expressed in precise domains along the A-P axis. The bithorax complex (BX-C) controls segment identity in the posterior half of the fly (Lewis, 1978; Bender et al., 1983; Karch et al., 1985). The BX-C executes this function with three homeotic proteins, Ubx, abdA and AbdB (Regulski et al., 1985; Sanchez-Herrero et al., 1985; Tiong et al., 1985), which are deployed differentially along the A–P axis. Ubx primarily controls the identities of the third thoracic and first abdominal segments and is expressed in parasegments 5 through 13 (White and Wilcox, 1985; Beachy et al., 1985), abdA controls the middle abdomen and is expressed in parasegments 7 through 13 (Karch et al., 1990; Macias et al., 1990) and AbdB controls the posterior abdomen and is expressed in parasegments 10 through 15 (Celniker et al., 1989; DeLorenzi and Bienz, 1990; Boulet et al., 1991). The restriction of Ubx, abdA and AbdB to their respective anterior boundaries is critical since development is perturbed in BX-C mutants that mis-express these products in more anterior locations (White and Akam, 1985; Karch et al., 1990; Celniker and Lewis, 1987; Celniker et al., 1990).

The genes of the BX-C are first transcribed in 2-hour blastoderm stage embryos (Akam and Martinez-Arias, 1985; Harding and Levine, 1988; Kuziora and McGinnis, 1988; Sanchez-Herrero and Akam, 1989). It is likely that the initial anterior expression boundaries are set by segmentation gene products such as hunchback and Kruppel (White and Lehmann, 1986; Harding and Levine, 1988; Irish et al., 1989; Reinitz and Levine, 1990; Qian et al., 1991). However, the early expression patterns of segmentation products decay by about 4 hours of embryogenesis. In contrast, the BX-C products are expressed continuously, within their proper boundaries, throughout subsequent larval and pupal development over a period of about 10 days (White and Wilcox, 1985; Brower, 1987). This expression within AP boundaries is required continually for proper segmental development (Lewis, 1964; Morata et al., 1983). Thus, a mechanism must exist to maintain the expression boundaries after they are set in early embryos.

The product of the Poly comb (Pc) gene (P.H. Lewis, 1947; E.B. Lewis, 1978; Duncan and Lewis, 1982) is a likely component of this maintenance machinery. Pc product is required in embryos and during postembryonic stages for proper control of the BX-C products (Struhl, 1981; Duncan and Lewis, 1982; Busturia and Morata, 1988). Embryos that lack zygotic Pc product die with all segments transformed towards the eighth abdominal segment (A8). The gene is named for the dominant adult phenotype, segmental transformation of second and third thoracic legs into first thoracic leg, causing duplications of the male sex comb (Duncan and Lewis, 1982). These phenotypes result from ectopic expression of homeotic products outside of their normal A-P boundaries. In particular, Pc mutant embryos misexpress the Antennapedia (Antp) and Sex combs reduced (Scr) genes of the Antennapedia complex (ANT-C) as well as Ubx, abdA and AbdB (Beachy et al., 1985; Wedeen et al., 1986; Carroll et al., 1986; Riley et al., 1987; Kuziora and McGinnis, 1988; Celniker et al., 1989). The recessive embryonic phenotype results primarily from the mis-expression of AbdB in anterior parasegments (Celniker et al., 1989). The dominant adult leg transformation is thought to result from posterior mis-expression of Scr (Glicksman and Brower, 1988). Molecular studies have shown that Pc protein binds to the ANT-C and BX-C loci in polytene chromosomes (Zink and Paro, 1989) although it fails to bind directly to DNA in vitro (Zink et al., 1991).

Another likely component of the maintenance machinery is the extra sex combs (esc) gene (Struhl, 1981, 1983). Embryos lacking both maternal and zygotic esc product show segmental transformations similar to those in Pc mutants. Animals lacking only zygotic esc product develop into adults bearing leg transformations like in Pc/+ adults (Struhl, 1981). esc mutants ectopically express Ubx and Antp in embryos as well as Ubx, Antp and Scr in imaginai discs (Struhl and White, 1985; Carroll et al., 1986; Glicksman and Brower, 1988, 1990). Careful analysis of the timing of Ubx misexpression in esc null embryos has shown that esc is required to maintain, but not to set up, the anterior Ubx expression boundary (Struhl and Akam, 1985).

A number of other mutations have been described that cause phenotypes similar to those in Pc and esc mutants (Gehring, 1970; Duncan, 1982; Ingham, 1984; Jiirgens, 1985; Breen and Duncan, 1986; Dura et al., 1987; Adler et al., 1989; Jones and Gelbart, 1990; Phillips and Shearn, 1990). The genes defined by these mutations have become known collectively as the Pc group (reviewed in Paro, 1990). It is generally assumed that these Pc group genes are trans-regulators of the ANT-C and BX-C genes and that, like Pc and esc, they maintain homeotic boundaries by repressing expression in inappropriate A–P axis positions. These assumptions were proved correct for the polyhomeotic and Enhancer of zeste loci, which are both required for proper boundaries of Scr and Ubx (Dura and Ingham, 1988; Jones and Gelbart, 1990). However, the extradenticle gene (Wieschaus et al., 1984), which could be classified in the Pc group on the basis of phenotype, is not required for the proper distribution of homeotic products (Peifer and Wieschaus, 1990).

A remarkable feature of at least some Pc group products is that they are required simultaneously to repress different homeotic products at multiple positions along the A–P axis. For example, Pc is required to maintain anterior boundaries of Antp in parasegment (PS) 3, Ubx in PS5, abdA in PS7 and AbdB in PS10. It is difficult to imagine Pc acting as a simple transcriptional repressor in many different parasegments at the same time, especially since Pc protein itself appears uniformly distributed along the A–P axis (Paro and Hogness, 1991). However, it is possible that Pc is a ubiquitously required subunit of the repression machinery and some of the other Pc group products are more specifically required for the regulation of certain homeotic products. For example, esc mutants show abundant ectopic expression of Antp and Ubx but Scr is only subtly mis-expressed (Glicksman and Brower, 1990).

To characterize further the roles of Pc group products in regulating multiple homeotic products, we have examined the distributions of abdA and AbdB proteins in eleven of the Pc group mutants. We report that both abdA and AbdB are ectopically expressed in embryos mutant for ten different Pc group genes. We also describe the patterning and kinetics of accumulation of this ectopic expression.

esc¹⁰ is a deficiency for the esc locus (Struhl, 1981, 1983). esc²and esc⁶ are apparent null alleles, based upon severity of embryonic phenotypes (Struhl, 1981, 1983). Batches of esc null embryos were collected as the progeny of esc¹⁰(esc² parents, esc null embryos were also collected as the progeny of esc⁶/esc⁶ parents. esc⁺/esc⁻ paternally rescued embryos were generated by crossing esc¹⁰/esc² females to ry⁵⁰² males.

E(z)^S2 is a temperature-sensitive allele that is null or nearly null for its homeotic function at the restrictive temperature, 29°C (Jones and Gelbart, 1990). Segmental transformations in embryos from hemizygous versus homozygous E(z)^S2 mothers are indistinguishable (R. Jones, personal communication). Batches of homozygous mutant E(z)^s2 embryos were collected at 29°C from a homozygous stock grown at 18°C. E(Z)^S2/+ paternally rescued embryos were collected at 29°C as the progeny of E(z)^s2 homozygous mutant females and ry⁵⁰² males.

Homozygous mutant embryos containing other Pc group alleles were collected as the progeny of heterozygous parents. The homozygous mutants constituted one-quarter of the embryos in mixed batches. Pc²²¹³⁰⁹ is a null allele that fails to make Pc protein (R. Paro, personal communication). Pc³ is an antimorphic allele that produces a phenotype stronger than the null (Lewis, 1978; Duncan and Lewis, 1982; Haynie, 1983). abdA and AbdB mis-expression in embryos mutant for either Pc allele were similar, ph⁵⁰³ is an apparent null allele affecting both coding units at the locus and Df(1)JA52 is a deficency for the ph locus (Dura et al., 1987; Perrimon et al., 1985). Mis-expression was similar in both ph alleles. Scm^H1 is an apparent null allele (Breen and Duncan, 1986). Scm^HI is a previously unreported Scm allele that is lethal over Scm null alleles (R. Jones, personal communication). Mis-expression was similar in both Scm alleles. Pcl^D5 is an apparent null allele (Breen and Duncan, 1986). Sce^D1 is the single allele for this locus (Breen and Duncan, 1986). It is a first instar larval lethal but its severity is not known. Asx^D1 is a hypomorphic allele (Breen and Duncan, 1986). Asx^XF23 behaves genetically as a null allele (G. Jürgens, personal communication). Df(2R)trix is a deficiency for the Asx locus (Breen and Duncan, 1986). Df(2R)trix and Asx^XF23 mutant embryos showed similar levels of mis-expression that were stronger than those seen in Asx⁰¹ embryos. The preferential mis-expression in the epidermis versus the CNS was seen with all three alleles. Psc^14,45 and Psc^11N48 are strong hypomorphs with respect to homeotic function (Jürgens, 1985; Adler et al., 1989; Lasko and Pardue, 1988; P. Adler, personal communication). Df(2R)vg^D is a deficiency that removes most of the Psc transcription unit (Brunk et al., 1991; Lasko and Pardue, 1988). Mis-expression in Psc^14,45 and Psc^IIN48 embryos was similar with slightly stronger mis-expression seen in vg^D embryos. pho^b (previously l(4)29^b) is an apparent null allele and Df(4)G is a deficiency for the pho locus (Breen and Duncan, 1986; Hochman et al., 1964). Similar subtle mis-expression was seen with both alleles, sxc¹ is an apparent null allele (Ingham, 1984). The Pc group mutants analysed are listed in Table 1.

Embryos were fixed and stained as described (Simon et al., 1990) using a polyclonal antibody against abdA (Karch et al., 1990) or a monoclonal antibody against AbdB (Celniker et al.,1989). Stained embryos were dissected and mounted as described (Simon et al., 1990), except for the embryo in Fig. 7A (See figure legend). Briefly, in germ band-extended (6h) embryos, the attachment of the posterior end of the germ band to the dorsal part of the head was cut and the germ band was flipped out to display the ectodermal surface in two dimensions. Later stage embryos were slit along the dorsal midline, the gut and visceral mesoderm were excised and the remaining tissues were flattened in two dimensions.

Embryos mutant for Polycomb group alleles were stained with antibodies to abdA (Karch et al., 1990) or AbdB (Celniker et al., 1989). In most cases, two or three mutant alleles for each gene were analyzed, including null alleles or strong hypomorphs (See Methods). The mutants are described in the order of most extreme to least extreme mis-expression of these two BX-C products. In general, esc and E(z) mutant embryos showed more extreme mis-expression of abdA and AbdB than other Pc group mutants. This is partly because many of the Pc group genes are expressed maternally (Denell, 1982; Haynie, 1983; Ingham, 1984; Breen and Duncan, 1986; Dura et al., 1988) and, except for esc and E(z), we have assayed embryos lacking only the zygotic contributions.

The wild-type expression patterns of abdA and AbdB have been described (Karch et al., 1990; Macias et al., 1990; Celniker et al., 1989; DeLorenzi and Bienz, 1990; Boulet et al., 1991). Examples of the wild-type patterns are shown here in Figs 1 and 2 for comparative purposes. abdA protein is first seen at about 4 hours of embryogenesis with a precise anterior boundary at the anterior edge of parasegment (PS) 7. This anterior boundary (Fig. 1A, B) persists throughout the rest of embryogenesis. AbdB protein is first detected at about 4 hours with an anterior boundary in PS13 which persists until about 6 hours in germ band-extended embryos (Fig. 2A). As embryogenesis proceeds, additional AbdB becomes detectable in parasegments 10 through 12 (Fig. 2B) with the final anterior boundary in PS10.

Embryos lacking both maternal and zygotic esc product were generated using apparent null alleles (Struhl, 1981, 1983). When abdA protein is first detected in these esc⁻ embryos the pattern appears normal (Fig. 3A), but between 5 and 6 hours ectopic expression is seen just anterior to the normal boundary in PS5 and PS6 (Figs IC, 3B). As embryogenesis proceeds, ectopic abdA spreads progressively into more anterior parasegments. Embryos between 6 and 7 hours often show intermediate spread, with abundant expression in PS5 and PS6 and weaker expression now detectable in PS3 and PS4 (Fig. 3C). Although there is some variability in the amount of ectopic abdA at this stage, the graded expression from posterior to anterior is reproducibly seen. Eventually, expression extends further anteriorly and into the head region (Fig. 3D) and by 9 hours the entire A–P axis shows strong accumulation of abdA (Fig. ID).

Although ectopic abdA eventually appears throughout the A–P axis, its cell-specific distribution is not uniform. Instead, ectopic abdA appears in the epidermis in a repeating pattern, high at the anterior margin and low at the posterior margin within each parasegment (Figs 1C, D, 3B–D). This cell-specific patterning resembles the wild-type distribution of abdA in parasegments 7–13 (Fig. 1A, B).

In addition to the anterior spread of abdA in esc⁻ embryos, there is also posterior spread. Normally the posterior boundary of abdA expression is in PS 13 (Karch et al. 1990). In 6 hour esc⁻ embryos, abdA accumulates in epidermal cells in PS14. By about 8 hours ectopic abdA is visible in PS14 and PS15, and weakly in the hindgut rudiment (Fig. 4A). By 9 hours, the expression in the hindgut is more intense and it includes the lateral processes of the developing Malpighian tubules (Fig. 4B).

Like abdA, ectopic AbdB does not appear at the same time in all parasegments. AbdB appears normal in esc~ embryos between 4 and 5 hours (Fig. 3E) but, during the next hour, ectopic AbdB is activated as far forward as PS3 (Fig. 2C, arrow in 3F). The earliest ectopic expression is pair-rule modulated with stronger expression in the odd-numbered parasegments. Additional AbdB then accumulates, primarily in posterior parasegments, so that by 6 hours graded expression is seen along the A-P axis and the pair-rule distribution is still visible (Fig. 3G). By 7 hours, most embryos have fairly uniform AbdB expression up to PS3, with less expression anterior to this boundary (Fig. 3H). Finally, between 7 and 9 hours, the most anterior region of the embryo accumulates AbdB, with 9 hour embryos showing a uniform distribution in virtually all epidermal cells (Fig. 2D). This uniform epidermal distribution is similar to the pattern of AbdB in PS13 of wild-type embryos (Fig. 2B).

In esc null embryos, all the segments are transformed into copies of the normal eighth abdominal segment (posterior PS13 plus anterior PS14) (Struhl, 1981, 1983). This is explained by the uniform distribution of AbdB protein (Fig. 2D) since AbdB predominates when it appears in the same parasegment with other homeotic products (Struhl and White, 1985; Busturia and Morata, 1988). If eggs derived from esc⁺/esc⁻ mothers are fertilized by sperm containing one copy of esc⁺, then less severe transformations are seen (Struhl, 1981). Such esc⁺/esc⁻ embryos show complete or partial transformation of thoracic and abdominal segments to eighth abdominal, but the head segments are unaffected or only partially transformed. Fig. 4D shows that AbdB expression in such partially paternally rescued embryos is quite uniform in much of the abdomen but much reduced relative to esc null embryos in the anterior thorax (PS3 and 4) and in the head (compare to Fig. 2D). The head expression appears spotty and random from embryo to embryo. The final AbdB distribution along the A-P axis in these esc⁺/esc⁻ embryos now resembles that in the 6–7 hour esc null embryos (Fig. 3G, H). Thus, the differential severity of the phenotype along the A–P axis correlates with the differential severity of AbdB mis-expression along that axis.

A similar partial spread of abdA occurs in these esc⁺/esc⁻ embryos. At 9 hours (Fig. 4C) and at subsequent stages, there is strong abdA expression in much of the thorax and abdomen but the anterior thorax and head region show weaker expression in isolated groups of cells. This contrasts to the strong abdA expression along the entire A-P axis in esc null embryos (Fig. ID). In fact, the final abdA pattern has a marked PS5 anterior boundary which mirrors the transient PS5 boundary seen at earlier times in esc null embryos (Fig. 3B). It seems that a partial amount of esc⁺ product, or its delayed appearance, causes a ‘freeze’ in the gradual anterior spread of abdA and AbdB in the intermediate stage.

Mutations that reduce or eliminate zygotic product of the Enhancer of zeste gene (Jones and Gelbart, 1990), also called polycombeotic (pco) (Phillips and Shearn, 1990), cause larval-to-pupal lethality. However, temperature-sensitive alleles have been used to severely reduce both maternal and zygotic E(z)⁺ product, resulting in embryos that die with extreme transformation of all segments towards A8 (Jones and Gelbart, 1990; Phillips and Shearn, 1990).

Homozygous mutant embryos containing the temperature-sensitive allele E(z)^S2 were collected at the restrictive temperature, 29°C, and examined for abdA and AbdB distributions. Although not null for other functions of E(z), the E(z)^S2 allele is null or nearly null with regard to homeotic function (Jones and Gelbart, 1990; R. Jones, personal communication, see Materials and methods). As in esc mutants, ectopic expression of abdA spreads forward gradually. At about 6 hours, ectopic abdA is most abundant in PS5 and PS6 (Fig. 1E). By 9 hours, abdA is expressed throughout the A–P axis including the head (Fig. 1F). In this same period of time, abdA also spreads posteriorly into PS14, PS15 and into the hindgut primordium. At all times, ectopic abdA is patterned within each parasegment with higher expression at the anterior margins.

The ectopic AbdB patterns in E(z) mutant embryos also resemble those seen in esc mutants. At 6 hours, AbdB is activated as far forward as PS3 in a pair-rule pattern (Fig. 2E). By 9 hours, AbdB is expressed throughout the A–P axis in all or nearly all epidermal cells (Fig. 2F).

If E(Z)^S2/+ embryos are collected at 29°C from homozygous mutant mothers and wild-type fathers, the embryos are still extremely transformed, but there is some phenotypic rescue, primarily in the head (Jones and Gelbart, 1990). In agreement with this, both abdA and AbdB still spread throughout the A-P axis in paternally rescued E(z)^S2/+ embryos, but the accumulation of AbdB in the head region now appears patchy (not shown).

Expression was examined in homozygous Pc embryos derived from heterozygous parents. Ectopic abdA is first seen between 5 and 6 hours of development in lateral epidermal patches in PS6. During the next 3 hours, abdA spreads into the more anterior parasegments. By 9–10 hours, expression is seen throughout the A–P axis and this pattern persists through 12 hours (Fig. 5B).

The earliest ectopic AbdB is seen in the pair-rule pattern described for esc and E(z) mutant embryos. Subsequently, the further accumulation of AbdB parallels that seen with these two other mutants except that the timing of the anterior spread is delayed by 2–3 hours, and AbdB fails to accumulate to high levels in all epidermal cells. In particular, patchy AbdB is seen in the head region in 9–12 hour embryos.

The patterns of ectopic abdA and AbdB in Pc mutants resemble those described for esc and E(z) with slightly reduced severity. Flowever, this abundant level of mis-expression occurs with removal of zygotic Pc product alone. This is consistent with phenotypic analyses showing that, although Pc is expressed maternally (Denell, 1982; Haynie, 1983) the loss of the zygotic contribution alone is sufficient to cause a severe transformation of most segments towards A8 (Lewis, 1978; Duncan and Lewis, 1982).

Ectopic abdA in ph mutant embryos is first seen at about 6 hours primarily in medial positions in parasegments 3 through 6. By 9–10 hours, abdA is present throughout the A–P axis (Fig. 5C). Ectopic expression is diffcult to assay beyond this time since extensive cell death occurs in the ventral epidermis (Dura et al., 1987).

AbdB mis-expression occurs at about 6 hours in a graded pattern with most ectopic expression in the abdomen, less in the thorax and none in the head. Pairrule modulation is not seen. By 9 hours, AbdB accumulates to high levels throughout the A-P axis including the head. AbdB appears at high levels in most epidermal cells although it fails to achieve the blanket uniform appearance seen in esc and E(z) mutants.

Mis-expression in homozygous mutant Scm embryos is less severe than in the above mutants. Ectopic abdA is first seen in lateral epidermal patches in PS6 (Fig. 1G). Subsequently, abdA is seen in more medial positions in parasegments 3 through 6, and by 9 hours expression has extended into the head (Fig. 1H). Although expression occurs throughout the A–P axis, the cellular distribution within each parasegment is much more limited than in esc or E(z) mutants (compare to Fig. ID, F). In the epidermis, ectopic abdA is preferentially expressed along both sides of the segmental grooves, marking the anterior portion of each parasegment. At later stages, about the same distribution along the A–P axis is seen, with abundant mis-expression in the CNS in a mottled pattern (Fig. 5D).

Ectopic AbdB is first detected at about 6 hours, extending a variable number of parasegments forward, but not anterior to PS3 (Fig. 2G). No pair-rule effect is seen. At these early times, there are more cells expressing AbdB ectopically in posterior abdominal segments than in thoracic segments, and AbdB is absent from the head. By 9 hours, AbdB spreads into the head in a spotty pattern, and graded expression along the AP axis is still seen (Fig. 2H). At 12 hours, expression appears more uniform, with particularly strong misexpression in the CNS (Fig. 6A).

Patterns of ectopic abdA and AbdB in Pcl mutant embryos are nearly identical to the patterns in Scm mutant embryos at all stages. This comparison is illustrated in Fig. 5E and 5D for abdA in 12 hour embryos. Consistent with this observation, the phenotypes in embryos lacking zygotic expression of Pci or Scm are very similar (Jürgens, 1985; Breen and Duncan, 1986).

Patterns of ectopic abdA and AbdB in See mutant embryos are similar to those in Scm and Pci embryos at all stages. Ectopic expression of both products first occurs at about 6 hours and at later stages it predominates in the CNS (Fig. 5F).

Ectopic abdA is first detected at about 6 hours in PS5 and PS6. Subsequently, abdA spreads as far forward as PS2 but little or none accumulates in the head. At 9 hours, ectopic abdA remains primarily in PS2 through PS6 in the epidermal cells that border the segmental grooves. Essentially the same distribution is seen at 12 hours, with widely scattered, mis-expressing cells seen also in the head (Fig. 5G). In contrast to most other Pc group mutants, ectopic abdA in Asx mutants is much more abundant in the epidermis than in the CNS.

The early stages of AbdB mis-expression resembles that in Scm mutants, with ectopic AbdB stronger in posterior parasegments. By 9–12 hours, this graded expression is still observed, but it is strikingly reduced in the CNS relative to that in Scm (compare Fig. 6B and 6A) and most other Pc group mutants.

The earliest ectopic abdA is seen at about 6 hours in lateral patches in PS6. By 9 hours, abdA spreads further forward, primarily in medial cells in PS2 through PS6, and spotty expression is seen in the head. Curiously, between 9 and 12 hours, ectopic abdA becomes weaker, with scattered expressing cells primarily in the brain lobes and in PS5 and PS6 in the CNS (Fig. 5H).

As in other Pc group mutants, the earliest ectopic AbdB spreads as far forward as PS3 but is largely absent from the head. By 9 hours, AbdB is detected throughout the A-P axis, although the number of misexpressing cells per parasegment is less than in most other Pc group mutants. Between 9 and 12 hours, ectopic AbdB becomes concentrated in the CNS.

pho [previously called 1(4)29] mutants die as pharate adults with segmental transformations (Gehring, 1970; Duncan, 1982). pho mutant embryos from heterozygous parents lack cuticle defects or transformations. However, the gene product clearly functions in embryos, since elimination of both maternal and zygotic product causes embryonic lethality and segmental transformations (Breen and Duncan, 1986).

pho mutant embryos from heterozygous parents show subtle defects in the control of abdA and AbdB. Ectopic abdA is detected as early as 6 hours in lateral patches in PS6, as is seen in other Pc group mutants (i.e. Scm, Fig. 1G), but the number of mis-expressing cells is much lower, usually only 5–10 cells in the entire parasegment. At later embryonic times, this very sparse mis-expression is still seen, primarily in the CNS in PS6 (Fig. 51, arrow), but occasionally in more anterior positions.

Ectopic AbdB is similarly sparse. At 6 hours, up to 10 mis-expressing cells per embryo are seen, primarily in the abdominal segments. At later stages up to 20 misexpressing cells per embryo are seen, again mostly in the abdomen.

Like pho mutants, sxc mutants die as pharate adults and survive to this stage due to maternally supplied product (Ingham, 1984). In contrast to pho, we did not observe ectopic abdA or AbdB in homozygous mutant sxc embryos from heterozygous parents.

The segmental transformation phenotypes seen in cuticles of Pc group mutant embryos result from misexpression of homeotic proteins in the epidermis. Misexpression in a number of Pc group mutants in another ectodermal tissue, the CNS, has been reported (Struhl and White, 1985; Wedeen et al., 1986; Dura and Ingham, 1988; Celniker et al., 1989; Jones and Gelbart, 1990) and we have extended that description here (Figs 5, 6).

Our analysis of dissected embryos shows that Pc group products also function in a number of other internal tissues, including mesodermal derivatives. Fig. 7A shows ectopic expression in the dorsal vessel or heart tube, which is a mesodermal tissue (Campos-Ortega and Hartenstein, 1985). abdA is normally expressed in the heart tube and pericardial cells with an anterior boundary in the fifth abdominal segment (PS 10) (Karch et al., 1990). In the Scm mutant embryo shown, abdA spreads far forward of the normal boundary (arrow, Fig. 7A) into the thoracic segments. Fig. 7B shows anterior spread of AbdB in the visceral mesoderm that surrounds the gut. Normally, the AbdB anterior boundary in the visceral mesoderm is in PS11 (DeLorenzi and Bienz, 1990; arrows in Fig. 7B). In the Asx mutant shown, AbdB spreads at least as forward as PS3. Ectopic expression of AbdB in the visceral mesoderm was also seen in esc, Pc, ph, Scm and Pcl mutant embryos.

The Pc group genes esc, E(z), Pc, Pci, ph, Scm, See, Asx, Psc and pho are each required to confine abdA and AbdB to their proper domains along the A-P axis. The amounts of ectopic expression varied widely among the different Pc group mutants (Fig. 5). Much of this variation is due to differences in the maternal expression levels of the Pc group genes (Breen and Duncan, 1986). Most of the mutant embryos examined were null or strong hypomorphs for zygotic product but maternal product was usually unaffected. Only in the cases of esc and E(z) mutants were the maternal and zygotic products both eliminated or severely reduced. Correspondingly, abdA and AbdB mis-expression was most severe in these esc⁻ and E(z)⁻ embryos, and in Pc mutant embryos, where the maternal contribution is small (Haynie, 1983). At the other extreme, the strong maternal components of sxc (Ingham, 1984) and pho expression (Breen and Duncan, 1986) are the likely reasons that abdA and AbdB appeared normal in sxc mutant embryos and were only subtly mis-expressed in pho mutant embryos (Fig. 5I). Embryos lacking this maternal sxc show segmental transformations towards eighth abdominal (Ingham, 1984), strongly suggesting that at least AbdB is under sxc control.

In the light of the large number of Pc group genes, it is important to determine if some products are preferentially involved in regulating certain homeotic genes but not others. In our analysis, there was little evidence for this since the general extent of abdA misexpression and AbdB mis-expression in particular Pc group mutants was similar. The relative roles of Pc group genes may also be investigated by comparing the precise tissue patterns of mis-expression in the different mutants. As mentioned above, the interpretation of these patterns is complicated by the perdurance of maternal product, except in the cases of esc and E(z). Figs 1 and 2 show that the timing and patterns of abdA and AbdB mis-expression in esc⁻ and E(z)⁻ embryos are nearly identical. This suggests that esc and E(z) perform similar molecular functions, perhaps as components of the same machinery. The tissue distributions in the other Pc group mutants are also consistent with common function, with the notable exception of Asx. Whereas most Pc group mutants cause abundant misexpression of abdA and AbdB in the CNS, Asx mutants show very little mis-expression in the CNS (Figs 5G, 6B). Thus, Asx may function primarily in the epidermis as opposed to the CNS. Alternatively, there could be differential decay of maternal Asx product in these two tissues.

Our analysis of abdA and AbdB shows that the Pc group products act as repressors in anterior parasegments. Some Pc group products may also act as positive regulators since Scr, Antp and Ubx appear repressed, rather than ectopically activated, in certain tissues in some Pc group mutants (Smouse et al., 1988; McKeon and Brock, 1991). Alternatively, these cases may be indirect consequences of repressive interactions among the multiply mis-expressed homeotic products. For example, abdA and AbdB normally act as repressors of Ubx in embryos (Struhl and White, 1985). Thus, a decrease in Ubx levels in a Pc group mutant might result from repression by ectopic abdA and AbdB, as has been shown in 12-14 hour esc mutant embryos (Struhl and White, 1985). Trans-repression of abdA by AbdB (Karch et al., 1990) may also explain why abdA fades in esc and E(z) mutants by 12 hours (not shown) and becomes more limited in Psc mutants by this time (Fig. 6H). We have attempted to minimize the complications of these secondary regulatory events in two ways. First, since patterns in late embryonic stages are more likely to be affected by the accumulation over time of other homeotic products, we have concentrated on the early patterns of ectopic expression in 5–7 hour embryos. Second, we have examined the distribution of AbdB, which is not known to be trans-repressed by any other homeotic product.

Initially, the on and off states of homeotic gene expression along the A-P axis are set by the combinatorial action of gap and pair-rule products (Duncan, 1986; Ingham and Martinez-Arias, 1986; White and Lehmann, 1986; Irish et al., 1989; Simon et al., 1990; Reinitz and Levine, 1990; Qian et al., 1991). For example, hunchback could set the initial abdA anterior boundary in PS7 by repressing abdA in more anterior parasegments. By 3–4 hours, the gap gene products decay (Gaul et al., 1987; Tautz, 1988) and this repression is then likely maintained by the Pc group products. Indeed, analysis of esc null embryos shows that esc is required for maintenance of Ubx expression within its proper A–P boundaries, but not for its initial activation (Struhl and Akam, 1985). Similar experiments show that E(z) is also required for maintenance but not initiation of Ubx (Jones and Gelbart, 1990). Likewise, we find that esc and E(z) are required for maintenance but not initiation of abdA and AbdB since the distribution of these two products is initially normal in esc⁻ and E(z)⁻ embryos (Fig. 3A, E). The possible role of other Pc group products in initiation has yet to be addressed. Here we show that they are each at least involved in maintenance of abdA and AbdB during embryogenesis. A recent report also implicates many of these Pc group genes in the maintenance of Ubx, Antp and Scr in embryos (McKeon and Brock, 1991).

These data, taken together, indicate that many of the Pc group products act at the same time at many positions along the A-P axis. Thus, esc, Pc, Pcl, Scm, Sce, Asx and ph are each necessary to repress simultaneously Antp, Ubx, abdA and AbdB anterior to their boundaries in PS3, PS5, PS7 and PS10, respectively. Clearly, the Pc group products are not acting as simple repressors with limited distributions, shutting off expression wherever they happen to be located. Instead, mechanisms for transcriptional repression must be considered that account for global function of Pc group products along the A–P axis.

These requirements are satisfied by envisioning that the Pc group products act by packaging portions of homeotic loci into an inaccessible or ‘closed’ configuration (Paro, 1990; Peifer et al., 1987). This model suggests that the Pc group products sense the initial on or off state of a homeotic gene and then compact the DNA into a stably repressed, heterochromatin-like structure in cells where the DNA was originally inactive (Paro, 1990). At later times, a large number of cellspecific activators direct the intricate patterns of homeotic expression, but this would occur only in parasegments where the factors could gain access to the DNA (Peifer et al., 1987). Thus, abdA remains stably off anterior to PS7 because its DNA regulatory regions would be inaccessible to positive factors in these anterior parasegments. In this way, the state of the chromatin would be fixed to ‘remember’ the initial positional information provided by gap and pair-rule products.

Our analysis of the kinetics of mis-expression unexpectedly revealed that ectopic abdA and AbdB do not appear simultaneously in all parasegments. Instead, in the strongest Pc group mutants, (esc, E(z) and Pc), abdA is first mis-expressed just anterior to the normal PS7 boundary in PS5 and 6 (Figs 1C and E, 3B). The more anterior thoracic and head regions accumulate abdA only after further time in embryogenesis (Figs ID, F, 3C, D, 5B). Similarly, ectopic AbdB arises in a nonuniform pattern. At first, this involves expression in a pair-rule pattern, as far forward as PS3 (Figs 2C, E, 3F). Subsequently, the pair-rule pattern fills in and AbdB expression eventually spreads forward into the head (Figs 2D, F, 3H). Similar gradual anterior activation of AbdB transcripts in Pc³ mutant embryos has been reported (Kuziora and McGinnis, 1988).

These early ectopic patterns can be explained in the context of the chromatin accessibility model for Pc group function. In wild-type embryos, AbdB expression in PS 13 in 2 hour embryos could involve positive control by the pair-rule product even-skipped (eve) and repression by gap gene products, including giant (Reinitz and Levine, 1990), in parasegments anterior to PS13. By 3–4 hours, the gap gene products decay and the task of repression in anterior parasegments is transferred to the Pc group products. However, in Pc group mutants this transition fails and in 4 hour embryos AbdB can be ectopically activated by any positive factor (such as eve) present in anterior locations and capable of binding to the AbdB regulatory regions. In fact, the earliest ectopic AbdB pattern at 5–6 hours (Figs 2C, E, 3F) is remarkably similar to the eve protein pattern at slightly earlier times (Frasch et al., 1987). At about 4 hours, the wild-type pattern of eve protein is in alternating stripes, stronger in the odd-numbered parasegments. Like ectopic AbdB, the most anterior eve stripe is in PS3, since the eve stripe in PSI decays by this time (see Fig. 8D in Frasch et al., 1987). The eve pattern in 4 hour E(z)⁻ embryos resembles this wild-type pattern (J.S. and W.B., unpublished). Thus, the early pattern of ectopic AbdB could be explained by a pair-rule factor gaining access to and activating AbdB in the wrong parasegments. The apparent gradient in AbdB along the A-P axis could be a vestige of control by gap products, expressed at earlier times in gradients (Stanojevic et al., 1989; Pankratz et al., 1990). Similarly, mis-regulation by a combination of gap and pairrule products could explain ectopic abdA appearing first in PS5 and 6.

Such a scenario may also explain two related observations. First, if a single copy of esc⁺ is provided paternally, the final abdA and AbdB patterns in the resulting heterozygous embryos (Fig. 4C, D) now resemble the transient patterns seen in the null. In this situation, like in the null, the transition to Pc group repression at 3–4 hours would be faulty, and early segmentation products would ectopically activate abdA and AbdB. However, zygotic expression of esc⁺ would eventually generate sufficient product to rescue Pc group repression and this would set in belatedly to fix permanently otherwise transient patterns. Second, the pair-rule distribution of AbdB is only seen in the strongest Pc group mutants, esc, E(z) and Pc. This is explained if the maternal contributions in the weaker mutants perdure long enough to allow the transition to Pc group repression to occur normally. Thus, when the maternal Pc group products eventually decay, pair-rule products are no longer present to influence the patterns of ectopic activation.

The cellular distribution of ectopic abdA within parasegments is highly patterned, with higher expression in the anterior portions of each parasegment (Fig. 1C–F). This patterning resembles the wild-type patterning of abdA normally restricted to PS7-13 (Fig. 1A, B). The wild-type abdA patterning is not due to trans-regulation by other homeotic products but rather reflects the intrinsic cell-specific controls that mediate abdA distribution (Karch et al., 1990). This patterning of ectopic abdA is most obvious in severe Pc group mutants, but it is also seen in the less severe mutants (Figs IG, H, 5B-G). Thus, although control along the A–P axis has broken down in these mutants, the cellspecific controls of abdA, even in the ectopic parasegments, still function. Again, the patterns of ectopic expression can be explained if abdA regulatory DNA is not adequately compacted in Pc group mutants. The normal cell-specific factors might then gain inappropriate access in PSI through PS6, resulting in ectopic abdA in a pattern that resembles the wild-type pattern. Although this ectopic transfer of cell-specific pattern is most obvious in embryos with abdA, similar effects have been noted for Scr and Ubx in imaginai discs (Jones and Gelbart, 1990; Glicksman and Brower, 1990). Similarly, the uniform blanket of ectopic AbdB in the strongest Pc group mutants (Fig. 2D, F) resembles the uniform pattern of AbdB normally seen only in PS13 and PS14 (Fig. 2B). In these cases, then, the defect in Pc group mutants involves transfer of normal homeotic pattern to inappropriate parasegments rather than indiscriminate homeotic activation. In summary, we suggest that the precise patterns of ectopic expression in Pc group mutants depends upon the distributions of normal activators in the ectopic parasegments.

Ten different Pc group products described here are required for both abdA and AbdB maintenance. Genetic data suggest that there may be as many as 40 members of the Pc group in total (Jiirgens, 1985). There are several possible explanations for the large number of Pc group genes. Individual Pc group products may interact independently with many different DNA sites within the large (>50 kb) regulatory regions of homeotic genes. The additional roles of some Pc group genes in diverse processes such as oogenesis, dorsal-ventral pattern formation and CNS development (Phillips and Sheam, 1990; Adler et al., 1989; Smouse et al.,1988) indicate that, at least in some instances, Pc group products can act independently from each other. Alternatively, some Pc group products may not directly affect transcription of homeotic loci but rather control transcription of other Pc products that are direct regulators. Pc protein itself is likely a direct regulator since it localizes to the ANT-C and BX-C loci on polytene chromosomes (Zink and Paro, 1989), but most of the other Pc group products have yet to be tested. As a third possibility, the Pc group products could act together in large multimeric complexes that compact ANT-C and BX-C DNA, as envisioned by the chromatin accessibility model. The recent identification of small DNA segments in the ANT-C (Zink et al., 1991) and in the BX-C (Simon et al., 1990; J.S. and W.B., unpublished) that mediate the response to Pc group products in vivo should help to address these issues.

We thank Sue Celniker for the gift of AbdB antibody. Mutant stocks and information about stocks were kindly provided by Paul Adler, Ian Duncan, Thomas Gutjahr, Rick Jones, Gerd Jürgens, Markus Noll, Renato Paro, David Smouse, Gary Struhl and Ting Wu. We especially thank Rick Jones and Mike O’Connor for numerous helpful discussions and Gary Struhl for suggesting the esc paternal rescue experiments. This research was supported by fellowships from the Jane Coffin Childs Fund and the Medical Foundation/ Charles King Trust to J.S. and by a grant from the NIH to W.B.