Role of the esc⁺ gene product in ensuring the selective expression of segment-specific homeotic genes in Drosophila

Insect segments come into being around the cellular blastoderm stage as a series of adjacent cell groups of similar size (reviewed in Lawrence, 1981). During subsequent development, these primordia diversify dramatically, giving rise to the unique segments of the larva and the adult. In Drosophila, this process requires a small number of ‘homeotic’ genes which act in combination to specify the particular developmental pathway followed by each segment (Garcia- Bellido, 1977; Lewis, 1978; Struhl, 1982).

Homeotic genes can be classified according to where, and when, their products are required. For example, several genes appear to function selectively in par- ticular segments, and are essential during both embryogenesis and subsequent development [called ‘selector’ genes by Garcia-Bellido (1975): viz, genes of the bithorax-complex (Lewis, 1963, 1978; Morata & Garcia-Bellido, 1976), the

Antennapedia+ (Antp+) gene (Denell, Hummels, Wakimoto & Kaufman, 1981; Wakimoto & Kaufman, 1981; Struhl, 1981b, 1982), and the Sex combs reduced+ (Ser+) gene (Lewis, Kaufman, Denell & Tallerico, 1980a; Lewis, Wakimoto, Denell & Kaufman, 1980Z?; Wakimoto & Kaufman, 1981; Struhl, 1982)]. A second class of genes is required in all segments during both embryonic and larval development, and appears necessary to maintain the correct expression of selec- tor genes [viz, the Polycomb+ (Pc+) gene (Lewis, 1978; Struhl, 1981a; Duncan & Lewis, 1982), the Polycomblike+ (Pci+) gene (Duncan, 1982), and possibly the Regulator of bithorax+ (Rg-bx+) gene (Ingham & Whittle, 1980; Ingham, 1981; Capdevila & Garcia-Bellido, 1981; Duncan & Lewis, 1982)]. Finally, the product of at least one gene, extra sex combs+ (esc+) has a discrete early role in the process of initiating the correct combinations of active and inactive selector genes in each segment (Struhl, 1981a; Struhl & Brower, 1982). These three classes suggest a hierarchy of genetic control in which the correct combinations of active and inactive selector genes are first established in each segmental primordium, and then maintained during subsequent development.

As a prerequisite to studying the roles of homeotic genes at the molecular level, it is helpful to understand as much as possible about the interactions between members of these different classes. Here, I examine the role of the product of the esc+ gene in the process of initially activating, or inactivating, other homeotic genes. In particular, I extend previous results (Struhl, 1981a) showing that the esc+ gene product is required for the correct expression of the genes of the bithorax-complex, and provide evidence that the esc+ gene product is also required for the correct expression of the Scr+ gene, the Antp+ gene, and possibly other segmental selector genes. Finally, I outline how the products of all these genes might act in concert to specify segmental determination throughout the body.

Eggs from appropriate parents were collected on agar plates over 12 h periods, and allowed to mature for a subsequent 24 h. They were then rinsed, dechorionated with dilute hypochlorite, fixed by incubation in glycerol-acetic acid (1:4) for 30 min at 60 °C, and then mounted in Hoyer’s mixture according to the procedure of Van der Meer (1977). After incubation of mounted prepara- tions overnight at 60 °C, most of the internal tissues are dissolved allowing the cuticular features to be studied under bright field, phase, Nomarski, or dark field optics. When necessary, embryos were dissected out from the vitelline mem- brane prior to fixing in glycerol-acetic acid.

The cuticle of the first instar larva is secreted approximately 12-14 h after fertilization (about halfway through embryogenesis), and has been described in fine detail by Lohs-Schardin, Cremer & Nusslein-Volhard (1979) and Turner & Mahowald (1979). The features which are of value for assessing segmental deter- mination are specialized sense organs unique to particular segments or sets of segments (e.g., the Keilin’s organs, and the ventral and dorsal pits found only on the thoracic segments), and the characteristic patterns of ventral hairs on the thoracic and abdominal segments. The assignment of cuticular structures to the thoracic and first seven abdominal segments (shown in Fig. 1) is widely accepted (e.g., Herth & Sander, 1973; Lohs-Schardin et al., 1979; Turner & Mahowald, 1979). The eighth abdominal segment is thought to bear the rectangular band of ventral hairs posterior to the seventh abdominal segment, but there is no clear consensus on whether the ‘terminalia’ (the posterior spiracles, anal pads, tuft, and associated sense organs) also belong to this segment or to other more pos- terior segments (e.g., Lohs-Schardin et al. 1979; Turner & Mahowald, 1979). Because homeotic mutations transforming the eighth abdominal segment alter not only the ventral and dorsal hair patterns but also the posterior spiracles and some of the sense organs (Lewis, 1978; Struhl, unpublished), whereas mutations which appear to affect the reverse transformation of more anterior segments to eighth abdominal segments add these structures back (Struhl, 1981a, Duncan & Lewis, 1982, and here (Figs 1, 2, 5 and 6)), I will assume that these structures normally derive from the eighth abdominal segment (the anal pads, tuft, and remaining sense organs are unaffected by either kind of mutation suggesting that they arise from other, more posterior segments, or are non-segmental). The segmental status of most of the head structures is uncertain. Fig. 1 shows the head structures, as well as tentative assignments of their segmental status according to Schoeller (1964), Lohs-Schardin et al. (1979) and Turner & Mahowald (1979).

In order to study the normal requirement for a particular gene, it is helpful to use mutations of that gene which remove most if not all of its wild-type activity. Aside from deficiencies which are visible cytologically, there are at present no direct genetic means of establishing that particular mutations are indeed null alleles. However, the mutations used here were isolated under conditions which should allow null alleles to be obtained, and at frequencies consistent with them being simple knock-out mutations in structural genes. Moreover, several muta- tions of each gene were obtained, and in each case, a subset of these mutations caused similar homeotic phenotypes at both 17 °C and 29 °C, whether in trans to a known deficiency of the locus or, when tested, in trans to each other. Such behaviour is expected of mutations which greatly reduce or eliminate a gene’s wild-type function. Consequently, I have treated these mutations as null muta- tions.

(i) esc⁻ mutations: The isolation and initial characterization of new mutations of the esc locus have been described previously (Struhl, 1981a). In this study, the mutations esc², esc⁵, esc⁶, and esc¹⁰ have been used interchangeably and are referred to as esc⁻.

(ii) Pc⁻ mutations: The dominant phenotype associated with apparent point mutations (Pc¹, Pc², and Pc³), and deficiencies of the Pc locus can be accounted for simply as a consequence of carrying only one, as opposed to the normal two, copies of the wild-type gene (Duncan & Lewis, 1982). During screens to generate new, EMS-induced mutations of the esc locus, I also obtained eight new alleles of the Pc locus (named Pc^{c1,c2, …,C8}) which show a dominant Pc phenotype in adults. These mutations arose with a frequency of approximately 1/2000 mutagenized chromosomes. Moreover, comparison of the embryonic pheno- types of each of these mutations (as well as of the Pc¹, Pc², and Pc³ mutations) in trans to a deficiency of the locus kindly provided by G. Jurgens did not reveal conspicuous differences whether at 17 °C or 29 °C. For these reasons, it seems reasonable to treat all of these mutations as null alleles. In order to study the phenotypes of esc∼ embryos carrying four copies of the Pc⁺ gene I used a tandem duplication of the Pc locus, Dp(3;3)C126, kindly provided by E. B. Lewis.

(iii) bμμorax-complex deficiencies, duplications, and mutations: All of the bithorax-complex deficiencies and duplications used here were kindly provided by E. B. Lewis (see Lindsley & Grell (1968), Lewis (1978,1980), and Morata & Garcia-Bellido(1976) for descriptions). During screens to obtain new EMS- induced mutations of the esc locus, four new alleles of the Ultrabithorax (Ubx) locus (named Ubx^C2,C3,C4’^C5) were also obtained. Embryos carrying each of these mutations in trans to Df(3R)P9 appeared indistinguishable in phenotype from each other and from Df(3R)bxd^1[®/Df(3R)P9 embryos. I therefore treat these mutations as null alleles of the Ubx locus.

(iv) Antp⁻ and Scr⁻ mutations: The isolation and characterization of apparent null mutations of the Antp locus used here have been described previously (Struhl, 1981b). Two mutations, Antp^{Ns + RC3} and Anip^{Ns + RC4}, were used inter- changeably.

In a systematic screen for third chromosome mutations altering the segmental pattern of the embryo, C. Nüsslein-Volhard, G. Jürgens, and E. Wieschaus obtained four new mutations of the Scr locus (Scr^7F,8B,9G,13A) they kindly provided for these experiments. All show the characteristic adult and embryonic phenotypes previously described for apparent null mutations of this locus (Lewis et al. 1980a,b;Wakimoto & Kaufman, 1981). Moreover, comparisons of the embryonic phenotypes of these mutations in trans to Df(3R)4Scb (a small deficiency eliminating both the Antp⁺ and Scr⁺ genes kindly provided by G. Jürgens) did not reveal conspicuous differences either at 17 °C or 29 °C suggest- ing that they eliminate most or all of the Scr⁺ gene function. Two mutations, Scr^9G and Scr^13A have been used interchangeably in this study. In order to study the phenotype of embryos lacking both the Antp⁺ and Scr⁺ genes, a chromosome carrying apparent null mutations at both loci was constructed as follows. Males carrying the Antp^{Ns+ RC3} mutation were mutagenized with EMS and outcrossed to wild-type females under conditions in which progeny carrying the Antp^Ns+RC3- bearing chromosome could be identified. Fl Antp^{Ns + RC3}/ + males were screened for the reduction of sex comb teeth associated with haplo-insufficiency for the Scr locus (Kaufman, Lewis & Wakimoto, 1980; Lewis et al. 1980a,b). One such male was obtained, and on subsequent outcrosses, passed on a new mutation at the Scr locus (Scr^C[). This mutation appears similar in its genetic properties to the other Scr mutations described above, and has been treated as a null mutation of the Scr⁺ gene.

(i) Unless otherwise stated, all mutations and chromosomal aberrations used are described in Lindsley and Grell (1968).

(ii) General procedure for generating esc⁻ adults: The phenotype of esc⁻ progeny depends critically on the genotype of the mother. When the mother is hemi- or homozygous for an esc⁻ mutation, esc⁻ zygotes develop into first instar larvae in which most of the segments appear transformed into eighth abdominal segments (such transformed embryos die as pharate first instar larvae). How- ever, if the mother is esc⁻/+, homozygous esc⁻ zygotes develop into normal first instar larvae which, in turn, are able to develop into adults. It became apparent during the course of experiments described here that such ‘rescued’ esc∼ adults were not always fully viable or fertile. This problem was overcome by generating homozygous esc⁻ adults from mothers which carried one mutant, and two wild- type copies of the esc locus. Either of two duplications of the esc⁺ gene were used for this purpose: Dp(2;3)17 of Ising which is an insertion of the region 31B-34D into 76C on the 3L (see Struhl, 1981a), and Dp(2;2)GYL (isolated by G. Yanopoulis and described by M. Ashburner) which is an insertion of the region 33B 1,2-3501,3 into 50A,B on the 2R. Because this study is concerned exclusive- ly with the cuticular phenotypes of esc⁻ progeny obtained from esc⁻ mothers, such embryos are here referred to simply as esc∼ embryos.

(iii) Generating esc⁻ embryos which are further transformed by other homeotic mutations: esc⁻ parents heterozygous for particular homeotic muta- tions on the third chromosome were generated by standard genetic methods (e.g., esc⁻, Df(3R)Ubx¹⁰⁹/ + parents were generated as follows. Virgin CyO, dp^1vesc²pr cn²/dp b cn Dp(2;2)GYL, esc⁺bw females were crossed to esc⁵cn/+’, Df(3R)Ubx¹⁰⁹/ + males, and male and virgin female CyO, dp^1v/esc²pr cn²/esc⁵cn; Df(3R)Ubx¹⁰⁹/ + progeny selected on basis of their Cy, esc, cn, and Ubx phenotypes.). Embryos derived from such parents were prepared as described in a) and embryos of the correct genotype (i.e., homozygous for the appropriate third chromosome mutations) were identified on the basis of their homeotic phenotype which differed significantly from that of standard esc⁻ em- bryos. Generating esc⁻ embryos carrying Df(3R)P9 which is thought to delete the entire bithorax-comptex (Lewis, 1978)) involves more complex genetic crosses to obtain the appropriate esc⁻ parents, because this deficiency must always be carried in trans to at least two copies of the bithorax-complex (Df(3R)P9/ + flies of both sexes are sterile).

(iv) Generating esc⁻embryos of known third chromosome genotype without reference to homeotic phenotype. In several cases, it was necessary to examine the phenotype of esc⁻ embryos carrying homeotic mutations in the third chromosome without reference to the resulting homeotic phenotype (e.g., when comparing the phenotype of esc⁻ embryos bearing different numbers of copies of the bithorax-complex, or of the Pc⁺ gene). To this end, esc⁻ parents were generated which were homozygous for the cuticle colour mutation yellow (y) and which carried the appropriate homeotic mutation or gene duplication in trans to a TM3, e^sSery⁺ chromosome (kindly provided by E. Wieschaus). For example, esc⁻ embryos homozygous for Dp(3;3)P5 (a tandem duplication of the entire bithorax-compiex (Lewis, 1980)) were prepared as follows: y; dp esc⁶stc pr en/ + ; Dp(3;3)P5/TM2, Ubx(tm) e^s males were crossed to virgin y; esc(tm) b pr/ In(2L)Cy, dp^u; TM3, e^sSer y⁺/Dp(2;3)17, esc⁺ females, and y; dp esc⁶stc pr en/esc¹⁰b pr, Dp(3;3)P5/TM3, e^sSery⁺ progeny selected on the basis of their y⁺, esc, pr, e⁺, and Cy⁺ phenotype. esc⁻ embryos obtained from such flies were prepared as described in a) and then screened initially under bright-field illumi- nation to identify phenotypically yellow, and hence genotypically Dp(3;3)P5/ Dp(3;3)P5, specimens. This method of independently marking esc∼ embryos bearing a particular third chromosome genotype was used for generating esc∼ embryos with the following third chromosome genotypes: Df(3R)P9/+, Dp- (3;3)P5/Dp(3;3)P5,Pc²/+,Pc²lPc²,Dp(3;3)C126/Dp(3;3)C126,Antp^{Ns+ RC3}l Df(3R)4Scb, and Set^13A//Df(3R)4Scb.

The major aim of these experiments is to examine the extent to which the product of the extra sex combs* (esc+) gene is required for the selective ex- pression of other homeotic genes. To this end, the phenotype of embryos lacking the esc+ gene product is described in detail, and then, this standard phenotype is compared with the phenotypes of esc+ and esc⁻ embryos which lack other homeotic genes, or which carry an altered number of copies of these other genes.

The phenotypes caused by recessive mutations of the esc locus depend critic- ally on the genotype of both the mother and the zygote. Homozygous esc⁻ embryos obtained from heterozygous females develop into normal first instar larvae which in turn are able to develop into adults. Such adults show a variable transformation of the distal second and third legs into distal first legs (Slifer, 1942; Tokunaga & Stern, 1965; Struhl, 1981a), as well as occasional, weak transformations of antenna to leg, and wing to haltere (Struhl, unpublished). In contrast, esc∼ embryos obtained from homozygous esc⁻ females develop into first instar larvae in which most of the segments appear similar to the eighth abdominal segment of wild-type larvae (Struhl, 1981a). The cuticle of the first instar larva is secreted about half way through embryogenesis; hence the cuticular pattern is treated as a direct reflection of the embryonic phenotype at the time of cuticle deposition. Because this study is concerned only with the cuticular phenotypes of esc⁻ progeny derived from esc⁻ mothers, such progeny will be referred to simply as esc⁻ embryos.

The phenotype of esc⁻ embryos is characterized by the transformation of at least one cephalic and two gnathal segments, all three thoracic segments, and the first nine abdominal segments into segments resembling to varying degrees the normal eighth abdominal segment. The wild-type eighth abdominal segment bears several unique characteristics which distinguish it from all other segments. These include (i) a square pattern of five to six rows of ventral hairs, (ii) posterior spiracles (both internal Filzkorper and external sensory structures), and (iii) unique sensory organs (see Methods and Fig. 1). In esc∼ embryos (Fig. 2), all three thoracic segments as well as the first seven abdominal segments bear the square pattern of ventral hairs, as well as some of the unique sensory organs characteristic of the eighth abdominal segment. In addition, rudimentary pos- terior spiracles (both internal and external portions) are frequently found on the more posterior abdominal segments, and sometimes on the more anterior abdominal and thoracic segments. The eighth abdominal segment itself appears indistinguishable from that of wild-type embryos. Thus, the cuticular patterns of the thoracic and first seven abdominal segments approach that of the normal eighth abdominal segment. In addition to these segments, four others also ap- pear to be transformed towards a normal eighth abdominal segment, but here, the transformation is less complete. These are a cephalic segment (tentatively called the antennal segment - see Fig. 5), two gnathal segments immediately anterior to the thorax (tentatively called the labial and maxillary segments - see below), and a ninth abdominal segment. The antennal segment is consistently transformed so that it bears a band of ventral-type hairs similar to that of the eighth abdominal segment. However, these hairs are formed on the dorsal side of the head suggesting the possibility that the transformed segment has rotated such that originally ventral cells now lie dorsally. Occasionally what appears to be a trail of these ventral-type hairs remains and links the dorsally positioned hairs to the ventral side of the embryo (Fig. 2). These remaining hairs lie im- mediately anterior to the remnants of the ventral portion of the cephalopharyn- geal skeleton. The extent to which the labial, maxillary and ninth abdominal segments are transformed is variable; the pattern of ventral hairs formed by these segments ranges from a small tuft of hairs to three to five rows of hairs arranged in a square pattern. None of these three segments exhibit posterior spiracles or sensory organs characteristic of the eighth abdominal segment. Although no sense organs unique to the thoracic segments remain in esc∼ embryos, remnants of most of the normal head structures (e.g., separate left and right halves of the labial sense organ, the maxillary cirri and associated sense organs, mouth hooks, maxillary and antennal sense organs, cephalopharyngeal skeleton, labral hatching tooth) can usually be identified. Because esc⁻ embryos fail to undergo head involution, these structures remain on the outside. Several structures appear un- affected in esc⁻ embryos. These are the anal pads, a tuft of ventral hairs posterior to the anal pads, and a set of sense organs immediately posterior to the anal pads.

esc⁻ embryos carrying one, or four copies of the bithorax-compXex have been generated, and their phenotypes compared with the standard esc⁻ phenotype in which two copies of the bithorax-compiex are present. esc⁻ embryos carrying only one copy of the complex show a less-extreme transformation than standard esc⁻ embryos (Fig. 3). For example, rudimentary Keilin’s organs and ventral pits remain in all three thoracic segments (these sensory organs are normally present in the thoracic segments of wild-type embryos, but absent in standard esc⁻ embryos). esc⁻ embryos carrying four copies of the bithorax-compXex show a more extreme transformation than that of standard esc⁻ embryos (Fig. 3). For example, rudimentary posterior spiracles appear in most or all of the abdominal and thoracic segments, and in addition, the remnants of the sense organs of the head appear further depauperated. Thus, the expression of the esc⁻ phenotype depends on the number of copies of the bμμorax-complex. As the number of copies rises from one to two to four, the transformation of all segments into eighth abdominal segments appears progressively more complete.

The dependence of the esc⁻ phenotype on the dosage of particular portions of the bithorax-complex has been studied by comparing the phenotypes of esc⁻ embryos carrying one or two copies of specific fragments of the complex. For example, Df(3R)Ubx¹⁰⁹ deletes a centromere-proximal portion of the complex, leaving the distal portion intact. esc⁺ embryos homozygous for this deficiency, or carrying it in trans to Df(3R)P9 (which is thought to be a deletion for the entire bithorax-complex (Lewis, 1978)) show a complete transformation of the metathorax and first four or five abdominal segments into mesothoracic seg- ments (Fig. 4). In addition, the remaining posterior segments appear partially transformed into thoracic segments, though the eighth segment still bears rudimentary posterior spiracles and lacks ventral pits and Keilin’s organs found on all of the more anterior abdominal segments (Lewis, 1978; Struhl, 1981A). The phenotype of esc⁻ embryos carrying Df(3R)Ubx^1G9 depends on whether or not the deficiency is hemi- or homozygous. In esc⁻; Df(3R)Ubx^W9/Df(3R)P9 embryos (Fig. 4), the thoracic and first seven abdominal segments are transfor- med into segments resembling the eighth abdominal segment of Df(3R)Ubx¹⁰⁹/ Df(3R)P9 embryos. The transformation is not complete; there are rudimentary Keilin’s organs and ventral pits in all of these segments. However, in esc⁻;, Df(3R)Ubx^W)IDf(3R)Ubx¹⁰⁹ embryos (Fig. 4) the Keilin’s organs and ventral pits are now absent in all of the thoracic and abdominal segments. Similar results have been obtained by comparing the phenotypes of esc⁻ embryos that are hemi- or homozygous for two other partial deficiencies of the foiμorax-complex (Df(3R)bxd¹⁰⁰, and Df(3R)P10). Thus, the esc phenotypes appear to depend on the numbers of copies of both the entire bithorax-complex as well as particular fragments of the complex.

Embryos homozygous for small deficiencies or point mutations of the Poly comb (Pc) locus show an embryonic lethal phenotype similar to, though less extreme, than that of the standard esc∼ phenotype (Lewis, 1978; Duncan & Lewis, 1982). However, unlike the esc⁺ gene product which appears to have a discrete early role (Struhl, 1981a ; Struhl & Brower, 1982), the product of the Pc⁺ gene appears to play the same role throughout development (Struhl, 1981a; Duncan & Lewis, 1982). The possibility that the product of the esc⁺ gene is required for the correct expression of the Pc⁺ gene has been examined by com- paring the phenotypes of esc⁻ embryos carrying zero, one, two, or four copies of the Pc locus. esc⁻ embryos carrying only one copy of the Pc⁺ gene show a more complete transformation of all the segments towards the eighth abdominal segment than that of standard esc⁻ embryos, approaching the phenotype of esc⁻ embryos carrying four copies of the entire bithorax-compiex (see above). The phenotype of esc⁻ embryos which are homozygous for an apparent null mutation of the Pc locus, Pc², show a more complete transformation similar to, if not more extreme than, that of esc⁻ embryos carrying four copies of the bithorax-compXex. In contrast, esc⁻ embryos carrying four copies of the Pc⁺ gene appear indistinguish- able from standard esc⁻ embryos. Thus, the esc⁻ phenotype depends on the number of copies of the Pc⁺ gene only within a defined range. As the number of genes decreases from two to one to zero, the transformation of all segments to eighth abdominal segments becomes more complete; however, adding extra copies seems to have no effect.

esc⁺ embryos homozygous for Df(3R)P9 show a homeotic transformation of the metathorax and first seven abdominal segments into mesothoracic segments, and of the eighth abdominal segment into a segment resembling in part an intermediate between the meso- and prothoracic segments (Lewis, 1978; Struhl, 1981a; Fig. 5). The eighth abdominal segment also carries bilaterally symmetric plates of sclerotized cuticle which have been interpreted as rudimentary mouth hooks (Lewis, 1978; Fig. 5). If the esc⁺ gene product were required for ensuring the correct expression of only the bithorax-complex genes, embryos lacking the entire bithorax-complex would be expected to develop identically whether or not they carried the esc⁺ gene product. However, esc⁻;, Df(3R)P9 embryos differ in phenotype from Df(3R)P9 embryos in that they show a transformation of the thoracic and abdominal segments as well as a head segment into a thoracic-like segment which appears to be an intermediate between the meso- and prothorax (Struhl, 1981a; Fig. 5). Also, these embryos lack the bilaterally symmetric plates of sclerotized cuticle present in the eighth abdominal segment of Df(3R)P9 embryos. Assuming that Df(3R)P9 does indeed remove the entire bithorax- complex, this second phenotype can be interpreted as a consequence of inappropriate expression of homeotic genes other than those of the bithorax- complex (Struhl, 1981a). To test whether the phenotype of esc⁻;, Df(3R)P9 embryos reflects the phenotype of esc⁻ embryos lacking the entire bithorax- complex, the phenotype of esc⁻;, Df(3R)P115 embryos has been examined (Df(3R)P115 deletes the salivary chromosome bands 89B-89E5,6 (Lewis, per- sonal communication, as quoted in Morata & Garcia-Bellido, 1976) and hence, unquestionably removes the bithorax-complex in its entirety). As in esc∼; Df(3R)P9 embryos, these embryos show a transformation of all thoracic and abdominal segments into intermediates between the meso- and prothorax, and lack the plates of sclerotized cuticle in the eighth abdominal segment. In addition, in favourable preparations it is possible to detect transformation of a head segment into a thoracic segment as in esc⁻; Df(3R)P9 embryos. Thus, the phenotype of esc⁻; Df(3R)P115 embryos provides further, though indirect, evidence that the esc⁺ gene product is required for the correct expression of homeotic genes outside the bithorax-complex.

The phenotype of embryos homozygous for Sex combs reduced (Scr) muta- tions has been described by Wakimoto & Kaufman (1981). Embryos homozygous for apparent null mutations of the Scr⁺ gene show a partial transformation of the prothoracic segment into the mesothorax, as well as a disruption of the larval mouthparts (e.g., Fig. 6). To their description, it should be added that the disruption of the larval mouthparts is associated with absence of the labium and presence of a second pair of maxillary sense organs lying between the normal mouthparts and the prothorax (Fig. 6). Finally, the mesothorax, metathorax, and abdominal segments appear normal. Thus, the Scr⁺ gene appears to be required in the prothorax to specify pro- and opposed to mesothoracic develop- ment, and in the labial segment in order to specify labial as opposed to maxillary development; however, it does not appear to be required in more posterior segments.

esc⁻ embryos which lack the Scr⁺ gene appear identical to standard esc⁻ embryos except in the following respect. In standard esc⁻ embryos, the remnants of the labial pits are found associated with a small tuft of ventral hairs, and anterior to these structures, a square band of three to five rows of ventral hairs is formed (e.g., Fig. 2). In esc⁻; Scr⁻ embryos, the labial pits are absent and replaced by a second pair of maxillary sense organs. In addition, the small tuft of ventral hairs lying between this second set of maxillary sense organs is enlar- ged to form a square band of three to five rows of ventral hairs. This phenotype suggests that the small tuft of hairs formed in standard esc∼ embryos corresponds to the labial segment, that the square band of hairs immediately anterior corres- ponds to the maxillary segment, and that in esc⁻, Scr⁻ embryos, the labial segment develops like the maxillary segment of standard esc⁻ embryos.

One method to examine whether or not the esc⁺ gene product acts to ensure the correct spatial expression of the Scr⁺ gene is to identify segments in which the Scr⁺ gene is normally not required, and then to test whether absence of the esc⁺ gene product leads to inappropriate activity of the Scr⁺ gene in these segments. As discussed above, the Scr⁺ does not appear to be required in the meso- or metathorax, or in the abdominal segments. However, in both esc⁺ and esc⁻ embryos, some or all of the bithorax-complex genes are active in these segments, thereby posing the possibility that activity of the bithorax-compiex genes might obscure a requirement for the Scr⁺ gene. Consequently, it is necessary to test for the presence or absence of the Scr⁺ gene activity in embryos lacking the bithorax- complex.

In Scr⁻ embryos lacking the entire bithorax-compiex (i.e., homozygous for Df(3R)P9y, referred to subsequently as BX-C⁻), the meso- and metathorax, as well as the first seven abdominal segments, all bear mesothoracic patterns of ventral hairs just as they do in BX-C⁻ embryos (Figs. 5, 6). In addition, the eighth abdominal segment appears identical in embryos of both genotypes (Figs 5, 6). Hence, even in the absence of the bithorax-compXex, these segments appear to develop identically whether or not the Scr⁺ gene is present¹. In contrast, the phenotype of these same segments in esc⁻; BX-C∼ embryos clearly depends on whether or not the Scr⁺ gene is present. Two differences are con- spicuous. First, in esc⁻; BX-C⁻ embryos, a second band of fine hairs normally found only on the prothorax is present in all of the transformed segments of the thorax and abdomen (Fig. 5). However, in esc⁻: Scr⁻BX-C⁻ embryos, this second band of hairs is reduced or eliminated, so that the thoracic and abdominal segments appear predominantly mesothoracic in character (Fig. 7). Second, rudimentary mouthpart structures, the cirri, and possibly other mouthpart struc- tures, are occasionally found in the thoracic and abdominal segments of esc⁻ ; Scr∼ BX-C⁻ embryos (Fig. 7), but were never observed in esc⁻ ; BX-C⁻ embryos.

Thus, the particular homeotic phenotype observed in the thoracic and abdominal segments of esc⁻; BX-C⁻ embryos depends on the presence of the Scr⁺ gene; however, no such dependence is observed in esc⁺ embryos similarly deficient for the bithorax-complex. These results suggest that absence of the esc⁺ gene product leads to inappropriate activity of the Scr⁺ gene in the thoracic and abdominal segments in question, and hence to a novel function of the gene in these segments. It should be noted that inappropriate expression of the Scr⁺ gene in these segments leads to a meso- to prothoracic transformation, as well as an apparent suppression of latent maxillary cirri. Both of these effects are consistent with the normal roles attributed to selective activity of the Scr⁺ gene in the prothoracic and labial segments.

Role of the esc⁺gene product in ensuring the selective expression of the Antp⁺gene The approach with the Antennapedia⁺(Antp⁺) gene is the same as in the preceding section dealing with the Scr⁺ gene; that is, to identify a segment in which the Antp⁺ gene is normally not required, and to test whether absence of the esc⁺ gene product leads to a novel function of the Antp⁺ gene in this segment.

The phenotype of esc⁺ embryos homozygous for Antp⁻ mutations has been described by Wakimoto & Kaufman (1981). They report that embryos homozygous for apparent null mutations of this locus develop into first instar larvae in which both the meso- and metathorax appear partially transformed into prothoracic segments. Since only the meso- and metathoracic segments of these embryos appear abnormal, it seems superficially that the Antp⁺ gene is normally required in only these segments.

To examine the possibility that the Antp⁺ gene is normally required not only in the meso- and metathorax, but also in other segments, the phenotypes of Antp⁻embryos lacking part or all of the bithorax-complex, the Scr⁺ gene, or both the Scr⁺ gene and the entire bithorax complex have been examined. The results obtained are described in Fig. 8 and can be summarized as follows. In the absence of both the Scr⁺ gene and the entire bμμorax-complex, the phenotypes of all three thoracic segments as well as the first seven abdominal segments appear to depend on the presence of the Antp⁺ gene; however, the phenotype of the eighth abdominal segment does not. This result can be interpreted as evidence that the Antp⁺ gene is normally active in the thoracic and first seven abdominal segments, but inactive in the eighth abdominal segment. Consequently, if loss of the esc⁺ gene product leads to inappropriate expression of the Antp⁺ gene in segments in which it is normally inactive, one might expect to see a novel function of the Antp⁺ gene in the eighth abdominal segment.

esc⁻; Antp⁻ embryos appear identical in phenotype to standard esc⁻ embryos, thereby providing no evidence that the esc⁺ gene product is required for the correct expression of the Antp⁺ gene. However, the results of deleting the Scr⁺ gene as well as the bithorax-complex in Antp⁻ embryos indicate that gene activity at either locus might obscure a requirement for the Antp⁺ gene. Hence, to test whether or not the esc⁺ gene product is required for the correct expression of the Antp⁺ gene, it is necessary to perform the test in the absence of both the Scr⁺ gene and the entire bithorax-complex.

The phenotypes of all of the thoracic and first eight abdominal segments of esc⁻ ; Scr⁻BX- C⁻ embryos clearly depend on whether or not the Antp⁺ gene is present. First, none of the thoracic or abdominal segments of esc⁻; Scr⁻BX-C⁻ embryos show bilaterally symmetric regions of sclerotized cuticle (Fig. 7); however, the thoracic and first seven abdominal segments of esc⁻ ; Scr⁻Antp⁻BX-C⁻ embryos bear such sclerotized regions (Figs 9, 10). Also, in many cases, these regions resemble mouth hooks of the wild-type larva and are often associated with other rudimentary mouthpart structures (e.g., rudimentary cirri and a ridged cuticle resembling a portion of the normal cephalopharyngeal skeleton (Figs 9, 10)). Curiously, the eighth abdominal segment of such embryos does not bear bilater- ally symmetric plates of sclerotized cuticle, though rudimentary cirri are often for- med (Fig. 10). Second, in esc∼; Scr∼ BX-C∼ embryos, the pattern of ventral hairs in each segment (including the eighth abdominal segment) appears to be primarily mesothoracic in character (Fig. 7) ; however, in esc⁻ ; Scr⁻Antp⁻BX-C⁻embryos, the pattern of ventral hairs appears to be partially prothoracic in character, most notably, a square band of hairs characteristic of the second (posterior) band of hairs of the normal prothorax, appears in each segment of the thorax and abdomen (Figs 9,10). Thus, the phenotype of the eighth abdominal segment depends on the presence or absence of the Antp⁺ gene in esc⁻; Scr⁻BX-C∼ embryos; however, no such dependence is observed in esc⁺; Scr⁻BX-C⁻ embryos. These results can be interpreted as evidence that in the absence of the esc⁺ gene product, the Antp⁺ gene functions inappropriately in the eighth abdominal segment.

Role of the esc⁺gene product in ensuring the selective expression of homeotic genes other than the bithorax-com/2/ex genes, and the Antp⁺and Scr⁺genes If the early role of the esc⁺ gene product was restricted to ensuring the correct expression of only the genes of the bithorax-complex, and the Antp⁺ and Scr⁺ loci, then one would predict that the presence or absence of the esc⁺ gene product would be irrelevant in embryos lacking all of these homeotic genes. However, esc∼ embryos lacking all of these genes differ in at least two respects from corresponding esc⁺ embryos (Figs 9,10). First, all of the thoracic and first seven abdominal segments, as well as the labial segment, bear rudimentary head structures such as the cirri, not found on corresponding esc⁺ embryos. In addition, the paired regions of sclerotized cuticle present in esc⁻ embryos now approach the appearance of mouth hooks. Second, as in esc⁻; Df(3R)P9 embryos, a seg- ment of the head (possibly the antennal segment) may be transformed into a thoracic-like segment, although this transformation is difficult to ascertain with certainty. Equivalent results were also obtained when a small deficiency deleting both the Antp⁺ and Scr⁺ gene, Df(3R)4Scb (Jurgens, unpublished), was used instead of the Anip^Ns+RC3 and Scr^13A mutations; however, the phenotypes obser- ved are complicated by the absence of the R14 gene (Lewis et al. l93Qa,b’, Wakimoto & Kaufman, 1981) which results in pairwise fusion of the segments.

Thus, the phenotypes of esc⁻ and esc⁺ embryos lacking the Antp⁺, Scr⁺, and all of the bithorax-complex genes are not identical, suggesting the possibility that other, as yet unidentified homeotic genes (perhaps genes involved in the deter- mination of head segments) may also require the esc⁺ gene product for their correct expression.

The process of segmental determination requires the selective activation and continuous activity of small numbers of homeotic genes such as the bithorax- complex genes in particular segments. However, the mechanism by which the correct combinations of active and inactive homeotic genes are initially established in each segmental primordium is mysterious. In a previous study (Struhl, 1981a), I described the existence of a gene product which plays a critical role in this process. Specifically, I showed that the product of the esc⁺ gene is required early in development for ensuring the correct pattern of expression of most, or all, of the bithorax-complex genes. Here, I examine whether the esc⁺ gene product has a more general role in ensuring the correct expression of all segment-specific homeotic genes. To this end, I have described the phenotype of embryos lacking the esc⁺ gene product, and then asked to what extent this phenotype depends on the presence of other homeotic genes. I discuss first the dependence of the esc∼ phenotype on the number of copies of the bithorax- complex and Pc⁺ genes, second, the evidence that the esc⁺ gene product is necessary for initiating the correct patterns of expression of homeotic genes other than those of the bithorax-complex (e.g., the Scr⁺and Antp⁺ genes), and third, a tentative outline of the roles of the products of all these genes in the process of segmental determination.

In embryos lacking the esc⁺ gene product, the extent to which most segments appear to be transformed towards the eighth abdominal segment increases as the number of doses of the bithorax-complex rises from one to two to four. More- over, a similar dose dependence is observed when only fragments of the bithorax- complex are used. Thus, the extent of the esc∼ phenotype appears to be directly proportional to the number of copies of the bithorax-complex whether above or below the normal two copies.

One explanation for this dose dependence is as follows. Previous studies of esc∼ embryos lacking part or all of the bithorax-complex have established that absence of the esc⁺ gene product causes many, if not all, of the bithorax-complex genes to become active inappropriately in segments in which they are normally inactive (Struhl, 1981a). However, these studies do not establish that the bithorax-complex genes are fully active in these segments, but only that they are at least partially active. Indeed, as described here in detail, most segments of esc’ embryos do not form perfect replicas of the eighth abdominal segment as might be expected if all of the bithorax-complex gene were fully expressed in all segments, but rather, approach the pattern of this segment only to varying degrees. It is therefore possible that in the absence of the esc⁺ gene product, the bithorax-complex genes are only partially derepressed in segments in which they are normally not active. Under these conditions, the degree to which anterior segments are transformed into eighth abdominal segments might be directly proportional to the amounts of bithorax-complex gene products synthesized in each segment, and hence, proportional to the number of copies of the bithorax- complex present. Interestingly, the eighth abdominal segment of esc’ embryos appears indistinguishable from that of wild-type embryos suggesting that all the bithorax-complex genes are expressed at their normal levels in this segment.

A second explanation which should be considered is that other homeotic genes outside of the bithorax-complex become active inappropriately in esc’ embryos, and hence modify the transformed phenotype resulting from inappropriate ex- pression of the uÍhorax-complex genes. If, for example, all segment determining genes become active in all segments, they may specify a novel segment type with characteristics influenced primarily, but not completely, by the bithorax-complex genes. Consequently, the degree to which the genes of the bithorax-complex in- fluence the determined state of the segment may be directly proportional to the abundance of their gene products relative to the products of the other homeotic genes. This second explanation, though it may in part explain the observed dosage dependence, cannot explain the varying extent to which different segments in esc’ embryos are transformed towards the eighth abdominal segment. This segment- specific variation suggests that all homeotic genes are not expressed equally in all segments of esc’ embryos, and hence, as proposed above, that some are only partially derepressed in segments where they are normally inactive.

The dependence of the esc’ phenotype on the number of copies of the Pc⁺ gene contrasts with the results obtained with the bithorax-complex in two respects. First, the mutant phenotype becomes progressively more extreme as the number of copies of the Pc⁺ gene falls from two to one to zero, and hence is inversely proportional to gene dosage. Second, esc’ embryos carrying four copies of the Pc⁺ gene appear indistinguishable from embryos carrying two copies; that is, dose dependence is only observed when the number of copies is less than or equal to the diploid complement of two. To interpret these results, it is helpful to compare them one at a time with the results of altering the number of Pc⁺ genes in otherwise wild-type flies. First, a minimum of two copies of the Pc⁺ gene are required for normal development. When only one copy is present, segments of the adult fly are transformed towards more posterior segments in a fashion suggesting that some of the bμμorax-complex genes are partially derepressed in more anterior segments (Lewis, 1978; Capdevila & Garcia-Bellido, 1981; Duncan & Lewis, 1982). When no copies are present, most segments of the embryo are transformed towards the eighth abdominal segment, a phenotype which appears to result from in- discriminate expression of the bithorax-complex genes in most or all segments (Lewis, 1978; Duncan & Lewis, 1982). Thus, in otherwise wild-type animals, the bithorax-complex genes become progressively derepressed as the number of Pc⁺ genes falls from two to one to zero. Returning to esc∼ animals, the results of lower- ing the number of Pc⁺ genes appears to follow the same pattern; that is, as the number of Pc⁺ genes falls from two to one to zero, the transformation of most seg- ments towards the eighth abdominal segments appears to become more extreme, suggesting that the extent to which the bithorax-complex genes are inappropriate- ly expressed increases. Second, raising the number of copies of the Pc⁺ gene from two to four in otherwise wild-type animals seems to have no effect (Duncan & Lewis, 1982). Assuming that the amount of Pc⁺ gene product is directly propor- tional to the number of copies of the Pc⁺ gene, this result indicates that superabun- dance of the Pc⁺ gene product does not in itself cause incorrect expression of the bithorax-complex genes. Again, returning to esc∼ embryos, the same result is ob- served; that is the mutant phenotype, and hence presumably the extent to which the bithorax-complex genes are incorrectly expressed, appears unaffected when the number of Pc⁺ genes is increased from two to four. Thus, in both esc⁺ and esc⁻ animals, the effects of altering the dosage of the Pc⁺ gene on the expression of the bithorax-complex genes appears to be the same. These results suggest that altering the levels of the esc⁺ and Pc⁺ gene products have additive and independent effects on the expression of the bithorax-complex genes, and hence, that these gene products have independent roles. This conclusion is not surprising given the dif- ferent temporal requirements, and hence different presumed roles, of the two gene products (Struhl, 1981a; Struhl & Brower, 1982).

Evidence that the esc⁺gene product is required for ensuring the selective ex- pression of the Scr⁺and Antp⁺genes, as well as at least one other homeotic gene outside of the bithorax-complex

Indirect evidence that the esc⁺ gene product is required for the correct expression of homeotic genes outside of the foiμorax-complex comes from studies of the phenotype of esc⁻ embryos lacking the entire bithorax-complex (Struhl, 1981a; here). In these embryos, most segments develop differently from their counterparts in esc⁺ embryos similarly deficient for the bithorax-complex (see Fig. 11), suggesting that absence of the esc⁺ gene product causes inappropriate expression of other homeotic genes in most or all of these seg- ments. To test this interpretation, I have examined whether the esc⁺ gene product is required for the correct expression of particular homeotic genes out- side of the bithorax-complex. The general approach has been as follows. First, identify those segments in which a given homeotic gene is normally not required. Then test whether absence of the esc⁺ gene product causes a homeotic phenotype in these segments which can be attributed to adventitious function of the gene in question. Such experiments have been performed for two genes of the Antennapedia-comptex, Scr⁺andAntp⁺. The pertinent results are summarized in Fig. 11 and discussed below.

The requirement for the esc⁺ gene product is clearest in the case of the Scr⁺ gene. Consequently, a thorough examination of the evidence in this case provides the best opportunity for assessing the validity of the experimental method and of the resulting conclusions. Both the embryonic (Wakimoto & Kaufman, 1981; here) and adult (Kaufman, Lewis & Wakimoto, 1980; Lewis et al. 1980a,6; Struhl, 1982) phenotypes of Scr mutations indicate that the wild-type gene product is normally required in the labial and prothoracic segments to specify their correct segmental states. However, no requirement for the Scr⁺ gene has been detected in the meso- or metathorax, or in the abdominal segments, even when other homeotic genes which normally function in these segments (the bithorax-complex genes and the Antp⁺ gene) are eliminated. In contrast, ab- sence of the esc⁺ gene product results in a detectable function of the Scr⁺ gene in these segments. The simplest interpretation of this result is that the Scr⁺ gene is normally active only in the labial and prothoracic segments, but that it becomes active inappropriately in all of the thoracic and abdominal segments when the esc⁺ gene product is absent.

One criticism of this interpretation is that it depends on the assumption that absence of a detectable requirement for the Scr⁺ gene in a given segment in- dicates that the Scr⁺ gene is not active in that segment. Yet, absence of a detect- able requirement is a negative result which may be misleading. For example, the number of cuticular structures which allow segmental status to be assessed in embryos is sufficiently small that some changes in segmental state caused by the absence of the Scr⁺ gene might go undetected. Alternatively, it is possible that the normal activity of the Scr⁺ gene in these segments might be masked or superfluous owing to the activity of other, as yet unidentified, homeotic genes. This criticism is countered by the nature of the novel function of the Scr⁺ gene which arises in the absence of the esc⁺ gene product. Normally, in BX-C⁻ em- bryos, most of the thoracic and abdominal segments appear identical to the normal mesothorax. In esc⁻; BX-C⁻ embryos, the determined state of these segments now appears to be shifted towards the prothorax; however, if the Scr⁺ gene is also absent, it is shifted back towards the mesothorax. Thus, in esc⁻ ; BX- C∼ embryos, the Scr⁺ gene functions abnormally to shift any segment which would otherwise have been a mesothoracic segment towards being a prothoracic segment. This abnormal function exactly parallels the normal function of the Scr⁺ gene in the prothorax of both wild-type and BX-C⁻ embryos, and hence, can be attributed to adventitious expression of the normal gene function in the wrong segments. Thus, the similarity between the normal and adventitious fun- ctions of the Scr⁺ gene observed in esc⁺ and esc⁻ embryos provides positive evidence that this function is normally absent in most of the thoracic and abdomi- nal segments.

A second criticism is that a novel function of the Scr⁺ gene is observed only when the bithorax-complex genes are eliminated. Thus, strictly speaking, the results indicate only that the esc⁺ gene product is required for the correct ex- pression of the Scr⁺ gene in the absence of the entire bithorax-complex. In principle, one could argue that absence of the bithorax-complex creates an artefactual situation in which the esc⁺ gene product is required for the correct expression of the Scr⁺ gene, whereas normally, no such requirement exists. However, a corollary of this argument is that activity of the bithorax-complex normally ensures the correct expression of the Scr⁺ gene. Because the esc⁺ gene product is required for the correct expression of the bithorax-complex, this argument leads ultimately to the conclusion that the correct pattern of expression of Scr⁺ gene would still depend, albeit indirectly, on the function of the esc⁺ gene product.

The evidence that the esc⁺ gene product is required for the selective expression of the Antp⁺ gene function is as follows. Based on the embryonic phenotypes of Antp∼ mutations alone, and together with Scr⁻ mutations, or various deficien- cies of the bithorax-complex, the Antp⁺ gene appears normally to be required in all of the thoracic segments and in the first seven abdominal segments, but not in the eighth abdominal segment. However in esc∼ embryos lacking both the Scr⁺ gene and the entire bithorax-complex, the homeotic phenotype of the eighth abdominal segment now depends on the presence or absence of the Antp⁺ gene. Subject to the reservations discussed above in the case of the Scr⁺ gene, these results suggest that the Antp⁺ gene function is normally not active in the eighth abdominal segment, but that it is expressed adventitiously in this segment if the esc⁺ gene product is absent.

The evidence that the esc⁺ gene product is required for the correct expression of at least one other homeotic gene outside of the bfiμorax-complex, and the Antp and Scr loci is identical to that initially provided for genes outside of the bithorax-complex. That is, in the absence of both the Scr⁺ and Antp⁺ genes, as well as the entire bithorax-complex, the homeotic phenotype still depends on the presence or absence of the esc⁺ gene product. This result suggests that at least one other homeotic gene also depends on the esc⁺ gene product for its correct expression. It is interesting that the effect of inappropriately expressing this gene (or genes) appears to be an enhancement of mouthpart structures appearing in most segments of Scr∼ Antp∼ embryos lacking the HÍμorax-complex. This phenotype suggests the possibility that the additional homeotic gene, or genes, may be involved normally in the specification of head segments.

Fig. 12 summarizes a tentative outline of the temporal and spatial require- ments for the homeotic genes discussed in this paper, and is based on embryonic phenotypes described here, and elsewhere (Lewis, 1978; Struhl, 1981a, Wakimoto & Kaufman, 1981; Duncan & Lewis, 1982). In this outline, the process of segmental determination is interpreted in terms of the compartment hypothesis (Garcia-Bellido, 1975; Crick & Lawrence, 1975; Lawrence, 1981). That is, during or soon after the cellular blastoderm stage, the embryo is sub- divided into a series of adjacent primordia of similar size. In response to their relative position in the anteroposterior axis, cells in each primordium initiate a particular combination of active and inactive homeotic genes (‘selector’ genes in Garcia-Bellido’s terminology). During subsequent development, the correct combinations of active and inactive selector genes are maintained continuously in the descendant cells of each segmental polyclone, thereby specifying the appropriate developmental pathway. It should be noted that this outline makes no allowance for the possibility that the products of some segmental selector genes may regulate the expression of other such genes, either in a quantitative or qualitative fashion. Such regulatory interactions may well occur (Struhl, 1982); however, they cannot be readily assayed until molecular probes for the expression of these genes are available. Consequently, the outline represents a formal description of the genetic results in which it is assumed that cross regula- tion does not occur between different members of the segmental selector genes except in those few cases where such interactions are supported by independent genetic evidence (Struhl, 1982). Similarly, this model makes no allowance for the possibility that the levels and kinds of gene products expressed by some segment- al selector genes may vary between segments in which these genes are normally active, or during the development of a single segment. Again, such variations in function in time and space may well occur (e.g., Morata & Kerridge, 1981; Kerridge & Morata, 1982; Struhl, 1981h, 1982) but they cannot be described in detail until molecular probes are available. Bearing these constraints in mind, the particular roles of the esc⁺, Pc⁺, and segmental selector genes can be inter- preted according to this model as follows.

As shown here and elsewhere (Struhl, 1981a), the esc⁺ gene product appears to be required early in development to ensure the selective expression of all of the segment-specific homeotic genes which have been tested. In the absence of the esc⁺ gene product, all of these genes appear to be expressed indiscriminately in most segments of the body. Finally, loss of the esc⁺ gene product after a discrete early period of embryogenesis appears to have relatively little effect on segmental determination (Struhl & Brower, 1982). Taken together, these results suggest that the esc⁺ gene product is part of a negative regulatory mechanism which keeps or turns off particular combinations of selector genes when the segmental polyclones are first established. For example, it could be part of a mechanism which measures the amount of a graded positional cue, and then dictates which segmental selector genes are to be inactivated, or not activated, accordingly. Alternatively, it may be involved in a subsequent step in which selector genes not to be expressed are labelled, or otherwise regulated, such that they remain inactive in all the descendant cells (see Struhl & Brower, 1982). Whatever the exact nature of the early role of the esc⁺ gene product, it is impor- tant to stress that it is part of the initial process by which the position of an embryonic cell is translated into a heritable commitment to develop as part of a particular segment.

As in the case of the esc⁺ gene product, the product of the Pc⁺ gene appears to be required in most or all segments to ensure the correct expression of the bithorax-complex genes, and possibly other homeotic genes (Lewis, 1978; Cap- devila Garcia-Bellido, 1981; Duncan & Lewis, 1982). Moreover, in the absence of the Pc⁺ gene product, selective expression of these homeotic genes yields to indiscriminate expression. However, absence of the Pc⁺ gene product during larval development appears to have the same effect as absence of the product during embryonic development; namely, the transformation of most segments towards a terminal abdominal segment (Struhl, 1981a; Duncan & Lewis, 1982). Thus, unlike the product of the esc⁺ gene, the Pc⁺ gene appears to play the same role in preventing indiscriminate expression of the segmental selector genes throughout development. These results can be interpreted as evidence that the Pc⁺ gene product is part of the mechanism which allows the correct pattern of homeotic gene expression to be inherited each time a cell divides. For example, just as the esc⁺ gene product may be required initially to label segmental selector genes which are not to be expressed, the Pc⁺ gene product may be required to propagate this label each time the genome replicates, or to prevent the ex- pression of segmental selector genes bearing such labels. Alternatively, stable expression of the correct segmental selector genes may not be controlled by a mechanism involving cis-acting labels, but rather by some form of trans-actmg regulation (e.g., once active, the products of the segmental selector genes might act as positive regulators of their own expression, creating a stable feedback loop). Accordingly, the Pc⁺ gene product may act to ensure that whatever control circuits are set up early in embryonogenesis remain stable during subsequent development. In either case, absence of the Pc⁺ gene product any time during development would result in indiscriminate expression of all seg- mental selector genes. It should be noted that this view of the role of the Pc⁺ gene is at odds with the proposition (Lewis, 1978; Duncan & Lewis, 1982) that the Pc⁺ gene product is distributed in a graded fashion both within segments and between segments, and that it directly controls the spatial pattern of expression of the μthorax-complex genes by binding with differential affinities to their promoters and thereby repressing their expression to different degrees. One weakness of this rather elaborate model is illustrated by experiments in which the number of copies of the Pc⁺ gene is raised or lowered. As discussed above, altering the number of copies of the Pc⁺ gene, whether in esc⁺ or esc⁻ animals, appears to affect the expression of the bithorax-complex when the number falls beneath two, but not when it is raised to four. Assuming that the amount of the Pc⁺ gene product is directly proportional to the number of copies of the gene, this result indicates that the expression of the bithorax-complex is not rigidly controlled by the amount of Pc⁺ gene product present. It therefore seems reasonable to adopt a simpler view of the Pc⁺ gene product as an essential component in the process of regulation for which there is an equal, minimum requirement in all segments.

Figure 12 presents a simplified view of the bithorax-complex drawn principally from the work of Lewis (Lewis, 1963,1964,1978,1981,1982; Duncan & Lewis, 1982). The complex comprises a series of genetic units each required in particular segments of the thorax and abdomen. With the exception of an isolated early function (Morata & Kerridge, 1981; Kerridge & Morata, 1982; here, see Fig. 8), none of the bithorax-complex genes appear to act in the mesothorax, suggesting that this segment represents a ground state (Lewis, 1978; Struhl, 1982). In more posterior thoracic and abdominal segments, the selective activities of particular combinations of h/tμorax-complex genes appear to specify the unique deter- mined states of each segment. Thus, in the metathorax, the activity of gene ‘0’ specifies meta- as opposed to mesothoracic development, in the first abdominal segment, the additional activity of gene T’ specifies first abdominal, as opposed to metathoracic development, and so forth till the eighth abdominal segment. Studies at both the genetic (Lewis, 1963; 1964,1978,1981,1982) and molecular levels (Bender et al. 1983) indicate that the situation is more complex than indicated by this simplistic model. However, the model may well be correct in broad outline.

As described above, the Scr⁺ gene appears to act only in the prothoracic and labial segments, specifying their particular determined states as opposed to mesothoracic and maxillary states respectively (Kaufman et al. 1980; Lewis et al. 1980fl,h; Wakimoto & Kaufman, 1981; Struhl, 1982). The role of the Scr⁺ gene in the prothorax may be seen simply as an analogue of the role of the bithorax- complex genes in the metathoracic and abdominal segments - namely, to specify a particular segmental type instead of the ground state, mesothorax (Struhl, 1982). However, it should be noted that even apparent null alleles of the Scr⁺ locus do not cause a clean transformation of the pro- to mesothorax in the embryo, but instead, transform the prothorax only partially towards the mesothorax (Wakimoto & Kaufman, 1981; this paper). This phenotype suggests the possibility that specification of the larval prothorax as distinct from the mesothorax may require the activities of other, as yet unidentified homeotic genes. The role of the Scr⁺ gene in the labial segment is also somewhat difficult to interpret. One possibility is that a set of presently unidentified genes specifies the unique developmental paths followed by the head segments in a fashion comparable to that of the bithorax-complex genes in the thorax and abdomen. Just as the Scr⁺ gene appears to raise the prothorax from an underlying, ground state towards the prothoracic level, it may similarly raise the labial segment from an underlying maxillary state. Presumably, the different responses of the labial and prothoracic segments to the loss of the Scr⁺ gene reflects the selective activity of at least one head-determining gene in the labial, but not the prothoracic, segment.

The role of the Antp⁺ gene is uncertain principally because the phenotype of Antp∼ embryos cannot be easily interpreted. As described above (see also Fig. 8), the phenotypes of Antp∼ embryos lacking the Scr⁺ gene, the entire bithorax- complex, or both, can be interpreted as evidence that the Antp⁺ gene is normally required in all three thoracic segments as well as the first seven abdominal segments. Absence of the Antp⁺ gene causes the transformation of any segment that would otherwise develop as a mesothoracic segment into a segment of novel character bearing in part a prothoracic pattern of ventral hairs and bilaterally symmetric regions of sclerotized cuticle (Wakimoto & Kaufman, 1981; here). This phenotype has been interpreted as a partial transformation of the meso- towards the prothorax, and hence, as evidence that the Antp⁺ and Scr⁺ genes have mutually exclusive functions and realms of activity, the Antp⁺ gene specify- ing meso- as opposed to prothoracic development, and vica versa, the Scr⁺ gene specifying pro- as opposed to mesothoracic development, (Wakimoto & Kauf- man, 1981). However, there are several problems with this interpretation. First, it predicts that the Antp⁺ gene should not be active in the prothorax; yet, the phenotypes of Scr⁻ and Scr⁻Antp⁻ embryos suggest that the Antp⁺ gene is normally active in the prothorax (see Fig. 8). Second, paired regions of sclerotized cuticle are not normally found in the prothorax of wild-type embryos; hence, their presence in the transformed mesothorax of Antp⁻BX-C embryos cannot be satisfactorily explained as a transformation of the mesothorax to prothorax. Third, if the Scr⁺ gene functions in an analogous fashion to the bziμorax-complex genes in specifying one segmental state (prothorax) as op- posed to the ground state (mesothorax), it should not be possible to obtain prothoracic development without the Scr⁺ gene. However, the apparent mesó- te prothoracic transformation caused by absence of the Antp⁺ gene appears identical whether or not the Scr⁺ gene is present. Finally, as described below, absence of the Antp⁺ gene in cells giving rise to the adult mesothorax does not result in meso- to prothoracic transformations, but rather in a transformation of some portions of the mesothorax into corresponding portions of the antenna. Thus, the interpretation of Wakimoto & Kaufman (1981) does not satisfactorily explain several aspects of both the embryonic and adult mutant phenotypes.

An alternative interpretation of the role of the Antp⁺ gene has been proposed on the basis of the adult phenotypes resulting from absence of the Antp⁺ gene (Struhl, 1981b; 1982). Absence of the Antp⁺ gene results in the transformation of some portions of the second leg to corresponding portions of the antenna. Conversely, dominant mutations of the Antp⁺ gene cause the reverse trans- formation of antenna to second leg (Lindsley & Grell, 1968; Kaufman et al., 1980). These results suggest that the Antp⁺ gene acts in the mesothorax to prevent inappropriate expression of genes specifying antennal development, but not in the antenna itself, where these ‘head-determining’ genes are normally required (Struhl, 1981b). Absence of the Antp⁺ gene in the pro- and metathorax can result in abnormal development of the first and third legs indicating that the Antp⁺ gene is normally required in these segments. As in Antp∼ embryos which also lack the Scr⁺ and hμμorax-complex genes, the ventral portions of all three adult thoracic segments develop alike when the Antp⁺, Scr⁺, and Ultrabithorax⁺(Ubx⁺) genes are removed together (i.e., all develop as second legs in which some portions are transformed into antenna). These results have led to the hypothesis that the Antp⁺ gene has the same role in all three thoracic segments in preventing the inappropriate expression of homeotic genes involved in speci- fying the head segments (Struhl, 1982). This interpretation is consistent with some aspects of the embryonic phenotype, notably (i) the requirement for the Antp⁺ gene in all three thoracic segments, (ii) the appearance of paired regions of sclerotized cuticle in the thoracic and abdominal segments of Scr⁻Antp⁻ embryos lacking the bithorax-compXex (these structures can be interpreted as rudimentary mouthpart structures resulting from inappropriate expression of one or more head determining genes) and (iii) the enhancement of mouthpart structures in these segments caused by removing the Antp⁺ gene from esc∼\ Scr∼ BX-C∼ embryos. However, it does not provide a satisfactory explanation for why the mesothorax appears partially transformed towards the prothorax in Antp⁻ embryos, or for why the reverse transformation of the prothorax towards the mesothorax is found in esc⁻; BX-C⁻ embryos, but not in esc⁻; Antp⁻BX-C∼ embryos (see Fig. 11). One major difficulty in interpreting these ‘prothoracic’ phenotypes is that, as described above, absence of the Scr⁺ gene does not com- pletely transform the prothorax of the embryo into the mesothorax. Although there are many explanations for this incomplete transformation, one possibility is that the genetic specification of the embryonic pro- and mesothorax differs from that of their adult counterparts. Hence, it is possible that the Antp⁺ gene has an additional early role in this particular process which is distinct from its postulated role in regulating the expression of head-determining genes. How- ever, whatever this role is, it cannot simply be to act in the mesothorax to specify meso- as opposed to prothoracic development. Thus, a consistent and satisfac- tory interpretation of the role of the Antp⁺ gene may require more information about the roles and regulation of genes controlling the difference between the meso- and prothorax of the embryo.

I thank P. A. Lawrence for his excellent advice and enthusiastic encouragement throughout the course of this work. Also, I thank G. Jürgens, J. Kimble and D. A. Goldberg for helpful dis- cussions, and M. Ashburner, E. B. Lewis, G. Jürgens, C. Nüsslein-Volhard, and E. Wieschaus for generously providing both published and unpublished mutations and chromosomal re- arrangements without which this work could not have been done. Finally, I thank the Nuffield Foundation, the Medical Research Council of Great Britain, the Thomas C. Usher Fund, the Harvard University Society of Fellows and a United States National Institutes of Health Program Project Grant (NIH 5 PO1GM29301) for financial support and research facilities.