Homeotic transformations of the abdominal segments of Drosophila caused by breaking or deleting a central portion of the bithorax complex

In Drosophila melanogaster, the morphological diversity of the thoracic and abdominal segments clearly depends on the normal activity of the bithorax complex (BX-C). In a series of genetic investigations spanning the past 30 years, Lewis (1951, 1954, 1955, 1963, 1964, 1967, 1978, 1981, 1982) has sought to define the individual genes within the complex and to determine their roles in specifying the unique developmental pathways followed by most of the thoracic and abdominal segments. These investigations have led to the isolation and characterization of (i) ‘viable’ mutations which cause homeotic transformations of the thoracic and abdominal segments of adult flies, and (ii) recessive lethal mutations and deletions which, when homozygous, result in homeotic transformations of the thoracic and abdominal segments of the larva. Each of these classes of mutations has provided valuable, though different, information about the wild-type functions of the complex. Thus, the many different recessive and dominant mutations affecting the adult have revealed the diversity of distinct functions encoded by the complex, while the deletions have revealed what happens when some or all of these functions are absent during embryogenesis. However, because the effects of viable and lethal mutations have generally been studied in different developmental stages, links between these two levels of analyses have depended on the few cases where genetic mosaics have been used to examine the phenotypes of lethal mutations in the adult cuticle (Lewis, 1963, 1964; Morata & Garcia-Bellido, 1976; Morata & Kerridge, 1981; Kerridge & Morata, 1982; Minana & Garcia-Bellido, 1982).

Putting together all the genetic and developmental information, Lewis (1978) proposed a general model for BX-C function which can be summarized as follows. The complete lack of BX-C function results in mesothoracic development, while full activity results in the pathway of development normally followed by the eighth abdominal segment. In between these extremes, the rest of the thoracic and abdominal segments are specified by particular combinations of active BX-C genes. This model predicts the existence of discrete genetic functions specifying the unique developmental paths followed by most of the thoracic and abdominal segments.

Until recently, most work on the bithorax complex has been focused on genes located in the centromere-proximal portion of the complex and known to control the development of the metathorax and parts of the mesothorax and first abdominal segment. Many dominant and recessive mutations, as well as rearrangement breakpoints and deletions have been obtained in these genes, and their phenotypic consequences in both the larva and adult have been described in detail. In contrast, our knowledge of genes in the more distal portions of the complex rests on the adult phenotypes caused by a small number of viable mutations, and the larval phenotypes of a few recessive lethal breakpoints and deletions (Lewis, 1978, 1981, 1982; Kuhn, Woods & Cook, 1981). Because the effects of the viable mutations on the functions of the altered genes are at present obscure (and likely to be minor) they are of limited value in assessing the normal roles of the wild-type genes. Consequently, it may be more informative to concentrate on phenotypic studies of breakpoints and deletions which are likely to greatly reduce or eliminate wild-type gene functions. Previous studies of such chromosomal aberrations have focused almost exclusively on the homeotic phenotypes observed in first instar larvae. An obvious extension of these studies would be to describe the adult transformations caused by these chromosomal aberrations and compare them with the larval transformations. However, this approach is confounded by the recessive lethality associated with virtually all breakpoints and deletions which disrupt the complex. One way around this problem is to study the phenotypes of somatic clones of cells that are mutant or deficient for the BX-C genes of interest and at the same time genetically marked. As Ripoll & Garcia-Bellido (1979) have shown, the majority of small chromosomal deletions are compatible with cell viability in somatic clones. Hence, by examining the phenotypes of mutant clones in all of the adult segments, it should be possible to obtain a piecemeal description of the effect of a particular BX-C deletion on the development of the adult fly. This method was first used by Lewis (1963, 1964) to study the effect of Ultrabithorax (Ubx) mutations in the adult cuticle and has been refined over the years (e.g., Morata & Garcia-Bellido, 1976). Because the generation of mutant clones rests on the elimination of the wild-type gene by X-ray-induced mitotic recombination, this method has the additional advantage that it permits the removal of the wild-type gene at virtually any time during development. It is therefore possible to detect gene functions which are only active during brief periods in early development (Morata & Kerridge, 1981).

Here, we describe the adult phenotypes associated with somatic clones that are mutant or deficient for genes within the central portion of the BX-C, and compare them with the phenotypes of larvae similarly mutant or deficient. Our results provide further evidence that the central portion of the BX-C contains a set of genetic functions which control the determined state of most of the abdominal segments. In addition, they show that these ‘abdominal’ functions are required in a cell autonomous fashion and that they play similar roles throughout development.

T(2;3)P10, Df(3R)Ubx¹⁰⁹ and Df(3R)P9 (called TP10, DfUbx¹⁰⁹ and DfP9 in the text) were isolated and described by E. B. Lewis (1978; 1980), and are illustrated in Fig. 1. Two other rearrangements (T(1;3)P115, and Tp(3;3)146) isolated and described by E. B. Lewis have been used. T(1;3)P115 is an insertional translocation of a portion of the third chromosome (89B9-89E6,8) containing the entire BX-C into the base of the X chromosome. Both the Dp(3;l)P115 and Df(3R)P115 segregants are viable. Tp(3;3)146 (referred to as Tp(3)S462 in Lewis, 1980) is a transposition of a larger portion of the third chromosome (89D1,2-9OD1) carrying the entire BX-C to the distal portion of the left arm of the third chromosome. Flies carrying a third chromosome bearing both Dp(3;3)P146 and Df(3R)P115 are viable and fertile.

The recessive mutations yellow (y), javelin (jv) and multiple wing hairs (mwh) affect respectively the colour and shape of bristles, and the number of hairs secreted by a single hair cell. Dp(l;3)sc^J4 is an insertion of a small portion of the distal X chromosome containing the y⁺ gene into the distal portion of the left arm of the third chromosome. M(3)i⁵⁵ is a dominant Minute mutation which reduces bristle size and causes mutant cells to proliferate more slowly than wild-type cells (Morata & Ripoll, 1975). TM1 and TM3 are third chromosome balancers. See Lindsley & Grell (1968) for further descriptions.

y; Dp(3;2)P10/+; Dp(l;3)sc^J4Dp(3;3)146 M(3)i⁵⁵Df(3R)P115/TM3 males were crossed to y; mwh jv Df(3R)Ubx¹⁰⁹/TM1 females, and their progeny X-irradiated at various times during development with 500 or 1000 rad. One eighth of the progeny from this cross should be genotypically y; Dp(3;2)P10/+; mwh jv Df(3R) Ubx¹⁰⁹/Dp(l;3)sc^J4Dp(3;3)146 M(3)F Df(3R)P115. Flies of this genotype are phenotypically normal (with the exception of the Minute (3)i^S5 phenotype) and hence distinct from the remaining progeny which are marked with the dominant mutations associated with the TM1 and TM3 balancer chromosomes, or with the dominant Ubx phenotype associated with haploinsufficiency for this gene. Fig. 2 illustrates the chromosomal constitution of these flies and shows that a mitotic recombination event occurring on the left arm of the third chromosome proximal to the M(3)i locus can generate a y; Dp(3;2)P10/+; mwh jv Df(3R) Ubx¹⁰⁹/mwh jv Df(3R)P115 daughter cell. All of the descendents of this cell would constitute a clone of cells hemizygous for the DpP10; DfUbx¹⁰⁹ deletion and marked by the y, mwh; and jv mutations. These cells also carry two copies of the M(3)C gene and therefore are fast-growing relative to the surrounding cells bearing the M(3)i⁵⁵ mutation (Morata & Ripoll, 1975).

The procedure is essentially the same as that for DpP10; DfUbx¹⁰⁹ clones except that DfP10 is used instead of DfUbx¹⁰⁹.

To control for the possibility that the homeotic phenotypes observed in mutant clones might have been caused by extraneous genetic factors unrelated to the loss of BX-C function, we produced TP10 clones exactly as described above except for the presence of an additional copy of the entire BX-C. To do so we irradiated larvae of the genotype y; Dp(3;l)P115/y; Dp(3;2)P10/+; mwh jv Df(3R)P10/ Dp(l;3)sc^J4Dp(3;3)146 M(3)i⁵⁵Df(3R)P115. Marked clones arising in these flies are genotypically identical to those arising in the TP10 experiment except that they also carry a wild-type copy of the BX-C on the X chromosome thereby covering the loss of BX-C function resulting from the TP10 breakpoint.

Eggs from appropriate parents were collected on agar plates over 12 h periods, and allowed to mature for 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 preparations 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, pharate first instar larvae were dissected out from the vitelline membrane prior to fixing in glycerol-acetic acid.

Each BX-C aberration was examined in the homozygous state. For example. DpP10; DfUbx¹⁰⁹ first instar larvae were obtained from a DpP10/DpP10; DfUbx¹⁰⁹/TM3 stock. Homozygous mutant larvae were recognized by a characteristic mutant phenotype distinct from the wild-type phenotype and were found in approximately the correct frequency relative to phenotypically wild-type larvae (e.g., about one quarter of the progeny from the DpP10; DfUbx¹⁰⁹/TM3 stock had the characteristic mutant phenotype showed in Fig. 8B).

A diagram of the genetic map of the BX-C (modified from Lewis, 1978) is shown in Fig. 1. The leftmost (centromere-proximal) portion of the complex contains at least five genes, anterobithorax (abx), bithorax (bx), Ultrabithorax (Ubx), bithoraxoid (bxd), and postbithorax (pbx), which are required for the normal development of particular portions of the meso- and metathorax, and the first abdominal segment (Lewis, 1963,1964,1967,1978,1981,1982). In general, abx, bx, bxd, and pbx mutations are viable when homozygous and transform particular anterior and posterior compartments of these segments towards the corresponding compartment in the adjacent anterior segment (Lewis, 1963, 1964; Morata & Garcia-Bellido, 1976; Lawrence, Struhl & Morata, 1979; Kerridge & Sang, 1981). Ubx mutations behave as if they lack the functions of the abx, bx, bxd, and pbx genes as well as an additional postprothorax (ppx) function (Lewis, 1963,1964,1978, Morata & Garcia-Bellido, 1976; Morata & Kerridge, 1981; Kerridge & Morata, 1982). For convenience, we will refer to the subset of BX-C genes which are inactivated by recessive lethal mutations of the Ubx gene as the ‘thoracic’ genes.

Most, if not all, of the BX-C genes which lie to the right of the thoracic genes appear necessary for the correct development of the abdominal segments and terminalia (Lewis, 1978,1981,1982; Duncan & Lewis, 1982). Here, we examine two rearrangements of the central portion of the BX-C which alter the function of a restricted subset of these ‘abdominal’ genes. The first rearrangement is the breakpoint associated with T(2;3)P10 (referred to subsequently as TP10) which is an insertional translocation that carries a centromere-proximal portion of the BX-C as well as an adjacent portion of the third chromosome tó the left arm of the second chromosome, leaving behind a complementary deletion (Lewis, 1978, 1980; see Fig. 1). The TP10 breakpoint within the complex behaves as a recessive lethal mutation which partially or completely inactivates at least one abdominal gene. The second rearrangement is a synthetic deletion, Dp(3;2)P10; Df(3R)Ubx¹⁰⁹ (referred to subsequently as DpP10; DfUbx¹⁰⁹), which begins at the TP10 breakpoint and extends rightwards until the distal breakpoint of DfUbx¹⁰⁹: this deletion eliminates one or more abdominal genes lying to the right of the TP10 breakpoint (Lewis, 1978, 1980, 1981,1982; see Fig. 1).

To clarify the functional relationships between the thoracic and abdominal genes we have constructed several genotypes involving mutations and break-points from both sets of genes. We find that recessive lethal Ubx mutations that completely inactivate the thoracic genes are fully complemented by recessive lethal breakpoints and deletions affecting the abdominal genes. For example, flies carrying the trans combination of Ubx¹³⁰ and TP10 are viable and phenotypically normal aside from the dominant Ultrabithorax phenotype resulting from haplo-insufficiency of the Ubx gene. Similarly, flies carrying the trans combination of Ubx¹³⁰ and DpP10; DfUbx¹⁰⁹ are viable and phenotypically normal (except for the dominant Ultrabithorax phenotype). Thus, the thoracic and abdominal genes appear to constitute separably mutable and functionally autonomous subgroups of the BX-C.

When either homo- or hemizygous, both TP10 and DpP10;DfUbx¹⁰⁹ cause lethality just before or after hatching of the first instar larva. Hence, to study the adult phenotypes associated with each rearrangement, we have used X-ray-induced mititoc recombination to generate somatic clones of mutant cells (Fig. 2).

Clones of both genotypes were produced at different developmental stages ranging from blastoderm to pupariation, and found in all cuticular structures of the fly which can be marked by the y,jv, and mwh mutations. Those found in the cephalic and thoracic segments as well as in the anterior compartment of the first abdominal segment and the terminalia (the genitalia and analia) differentiate normally. Moreover, they are similar both in size and frequency to clones generated in control flies. Thus, neither aberration appears to alter the development of imaginal cells giving rise to the adult head, thorax, terminalia, or first abdominal segment (anterior compartment). In contrast, clones in the remaining abdominal segments show homeotic transformations. Before considering these phenotypes, it should be noted that even though each abdominal segment may be formed by an anterior and posterior compartment (Korn-berg, 1981), the posterior compartments construct disproportionately small parts of each segment which generally cannot be marked by the y, jv, or mwh mutations. Our results therefore pertain only to the anterior compartments of the abdominal segments.

As described above, clones found in the first abdominal segment appear normal. Virtually all of the clones in abdominal segments 2–5 are clearly transformed to segment 1 (Table 1; Figs 3, 4A, B); the few cases of untransformed clones are probably rare recombination events producing marked clones not deficient for the BX-C. The transformation appears to be complete in both the dorsal (tergites) and ventral (sternites) regions although some very rare cases (3) were found in which the transformation seems to be partial. Clones found in the tergites of segments 6 and 7 in females and of segment 6 in males (wild-type males do not have a seventh tergite) show a phenotype different from that found in more anterior segments. In most cases they produce only hairs but sometimes one or two bristles are also found. However, these hairs and bristles are different from those normally found in the first abdominal segment; they appear more like those of segments 2–5. In general the clones in tergites 6 and 7 are smaller than those produced in more anterior abdominal segments and are sometimes associated with abnormal cuticle differentiation. Since the number of clones found in these segments is low (Table 1), it is very likely that many do not develop or are eliminated. Clones in the sternites of segment 6 behave like sibling clones in the sternites of segments 2–5; namely they appear transformed into the first abdominal segment in that they do not produce bristles (Fig. 4C, D) and have a pattern of hairs like that of the first sternite. Frequently they extend from the sternite to the pleura (Lawrence, Green & Johnston, 1978). Note that clones in the sixth sternite behave differently from clones in the sixth tergite which differentiate structures appropriate to abdominal segments 2–5.

Three additional features of these clones are worthy of note. First, only cells marked with the y, jv, and mwh mutations, and hence unquestionably belonging to the mutant clone, show homeotic transformations. Taken together with the finding that virtually all the clones found in abdominal segments 2–7 are clearly transformed, this result indicates that the mutant phenotype is expressed in a cell autonomous fashion. Similar results have been obtained with mutations of the thoracic genes of the BX-C (Lewis, 1963, 1964; Morata & Garcia-Bellido, 1976) suggesting that BX-C genes are generally required cell by cell in the segments in which they act. Second, clones in the tergites of segments 2–5 differentiate fine short bristles characteristic of the first abdominal tergite which are thinner and longer in the centre of the tergite than they are at the sides. This gradual change in bristle morphology according to mediolateral position is also found in the normal first abdominal segment. Similarly, another first tergite structure differentiated by the clones, the archlike putative apodeme, is also found laterally in segments 2–5, just as it is in segment 1 (e.g., Fig. 4B). These observations suggest a point-by-point spatial correspondence between the cuticular derivatives of the first abdominal segment and the derivatives of the next four abdominal segments. Similar correspondences have been observed between the derivatives of other adult segments (Postlethwait & Schneiderman, 1971; Morata & Garcia-Bellido, 1976; Morata & Lawrence, 1979; Lawrence et al. 1979; Struhl, 1981b) and have led to the supposition that the unique patterns of each segment reflect different interpretations of a common field of positional information reiterated in each segment. Third, wild-type bristles not belonging to the clones are frequently present inside the clone surrounded by marked, mutant hairs (Fig. 4B). Conversely, mutant bristles belonging to the clone sometimes appear to have migrated outside of the clone territory (Fig. 4A). These results indicate that cells set aside to become bristles often move away from sibling, hair-forming cells during the process of bristle patterning. Similar observations have been made for the sex comb teeth (Tokunaga, 1962) and tarsal bristles (Lawrence et al. 1979) of the legs and for wild-type bristles found in the abdominal tergites (Garcia-Bellido & Merriam, 1971). Yet, here, the bristle-determined cells intermix with surrounding cells that have a different segmental specification. This result supports previous suggestions that affinities are similar between cells in neighbouring abdominal segments (Wright & Lawrence, 1981).

A special experiment was performed to find out the temporal limit of the requirement for the deleted genes. The mutant clones were produced at different times after pupariation at intervals of 4h. Even those generated as late as 16 h after pupariation show the characteristic segmental transformations observed in clones generated during larval development, indicating that the normal gene functions are required well into the pupal period.

As in the previous experiment clones in abdominal segment 1 appear normal, in segments 2–5 they appear transformed towards segment 1, and in segments 6 and 7 they appear transformed towards abdominal segments 2–5. However, in contrast to the results with DpP10; DfUbx¹⁰⁹ clones, the transformations caused by TpP10 clones appear incomplete. For example, many of the clones in both the tergites and sternites of segments 2-5 do not show a detectable transformation to segment 1, or differentiate both transformed and untransformed structures (Table 2; Fig. 5A). Note that the degree of incompleteness of the transformation depends on the particular segment in which the clone occurs: those in segment 2 are nearly always transformed, whereas those in segment 5 are almost never transformed (Table 2). In parallel with the DpP10; DfUbx¹⁰⁹ experiment, TP10 clones in abdominal tergites 6 and 7 do not develop bristle or hair patterns characteristic of the first abdominal segment, but instead differentiate patterns resembling those of abdominal segments 2–5 (Fig. 5B). However, as in the more anterior abdominal segments, the transformations are not complete (Table 2).

It is possible, in principle, that some of the homeotic transformations we observed in mutant clones could be due to extraneous genetic factors present on the chromosomes used for the mosaic analysis. To control for this possibility we generated clones in flies which were genotypically identical to flies in the TP10 experiment except that they carried an extra dose of the entire BX-C (see Materials and Methods). In this case, the y,jv, mwh, M(3)i⁺ clones are expected to be normal in all segments if all the effects observed in previous segments were due exclusively to the partial loss of BX-C function. In a sample of 67 flies irradiated as late second and early third instar larvae (72–96 h after egg laying) we found 20 clones in the second, third, fourth, and fifth tergites, 11 clones in the sixth and seventh tergites, and 17 clones in the sternites and pleura. None showed any sign of homeotic transformation.

The TP10 and DpP10; DfUbx¹⁰⁹ rearrangements consist of the centromereproximal fragment DpP10 carrying the thoracic genes and the centromere-distal fragments resulting from the DfP10 or DfUbx¹⁰⁹ deletions which remove all of the thoracic genes and partially or completely inactivate one or more of the abdominal genes. In order to assess the independent contributions of the BX-C material present in each of these fragments, we examined first the phenotypes caused by DfP10 and DfUbx¹⁰⁹ alone, comparing homozygous larvae with wild-type and BX-C∼ larvae (Figs 6, 7). Then, we compared these phenotypes with the phenotypes of homozygous DfP10, DfUbx¹⁰⁹, and BX-C⁻ larvae which also carried two copies of the DpP10 fragment (Figs 7, 8). Detailed descriptions of these phenotypes are provided in the legends to Figs 6–8. Here we summarize the principal results derived from this analysis.

The head and thorax of TP10 larvae are indistinguishable from that of wild type; however the first eight abdominal segments show segmental transformations (Figs 6A, 7A). These are (i) the anterior halves of the first and second abdominal segments appear similar to the anterior half of the normal first abdominal segment, but the posterior halves of these segments appear thoracic in character, and (ii) beginning in segment 3 and proceeding until segment 8, the segments display a graded sequence of novel patterns which appear inter-mediate between the first abdominal segment and the normal segment (compare Figs 6A, 8A). Because the characteristic patterns of the wild-type abdominal segments differ from each other principally in the arrangement of otherwise similar hairs, we cannot say whether the intermediate patterns are mosaics of cells of different segment types, or alternatively, whether they comprise cells which themselves have intermediate segmental characteristics. To a first approximation, the phenotype of TP10 larvae appears to reflect the additive effects of the independent contributions of the DpP10 fragment and the centromere-distal fragment resulting from the DfP10 deletion. Thus, in the absence of all the BX-C genes, most of the abdominal segments develop like mesothoracic segments (Fig. 6C). The DpP10 fragment appears to raise most of the abdominal segments from the mesothoracic state towards the first abdominal state (Fig. 8C), whereas the distal fragment complementary to the DfP10 deletion plays no apparent role in the development of the first abdominal segment, but appears to raise abdominal segments 2-8 from the mesothoracic state to a state intermediate between that of the mesothorax and that of the normal segment (Fig. 6B). Together, the fragments appear to raise all the abdominal segments to intermediate states between that of the first abdominal segment and that of the normal segment (Fig. 8A).

The phenotype of DpP10; DfUb¹⁰⁹ larvae differs from that of TP10 larvae chiefly in the following respects: (i) abdominal segments 3 – 7 appear similar to segments 1 and 2 (i.e., the anterior halves develop as in the first abdominal segment, while the posterior halves develop as in the meso- or metathorax), and (ii) the pattern of ventral hairs in segment 8 appears further transformed towards the normal first abdominal segment (compare Figs. 8A, B). As in the case of TP10 larvae, the phenotype of DpP10: DfUbx¹⁰⁹ larvae appears to a first approximation to reflect the additive contributions of the two BX-C fragments involved (see Figs. 7B, C and 8B, C).

Bearing in mind that we probably examined only the anterior compartments of the adult abdominal segments, we find that the larval and adult transformations are generally in very good agreement. Neither rearrangement affects the development of the head, thorax, anterior first abdominal segment, or terminalia. In both TP10 larvae and adults, abdominal segment 2 developed like segment 1, segments 3 – 5 developed as a graded series of novel segment types intermediate between segment 1 and segments 2 – 5, and segments 6 and 7 appeared as intermediates between segments 2 – 5 and segments 6 and 7. Similarly, in both DpP10; DfUbx¹⁰⁹ larvae and adults, the anterior portions of segments 1 – 5 developed as in segment 1, as did the ventral anterior portion of segment 6. In addition the dorsal portions of segments six and seven appear to have characteristics of abdominal segments posterior to the first abdominal segment.

We have described the larval and adult phenotypes resulting from a breakpoint and an internal deletion in the central portion of the BX-C. Here, we discuss our results in the light of current views of the organization and function of the different genes of the complex.

As described in Fig. 1 and Results, the bithorax complex can be subdivided into centromere-proximal and centromere-distal portions which appear to carry functionally distinct sets of genes. The set of genes located in the proximal portion appears to be involved principally in the development of the meso- and metathorax, and the first abdominal segment. This set of genes includes the abx, bx, Ubx, bxd, and pbx genes, and possibly the ppx gene (if it exists as a separate gene). Because mutations in all of these genes have the general property that they transform portions of the meso- and metathorax, and the first abdominal segment into corresponding portions of an anteriorly situated thoracic segment, we refer to them as the ‘thoracic’ genes. Mutations of the Ubx gene behave as if they partially or completely inactivate all of the thoracic genes and hence allow us to define their realm of action. In adults, all of the Ubx mutations which have been studied to date (Lewis, 1963,1964; Morata & Garcia-Bellido, 1976; Morata & Kerridge, 1981; Kerridge & Morata, 1982; Minana & Garcia-Bellido, 1982; Sanchez-Herrero & Morata, unpublished) affect at most the posterior compartment of the mesothorax, the entire metathorax, and the anterior compartment of the first abdominal segment; no phenotype is observed in any other segment. It is not yet clear why the limits of Ubx gene function in the adult coincide with anteroposterior compartment boundaries within segments rather than with segmental boundaries. However, mutations in most of the subsidiary thoracic genes (e.g., bx, bxd, pbx, and ppx) seem to affect only anterior or posterior compartments within a segment, suggesting that BX-C genes may be deployed compartment by compartment, rather than segment by segment. To a first approximation, the phenotype of Ubx larvae reflects the adult phenotype; however, there are also slight, but significant, segmental transformations in abdominal segments 2 – 7 (Lewis, 1978; Struhl, submitted). Thus, at least some thoracic genes appear to act in cells giving rise to most of the abdominal segments of the larva.

The ‘abdominal’ genes present in the complementary distal portion of the complex are considerably less well characterized than the thoracic genes. The majority of mutations affecting these genes are dominant or weak recessive mutations which appear to partially deregulate, or partially inactivate, one or more genes (Lewis, 1978,1981,1982; Kuhn et al., 1981). Both rearrangements we have studied retain all of the thoracic genes but break in or delete some of the abdominal genes. Though the breakpoint itself may not completely inactivate any of the abdominal genes (Lewis, 1978,1981,1982; here, Fig. 6A,B), it is virtually certain that the deletion does. Here we show that in both the larva and adult, neither aberration affects the development of the head, thorax, or anterior compartment of the first abdominal segment. However, both rearrangements cause anteriorly directed segmental transformations in all the remaining abdominal segments (excluding the terminalia). These results establish that abdominal genes of the BX-C are not required for the normal development of the head and thorax.

Thus, in the adult, the thoracic and abdominal genes appear to be required in mutually exclusively domains of the thorax and abdomen. Moreover, these domains appear to meet in the middle of the first abdominal segment, possibly at the anteroposterior compartment boundary. Though the distinction is not quite so clear in the larva, the anteroposterior compartment boundary in the first abdominal segment again seems to represent a boundary of some kind between thoracic and abdominal gene function (Struhl, submitted).

One perplexing result we obtained is that in a few cases the segmental transformations caused by either the breakpoint or the deletion appeared to be out of register in the dorsal and ventral derivatives of a given segment (e.g., clones of the deletion caused the sixth tergite to be transformed towards the tergites 2 – 5, whereas they caused the sixth sternite to be transformed towards sternite 1). We do not understand why these differences occur. They may reflect that the entire segment is of intermediate segment type which has properties of different segments in different regions. Alternatively, it is possible that the dorsal and ventral realms of action of some of the abdominal genes may be out of register by a segment.

The segments of the adult abdomen develop in a markedly different fashion from the segments of the adult head and thorax. Whereas the latter derive from imaginai disc cells which proliferate during most of the larval period, the abdominal derivatives are formed by the descendents of small nests of histoblasts which form part of the larval epidermis and do not divide until the pupal period. Clones of cells hemizygous for the deletion consistently showed segmental transformations even when induced 16 h after the onset of the pupal period at which time the histoblasts would already have undergone several rounds of proliferative divisions (Madhavan & Madhavan, 1980). Moreover, most if not all of the cells belonging to the clones were transformed whereas surrounding BX-C⁺ cells were invariably normal. These results establish that the deleted BX-C genes are normally required in proliferating histoblasts during the pupal period, moreover, they indicate that these genes are required autonomously in all of the cells in the affected segments which could be marked by the clones. Finally, the larval and adult phenotypes of the breakpoint and the deletion were in close accord, suggesting that the broken or deleted genes in each case played similar roles during embryogenesis and metamorphosis. Thus, in terms of their cellular and temporal requirements, the abdominal genes appear to be analogous to the thoracic genes which are required continuously and autonomously in all cells of the appropriate compartments and segments.

Fig. 9 shows a simplified view of the organizaton of the BX-C genes along the chromosome and their corresponding realms of action along the body axis, and is intended to illustrate the main features of Lewis’s model for the complex (Lewis, 1978, 1981, 1982). These are: (i) that the BX-C consists of a series of discrete genes which are active only in particular segments, (ii) that the order of these genes on the chromosome directly corresponds with the anteroposterior order of the segments in which they are active, (iii) that a given gene (say gene 4) is active in all segments posterior to the most anterior segment in which it is active (i.e., abdominal segments 4 – 8), and (iv) that the development of each segment is dictated either by the particular combination of active BX-C genes, or by the highest numbered gene which is active (e.g., gene 4 in abdominal segment 4). For the sake of the argument, let us assume that the TP10 breakpoint inactivates gene 2 and the DpP10; DfUbx¹⁰⁹ deletion inactivates genes 2-5, but that all other BX-C genes function normally. Fig. 10 shows the combinatorial code words of active and inactive BX-C genes which would be present in each segment. In the case of the breakpoint, all segments anterior to the second abdominal segment would have the normal code word, segment 2 itself would have the normal code word for segment 1, and the remaining posterior segments would have novel code words not normally found in any segment. If it is the combination of active genes rather than the highest numbered gene which is active that determines segmental stage, one might predict that all segments anterior to abdominal segment 2 would be normal, segment 2 would develop like segment 1, and the remaining posteriorly situated segments would be transformed into novel segment types intermediate in character between what they would normally be and the first abdominal segment. The predictions for the deletion would be similar except that abdominal segments 2 – 5 would all develop like the first abdominal segment. In general these predictions are borne out by our results in both the larva and the adult thereby providing further support for the general features of Lewis’s model. However, when examined in detail, they raise several questions about more specific features of the model some of which we consider below.

First, a major unresolved question is whether the BX-C genes work in a combinatorial fashion as described above, or alternatively, in a hierarchical fashion such that the highest numbered gene active in a segment dictates the development of that segment irrespective of the activity of the lower numbered genes. Several aspects of our results can be interpreted as support for a combinatorial mechanism. One of the most striking results we observe is that all of the rearrangements we describe cause abdominal segments which one would predict are expressing a novel combination of BX-C genes to develop into intermediate segment types. Moreover, we show that the particular intermediate pattern formed in a given segment depends on which, and how many, of the BX-C genes are present. These additive phenotypes suggest the possibility that each BX-C gene normally active in a segment influences the development of that segment in a qualitative fashion, and that the combination of these discrete influences normally dictates the particular pattern differentiated by the segment. Finally, we find that in both the larva and the adult, the intermediate segments form a graded series in which each mutant segment appears more like the normal abdominal segment than the preceding anterior segment. These phenotypic gradations may reflect the fact that each segment would be expected to approach the normal combination of active genes more closely than the preceding anterior segment.

If one could make the simplifying assumption that breakpoints and deletions of the BX-C affect the expression of only the genes actually broken or deleted, it is clear that these results would rule out a hierarchical model for BX-C gene function. However, this assumption is not valid because several examples of position effects between mutations in separable BX-C genes have been described (Lewis, 1954,1955,1963,1964,1978,1981,1982). Thus an alternative possibility is that the BX-C genes function in a hierarchical fashion, but that breakpoints or deletions lower the activity of distally situated genes in a polar fashion which decreases with distance. Accordingly, intermediate phenotypes would result from partial activity of the highest numbered gene which is active in a given segment. Fig. 10 illustrates how all of the particular phenotypes which can be accounted for by a combinatorial model can be equally well explained by a hierarchical model in which chromosomal rearrangements cause position effects on distal genes.

Recently Lewis (1981) has argued that the BX-C genes may function combinatorially by each specifying a particular subset of pattern elements; accordingly, the unique pattern of each segment would reflect a literal reading out of the cuticular elements specified by the active BX-C genes. If this view is correct, one would predict that the segmental patterns specified by two non-overlapping fragments of the complex would be strictly additive. However, we find that many of the intermediate segmental patterns resulting from TP10 and DpP10; DfUbx¹⁰⁹ are not strictly additive element by element (e.g., compare the eighth abdominal segments of DfP10, DpP10; DfP9, and TP10 larvae: Figs 6B, 8C and 8A). We therefore argue that the BX-C genes do not act by specifying subsets of discrete pattern elements, but that they discretely influence the developmental pathways of each segment as a whole. Hence the final patterns may be viewed as Gestalten with field properties rather than as assemblages of particular combinations of discrete elements.

We thank P. A. Lawrence for his constructive criticisms of the manuscript and E. B. Lewis for generously providing the many BX-C aberrations we have used. G. M. and J. B. gratefully acknowledge support from Comision Asesora del Descuento Complementario; S. K., from the European Molecular Biology Organization; and G. S. from the Thomas C. Usher Fund, the Harvard Society of Fellows, and a United States National Institutes of Health Program Project Grant (NIH 5 PO1 GM29301).