Contrabithorax and the control of spatial expression of the bithorax complex genes of Drosophila

The bithorax complex (BX-C) is a group of homeotic genes that specify the development of thoracic and abdominal segments of Drosophila (Lewis, 1963, 1978). When the entire BX-C function is eliminated, segmentation as such is not impaired, but all segments posterior to mesothorax develop equally and uncharacteristically as a mosaic of anterior mesothorax and posterior prothorax (Lewis 1978; Morata & Kerridge, 1981; Kerridge & Morata, 1982; Hayes, Sato & Denell, 1984). Genes outside the BX-C such as extra sex comb (esc) and Poly comb (Pc) appear to act as regulators of BX-C function; loss of function of esc⁺ or Pc⁺ results in inappropriate expression of BX-C genes in all body segments (Lewis, 1978; Struhl, 1981a, 1983; Lewis, 1981,1982; Duncan & Lewis, 1982);.under these conditions all segments’ primordia are able to react appropriately to any combination of BX-C genes. However, in wild-type flies each segment develops differently and for the thoracic and abdominal ones the identity is determined in each case by a specific combination of active BX-C genes (Lewis, 1978). It thus follows that there must be a mechanism to account for the precise positional deployment of BX-C genes on each segment.

Little is known of this mechanism although genes such as esc and Pc are probably involved. There is nonetheless an important observation already emphasized in several reports (Lewis, 1981; Duncan & Lewis, 1982; Bender et al. 1983; Morata, Botas, Kerridge & Struhl, 1983; see Lawrence & Morata, 1983, for a review): there is correspondence between the genetic order of the different elements within the BX-C and the sequence of segments on which those elements act. This suggests that the linear arrangement of its genes is a component of the mechanism of spatial expression of the BX-C; that is, the genetic architecture of the complex may itself be part of the regulatory mechanism to ensure the correct spatial expression of the BX-C.

One way to study this phenomenon would be to alter the relative position of individual functions within the BX-C and to define the change -if any -in the BX-C expression. The recent cloning of BX-C DNA has allowed the molecular description of its mutations (Bender et al. 1983). One of these, Contrabithorax (Cbx¹) appears to be a change of the relative position of one of the BX-C functions; a DNA fragment of 17 kb has been transposed proximally a distance of about 40 kb and inserted in the area where the Ubx mutants (other members of the BX-C) map. There is no loss of genetic material in Cbx¹, except perhaps at the insertion point.

Thus Cbx¹ is a case of a change in the linear order of the BX-C functions. With this idea in mind we have analysed the different phenotypic components of Cbx¹. We find that the insertion of the pbx⁺ fragment in the Ubx area has two effects; a Ubx phenotype, caused by the breakpoint, and an alteration in the anatomical domain of pbx⁺ function, presumably caused by the change of genetic order.

We have used a number of chromosomes carrying mutations or rearrangements affecting the BX-C. Most of them have been isolated by E. B. Lewis and are described in several recent reports (Lewis, 1978; Struhl, 1981a; Morata et al. 1983), therefore, except when important for this paper, only a brief description of them will be given here. T(1;3) P115 is an insertional translocation of a portion of the third chromosome containing the entire BX-C to the base of the X chromosome. The two aneuploid segregants arising from this translocation, Df(3R) P115 and Dp(3;1)P115 are viable. We have used the fragment Dp(3;1)P115 to generate flies with supernumerary doses of BX-C in the X chromosome. T(2;3) PIO is a translocation carrying all the thoracic BX-C genes to the second chromosome. The breakpoint within BX-C affects the abdominal but not the thoracic genes (Morata et al. 1983). The two aneuploid segregants Df(3R) P10 (lacking the BX-C thoracic genes) and Dp(3;2) P10 are viable. Dp(3;3) bxd¹⁰⁰ is a segregant of Tp(3;3) bxd¹⁰⁰ (Lindsley & Grell, 1968) carrying part of the thoracic BX-C genes to the left arm of the third chromosome. As the duplicated fragment carries a full bx⁺ function, this has been used to discriminate the part of a given phenotype attributable to a loss of bx⁺ function. Dp(3;3) P146 is a segregant of Tp(3;3) S462 (Lewis, 1980), carrying a full set of BX-C genes to the left arm of the third chromosome. Df(3R) P9 is a deletion of all the BX-C genes. Df(3R) C4 is a deletion of the distal part of BX-C removing some abdominal genes. Dp(3;3) P5 is a tandem duplication of the entire BX-C. Throughout the text all the duplications and deficiencies are written abbreviated, e.g. Dp(3;3) P115 is called DpP115. Df(3R)P10 is DfPIO and so on. Ubx¹ and Ubx¹³⁰ are Ultrabithorax mutations (Lindsley & Grell, 1968). The abd-A^M1 is an EMS induced BX-C mutation (Sánchez-Herrero, Vernós, Marco & Morata, 1985) affecting the development of abdominal but not the thoracic segments.

The stock carrying the mutation Rgpbx was kindly provided by Mr Hilfiker and Professor Nöthiger (Zürich University). It is associated with the translocation T(3;Y) P92 generated by Ed Lewis and studied cytologically by Duncan & Kaufman (1975). The mutant syndrome is produced by some factor located in the fragment 84D10-11; 85A1-3 translocated to the Y chromosome. This fragment is viable and fertile as a duplication and apart from producing the pbx phenotype, carries the wild-type genes of dsx, Dfd and p^F (Duncan & Kaufman, 1975).

The homeotic transformations described in this paper affect the mesothoracic (T2), metathoracic (T3) and first abdominal (Al) segments. Although in general the phenotypes can be observed in the totality of the segment, they are much more clearly seen in the dorsal sides, that is wings, halteres and first tergite, consequently we have used these to measure the transformations. The change of, for example, wing into haltere is a complex phenomenon that affects the size and shape of the appendage as well as the presence or absence of specific structures such as bristles or veins. Even if structures in common such as trichomes are affected, wing trichomes being larger, they will be reduced in size if transformed towards haltere. These changes are not easily added up to quantify the transformation; however, we have observed that in general all these parameters roughly correlate with each other and then we have (except in the case of the experiments with the Rgpbx) used the number of bristles as indicative of the transformation. The increment, positive or negative, of the number of bristles can be used as indicative of the homeotic change. In the dominant transformation, the wing (that contains bristles in the anterior and posterior margins) is transformed towards haltere (that contains no bristles): the reduction in the number of bristles measures the homeotic effect. For the recessive transformation in the anterior haltere it is exactly the reverse, the presence of bristles and their number define the transformation. The transformation in Al results in Al (with over 100 bristles) partially developing as proximal part of T3 (with no bristles) so the number of bristles in Al diminishes. In all cases the values are given as the ratio of the increment in the number of bristles to the average number of the corresponding normal structure. The transformation produced by the Rgpbx was measured differently because it includes transformed structures such as postnotum that are devoid of bristles. As is apparent from the main text, it was important to distinguish the penetrance in the different regions. We considered the presence of four characteristic T2p elements along the proximodistal axis of the dorsal T3p: postnotum, axillary cord, alula and distal wing. The number of times these elements appear transformed represents the penetrance for the corresponding region and are given as percentages.

The dominant Cbx¹ mutation was found after X-ray treatment as a fly showing a dominant transformation of wing towards haltere (Fig. 1). By subsequent recombination, Lewis (1954) found that the Cbx¹ chromosome also carried a recessive mutation, pbx¹, which transforms posterior haltere towards posterior wing. That is, Cbx¹ and pbx¹ produce opposite transformations in different sets of cells. The two mutants’ phenotypes are closely linked but separable and can be expressed independently of each other, e.g. Cbx¹pbx⁺/Cbx¹pbx⁺ flies show the Cbx¹/Cbx¹ phenotype and Cbx⁺pbx¹/Cbx⁺pbx¹ only a pbx phenotype. Cbx¹ is a gain-of-function mutation (Lewis, 1978) as indicated by the following observations: (1) it is dominant and the phenotype of the homozygotes is stronger than, but of the same quality as, that of the heterozygotes; (2) the deletion of the chromosomal region where Cbx¹ is located does not have the dominant phenotype of Cbx¹, indicating that the mutation is associated with an extra function not present in normal flies.

In addition, Lewis (1954, 1963) made an important observation: the mutation Cbx¹ suppresses pbx¹ when both mutations are in cis; Cbx¹pbx¹/Cbx¹pbx¹ shows no pbx phenotype and, therefore, Cbx¹ must provide whatever wild-type function is lacking in pbx¹. However, the expression of the Cbx¹ phenotype is not affected by having pbx¹in cis. An explanation for the origin and genetic characteristics of Cbx¹ has been provided by the finding of Bender et al. (1983), that Cbx¹ is a transposition of 17 kb pbx⁺ DNA from its normal position to the Ubx area, about 40 kb towards the centromere. Subsequent recombination separated the two elements of the transposition (Fig. 2) generating a 17 kb duplication (Cbx¹) and the corresponding deficiency (pbx¹). This result explains the common origin of the two mutations. Additionally it suggests possible models to account for the dominant transformation of Cbx¹ and for the suppression of pbx¹ by Cbx¹, proposed by Lewis (1982) and Bender et al. (1983). The sequence inserted at –44 kb may be functional in T3p, therefore able to suppress the pbx¹ phenotype, and also in T2 where it promotes T3 development. Alternatively it is possible that the insertion causes the inappropriate expression of bx⁺ or Ubx⁺ which may be responsible for the dominant transformation in T2 and substitutes for the loss of pbx⁺ function in Cbx¹pbx¹ homozygous flies.

The entire phenotype of Cbx¹ can be subdivided into two sets of effects: first is the dominant transformation of wing into haltere and more rarely of mesonotum towards metanotum; this is very conspicuous and has been described in several reports (Lewis, 1955, 1963; Morata, 1975). It is variable, and affects predominantly the posterior compartment of the wing although the anterior compartment is also frequently transformed (Fig. 1). The transformation is generally restricted to the dorsal part of the T2 segment, the ventral part (second leg) is little or not affected (Lewis, 1982).

Second is a set of recessive homeotic transformations that affect segments T2, T3 and Al. These transformations are weak and cannot be detected in homozygous Cbx¹ flies, but only in trans combinations of Cbx¹ with deletions of the BX-C that eliminate the Ubx gene, or with Ubx mutations.

In segment T2, only the posterior compartment of the leg is affected and shows a slight transformation towards the posterior compartment of the prothoracic (Tl) leg. The criterion for this transformation is the appearance of the typical long bristles in the femur of the posterior Tl leg that appear in T2p. This transformation is expressed more strongly at 17°C; approximately 10% of the flies show signs of prothoracic transformation.

In segment T3 the halteres become wing-like and show typical wing structures such as marginal bristles (Fig. 3). This transformation was noted earlier by Lewis (1963, 1964), and has also been described in detail by Kauffman (1981) who showed that the expressivity of the transformation is affected by temperature; at 17°C the expression is the strongest. We have studied and quantified it in a number of genotypes; it appears in any trans combination of Cbx¹ with mutations or deletions of the Ubx⁺ gene. It does not appear in combinations with deletions of mutations of BX-C genes affecting abdominal segments (Table 1).

Finally we have observed in segment Al that the cuticle becomes thoracic-like and there is a clear reduction in the number of bristles (Fig. 3, Table 1). This transformation appears not only in trans combinations with Ubx but also with strong bxd mutations.

Since Cbx¹/Ubx^-flies show a set of recessive transformations not present in +/Ubx^-, the preceding observations indicate that Cbx¹ contains recessive mutation(s) responsible for these effects. These transformations are very similar to those produced by weak mutations at the bithorax (bx), postbithorax (pbx) and bithoraxoid (bxd) loci; bx and pbx would transform anterior and posterior haltere compartments respectively into the homologous ones in the wing, and bxd would transform first abdominal segment into a thoracic state. Also the prothoracic transformation observed in T2p leg is similar to that observed in Ubx^- cells (Morata & Kerridge, 1981; Kerridge & Morata, 1982) and in the anterobithorax (abx) mutations (Casanova, Sánchez-Herrero & Morata, 1985). The quadruple combination for abx, bx, bxd and pbx is characteristic of Ubx mutations (Lewis, 1981; Casanova et al. 1985), therefore suggesting that Cbx¹ behaves as a weak mutant allele of the Ubx* gene. Indeed the recessive transformations observed in Cbx¹/ Ubx^- are indistinguishable from those of weak viable Ubx alleles recently isolated (Sánchez-Herrero et al. 1985). Further support for the hypothesis of Cbx¹ being a mutant allele of Ubx⁺ was obtained from flies carrying a duplication of the Ubx⁺ gene and therefore expected to cover any part of the Cbx¹ phenotype attributable to a defect in Ubx⁺ function. Flies Dp P115;Cbx¹/DfP9 present a total suppression of the Ltox-like phenotype of Cbx¹ although the dominant transformation is not affected (Table 1). The same phenotype is observed in the genotype DpP10;Cbx¹/DfP9 where Dp P10 carries only one intact BX-C gene, Ubx+. These observations also establish that the two phenotypical components of Cbx¹ are independent.

The idea that Cbx¹ is a weak allele of Ubx⁺ is also enforced by the results obtained combining Cbx¹ with chromosomes partially defective in Ubx⁺ function; we find that like regular Ubx alleles (Casanova et al. 1985) the recessive phenotype of Cbx¹ can be subdivided into at least two components, attributable to the abx-bx and bxd-pbx subunits respectively. For example, the combination DfP9/Ubx¹ exhibits a strong set of abx, bx, bxd and pbx transformations (Morata & Kerridge, 1981). However, adding the fragment Dpbxd¹⁰⁰suppresses the abx-bx transformations so that Dpbxd¹⁰⁰DfPO/Ubx¹ flies only show a bxd-pbx phenotype. Similarly Dpbxd¹⁰⁰DfP9/Cbx¹ only shows bxd-pbx phenotype (Fig. 3). It is worth noticing that Dpbxd¹⁰⁰carries all the centromere-proximal DNA to the point 4 kb in the physical map (Bender et al. 1983), therefore including the insertion site of the pbx⁺ insert and also covering the site of the small insertion causing the Ubx¹ mutation. It is a general feature of all Ubx mutations that their effect extends tens of kilobases from the site of the mutation. Since the transposition of the pbx⁺ fragment has generated a breakpoint in the Ubx DNA and many breakpoints in this region have a Ubx phenotype it strongly suggests that the insertion breakpoint is responsible for the Ubx phenotype of Cbx¹.

In summary, the entire phenotype of Cbx¹ can be described as the sum of a weak Ubx phenotype and a dominant transformation of wing into haltere.

Previous observations (Lewis, 1963, 1964, 1978, 1982; Morata, 1975) indicate that the wing-to-haltere transformation is probably due to a derepression of some BX-C⁺ activity in the dorsal part of segment T2 where normally the BX-C does not function. To gain information about the nature and causes of this transformation, we have performed experiments to determine the ways in which this transformation can be altered and also the interactions of Cbx¹ with the mutation Rgpbx that causes a defect in pbx⁺ function.

(a) Factors affecting the dominant transformation

The expression of the transformation is significantly decreased in trans combinations of Cbx¹ with Ubx mutations or BX-C deletions as observed previously by Lewis (1963, 1964) and Kauffman (1981). In addition to this, we have found that this reduction occurs exclusively with deletions of the Ubx⁺ gene and not with mutations or deletions of BX-C abdominal genes; in Cbx¹/Df P10 the dominant transformation is decreased when compared to Cbx¹/+, but the latter is indistinguishable from Cbx¹/abd-A^M1 and Cbx¹/Abd-B^M1 (see Sánchez-Herrero et al. 1985 for a description of these mutants). These observations indicate that the ectopic BX-C⁺ function, responsible for the dominant transformation, resides in the Ubx region of the BX-C and that to a certain extent it requires the co-operation of the homologous chromosome. Trans interactions of this type have been reported for BX-C functions depending on the Ubx⁺ gene (Lewis, 1954, 1963; Kerridge & Morata, 1982).

We checked for the effect of additional copies of the BX-C⁺. The presence of up to five doses of the BX-C⁺ does not increase the amount of transformation (Table 2). This excludes any trans interaction between the BX-C of Cbx¹ and normal doses of the BX-C and indicates that the transformation is solely caused by the ectopic activity of the rearranged BX-C of Cbx¹. This argument is further reinforced by the result shown in Fig. 4. As described above, Cbx¹/Ubx^- flies show a reduced level of dominant transformation. This level of transformation is not affected by additional copies of BX-C⁺ that nevertheless eliminate the slight Ubx phenotype that Cbx¹/Ubx^- flies show in T3 and Al segments.

The Regulator of postbithorax (Rgpbx) is a dominant mutation found by Lewis (1968) that produces a phenotype of the same quality and often indistinguishable from that of pbx mutations, that is, the posterior compartment of T3 becomes like that of T2. The transformation affects both the dorsal and the ventral regions of the segment. The penetrance is high (Table 3); most of the halteres show at least some transformation towards wing. Cytological analysis (Duncan & Kauffman, 1975) showed it is associated with a duplication of the chromosomal fragment 84D10-11–85A1-3, well separated from the BX-C, located at 89E1-5. The nature of the connection between the mutation and the BX-C is unclear; the different location and similar phenotype to pbx suggests a regulatory role, hence the name Rgpbx. The evidence that Rgpbx interacts with the BX-C comes from dosage experiments; though Rgpbx produces a dominant pbx phenotype even in the presence of up to five doses of the BX-C⁺ (Lewis, 1968), the penetrance and expressivity of the transformation decreases with the number of doses of the wildtype genes (Lewis, 1968, also Table 3).

We have combined the Rgpbx with several partial duplications of the BX-C. Supernumerary doses of the BX-C genes included in the Dp P10 decrease the penetrance and expressivity of the Rgpbx. Not only the penetrance of the transformation is consistently lower in flies carrying three doses of pbx⁺ as shown in Table 3 but also the amount of transformed tissue in each region is clearly smaller. This indicates that the Dp P10 contains the gene(s) inactivated by the Rgpbx. It was expected because Dp P10 carries Ubx⁺ and pbx⁺ which are both necessary for the normal development of T3p.

By using Dpbxd¹⁰⁰ we can subdivide the BX-C region defined by Dp P10 in two parts; one including abx⁺, bx⁺ and Ubx⁺ DNA and the other containing bxd⁺ and pbx⁺ (Fig. 5). We find that three doses of abx⁺, bx⁺ and Ubx⁺ do not reduce the penetrance of Rgpbx (Table 3), thus suggesting that it is the bxd⁺-pbx⁺ DNA fragment that is inactivated by the Rgpbx. Since Rgpbx flies do not show any detectable bxd phenotype, it appears that only the pbx⁺ function encoded in this fragment is affected.

This finding prompted us to study the interaction of Rgpbx with Cbx¹ in which pbx⁺ DNA has altered its position. The result is a suppression of Rgpbx by Cbx¹. The suppression is virtually complete in distal structures so that the haltere becomes almost normal, but in the proximal region the transformation of meta-notum towards mesonotum usually remains (Table 3, Fig. 6). This is consistent with the dominant effect of Cbx¹ in the homologous areas of the T2 segment where it exhibits a strong transformation in the appendage but weak in the proximal structures (Lewis, 1963, 1982; Morata, 1975).

It is worth noting that the Rgpbx is able to inactivate up to five doses of pbx⁺ in normal position (Lewis, 1968) and that the presence of three doses of normally functioning Ubx⁺ (Table 3) does not reduce the penetrance of Rgpbx. Yet one dose of Cbx¹ rescues it completely in those regions where Cbx¹ acts dominantly. This experiment indicates that the presence of pbx⁺ insert substitutes for the loss of pbx⁺ function in Rgpbx. The experiment also suggests that the Rgpbx does not recognize the identity of the pbx⁺ DNA but its position within the BX-C.

The molecular phenotype of Cbx¹ consists of an insertion in the Ubx area of the BX-C of a 17 kb fragment normally located distally that carries totally or in part the DNA corresponding to the pbx⁺ DNA. There are two aspects of this rearrangement worth considering: (1) the insertion has generated an interruption (breakpoint) in the Ubx⁺ DNA and (2) the quality and new position of the inserted DNA.

Morphologically there are two distinct components of the Cbx¹ phenotype: one recessive and one dominant. These are independent; by adding intact copies of the BX-C⁺ we can suppress the recessive set of transformations without affecting the expression of the dominant transformation.

The characteristics of the recessive component resemble, although in a mild form, the phenotype of Ubx mutants: the series of compartments T2p-T3a-T3p-Ala develop like Tlp-T2a-Tlp-T2a. Normally Ubx mutants present a very strong transformation of this type and they are zygotic lethal (Lewis, 1963; Kerridge & Morata, 1982) but there are a number of weak Ubx alleles recently isolated (Sánchez-Herrero et al. 1985) that are hemizygous viable and show a phenotype very close to that of Cbx¹/Ubx^-. Lethality is not a necessary feature of Ubx mutations; it more probably reflects the strength of the mutant syndrome. Since in Cbx¹ there is a breakpoint in the Ubx area of the BX-C and many Ubx mutations have arisen from breakpoints in this region (Bender et al. 1983) we conclude that the insertion breakpoint is responsible for the Ubx phenotype. There is one example of a regular Ubx mutation caused by an insertion: Ubx¹ is associated with a Doc insert at –32 kb (Bender et al. 1983).

The dominant transformation is as Lewis (1978) has pointed out a gain of function. The phenotype in wing and mesothorax is attributable to an adventitious expression of some BX-C⁺ wild-type function normally not active in these regions.

The dose experiments (Table 2) show that supernumerary copies of the BX-C⁺ do not affect the dominant transformation, indicating that this is caused exclusively by an abnormal activity of the BX-C of Cbx¹. In this chromosome, in spite of the debilitation of the Ubx⁺ function resulting from the insertion breakpoint, there is enough wild-type BX-C activity to generate a normal segmental pattern posterior to T2.

It then appears that the rearrangement of pbx⁺ has resulted in an abnormal spatial expression of some BX-C function which affects thoracic segments. The homeotic effects observed indicate that the function involved is dependent on Ubx⁺ activity. This is also supported by the observation that the Ubx¹ mutation suppresses the dominant phenotype of Cbx¹(Cbx¹Ubx¹/+A has virtually wild-type wings, Lewis, 1963).

At this point it is pertinent to consider some recent results on the control of expression and the molecular biology of the Ultrabithorax gene (Ingham, 1984; Beachy, Helfand & Hogness, 1985). There are two transcription units, the Ubx unit and the bxd-pbx unit. While the Ubx unit encodes for a family of related protein products, the bxd-pbx unit appears to have a regulatory role (Beachy et al. 1985), presumably to ensure the presence of Ubx products in compartments T3p-Ala. Based on genetic evidence, Ingham (1984) has also proposed that the pbx+ function regulates the expression of bx⁺ (or Ubx⁺) in compartment T3p. It is possible that the abx-bx region (mostly or totally intronic, Beachy et al. 1985) have a similar regulatory role in compartments T2p-T3a. However, to account for the characteristic morphological pattern of T2p-T3a it is assumed that a specific combination of Ubx products is expressed here although some of them may be common with those expressed in T3p-Ala.

In Cbx¹, the change in position of the pbx⁺ DNA probably results in an alteration of the regulatory function of pbx* leading to inappropriate expression outside their normal anatomical domain of those members of the Ubx protein family characteristic of the T3p compartment. In addition to T3p, they are expressed in T2a, T2p and T3a. In compartments T2a and part of T2p where there is normally no BX-C activity they produce the homeotic transformation dictated by their function. In T3a there is normally Ubx⁺ activity mediated by abx⁺-bx⁺ function but in Cbx¹ flies there is an extrafunction in this compartment as indicated by the observation that in bx³Cbx¹/bx³Cbx¹ flies there is a partial suppression of the bx³ transformation at T3a.

In our view the phenotypical components of Cbx¹ are the result of two distinct and in part contradictory effects: on the one hand there is a general but slight debilitation of the Ubx⁺ activity causing the recessive and weak Ubx phenotype, and on the other hand the characteristic Ubx products of T3p are expressed adventitiously outside this compartment. This interpretation explains the genetic and developmental data on Cbx¹; the dominant transformation at T2p would be caused by the expression there of the T3p Ubx products. Homozygous Cbx¹/Cbx¹ flies exhibit a stronger dominant transformation as both third chromosomes produce extrafunction (pbx is cis-acting, Beachy et al. 1985). It also explains the suppression of pbx⁺ and of Rgpbx as essentially Cbx¹ contains a normal pbx⁺ activity with an extended realm of action but including the normal one. As the dominant transformation results from the activity generated by the pbx⁺ insert it is independent of the number of copies of the BX-C⁺ in normal topological order.

There are, however, two observations that at first sight do not fit our interpretation. The first one is that the dominant transformation of Cbx¹ is not restricted to the posterior compartment of the wing, but extends also to the anterior one (Lewis, 1963,1964; Morata, 1975). Our view that this transformation results from ectopic function derived from pbx¹+ activity is in apparent contradiction to the well-established observations (Lewis, 1963, 1978; Garc ía-Bellido, Ripoll & Morata, 1973, 1976; Morata & Lawrence, 1977) that bx and pbx mutations affect respectively and exclusively the anterior and posterior haltere compartments; a derepression derived from pbx¹+ activity should therefore affect exclusively the posterior wing compartment. Second is the partial suppression of bx³ by Cbx¹ (Lewis, 1955). Again it is difficult to see how pbx⁺-derived function should rescue the lack of function of bx⁺ which is specific to the anterior compartment. Our explanation is that bx⁺ and pbx⁺ are similar developmental functions. The evidence for this (Morata, 1982) is based on the interpretation of double mutant combinations including bx, pbx and engrailed (en). This latter mutation transforms posterior compartments of thoracic and head segments toward the corresponding anterior ones (García-Bellido & Santamaría, 1972; Morata & Lawrence, 1975,1979; Struhl, 1981b). The critical observation is that the absence of en⁺ activity reveals that the developmental effects of bx and pbx mutations are similar, although affecting different sets of cells. For example, in en¹; bx³the anterior and in en¹;pbx¹ the posterior haltere (T3) develops similarly as anterior wing (T2). In a haltere deficient for engrailed the posterior compartment develops in part as anterior even though there is pbx⁺ function. The acquisition of the characteristic posterior morphology therefore does not depend on bx⁺ or pbx⁺ but on en⁺.

In molecular terms it would indicate that the combination of Ubx products in T3a is not too different from those in T3p and if present in T2a (where en⁺ function is absent, Morata & Lawrence, 1975; Fjose, McGinnis & Gehring, 1985; Kornberg, Siden, O’Farrell & Simon, 1985) they may promote T3a development. This would also explain the partial suppression of bx³ by Cbx¹in cis. These pbx⁺-mediated Ubx products may be more efficient acting in posterior than in anterior compartments and hence the effect on T2p and T3p (total suppression of pbx¹) is stronger than in T2a and T3a (partial suppression of bx³).

There is the alternative hypothesis that the insertion of the pbx⁺ DNA produces an overall derepression of Ubx⁺ activity. It would explain the partial suppression of bx³ by Cbx¹ and by the same-argument we made above that bx⁺ and pbx⁺ are functionally equivalent it would explain the suppression of pbx¹ and of Rgpbx. However, we do not favour this hypothesis for two reasons: (1) the partial suppression of bx³ by Cbx¹in cis would demand a derepression of Ubx⁺ in T3a of Cbx¹ flies. Similarly, the suppression of the Rgpbx by Cbx¹ would imply a level of Ubx+ activity equivalent to more than 3 doses of the normal gene (Table 3). Yet we have shown that in a Cbx¹ chromosome there is less Ubx activity than the one corresponding to one dose of Ubx⁺(Cbx¹/ Ubx^- flies are of weak Ubx phenotype (Fig. 3) whereas Ubx^-/+ are essentially normal). (2) The derepression of the full Ubx⁺ activity would release the pattern determinants corresponding to Ala in T2a and T3a that should therefore exhibit abdominal transformation (which is developmentally possible as Hab mutants show it, Lewis, 1978) and this does not appear in Cbx¹ flies.

We thank R. González for technical assistance, Margrit Eich for the photography and Ursula Nóthiger and M^a Teresa Aguado for typing the manuscript. The part of the work carried out in Madrid was supported by the CAICYT and F.I.S. In Zürich by the Swiss National Foundation Nr. 3237.0-82.