Function of an Ultrabithorax minigene in imaginal cells

The Ultrabithorax (Ubx) gene belongs to the group of homeotic genes in Drosophila whose function it is to control the morphogenesis of various position-specific structures in the embryo, the larva and the adult (Lewis, 1963, 1978; Sánchez-Herrero et al., 1985a; Teugels and Ghysen, 1985; Hooper, 1986; Bienz and Tremml, 1988). It is one of three genes located within the bithorax complex (Sánchez-Herrero et al., 1985b) and it extends over more than 130 kb (Bender et al., 1983). Ubx transcripts first appear in the early embryo within a domain restricted along the anteroposterior axis (Akam and Martinez-Arias, 1985). Although the pattern of Ubx expression becomes modified subsequently, its limits along the axis are maintained throughout development (Akam, 1983; White and Wilcox, 1984, 1985a; Beachy et al., 1985).

The main realm of Ubx function comprises parasegments (ps) 5 and 6 in the epidermis (Lewis, 1963, 1978; Morata and Garcia-Bellido, 1976; Morata and Ker-ridge, 1981; Struhl, 1981, 1984; Hayes et al., 1984; Sánchez-Herrero et al., 1985a; Casanova et al., 1985) and in the nervous system (Teugels and Ghysen, 1985; Weinzierl et al., 1987). Minor effects of Ubx mutation can be observed also in more posterior segments of the larval epidermis (Lewis, 1978; Bender et al., 1983; Struhl, 1984). Internally, Ubx exerts a control function in abdominal segments 1 –5 in the larval somatic mesoderm (Hooper, 1986), in ps7 in the embryonic visceral mesoderm (Bienz and Tremml, 1988) and, indirectly, in the endoderm (Immerglück et al., 1990). Absence of Ubx function is not lethal for individual cells, though animals lacking Ubx function usually die as young larvae (Lewis, 1978), perhaps due to cumulative defects in different germ layers. However, clones of homozygous Ubx⁻ cells survive till adulthood and form adult cuticular structures (Lewis, 1963; Morata and Garcia-Bellido, 1976; Morata and Kerridge, 1981; Minana and Garcia-Bellido, 1982). From studies of these Ubx⁻ clones, it was concluded that, in the absence of Ubx function, adult cuticular structures belonging to ps5 and 6 (T2p-T3a-T3p-Ala) were transformed into those of ps4 (Tlp-T2a; reviewed by Sánchez-Herrero et al., 1985a; T and A, thoracic and abdominal segments; a and p, anterior and posterior compartments). As a rule, Ubx function is required continuously throughout development (Morata and Garcia-Bellido, 1976), except for T2p and, to some extent, T3p in which Ubx function becomes dispensable after an early embryonic stage (Morata and Kerridge, 1981; Casanova et al., 1985). This early Ubx requirement in T2p and T3p was termed postprothorax (ppx⁺) function (Morata and Kerridge, 1981).

In an attempt to reconstruct the embryonic Ubx expression pattern, we identified three key control regions in remote areas of the Ubx gene which, upon combination, are capable of directing a Ubx-like β-galactosidase (β-gal) expression pattern in transformed embryos (Müller and Bienz, 1991). We found that these control regions, if linked to a Ubx cDNA in a minigene construct, confer Ubx function in the larval epidermis. Here, we ask whether these control regions are also capable of supporting β-gal expression in imaginal disc cells. We test whether the same minigene can provide Ubx function in the adult epidermis.

The basic β-gal and Ubx minigene constructs used for transformation were previously described (Bienz et al., 1988; Müller and Bienz, 1991; Müller, 1991). Details of maps are available on request.

cn; ry⁴² flies were used for P-element transformation, and transformant lines were obtained as described (Bienz et al., 1988). In 2 of the 11 newly isolated BPuA lines (T2 and T8), the minigene transposon is inserted in the chromosomal Ubx gene, disrupting endogenous Ubx function (to be described elsewhere). The U12 transposon is inserted near the fushi tarazu gene (D. Yen and J. M., unpublished results); however, none of the other transposons analysed is inserted near a homeotic gene complex (U81 maps to the third chromosome at a distance from the Ubx gene of at least 50 Morgans; T7 and T10 map to the second chromosome).

Recombinant chromosomes bearing a minigene transposon and the Ubx¹ mutation were constructed (using a ry⁴²Ubx¹ recombinant in which at least 3/4 of the original Ubx¹ chromosome was replaced), and balanced strains were established by standard genetic procedures. These were crossed with balanced strains containing various Ubx mutations (see text), and non-balancer offspring flies (Ubx homozygotes or transheterozygotes bearing a minigene) were analysed. Control crosses were done with a Ubx¹/TM1 strain whose original Ubx¹ chromosome was “cleaned up” by recombination (J. C.-G., unpublished). All Ubx mutations have been described (Lewis, 1963, 1982).

To determine the lethality of Ubx¹ mutants, crosses were done, as described above, and homozygous Ubx larvae, recognisable by their extra spiracles (Lewis, 1978), were counted at various stages. We could not find any of these homozygous Ubx⁻ larvae which survived beyond the early first instar. Homozygous Ubx larvae bearing a minigene can be recognised by their denticle belt pattern (Müller and Bienz, 1991).

Inverted anterior halves of third instar larvae (typically homozygous for their transposon insert) were prepared in Ringer’s solution, rinsed in phosphate-buffered saline (PBS) and fixed for 2 minutes in 1% glutaraldehyde in PBS. After thorough rinsing in PBS, they were incubated in β-gaI staining solution (Bienz et al., 1988) for two hours. Stained discs were separated from the larval remains, dehydrated, passed through methylsalicylate and mounted as described for embryos (Bienz et al., 1988). The location of β-gal staining in posterior parts of discs was confirmed by β-gal staining of larvae obtained from a cross between ABP transformants with an engrailed/ β-gal transformant line (in the latter, β-gal staining is restricted to posterior compartments; Hama et al., 1990). Staining with a monoclonal antibody against Ubx protein (White and Wilcox, 1984) was done as described (Castelli-Gair et al., 1990). Analysis of adult cuticle was done by following standard procedures (Morata and Garcia-Bellido, 1976).

In this study, we use the same Ubx fragments from the PBX, ABX and BXD control regions whose maps and regulatory properties we have described (Millier and Bienz, 1991). Briefly, the PBX and BXD fragments are derived from remote regions upstream, the ABX fragment from an intronic region downstream of the Ubx transcription start site, as sketched out in Fig. 1. Of these fragments, PBX and ABX direct an early pattern of β-gal expression which is restricted to the Ubx domain along the anteroposterior axis of the embryo and strongest anteriorly within ps6 and ps5, respectively. The BXD fragment is activated at a later embryonic stage to direct a pattern extending from head to tail which resembles the Ubx expression pattern in abdominal segments.

Most transformants used for this analysis have been isolated previously (including the minigene-bearing transformant lines U12 and U81; Müller and Bienz, 1991). In addition, we constructed a second type of minigene in which the ABX fragment is placed downstream of the coding region (BPuA; Fig. 1). We isolated 11 individual lines of which 5 showed a strong CbrMike phenotype (Fig. 3D), like the U12 and U81 lines, a phenotype that we presumed to be caused by minigene-derived ectopic Ubx⁺ activity (Müller and Bienz, 1991). We took this phenotype as an indication of high minigene efficacy and chose two of these lines (T7 and T10) for further analysis.

We stained imaginal discs from late third instar larvae of PBX, ABX and BXD transformants for β-gal activity. We found that neither the Ubx proximal promoter by itself (not shown) nor the BXD fragment linked to the Ubx proximal promoter (Fig. 2A,B) is capable of directing any substantial β-gal expression in imaginal discs. In contrast, both the ABX (Fig. 2C-F) and the PBX fragment (Fig. 2G-K) linked to the Ubx proximal promoter confer patterns of strong β-gal expression in individual imaginal discs as follows.

β-gal expression in ABX transformants is restricted to the dorsal discs, the wing (Fig. 2C) and the haltere disc (Fig. 2D); neither of the leg discs (Fig. 2E,F) nor any other disc show β-gal staining. Staining in the wing and the haltere disc is very strong in their centers, their “pouch” regions, which give rise to the most distal derivatives of these discs, the wing blade and the haltere capitellum (reviewed by Bryant, 1978). In some but not all transformant lines, we also see strong β-gal expression in the stalk region of the wing disc (Fig. 2C) and, very occasionally, a strong β-gal spot in the stalk region of the haltere disc, the regions giving rise to the meso- and metanotum. Expression in the “pouch” regions of the two discs is regularly observed and equally strong in virtually all ABX lines, although the extent of the “pouch” expression varies slightly (in some lines restricted to the most central regions; cf. Fig. 3A). We shall refer to this expression in the wing and haltere “pouches” as the imaginal ABX pattern. We observe it independently of the type of ABX construct, i.e. whether the latter contains the large genomic 3.0 kb or the minimal 0.6 kb ABX fragment, and whether the minimal fragment is located upstream or downstream of the fusion gene (B βA transformants; cf. Müller and Bienz, 1991).

The imaginal ABX pattern shows very little resemblance to Ubx protein expression in imaginal discs (White and Wilcox, 1984, 1985a). However, the staining in the wing “pouch” is reminiscent of the ectopic Ubx expression pattern in wing discs of Hm and Cbx^M1mutants (Fig. 3A,E; Cabrera et al., 1985; González-Gaitán et al., 1990). In Hm (Bender et al., 1983) and probably in Cbx^MI mutants (González-Gaitán et al., 1990), most of the upstream flanking region of the Ubx gene is lacking and, thus, the mutant Ubx expression patterns may reveal the intrinsic activity of control regions downstream of the promoter (e.g. of the ABX fragment), a notion strongly supported by our imaginal ABX pattern.

We initially obtained only one transformant line with the large genomic PBX fragment. Two further lines were generated by “jump-starting” (Robertson et al., 1988) ; however, one of these lines showed β-gal staining in all discs, whereas the other showed none. We shall refer to the β-gal staining pattern in the initial line as the imaginal PBX pattern since we observe virtually the same pattern in other transformants carrying part or the whole of the large PBX fragment (see below). The imaginal PBX pattern consists of strong β-gal expression in the wing and in the haltere discs (Fig. 2G,H) and of a comparatively low but significant level of β-gal expression, evenly distributed, in all three types of leg discs (Fig. 2H-K). There is also β-gal expression in the eye and in the humeral disc (not shown). β-gal staining in the haltere disc is usually observed throughout the disc (Fig. 2H), with occasional white patches in the anterior part. In the wing disc, β-gal staining is confined mostly to the posterior part, with an anterior expression boundary forming a striking line across the disc (Fig. 2G), a line reminiscent of the anteroposterior compartment boundary (Brower, 1986). The β-gal pattern varies somewhat between individuals as we frequently observe β-gal patches in the anterior part (Fig. 3B). On occasions, the whole disc is stained except for a white line approximately following the anteroposterior compartment boundary (Fig. 3C). Surprisingly, we do not see the imaginal PBX pattern in transformants containing the minimal PBX control region (between the SamHI and the Psrl site; see Fig. 1) which is sufficient to confer the embryonic PBX pattern (Müller and Bienz, 1991), but we find a very similar pattern in imaginal discs of transformants containing a more proximal genomic fragment (3104; see Fig. 1) which partly overlaps the large PBX fragment. Thus, the regulatory sequences conferring the imaginal PBX pattern are closely linked to, but separable from (and downstream of) the sequences conferring the embryonic PBX pattern.

The imaginal PBX pattern resembles the Ubx protein pattern in Cbx¹ mutants, particularly within the wing disc (Fig. 3; White and Akam, 1985; Cabrera et al., 1985; see also Castelli-Gair et al., 1990). Variable patterns of Ubx protein expression are observed between individual Cbx¹ mutants (Fig. 3F,G), very similar to the variable β-gal patterns of PBX transformant individuals (Fig. 3B,C). The Cbx¹ mutant chromosome bears a duplicated copy of an upstream genomic fragment at an intronic location closer to and downstream of the Ubx transcription start site (∼1-13 kb; Bender et al., 1983). The transposed piece of genomic DNA contains our large PBX fragment. Evidently, the pattern of ectopic Ubx protein expression in Cbx¹ wing discs reflects the intrinsic capability of the PBX control region to direct this particular pattern of expression Finally, we examined transformants of a triple combination construct, bearing all three control regions upstream of the Ubx proximal promoter (ABP; Müller and Bienz, 1991). The majority of these ABP lines show a ∼-gal-staining pattern in the dorsal imaginal discs which corresponds to a PBX pattern superimposed on an ABX pattern (Fig. 2L,M), best visible in the wing disc where ∼-gal staining in the “pouch” protrudes across the PBX expression boundary into the anterior part of the disc. Compared to the wing disc pattern in PBX transformants which extends variably into the anterior compartment, we note that β-gal staining in ABP transformants is more regularly restricted to the posterior part of the disc, reflecting some degree of nonadditivity of patterns. More striking non-additivity is observed in the leg discs: the first leg discs in ABP transformants do not stain at all (Fig. 20), and β-gal staining in the second leg discs is restricted to the posterior half (Fig. 2N). There is moderately strong β- gal staining in the whole of the third leg discs (Fig. 2M). As in the wing disc, the β-gal expression boundary in the second leg disc apparently follows the anteroposterior compartment boundary (cf. Steiner, 1976). There is virtually no /kgal expression anterior to the anteroposterior compartment boundary in T2 discs. This boundary corresponds to the anterior boundary of ps5 and thus to the anterior limit of Ubx expression.

The imaginal ABP pattern closely resembles the pattern of Ubx protein expression in imaginal discs (White and Wilcox, 1984, 1985b), although it differs from the latter in two main respects. Firstly, there is ectopic β-gal expression in the wing disc which probably causes the Cbx¹-like phenotype of our minigene transformants. Secondly, the levels of β-gal staining in the leg discs are somewhat low, suggesting that the-ABP construct lacks enhancer sequences required for efficient expression in ventral discs.

Since an ABP minigene (all three fragments linked to a Ubx cDNA; Müller and Bienz, 1991) is capable of conferring a Cbx¹ phenotype through its activity in the wing disc (Fig. 3D), we wondered whether this minigene might be sufficiently active in other imaginal disc cells to rescue aspects of the adult Ubx phenotype. For this, we established balanced strains containing a minigene transposon (U12, U81, T7 or T10) and the Ubx¹ mutation (an antibody-negative null mutation; Weinzierl et al., 1987; see Material and Methods). We first tested the function of these minigenes in abx, bx, pbx and bxd mutants. Flies carrying either of these mutations in combination with Ubx¹ are viable; however, they show transformations of individual compartments in their epidermis: T3a is transformed into T2a in abx/Ubx or in bx/Ubx mutants, T3p is transformed into T2p in pbx/Ubx mutants, whereas both T3p and Ala are transformed into T2p and T3a in bxd/Ubx mutants (reviewed by Sánchez-Herrero et al., 1985a; the T2p>Tlp transformation caused by abx¹/Ubx¹ or bx³/Ubx¹ mutation cannot be discerned in the above allelic combinations).

Homeotic transformations in the haltere disc derivatives (dorsal T3a or T3p) which are caused by abx¹bx³ (Fig. 4A) or by pbx¹ mutation (Fig. 4C) are mostly and, in the case of pbx¹ mutation, virtually fully rescued by one copy of either of the minigenes (Fig. 4B,D; U12 and U81 minigenes show slightly better rescue activity than T7 and TIO minigenes). The rescue activity is particularly strong in the distal region of T3, the haltere, but there is also some rescue activity in the proximal region, the notum. This activity is particularly striking in the case of the four-winged triple mutant abx¹bx³pbx¹ (Fig. 4E) whose second pair of wings are reverted towards normal halteres and whose duplicated mesonotum is partly suppressed, due to minigene function (Fig. 4F). However, these flies have two pairs of T2 legs (not shown) and, thus, we do not observe full rescue function of the minigene in the third legs. This is somewhat surprising, given that there is β-gal expression in the third leg discs of ABP transformants (Fig. 2M). It is possible that the level of expression in the third leg discs is too low, due to lack of enhancer sequences required for high levels of ventral disc expression.

In addition to affecting the haltere disc derivatives (T3p>T2p), bxd¹/Ubx¹ mutant flies show an A1a>T3a transformation, having an additional pair of legs (Fig. 4G), and lacking an Ala tergite. If these flies carry one of our Ubx minigenes, their T3p>T2p transformation is rescued (not shown) and their extra pair of legs is completely suppressed (Fig. 4H). However, they still lack the Ala tergite and, thus, the minigenes apparently lack sequences mediating Ubx expression necessary for tergite development.

The recombinant strains containing both a Ubx¹ mutation and a Ubx minigene allowed us to test the rescue activity of this gene in Ubx¹ homozygotes. Homozygous Ubx¹ larvae, obtained from a control cross, died shortly after eclosion from the embryonic membranes (see Materials and methods; cf. also Lewis, 1978; Frayne and Sato, 1991). However, we find that one copy of a Ubx minigene can support development of Ubx¹ homozygotes throughout larval life: a relatively low, but significant fraction of these homozygotes (more individuals in the case of T7 and T10, compared to U12 and U81) survive to the end of pupation and secrete adult epidermis which is fully differentiated. The morphological features of these pharate adults can be examined. Rescue of the larval lethality of Ubx null mutation demonstrates that our Ubx minigenes are capable of providing all Ubx function required for larval and pupal viability.

We peeled from their pupal case Ubx¹ homozygous or Ubx¹/Ubx¹⁹⁵ pharate flies containing one of our minigenes and prepared their cuticle for microscopic analysis (Fig. 5; there was no difference in rescue activity in the two cases of allelic Ubx combinations). In agreement with our previous results, the T3>T2 transformations of the haltere derivatives are partially rescued by the minigenes (Fig. 5A,C,E), particularly in the distal region, and the fourth pair of legs is completely suppressed (Fig. 5D,F). Again, the T3>T2 transformation in the legs is not rescued (Fig. 5D,F), and the Ala tergite is also lacking (Fig. 5C,E). However, we find that the ppx phenotype (the T2p>Tlp transformation) in the legs is fully rescued (Fig. 5B,D,F), showing that our minigenes provide ppx⁺ function.

We have analysed expression and function of two types of Ubx minigenes in imaginal disc cells. These minigenes contain the same three Ubx control fragments which, upon combination, confer a Uhx-like expression pattern in the embryo as well as Ubx function in the larval epidermis (Müller and Bienz, 1991). We now find that two of the three fragments also direct distinct patterns of expression in imaginal discs, and that their combination results in a pattern resembling Ubx expression in these discs. Both types of Ubx minigenes are capable of providing all Ubx function needed for larval viablity and pupal development as well as partial Ubx function in the adult epidermis throughout the Ubx domain. Certain control elements required for adult Ubx function are still missing from these minigenes, which is not surprising as the initial screen for control sequences was based entirely on their activity in the embryo. Nevertheless, at least some of the control sequences conferring adult Ubx function are evidently closely linked in the chromosomal Ubx gene to those conferring larval Ubx function.

It has been proposed that the gain-of-function phenotype of the Cbx¹ mutation is due to ectopic Ubx expression in the wing disc (Lewis, 1982), a prediction borne out by Ubx antibody staining results (White and Akam, 1985; Cabrera et al., 1985; González-Gaitán et al., 1990). There remained the question as to why, upon transposition, an upstream cis-regulatory element expected to be active in ps6 and conferring pbx function in T3p (Bender et al., 1983) should be activated ectopically in ps5 in Cbx¹ mutants. To explain this, Peifer et al. (1987) proposed in their “open for business” model the existence of parasegment-specific chromosomal domains which need to be “opened up” in a parasegment-specific way for enhancers located within these domains to become active. According to this model, enhancers from the pbx region are active ectopically in ps5 in Cbx¹ mutants, because of their new location within a chromosomal domain which is “opened up” in ps5.

Our results seem to argue against this view. The Ubx expression pattern in Cbx¹ mutant imaginal discs closely resembles the β-gal staining pattern conferred by the PBX fragment (which is contained within the transposed DNA in Cbx¹ mutants) and thus apparently reflects intrinsic PBX enhancer activity. This activity is overt in all our PBX-bearing transformants and therefore does not seem to require any particular chromosomal context (nor does it depend on the presence of the ABX control region). In the chromosomal Ubx gene, and also in a Ubx/β-gaX construct containing 35 kb of Ubx upstream flanking sequence (Irvine et al., 1991), the intrinsic PBX activity in ps4 and ps5 is evidently suppressed. We suggest that there is an upstream repressor element in the Ubx gene, located between our PBX fragment and the Ubx promoter, which acts to suppress the PBX enhancer activity in ps4 and ps5. Whether this activity is suppressible by the upstream repressor may be determined by the position of the two regulatory elements with respect to each other and to the Ubx promoter, e.g. whether they are located on the same or opposite sides of the promoter, or perhaps by their relative distance from the promoter (for a discussion of distance affecting competition for promoter interaction between remote transcription factors, see Hanscombe et al., 1991). We previously found a similar repressor element within the PBX fragment itself (mediating long-range repression anterior to ps6; Müller and Bienz, 1991), but the latter is evidently only effective in the embryo and not in imaginal cells.

Hm and Cbx™¹ mutants show ectopic Ubx protein expression in the wing disc “pouch” (White and Akam, 1985; Cabrera et al., 1985; González-Gaitán et al., 1990) which resembles the β-gal expression pattern conferred by the ABX fragment. It is likely that the mutant ectopic Ubx expression reflects the intrinsic enhancer activity of the ABX fragment, and that this intrinsic activity is normally suppressed by an upstream repressor element (which is uncoupled from the Ubx promoter in the mutants). This repressor element may be the same as the one mentioned above that suppresses the PBX enhancer activity in ps4 and ps5.

We found that the imaginal ABP pattern does not correspond to a simple superimposition of imaginal ABX and PBX patterns: the anterior ABP expression limit to a large extent follows the anterior boundary of ps5, although the PBX fragment on its own mediates expression anteriorly to this limit, especially in the leg discs. It thus appears that the ABX fragment contains a long-range repressor element (suppressing PBX enhancer activity in the leg discs and perhaps in the wing discs anteriorly to ps5). Again, we found that the ABX fragment does contain such a repressor element which, in the embryo, mediates an anterior expression boundary between ps4 and ps5 (Muller and Bienz, 1991) . Although the imaginal disc repressors may be different from the embryonic repressors (one of which is the hunchback protein; Zhang et al., 1991; Zhang and M.B., unpublished), their action may depend, directly or indirectly, on the preceding action of embryonic repressors.

Two features of the adult Ubx phenotype are apparently fully rescued by our Ubx minigenes: the ppx phenotype (the T2p>Tlp transformation) and the formation of a fourth pair of legs. We propose, for the following reasons, that rescue activity in these cases reflects minigene-dependent deployment of Ubx in the embryo, either within imaginal disc primordia or suppressing the formation of these.

The ppx⁺ function is a subfunction of Ubx which is required exclusively early in development to control the morphogenesis of T2p and in T3p in the adult (Morata and Kerridge, 1981; Casanova et al., 1985). In the absence of ppx⁺ function, i.e. if Ubx product is eliminated within the first 6 hours of embryonic development, these two posterior compartments are transformed into Tip, a transformation rescued by our Ubx minigenes. We note that the imaginal domains within which ppx⁺ function is required correspond to the embryonic domains within which ABX- and PBX-mediated expression is highest (anterior part of ps5 and ps6, corresponding to T2p and T3p; Müller and Bienz, 1991). Our evidence strongly suggests that this expression anteriorly within ps5 and ps6 is due to direct activation of the ABX and PBX control regions by the segmentation products even-skipped (eve) and fushi tarazu (ftz), respectively (Müller, 1991; Müller and Bienz, in preparation). It therefore appears likely that eve- and /iz-mediated Ubx expression in the early embryo confersppx⁺ function. Among the cells that we expect to be supplied with this pulse of eve- and ftz- mediated Ubx expression are those forming the prospective posterior compartments of the T2 and T3 imaginal disc primordia as the latter become visible at this time, straddling parasegment boundaries (Cohen, 1990; Cohen et al., 1991). This early Ubx protein may fade later to become virtually undetectable in the wing discs (White and Wilcox, 1984, 1985b) as these are the most actively proliferating discs (Bryant, 1978).

A fourth pair of legs is formed in bxd mutants (Lewis, 1963), apparently reflecting establishment of an additional pair of disc primordia in the embryo (Cohen et al., 1991). This extra pair of legs is completely suppressed by our Ubx minigenes. The minigenes also suppress the extra pair of Keifin’s organs (corresponding to rudimentary larval legs) due to Ubx mutation, an activity which we ascribe to the BXD control element (Müller, 1991). It therefore appears that BXD-mediated Ubx expression may suppress the establishment of larval and adult leg primordia in the embryo. The latter do not seem to be suppressed by PBX- and ABX-mediated Ubx expression (see above); however, the PBX and ABX control regions are active at different times and levels as well as in different cells of the embryo than the BXD control region (Müller and Bienz, 1991).

Our Ubx minigenes are not capable of supporting development of Ala tergites which are missing in Ubx mutants or in bxd mutants lacking Ubx activity in ps6. These mutants lack abdominal histoblasts, the tergite precursor cells, in Al at young larval stages; instead, they show buds of dorsal and ventral imaginal discs (Kerridge and Sang, 1981; Frayne and Sato, 1991). Although the Ubx minigenes apparently suppress the disc primordia in Al (see above), they obviously do not supply sufficient Ubx function to support development of abdominal histoblasts in Al. It is conceivable that primordial histoblasts never form in minigene transformant embryos. We think it more likely that these form, but that they cannot develop and proliferate (cf. Frayne and Sato, 1991), due to insufficient minigene function at later stages.

Homeotic gene complexes are strikingly conserved between flies and mammals (Duboule and Dolle, 1989; Graham et al., 1989). It has been speculated that this conservation reflects functional significance (cf. Akam, 1989), though in the case of the bithorax complex, integrity of the complex is not necessary for bithorax gene function (Struhl, 1984). Functional importance has also been attributed to the size of homeotic genes in Drosophila which are unusually large, apparently due to the complexity and redundancy of cri-regulatory elements (Bender et al., 1985; Simon et al., 1990; Irvine et al., 1991). The latter may be needed to generate the complex pattern of Ubx expression, some aspects of which however are functionally insignificant (González-Reyes and Morata, 1990).

Our results seem to indicate that there is no absolute requirement for homeotic genes to be located within homeotic gene complexes: Ubx minigenes are capable of functioning, out of context, in a number of different chromosomal locations to provide all Ubx functions needed for embryonic and larval life as well as partial Ubx function for adult morphogenesis. The minigenes function to a perhaps surprising level, despite containing few and comparatively short control regions. Their functional shortcomings in certain imaginal cells (in the legs and the Ala tergite) are probably due to lack of the corresponding control elements which might be identifyable by appropriate searching. Finally, our minigenes do not function reliably from individual to individual, as judged by the varying rescue activities observed among individuals in any one transformed line. Thus, large control regions and/or redundancy of control elements may be required in some way to ensure reliability of the control process.

We thank Gines Morata, Ernesto Sánchez-Herrero, Jordi Casanova, Maria Paz Capdevila and Antonio Garcia-Bellido for comments on the manuscript. We also thank Antonio Garcia-Bellido for support while this work was completed in his laboratory, and Charo Hernández and Ana López for help with the cuticle preparations. This work was funded by EMBO (short-term fellowship awarded to J. C.-G.) and by the Swiss National Science Foundation (grant nr. 31-26198.89 to M. B.).