A clonal analysis of the requirement for the trithorax gene in the diversification of segments in Drosophila

Differential expression of selector genes (Garcia-Bellido, 1975, 1977) is the primary cause of segment diversification in Drosophila. Selector genes are defined by mutations which result in homoeotic transformations of different body segments. Bithorax complex (BX-C) mutations cause transformations of thoracic and abdominal segments (Lewis, 1978), from which it is inferred that these genes normally control the diversification of these body segments. Antennapedia complex (ANT-C) mutations, on the other hand, transform structures deriving from the head and thoracic segments (Kaufman, Lewis & Wakimoto, 1980) implying that the ANT-C controls the development of the anterior part of the fly.

In addition to those of the BX-C and ANT-C, a considerable number of other homoeotic mutations have been described in Drosophila melanogaster. Recent studies have provided evidence that several of these mutations define genes which function in controlling the expression of the ANT-C and BX-C. The majority of these appear to exert a negative control over the two complexes (Lewis, 1978; Struhl, 1981a; Duncan & Lewis, 1982; Duncan, 1982; Ingham, 1984; see also review of Ingham, 1985), whereas one, defined by the Regulator of bithorax (Rg-bx) (Capdevila & Garcia-Bellido, 1981) and trithorax (trx) (Ingham & Whittle, 1980) mutations appears formally equivalent to a positive regulator of the BX-C.

Much interest centres around the developmental stage at which these genes act; in particular which, if any, are involved in the initiation of selector gene expression. Capdevila & Garcia-Bellido (1981) argued that Rg-bx defines a gene required solely for the initiation of BX-C expression. A preliminary analysis of two trx alleles, which fail to complement Rg-bx, produced results inconsistent with this interpretation (Ingham, 1981). Here I present the results of an extensive clonal analysis of both the Rg-bx and trx mutations. The data confirm that the mutations define a single gene, trithorax, which is required throughout the larval stage for the appropriate differentiation of adult derivatives of thoracic, abdominal and head segments.

trx², trx³ are recessive lethal alleles which fail to complement trx¹ (see Ingham, 1981). Rg-bx is a recessive lethal which also fails to complement trx¹, trx² and trx³ (Ingham, 1981). The allelism of Rg-bx and trx has been questioned by Capdevila & Garcia-Bellido (1981) but is confirmed by results of this analysis. Since the trx symbol conforms to the convention for Drosophila gene nomenclature I here refer to Rg-bx as trithorax of Lewis (trx^L). Dp(3:1)kar⁵¹ carries a copy of the trx locus on the×chromosome, which covers all trx alleles. Two crosses were employed to generate animals bearing trx mutant clones:

(1) ♀ ♀y/^36a; mwh red trx²/TM2× ♂ ♂3Dp(3:l)kar⁵¹M(1)O^Sp; red trx³/DP(1:3)A59.

(2) ♀ ♀y/^36a; mwh red trx²/TM2× ♂ ♂Dp(3:l)kar⁵¹M(l)O^Sp; mwh trx^L/ Dp(l:3)A59. y,f^36a, and mwh are cuticular markers; these, and red and TM2, are described in Lindsley & Grell (1968).

The dominant M(1)O^Sp mutation causes cell autonomous slow growth. Dp(l:3)A59 carries a wild-type copy of the M(1)O^Sp locus on chromosome 3 (see Kerridge & Morata, 1982).

Clones of marked cells in trx¹/trx² flies were induced in larvae from the cross ♀ ♀y;mwh st trx¹× ♂ ♂y;Dp(l:3)sc¹⁴, y⁺M(3)i⁵⁵ red trx²/TMl.

Flies were reared at 25°C on a yeast/molasses/cornmeal/agar medium.

Eggs were collected from the appropriate crosses over 12 h periods in split glass bottles containing standard medium. The resultant larvae were subsequently irradiated using a Philips Constant Potential MG102 X-ray machine. A dose of 1000 rad was delivered in 3 min 14 s at 4mamp, 100kV. Larvae were irradiated at either 48±6h after egg laying (AEL) or 84±6h AEL, corresponding to the first and second larval instars. Adult progeny of the appropriate genotype were collected and preserved in 70 % alcohol.

In the case of flies irradiated during the first instar, most induced yf^36a clones are expected to be sufficiently large to be detectable at low (×25) magnification. Accordingly, animals irradiated at 48 ± 6h AEL were screened under the dissecting microscope for clones in the head, dorsal thorax and abdomen. Animals bearing clones were dissected and mounted for examination at higher magnification. Since y f^36a clones in the legs are not so easily detectable, all the legs from these animals were mounted and screened at high magnification for clones. All parts of flies irradiated at later stages were similarly screened at high magnification. This screening procedure means that some clones induced during the first instar may have gone undetected, especially if they had not resulted in transformations. However, judging from their frequency amongst later stage irradiations, the number of such untransformed clones would have been small.

The phenotypes of the clones in each segment are described in Results. The detailed morphology of adult structures have been described by various authors. This study follows the descriptions of Struhl (1981b) for the proboscis; Morata & Lawrence (1979) for the eye antenna; Roseland & Schneiderman (1979) for the abdomen; Schupbach, Wieschaus & Nothiger (1978) for the genitalia; sources of descriptions of thoracic structures are as previously cited (Ingham, 1981).

The mutant alleles trx², trx³ and trx^L are all homozygous lethal; they fail to complement the homoeotic phenotype of trx¹, as well as each others lethality. Genetic analysis suggests that all three alleles abolish most, if not all, of the wild-type function of the trx locus (Ingham, 1981); in addition, there is some evidence that trx³ may have an antimorphic component (unpublished observations).

Clones trans-heterozygous for pairwise combinations of these alleles allow an investigation of the requirement for trx wild-type activity in cells of different segments and at different developmental stages. It has previously been shown that cells trans-heterozygous for trx² and trx³ are viable in clones in the adult thoracic cuticle (Ingham, 1981). Here the viability and differentiative potential of cells lacking trx activity has been investigated in the adult derivatives of head, thoracic and abdominal segments. To rule out allele-specific effects, two independent series of clones were induced, genotypically either yf^36aM(1)O⁺;trx²/trx³ or y f^36aM(1)O⁺;trx²/trx^L. The Minute technique (Morata & Ripoll, 1975) was employed, in order to generate large clones.

In all regions of the animal, clones of these two genotypes are phenotypically indistinguishable. Thus in the following description the two series of clones have been pooled and will be referred to simply as trx⁻ clones. The characteristics of clones in each adult structure are described below:

A total of 79 antennal clones and 49 clones in the eye or head capsule induced during the first and second instars (see Table 1) were examined; the majority of clones exhibit the following phenotype.

Clones in the antenna result in the transformation of distal structures (arista and third segment) to homologous leg structures (the tarsal segments) bearing characteristic structures such as bracted bristles, claws and pulvilli (Fig. 1B). In contrast, the proximal regions of the antenna develop normally in the absence of trx⁺.

Clones in the eye were detected since they result in an outgrowth of bristle-bearing cuticle (see Fig. 1C). This cuticle also bears trichomes; the identity of the bristles is not clear, but they are of a size comparable to those found on the notum. Clones in other regions of the head are integrated into the head capsule but cause disruption of the normal pattern. In particular, clones on the top of the head capsule result in the absence of the ocelli and of the orbital bristles (Fig. 1A).

27 clones induced in the first instar were examined. Of these, sixteen are located in the anterior compartment whilst eleven are of posterior provenance (see Table 1).

All the anterior clones show evidence of transformation of the palps to leg tissue. In most cases this is indicated by the presence of an apical bristle (characteristic of the second leg); some clones also differentiate claws and pulvilli (Fig. 2A). In one exceptional case, the clone bears both bracted bristles and a structure resembling an arista (Fig. 2B). Posterior clones in the prementum often cause unusual bristle arrangements. These cannot be identified as being characteristic of any particular tissue (Fig. 2C); however they are similar to those caused by clones of lethal engrailed alleles (see fig. 1 in Lawrence & Struhl, 1982). In the posterior region of the palps, clones differentiate large bristles, similar to apical bristles. Though apical bristles are characteristic of the anterior leg compartment, similar bristles are also found in trx⁻ clones in the posterior leg compartment (see below).

A total of 217 clones induced in the three thoracic legs during the first larval instar were examined. The number and distribution of clones is presented in Table 2.

A further 35 clones induced during the second instar in the first and third thoracic legs were found to have a similar phenotype. As previously reported (Ingham, 1981) clones lacking trx⁺ activity in the prothoracic and metathoracic legs exhibit changes in pattern consistent with a transformation towards mesothoracic leg. These changes include the presence of stenopleural bristles in the presumptive propleural (Fig. 3A) and metapleural regions and of apical bristles on the distal tibia (Fig. 3B). The pattern differentiated by these clones is not however strictly that of a normal mesothoracic leg; furthermore, clones in the mesothoracic leg also differentiate abnormally. In the tarsal segments of all three legs, clones bear larger than normal bristles at the distal tip of each tarsal segment (Fig. 3B), whilst those including the distal anterior tibia frequently bear a duplicated apical bristle (Fig. 3B) [(see also fig. la, Ingham, 1981)].

Clones in the posterior tibia also bear a larger than normal bristle at the distal tip, adjacent to the (anterior) pre-apical bristle. In the posterior femur of all three legs, clones differentiate some large bractless bristles in an arrangement not typical of any leg (Fig. 3C,D and E). Similar ectopic bristles are also produced by clones of en lethal alleles in the posterior femur (Lawrence & Struhl, 1982). In other respects, posterior clones in the first and third legs exhibit a clear transformation to second leg; this is indicated in the first leg by the reduction in size of the typical proximal distal row of large femoral bristles. In the posterior third leg, clones bear typically mesothoracic bristles in the coxa (e.g. the BH⁻ bristle), trochanter and femur. In addition, the transverse rows of bristles unique to the posterior metathoracic tibia and tarsal segments are removed.

A total of 23 humeral and 35 notal clones were examined. None of these show any evidence of transformation or other abnormalities.

144 clones induced in the wing during the first and second instars were examined.

In the posterior compartment (54 cases) most clones exhibit a consistent abnormal phenotype. Large socketed bristles are formed along the posterior margin of the wing blade (Fig. 4B). In the wild-type wing, bristles are present along the anterior margin but not the posterior margin. The bristles formed by clones in the proximal region of the posterior margin resemble those of the costal region of the anterior wing. However, in clones extending distally along the margin, the bristles do not form the triple row arrangement characteristic of the homologous region of the anterior margin; rather, even in this region the bristles resemble those of the anterior costa. The clones frequently cause considerable distortion of the posterior wing, increasing its size relative to the anterior compartment (see Fig. 4B). Bristles are also formed within the wing blade itself whilst the posterior wing veins (L4 and L5) are disrupted. Most of the 90 anterior clones also exhibit an abnormal phenotype. Venation is disrupted (Fig. 4A) as is the triple row of bristles along the anterior margin. This is replaced by an arrangement of bristles which, like those differentiated by clones along the posterior margin, resemble those of the costa. Clones extending along the length of the anterior margin exhibit a reiterated pattern of such bristles (Fig. 4A). This pattern bears a striking resemblance to the phenotype of the Dipr mutation (Kerridge, 1981).

34 first instar clones were detected in the haltere on the basis of their being transformed to wing tissue (Fig. 5). Only anterior compartment structures (i.e. bristles) could be unequivocally identified; no clones were found which differentiated posterior wing elements such as the axilliary cord or alula. This is consistent with the observation that trx⁻ clones in the posterior wing result in a transformation toward anterior compartment structures.

Clones in the tergites of abdominal segments one to seven (A1-A7) of animals irradiated during the first and second instars differentiate normal patterns (Fig. 6A). There is no evidence of transformation towards segment Al in clones in segments posterior to Al. This is in contrast to the transformations of tergites of animals homozygous or hemizygous for the hypomorphic trx¹ allele (Ingham & Whittle, 1980). Transformation of abdominal clones towards mesothorax might be expected to result in the loss of the transformed clone as happens with bxd~ clones induced in Al (Morata & Garcia-Bellido, 1976; Kerridge & Morata, 1982). A comparison of the frequency of abdominal clones in control and experimental animals irradiated during the second instar (Table 3) provides no evidence for such loss. A possible explanation of this paradoxical lack of effect of removal of trx⁺ on abdominal development will be discussed below.

In contrast to clones in segments Al to A7, clones in the genitalia (which derives from A8, Schupbach et al. 1978) are transformed to tissue which bears pattern elements resembling those of the thorax. 19 clones in the genitalia induced during the first instar were examined. In each case tissue bearing yellow forked bristles is associated with the absence of part or all of a vaginal plate (Fig. 6B,C). Thus trx⁻ clones cause abnormal differentiation of cells which would normally contribute to this structure. The tissue differentiated by the clones is usually disorganized; however in several cases evidence of segmentation, similar to that of the tarsus, is visible. In addition, some of the bristles formed by the clone are associated with bracts (Fig. 6C,D). Bracts are normally found in association with bristles only in the legs and a very small region of the wing. Given the ventral origin of the vaginal plate a transformation of this structure to leg, as suggested by the bracted bristles, would not be inconsistent. However it should be emphasized that many of the bristles within the clones are not associated with bracts. Occasionally other tissue, bearing some resemblance to the wing hinge/notum region is associated with these clones.

Animals homozygous or hemizygous for the hypomorphic allele trx¹ exhibit partial transformations of many adult body segments. In the prothoracic segment, this results in a foreleg which is a mosaic of prothoracic and mesothoracic structures (Ingham & Whittle, 1980). The early phenocritical period of this allele suggested that these transformations might arise as a result of the failure of the gene to function during early embryogenesis rather than at a later stage of development. To test this, the fate of the progeny of single cells marked at different developmental stages in mosaic forelegs was analysed.

Clones of cells marked with yellow and multiple wing hairs were induced in the first legs of trx¹/trx² flies during the first and second larval instars (see Materials and Methods). Clones were given a growth advantage by using the Minute technique (Morata & Ripoll, 1975). The numbers and locations of clones induced are shown in Table 4.

Amongst thirteen clones induced during the first instar, eight marked regions of the partially transformed legs bearing only structures characteristic of the prothoracic leg. The remaining five marked both prothoracic and mesothoracic structures, principally the sex comb and the apical and/or pre-apical bristles. The expected number of legs bearing two independently arising clones in such a sample is 0·6. Therefore most or all of these marked regions must represent the descendants of single cells.

Amongst 30 clones induced during the second instar nine include both prothoracic and mesothoracic structures. The expected number of legs bearing two independent clones in this sample is five.

The identity of all segments posterior to the mesothorax depends upon a successive increase in the expression of BX-C elements (Lewis, 1978, 1981a,b). Partial loss of function of the trx gene results in homoeotic transformations corresponding to a partial loss of BX-C activity. To test whether there is an absolute requirement for the trx gene for BX-C expression, trx⁻ clones have been induced in all the segments in which the BX-C is required.

As described in a previous report, (Ingham, 1981) trx⁻ clones in the metathoracic (T3) segment are transformed to mesothorax (T2). This is consistent with absence of trx⁺ causing a loss of function of the Ubx gene. The BX-C is postulated to be fully expressed in the eighth abdominal segment (Lewis, 1978,1981a,b), the segment from which the female genitalia is derived (Schupbach et al. 1978). Since trx⁻~ clones cause an apparent transformation of female genital structures to mesothoracic tissue, it appears that absence of trx⁺ from segment A8 prevents expression of most or all of the BX-C in segment A8.

Surprisingly, trx⁻ clones in abdominal segments one to seven differentiate normally. Complete removal of the trx gene would be expected to have at least an equal effect to that caused by the hypomorphic trx¹ allele (Ingham & Whittle, 1980). A probable explanation of this lack of transformation depends upon the growth dynamics of these segments. Unlike the cells of most imaginal discs, abdominal histoblasts grow without dividing throughout the larval stages (Madhavan & Schneiderman, 1977). Hence recombinant chromosomes induced during this period do not segregate to give rise to mutant clones in the abdominal segments until the pupal stage. One possibility is that sufficient trx⁺ product is produced prior to clone formation to support normal development even in the subsequent absence of the trx⁺ gene. This could be tested by inducing trx⁻ clones at the blastoderm stage.

The transformation of trx⁻ clones in the ventral prothorax (Tl) to T2 is consistent with a role for trx in the expression of the Scr gene, which promotes the prothoracic developmental pathway (Lewis, Kaufman, Denell & Tallerico, 1980a; Lewis, Wakimoto, Denell & Kaufman, 1980b).

The dorsal derivatives of segment Tl are the subject of some debate. Gynandromorph fate mapping suggests that the humerus is derived from dorsal Tl. However, amongst trx¹ homozygotes or hemizygotes, prothoracically located mesonotum and wing tissue is always additional to and not in place of the humerus. One explanation of this finding depends upon the proposal that the humerus is derived wholly from the posterior compartment of this segment (Lawrence & Struhl, 1982). In this case it is possible that the ectopic mesonotum and wing tissue may be exclusively of anterior origin. trx⁻ clones in the humerus might thus be expected to be transformed to posterior T2, whilst clones in the anterior compartment of Tl should result in additional notal tissue adjacent to the humerus. No examples of this latter type of clone were observed, whilst all clones in the humerus are morphologically normal. The explanation for this lack of effect of trx in dorsal Tl is unclear. It is possible that trx is only required in this compartment during early development. Alternatively, at least in the case of the humerus, it may reflect the delay between the time of irradiation and the subsequent segregation of the recombinant chromosomes, since, like the abdominal histoblasts, the humeral disc cells do not begin dividing until the pupal stage (Madhavan & Schneiderman, 1977).

trx⁻ clones have revealed a requirement for the gene in derivatives of the head segments unaffected by the trx¹ mutation. Absence of trx⁺ in the proboscis and antenna results in a transformation of these structures to mesothoracic leg. In both cases, this transformation is incomplete in that clones form only distal leg structures (tarsal segments and tibia) whilst the proximal region of both the proboscis and antenna appear untransformed.

The genetic control of development of the head segments is less well understood than that of the thoracic and abdominal segment. Dominant Antp mutations result in transformation of the antenna to mesothoracic leg, which may be more or less complete. The Antp gene is not essential for antennal development (Denell, Hummels, Wakimoto & Kaufman, 1981; Struhl, 1981c). Rather, it is required in the mesothoracic segment to suppress antennal development (Struhl, 1981c, 1983). A likely interpretation of the dominant Antp phenotype therefore is that it is the consequence of inappropriate expression of Antp⁺ in the antennal segment, resulting in the repression of an unspecified antennal gene or genes. The identity of these latter genes is uncertain, spineless anstapedia (ss^a) mutant alleles result in a partial transformation of the antenna to leg; interestingly, like trx⁻ clones, the effects of these mutations are limited to the distal-most regions of the antenna. Hence the effect of absence of trx⁺ in the antennae is equivalent to loss of expression of the ss^a gene. Two genes are known to be required for the correct developmental programme of cells of the labial segment, from which the proboscis is derived. In the absence of the Scr gene the proboscis is transformed to maxillary palps (Lewis et al. 1980a,b) whilst absence of the proboscipedia gene, on the other hand, transforms the proboscis to prothoracic leg (see Kaufman et al. 1980). Since trx⁻ clones in the proboscis are most likely transformed to second rather than first leg, their phenotype corresponds to the loss of neither Scr norpb alone. However, the transformation is consistent with the combined effect of loss of Scr⁺ from a proboscis which is also pb⁻.

In addition to the requirement for the ss^a and pb genes in adult head development, it is expected that other ‘head’ genes are involved in controlling the development of derivatives such as the eye and head capsule. trx⁻ clones cause transformations of these structures to tissue resembling mesonotum. Thus it is clear that the trx gene is required during the larval stages for the normal development of much of the head. It seems likely, that by analogy to its effects on BX-C expression, trx is essential for the correct expression of several ‘head’ genes (Struhl, 1983) in imaginal cells.

Since animals homozygous for trx null alleles or deficiencies die at the end of embryogenesis, it is clear that there is an early requirement for trx. Capdevila & Garcia-Bellido (1981) proposed that the activity of trx is indeed restricted to embryogenesis, its function being the initiation of the selective expression of the BX-C genes in different segment primordia. This has been refuted by the demonstration that the requirement for trx in the thoracic segments persists beyond the completion of embryogenesis (Ingham, 1981). The results of the present analysis show that there is a continuing requirement for trx in many adult tissues at least until the early third instar. Paradoxically, whilst removal of the trx⁺ gene results in the transformation of adult derivatives of thoracic, some head, and at least one (but probably all) abdominal segments to mesothorax, a corresponding transformation of larval segments does not occur (Duncan & Lewis, 1982; Ingham, 1983). Thus BX-C expression is totally suppressed in the imaginal A8 segment, but only slightly reduced in its larval counterpart. This difference in requirement between larval and adult cells suggests that the trx gene may play a crucial role in the maintenance of expression of the BX-C and other selector genes during the proliferation of the imaginal cells.

This interpretation is supported by the analysis of cell lineage of partially transformed forelegs in trx¹/trx² flies. Cells marked during the first larval instar can still contribute to both normal prothoracic and transformed mesothoracic tissue. This suggests that the trx¹ mutation causes an instability in the heritability of the decision to become prothoracic, such that some cells may revert to the mesothoracic level. This is not inconsistent with the early temperature-sensitive period and maternal effect of the trx¹ mutation (see Ingham & Whittle, 1980). These may reflect a particularly high requirement for the trx gene early in embryogenesis but they do not rule out a subsequent continual function for the gene. If trx were autogenously regulated, either directly or via an intermediate such as a BX-C product, then a shortfall in trx activity early in embryogenesis could result in a reduction in activity of the gene thereafter.

trx⁻clones in the posterior wing cause the differentiation of structures resembling those of the anterior wing. This phenotype is similar to that caused by mutations of the engrailed gene, which controls the development of the posterior compartments (Morata & Lawrence, 1975; Lawrence & Morata, 1976). Though analagous posterior to anterior transformations are not observed in other structures, it is interesting that trx⁻ clones in the posterior compartments of the legs and proboscis exhibit some abnormalities characteristic of clones of lethal engrailed alleles. Taken together, these effects of loss of trx⁺ on posterior structures suggest a role for the gene in maintaining the anterior/posterior distinction. It is possible that trx is required for expression of the en gene itself; however it should be noted that no evidence of trx⁻ clones transgressing compartment boundaries was found.

This work was supported by fellowships from the Royal Society (European Exchange Programme) and EMBO. I thank Pat Simpson for her help, Antoine Cassab for suggestions and Cathie Carteret for technical assistance.