Rescue of Drosophila engrailed mutants with a highly divergent mosquito engrailed cDNA using a homing, enhancer-trapping transposon

Complementation of a genetic defect caused by loss of a vital protein function often requires correct temporal and spatial expression of the therapeutic transgene. Although correction of the mutation by homologous recombination is ideal, it is difficult to achieve in vivo. The random incorporation into the genome of a vector containing the therapeutic gene is easier but has three significant limitations: (1) the regulatory sequences for the gene must be known and incorporated into the vector, (2) expression levels and patterns are often influenced by the insertion site, and (3) insert sizes in most vectors are limited. An alternative approach to obtaining the correct expression pattern was suggested by the discovery of enhancer trapping. In enhancer trapping, a transposon-encoded reporter gene expressed from a minimal promoter is regulated by genomic sequences flanking the insertion site (O’Kane and Gehring, 1987; Fasano and Kerridge, 1988; Bellen et al., 1989; Bier et al., 1989). The expression pattern of the reporter often mimics that of a nearby gene, because the transposon has come under the control of the same regulatory influences as the endogenous gene. This suggests that addition of a transposon containing a functional coding region to an endogenous locus that encodes a nonfunctional protein might complement the mutant phenotype even though the genomic location would be altered by the addition of a transposon. To test whether enhancer-trapping could be used to express a therapeutic gene, we have taken advantage of a method for selective insertion of a Drosophila P-element vector.

Although most P-element derived vectors are thought to insert in the genome in a fairly non-selective manner, there are exceptions to this; specific fragments of regulatory DNA from the Drosophila genes engrailed (en) (Hama et al., 1990; Kassis et al., 1992), polyhomeotic (Fauvarque and Dura, 1993), and Ultrabithorax (Hudson et al., 1995) cause P-transposons to insert in the genome in a selective manner, often near the endogenous gene. This phenomenon of insertion near the parent gene has been called ‘homing’ (Hama et al., 1990). A 2.6 kb fragment of en regulatory DNA (from −2.4 kb to +188 bp) that acts as a minimal promoter for enhancer detection also causes homing of the transposon near en (Kassis, 1990; Kassis et al., 1992).

The en gene is important for many aspects of Drosophila development, including segmentation of the embryo, neurogenesis and formation of the posterior compartments in the adult cuticle (Kornberg, 1981; Morata and Lawrence, 1975). Although en mutations span a genomic region of at least 70 kb of DNA, the transcription unit occupies only about 4.5 kb (Drees et al., 1987). Presumably, the bulk of the DNA in the en locus is required for the regulation of expression in precise patterns throughout development. In addition to the en gene, the Drosophila genome contains a related gene, invected (inv). inv is located adjacent to en, is expressed in a pattern very similar to en, contains a closely related homeodomain, and provides a function redundant to that of en (Coleman et al., 1987; Gustavson et al., 1996).

The en gene is evolutionarily conserved over a diversity of phyla (Joyner et al., 1985; Dolecki and Humphreys, 1988; Fjose et al., 1988; Walldorf et al., 1989; Kamb et al., 1989; Holland and Williams, 1990; Wedeen and Weisblat, 1991; Hui et al., 1992; Webster and Mansour, 1991; Brown et al., 1994). En proteins are distinguished by a C-terminal 99 amino acid domain that includes a homeodomain and flanking amino acids (Joyner and Martin, 1987). We have been studying the en gene from the mosquito Anopheles gambiae, which is estimated to have diverged from Drosophila 180 million years ago. Although the overall sequence identity of Drosophila and mosquito en coding regions is only 35%, there is remarkable conservation within the homeodomain and five other domains. Because of the conservation of these domains, we believed that the mosquito en protein might functionally replace its Drosophila homologue. Therefore, we used the mosquito en cDNA to test the hypothesis that a homing transposon encoding a functional gene could rescue an en mutant. The mosquito en protein is able to rescue en mutant phenotypes when inserted into the inv or en loci. Complete rescue to viable adults is only obtained when the homing transposon is inserted near the endogenous en promoter, suggesting that rescue is achieved only when mosquito en expression precisely mimics Drosophila en expression.

The mosquito en genomic clone was obtained from an Anopheles gambiae library by low stringency hybridization (20% formamide, 2× SSC, 42°C; washing conditions were 2×SSC, 0.1% SDS, 50°C) with a Drosophila en homeobox probe. Approximately 7.5 kb of sequence were generated from two EcoRI fragments cloned into M13 in two orientations. The cDNA was obtained by RT-PCR using RNA from mosquito embryos, and primers from the genomic sequence. Several clones were sequenced to ensure that no PCR errors had occurred.

P[mos-en] was prepared by sequential additions of fragments to pCaSpeR (Pirrotta, 1988). The en promoter and homing fragment (from −2.407 to +212 bp) was amplified using the primers CTCGAATTCCGTTGATATGATCGCA and CTCGGATCCTGGTTTC-GACTTGGTGCTAAATC and cloned as a EcoRI-BamHI fragment. Next, the mosquito en cDNA was cloned from M13 replicative form as an BamHI fragment. Finally 3′ sequences from Drosophila en were amplified using the primers GTGTCTAGAGCGCTGACCAATGGT-TACCA and GTGTCTAGATTTCCCATAATTGTTTTATTTTATTT and cloned as a XbaI fragment.

Constructs were injected into homozygous y w^67c2 (referred to as y w below) embryos using standard procedures with 300 ng/μl of P[mos-en] and 150 ng/μl of pπ25.7wc (Spradling, 1986). Approximately 400 injected embryos gave 51 fertile adults. From these fertile adults, three transgenic flies were obtained. One had an insertion on the X chro-mosome, one had an insertion on the third chromosome, one had an insertion on the second chromosome. The second chromosome insertion (line las1G) was located about 10 kb distal to the en genetic region. P[mos-en] in las1G was mobilized to en mutant chromosomes by two different schemes. For mobilization onto the en^B86 chromosome, progeny of y w; P[mos-en]/en^B86; P[ry⁺, △2-3](99B) (which contains the P-element transposase; Robertson et al., 1988) males crossed to y w virgins were scored for a change in eye color. Flies with a change in eye color were assayed for the presence of the en^B86 mutation by PCR (see below). We examined approximately 3200 flies, selected 39 flies with a change in eye color and recovered 3 insertions on the en^B86 chromosome. From these three insertions, we derived lines 6B and 10B at 48A (Fig. 2B) and line 9 at 47F. For mobilization onto the en^IM199 chromosome, the same general scheme was used except that the en^IM199 chromosome carried three visible markers (Sp Bl Pin^B). Colored-eyed flies that were Sp Bl Pin^B were selected. From approximately 2500 flies, we selected 11 flies and recovered 5 insertions on the en^IM199 chromosome. The presence of the en^IM199 mutation was confirmed by PCR (see below).

P[mos-en] in line 12A apparently disrupted inv expression (see Results). In order to generate a derivative of line 12A which no longer expressed mosquito en, internal deletions of P[mos-en] were obtained by crossing line 12A to flies which express the P-element transposase (Robertson et al., 1988) and selecting for loss of eye color. In one line, 12A-11, the white gene and the mosquito en gene were deleted (as determined by PCR), but the P-element ends were intact and had not moved from the 12A insertion site (as determined by inverse PCR).

Individual Sp Bl en^IM199P[mos-en]-12A Pin^B/SM6a; P[ry⁺,△2-3]99B males were crossed to w; en^B86/SM6a, en^LA10/CyO, or en²⁴/CyO virgins. Vials were scored for the presence of progeny that did not carry SM6a or CyO. These were potentially flies that produced no functional Drosophila en, but were rescued by mosquito en. Such flies were obtained in about one of every ten vials (25-100 total progeny scored per vial). These flies were then further tested by crossing to a variety of en mutants and by PCR to confirm the presence of the en^IM199 mutation.

Individual en^B86P[mos-en]-10B/CyO, P[ry⁺△2-3]99B males or females were crossed to w, en²⁴/CyO flies. Vials were scored as above. Although a similar frequency of rescue was obtained, only one line is discussed here.

The primers used for amplification of DNA flanking the en^IM199 mutation were GCGAGGCGTTGATGGCCCCACTCC and TTCAT-GCGCATCTCGAGCT. PCR was for 35 cycles. Products were phenol-chloroform extracted, ethanol precipitated, resuspended, cut with HhaI, and run on 1.3% agarose gels. HhaI digestion yielded four bands in wild-type en and three bands in en^IM199. The primers used for amplification of DNA flanking the en^B86 mutation (a 53 bp deletion in the first exon of en; Gustavson et al., 1996) were ATGGCCCTGGAGGATCGCTGC and GCAGGAT-GCGGGGCGATCCACT. PCR was for thirty-five cycles (94°C for 1 minute, 72°C for 3 minutes, 55°C for 1.5 minutes) using Buffer D or J from the INVITROGEN PCR optimizer^TM Kit. PCR products were separated on 1.6% agarose gels.

The insertion sites of P[mos-en] were identified by amplifying flanking genomic DNA via inverse PCR (as described by Whiteley et al., 1992), ³²P-labelling of the product, and hybridizing to lambda clones from the 48A region of the chromosome (Kuner et al., 1985). Insertions were further localized by hybridization to subclones from the region. The locations of P[mos-en] near the en promoter in lines 10B-5 and 12A-5R were determined by direct PCR with primers from P-element ends and primers from the first en exon and from upstream of the en transcription start site. In both l0B-5 and 12A-5R, a primer from the 5′ P-element end combined with primers from the first exon gave PCR products and a primer from the 3′ P-element end combined with primers from en upstream sequences gave PCR products.

In situ hybridization was performed essentially as described by Tautz and Pfeifle (1989). Digoxigenin-labelled probes were prepared from PCR products from the first exons of mosquito and Drosophila en. The antibody staining protocol was essentially as described by Patel (1994). Rabbit anti-en antibody (a gift from Charles Girdham and P.H. O’Farrell) was used at a dilution of 1:600. Peroxidase-labelled, anti-rabbit secondary antibody was obtained from Jackson laboratories and used at a concentration of 1:300. The monoclonal antibody 4D09 (Patel et. al., 1989) was used at a concentration of 1:2. A Vectastain elite ABC kit (Vector Labs) was used for detection of the 4D09 antibody as described by Zhang et al. (1996).

The Anopheles gambiae en gene was isolated from a genomic library by low stringency hybridization with a Drosophila en homeobox probe. Two overlapping lambda clones encompassing about 25 kb of DNA were isolated (not shown). From these clones, approximately 7.5 kb of continuous sequence was obtained including the entire mosquito en transcription unit (Fig. 1). This sequence includes about 800 bp of sequence upstream of the presumed translation start site, the entire coding region and two introns. The details of the sequence have been lodged with GenBank (Accession number U42429). The intron/exon organization of the mosquito en gene was determined by comparing the sequences of cDNAs (obtained by RT-PCR using mosquito embryonic RNA) to the genomic sequence. Like Drosophila en, mosquito en contains two introns, located in identical positions in the coding sequence (Fig. 1A). Intron 1 is significantly larger in mosquito en, 4.5 kb, compared with 1.1 kb in D. melanogaster.

The predicted protein sequence of mosquito en was determined from the cDNA. The mosquito en protein shows significant divergence from the Drosophila protein. The overall sequence identity is only 35% and is confined to seven domains (Fig. 1). Four of these domains, the en/inv domains, are found in both the Drosophila en and inv proteins and in all en-class proteins including those from Bombyx mori, Tribolium and in the mouse genes En-1 and En-2. Two domains, the en-specific domains, were previously identified as conserved between Bombyx mori and Drosophila en, but are not present in the Drosophila inv protein. Finally, we identified a new domain, which we refer to as ‘dipteran-specific en’ since it is found in the en gene from the dipteran mosquito and Drosophila, but not in other en or inv genes. Because of the strong conservation of these domains, we believed that the mosquito en protein would be able to substitute for the Drosophila en protein. At the very least, since early mosquito development is similar to early Drosophila development, we reasoned that mosquito en would be able to substitute for Drosophila en during that stage.

melanogaster en fragments extend from −2.4 kb to +188 bp (the fragment that leads to homing; Kassis et al., 1992) and bp +5100 to −5600 (the 3′ untranslated leader and polyadenylation site). (B) P[mos-en] insertions on en mutant chromosomes. The thin line indicates DNA in polytene subdivison 48A. Arrows indicate the position of insertion of P[mos-en] in 7 lines (12A, 12A-5R, 10B-5, 11A, 6B, 10B and 19A). The names of the original insertions are shown in bold. 12A-5R was generated from a remobilization of the 12A insertion and contains the original insertion at 12A and a new insertion (*). 10B-5 was generated from remobilization of 10B and contains the original insertion at 10B and a new insertion (*). The thick black line indicates the extent of the en gene as determined by breakpoint mutations (Kuner et al., 1985). The black boxes indicate the approximate locations of the transcription units of en, inv and three uncharacterized genes (Drees et al., 1987; Kassis, unpublished). The only intron shown is the large inv intron. Somewhere within this intron is a 9 bp microexon (Coleman et al., 1987). The insertion of P[mos-en] in line 12A disrupts inv protein accumulation (see Fig. 3 and 4), possibly by disrupting the splicing of this microexon. 6B, 10B and 10B-5 are inserted on a chromosome which carries the mutation en^B86. 12A, 12A-5R, 11A, and 19A are inserted on a chromosome which carries the

The P-transposon (P[mos-en]) used for these experiments is shown in Fig. 2A. It contains the Drosophila en homing fragment and promoter, the coding region from the mosquito en gene, and the 3′ region of the Drosophila en transcription unit. P[mos-en] was mobilized from a wild-type chromosome onto en mutant chromosomes in a transposase-producing strain (Robertson et al., 1988). Of 8 independent insertions on en mutant chromosomes, 5 were at polytene chromosome subdivision 48A (the genomic location of en) and one was at 47F. The location of the original five insertions at 48A are shown in bold in Fig. 2B. The expression pattern of mosquito en RNA was examined in all lines using in situ hybridization with a mosquito en specific probe. There was no expression of mosquito en in the two lines in which P[mos-en] was inserted outside of the 47F-48A region. In lines 6B, 10B, 11A, and 19A, mosquito en RNA is expressed in very weak en-like stripes in germ band shortening embryos (not shown). This striped expression pattern appeared much later than Drosophila en, and there is no rescue of the en phenotype in any of these lines. In contrast, in line 12A, which has P[mos-en] inserted into the large intron of the inv gene, mosquito en RNA was expressed in a pattern similar to Drosophila en, and many aspects of the en embryonic phenotype were rescued (see below).

The Drosophila en gene is necessary for normal segmentation of the embryo. For this study we have used two en mutant alleles, en^IM199 and en^B86. en^IM199 contains a single base pair change which causes the protein to prematurely terminate within the homeodomain at codon 497 (Gustavson et al., 1996). en^B86 has a 53 bp deletion at codon 75, which causes a frameshift and generates a 104 residue protein (Gustavson et al., 1996). Both mutations produce strong en phenotypes. The phenotype of an en^IM199 embryo cuticle is shown in Fig. 3B. There are pair wise fusions of the abdominal segments, the head fails to involute and there is extensive cell death which leads to a shortened cuticle. In contrast to the severe effects of en mutations, it has been recently shown that flies with mutations in the inv gene are homozygous viable and have no abnormalities (Gustavson et al., 1996). Despite this, inv-en double mutants have a stronger phenotype than single en mutants (Gustavson et al., 1996). During the course of these experiments, a P[mos-en] insertion occurred within the large inv intron (on an en^IM199 chromosome, line 12A), disrupting inv expression and function, creating an inv-en^IM199 double mutant. The evidence for this follows.

The monoclonal antibody 4D09 recognizes a conserved epitope present within the homeodomain of Drosophila en and inv and in the mosquito en protein (Patel et al., 1989). Fig. 4 shows 4D09 antigen in wild-type (A,B), en^IM199 (C,D), and in a derivative of line 12A, 12A-11, which contains an internal deletion of the P[mos-en] transposon and no longer produces mosquito en protein (see Materials and Methods). In wild-type embryos, 4D09 recognizes both en and inv. In en^IM199 embryos, the en protein is truncated and the 4D09 antibody recognizes only the inv protein. This is detected strongly in the hindgut and nervous system in homozygous mutant embryos (Fig. 4C,D). In 12A-11 embryos, 4D09 also sees only the inv protein, due to the presence of the en^IM199 mutation and the deletion of the mosquito en cDNA in this line. 12A-11 embryos make very little 4D09 antigen compared to en^IM199 embryos (compare Fig. 4C,D with 4E,F), suggesting that very little inv protein is made in these embryos. Likewise, the cuticles of 12A-11 embryos are more disrupted than en^IM199 embryos (compare Fig. 3C and 3B). Embryos are shorter and segment fusions are more severe, consistent with the observation that en-inv double mutants have a more severe phenotype than single en mutants. The 12A-11 phenotype is less severe than that seen for a deletion of en-inv region (Gustavson et al., 1996), indicating it is not a complete loss of function. This is consistent with the low level of inv protein present in these mutants.

In 12A homozygotes, which carry the mutation en^IM199 and an intact P[mos-en], 4D09 could recognize both mosquito en and the Drosophila inv proteins if both were expressed. In 12A homozygotes late in embryogenesis, 4D09 antigen is detected in the posterior spiracles, in the last abdominal stripe, and only very weakly in the hindgut (Fig. 4H, compare with a wild-type embryo of the same stage, Fig. 4G). This suggests that the level of inv protein is very low in these embryos, similar to its low level in 12A-11 embryos.

Mosquito en RNA in line 12A was first evident at the beginning of gastrulation (Fig. 5A), a few minutes after Drosophila en could be detected. Throughout embryonic development, mosquito en continued to be expressed in a pattern similar to Drosophila en (Fig. 5B,C), however its expression was weaker than Drosophila en, and some stripes contained single cells or groups of cells which showed no expression (see arrows in Fig. 5B,C). The difference in expression level between Drosophila en and mosquito en was particularly evident when examining the protein level with the monoclonal antibody 4D09. As stated above, 4D09 recognizes mainly the mosquito en protein in line 12A, and both inv and en in wild-type embryos. Fig. 5D shows a wild-type and a homozygous 12A embryo collected and stained at the same time. The pattern of expression is remarkably similar in the two embryos, but the 12A homozygote has a much lower level of expression and some stripes are narrower than wild-type en stripes (see arrows). In addition, while ectodermal en stripes are maintained throughout embryonic development in wild-type embryos, no ectodermal stripes are detected late in development in 12A embryos (except the last stripe, see Fig. 4H vs. 4G).

We examined the effect of mosquito en in line 12A on two different en mutant phenotypes. Functional en protein is required to maintain en ectodermal stripes throughout embryonic development. Thus, in an en point mutant, where en transcript is made, but functional en protein is not, en RNA expression is normal early in development, but fades at about 6.5 hours after egg laying (AEL) (Heemskerk et al., 1991). Fig. 6 shows en RNA expression in a wild-type, en^IM199 mutant, and a 12A embryo at about 6.5 hours AEL. The mosquito en protein in line 12A is able to restore Drosophila en ectodermal stripes in these embryos. The cuticles of homozygous 12A embryos were also examined. Homozygous 12A cuticles looked normal (Fig. 3D) except for some missing denticles, mostly in the first row of each segment (detail not shown). This denticle loss may result from the lack of mosquito en expression in cells within some stripes, or to the absence of mosquito en stripes late in development. Despite the normal appearance of the cuticles, most homozygous 12A embryos fail to hatch and 12A could not complement any of the en alleles tested (Table 1). Since the lethality of homozygous 12A embryos could be due to another embryonic lethal present on the same chromosome, we examined the lethal period of 12A/en^B86 embryos. 12A/en^B86 embryos hatch but die sometime during the larval period. The cuticles of 12A/en^B86 embryos look similar to those from 12A homozygotes (data not shown).

In order to obtain P[mos-en] lines that rescue en mutants to viable adults, we took advantage of the phenomenon of local hopping (Tower et al., 1993; Zhang et al., 1993). When P-elements transpose, they often insert nearby in the genome. P[mos-en] in line 12A was mobilized, and chromosomes that produced viable adults over en null mutations were selected. Six chromosomes were obtained. Five of these carry deletions and rearrangements of DNA flanking the 12A insertion site; these will be described elsewhere. The sixth chromosome, 12A-5R, contains two insertions of P[mos-en], one at the original 12A site and a second insertion 300 bp upstream of the en transcription start site. Mosquito en expression in line 12A-5R was indistinguishable from Drosophila en expression in wild-type embryos in both pattern and level (as judged by reaction with the 4D09 antibody; data not shown). The cuticles of 12A-5R embryos look normal, showing that the mosquito en protein can fully rescue this phenotype when correctly expressed. Finally, the P[mos-en] insertions on chromosome 12A-5R were able to complement many different en alleles: en^B86, en^LA10 and en^IM199 (which contain stop codons within the en-coding regions; Gustavson et al., 1996); breakpoint mutations en^CX1 (which breaks in the first en intron or exon) and en²⁴ (which breaks 20 kb upstream of the en transcription unit). Flies obtained from these crosses look normal, except for minor defects in the posterior compartment of the wings (Fig. 7B,C), and were fertile. In addition, 12A-5R is able to survive over en-inv double mutations including Df(2R)enE, a 50 kb deletion that removes the entire en transcription unit and the 3′ end of the inv transcription unit (Gustavson et al., 1996). 12A-5R/Df(2R)enE flies have wings with severe defects in the posterior compartments (Fig. 7D,E) and both sexes are sterile.

The parental chromosome for 12A-5R carries the en mutation en^IM199, which contains a single base change within codon 497 changing a CAG to TAG (Gustavson et al., 1996) causing the loss of a HhaI site that can be easily detected after amplification of flanking DNA using PCR. DNA samples from adult flies of the genotypes 12A-5R/en^IM199, 12A-5R/Df(2R)enE, Oregon R (wild-type), and en^IM199/SM6a were used for PCR. SM6a is a balancer chromosome that contains a wild-type en gene. HhaI digestion of a 1042 bp PCR fragment including the mutated region gave four bands in wild-type DNA (Fig. 8A). Five bands were generated in en^IM199/SM6a DNA corresponding to both mutant and wild-type en, and only the 3 mutant bands were seen in 12A-5R/en^IM199 and 12A-5R/Df(2R)enE DNA. This confirms that no wild-type allele of en was present in flies of these latter genotypes.

Local hopping was also used to generate P[mos-en] insertions on a chromosome that contained the mutation en^B86. In this case, local hopping of P[mos-en] in line 10B gave rise to a chromosome with the original 10B insertion and an insertion very near the en transcription start site (10B-5). 10B-5 made no Drosophila en, as evidenced by no reaction with an antibody directed against the N-terminal region of en (Fig. 9); in contrast, a strong reaction with 4D09 indicates that mosquito en proteins were made at a high level (not shown). 10B-5 rescued adult viability of all en mutations tested (Table 1), including en-inv double mutations and a deletion, Df(2R)enX31, that removes about 200 kb of DNA (Kuner et al., 1985). In addition, 10B-5 homozygotes survived and, though sterile, look wild-type. en^B86 contains a 53 bp deletion of DNA within the first exon of en. PCR of DNA flanking this deletion confirmed that no wild-type Drosophila gene was present in 10B-5/Df(2R)enX31 adults (Fig. 8B). The wing phenotypes observed in 10B-5/Df(2R)enE flies (Fig 7B,C) were less severe than those observed in 12A-5R/Df(2R)enE (Fig. 7D,E), indicating a more complete rescue with 10B-5 than with 12A-5R. We believe that the inv gene is inactivated in 12A-5R, due to the presence of the 12A insertion and that the intact inv gene on the 10B-5 chromosome leads to more complete rescue.

The 10B-5 chromosome rescued both Df(2R)enE and Df(2R)enX31, while 12A-5R only rescued the former (Table 1). Both deletions lead to a complete loss of en and inv proteins, however, Df(2R)enE leaves 40 kb of en regulatory DNA intact. We suggest that en regulatory DNA on the Df(2R)enE chromosome can activate the expression of P[mos-en] in 12A-5R by a transvection phenomenon (Judd 1988), leading to rescue. This regulatory DNA is not present on the Df(2R)enX31 chromosome.

This study yielded three principle findings. First, gene expression by enhancer-trapping is precise enough to restore viability to Drosophila en mutants. Second, the highly divergent mosquito en protein is able to substitute for the Drosophila en protein during all stages of development. Finally, our results suggest that inv can play a role in proper development. This conclusion is consistent with that reached by Gustavson et al. (1996).

The en gene of Drosophila is a member of the segment-polarity class of developmental genes. It is expressed throughout development, in a dynamic pattern, and as much as 70 kb of DNA may contribute to its proper expression pattern. Although some of the sequences necessary for correct regulation have been identified, only partial rescue of the en phenotype has been achieved (Smith and Jaynes, 1996). In this report we have restored adult viability to two different lethal en mutants by combining the techniques of enhancer-trapping and P-element homing.

A fragment of en DNA has been identified which acts as both a minimal promoter for enhancer-trapping and causes P-elements to ‘home’ to polytene chromosome bands 47F/48A, the location of the endogenous en gene and the related gene inv (Hama et al., 1990; Kassis et al., 1992). Although the mechanism for homing is unknown, we hypothesize that proteins bound to en regulatory DNA present in the transposon interact with proteins bound to the endogenous en locus, concentrating the transposon in that region of the genome. Transposition then occurs locally. In this study we used the vector P[mos-en], which contains the mosquito en cDNA cloned behind the en fragment which contains a minimal promoter and causes P-element homing. We obtained eight P[mos-en] insertions in 48A and one in 47F. Like previous studies on P-element homing, we note that insertion at 48A is not site-specific, but occurs over a large region: insertions are distributed over about 200 kb. In all nine lines, mosquito en expression resembled Drosophila en at some stage of development. Thus, we suggest that this 200 kb region represents a chromosomal domain: an en promoter in P[mos-en] present anywhere in this domain can interact with at least some en enhancers. Mosquito en was not expressed in an en-like pattern when P[mos-en] was inserted elsewhere in the genome.

Although it was easy to obtain insertions of P[mos-en] into polytene chromosome band 48A, rescue of the en phenotype was only achieved when P[mos-en] was inserted in the inv or en gene. Further, the insertion in the inv gene led only to embryonic rescue. Complete rescue to adulthood was only achieved by insertions of P[mos-en] just upstream of the en promoter. We suggest that there may be constraints on promoter-enhancer interactions in vivo which only allow accurate, high level expression when the transposon-encoded promoter is located in the same place as the endogenous promoter. We are currently trying to move P[mos-en] to different locations within the en locus to test this hypothesis.

Rescue by enhancer-trapping could be tried for a few other genes in Drosophila. DNA from both the polyhomeotic and the Ultrabithorax loci have been shown to mediate P-element homing (Fauvarque and Dura, 1993; Hudson et al., 1995). In addition, the same en fragment that causes transposons to home to polytene chromosome band 48A also seems to direct P-transposons to the escargot locus (Whiteley et al., 1992; Kassis, 1994). Genes which are ‘hot-spots’ for P-element insertion are also candidates for this approach. While a large number of lines might need to be screened to obtain the line of interest, it should be noted that insertion at some sites in the genome will not be allowed because mis-expression will lead to embryonic lethality.

The en gene encodes a transcription factor with a highly conserved homeodomain. D. melanogaster contains two en-class genes, en and inv, which are expressed in similar patterns during development (Coleman et al., 1987; Gustavson et al., 1996). Some insects such as the silk moth Bombyx mori and the honeybee also contain two en-related genes (Hui et al., 1992; Walldorf et al., 1989), whereas the more primitive insects, grasshopper and Tribolium contain only one en-related gene (Patel et al., 1989; Brown et al., 1994). Sequence comparisons between the Drosophila and the Bombyx genes have identified 4 domains shared by inv and en, two en-specific domains and one inv-specific domain (Hui et al., 1992). Vertebrate species also have two en-related genes (Mouse En-1 and En-2 are shown in Fig. 1B), which contain only those domains shared by inv and en. en/inv domain 1 has recently been shown to mediate active transcriptional repression and is also present in some other homeodomain-containing proteins (Smith and Jaynes, 1996). We have sequenced the en gene from a mosquito, Anopheles gambiae. We chose the mosquito because, though evolutionarily very divergent, early mosquito development is similar to early Drosophila development and we reasoned that the function of the en protein would be conserved. We believe that the gene we have cloned is the true en homologue since it contains the two en-specific domains identified by Hui et al. (1992) but not the inv-specific domain. Although we did not identify an inv homologue in this work, our genomic Southerns suggest that the Anopheles gambiae genome contains two en-related genes (data not shown). In addition to the domains previously identified, one additional domain was conserved between mosquito en and Drosophila en which we refer to as Dipteran specific.

The mosquito en protein functions remarkably well in Drosophila. Table 1 shows that two rescue chromosomes obtained, 12A-5R and 10B-5, rescue adult viability of many different en mutants, including Df(2R)enE, a deletion which removes both en and inv. 10B-5 rescued more completely, yielding fertile 10B-5/Df(2R)enE adults with only slight defects in the wing veins of the posterior compartment. Further, 10B-5 survives as a homozygote and has phenotypically wild-type wings. However, the mosquito en protein probably does not function in Drosophila as well as the Drosophila en protein. There are no defects in the wings of wild-type en/Df(2R)enE flies. However, since we have not rescued a Drosophila en mutant with a transposon-encoded Drosophila en gene, we cannot rule out the possibility that the transposonencoded gene does not function as well as the endogenous gene because of slight differences in expression pattern or level.

Our result on the conservation of en function is consistent with the results of Hanks et al. (1995) who reported that the mouse En-1 protein can be completely substituted by the mouse En-2 protein, even though the proteins from these genes share only 55% amino acid identity. The Drosophila and mosquito en proteins share only 35% identity, emphasizing the flexibility of sequences outside of the conserved domains. Given these results, it is interesting to speculate about whether a mouse en homologue could function in Drosophila. Examination of the conservation of the en/inv domains between Drosophila, mosquito and mouse (Fig. 1B) suggests that a mouse en protein may not function as well in Drosophila as the mosquito protein does. Although the homeodomain is highly conserved between these three diverse organisms, the other domains are much more highly conserved between Drosophila and mosquito, perhaps reflecting more common function of the Drosophila and mosquito en proteins.

There have been several reports of rescue of various Drosophila developmental mutants with highly divergent proteins (Muller et al., 1995; Lutz et al., 1996; Xue and Noll, 1996). In one extreme example, Xue and Noll (1996) report that the Drosophila gooseberry and the mouse Pax3 proteins can partially substitute for the Drosophila protein, paired. Remarkedly, two copies of a Drosophila gooseberry transgene can restore adult viability to some embryos with the lethal mutation paired. They suggest that it is the evolution of new cis-regulatory regions rather than the divergence of protein coding regions which has led to the diversification of gooseberry, paired and Pax3 function. A similar conclusion was reached by Hanks et al. (1995) on the ability of En-2 to replace En-1 in mice. Our results reinforce the view that changes in cis-regulatory regions may play a bigger role in evolution than protein divergence, at least for homeodomain-containing proteins. In the course of these experiments, we obtained a number of insertions of P[mos-en] which cause dominant defects in the adult cuticle. We presume these defects are caused by expression of mosquito en in tissues where Drosophila en is not normally expressed. Thus, the en protein has acquired new activities, by the acquisition of a new cis-regulatory regions via enhancer-trapping.

The inv gene has recently been shown to be dispensable for viability of D. melanogaster, yet en-inv double mutants have a more severe phenotype than en mutants (Gustavson et al., 1996), suggesting that inv can play a role in development. In this study we generated a P[mos-en] insertion into the large intron of inv (in line 12A), creating an inv mutation on an en^IM199 chromosome. inv protein levels are severely reduced in 12A embryos (Fig. 4H). Deletion of the mosquito en cDNA from line 12A led to the creation of an inv/en double mutant chromosome (line 12A-11). The cuticle phenotype of this double mutant is more severe than en^IM199, but less severe than that reported for Df(2R)enE, a deletion of inv and en. In fact, some inv protein is made in 12A-11 embryos (Fig. 4E,F). It is not known how the P[mos-en] insertion in the inv intron disrupts inv expression. In situ hybridization of 12A embryos with a probe to the first exon of inv did not show any decrease in the transcript levels (data not shown). The inv transcript contains a 9 bp microexon at an unknown location within the large intron (Coleman et al., 1987). It is possible that P[mosen] in 12A disrupts the splicing of this exon, and that either there is a minor class of inv transcript which does not contain this exon or that a low level of splicing can occur in 12A and 12A-11 embryos.

The presence of an intact inv gene seems to contribute to rescue by the mosquito en gene. Line 12A-5R, which contains P[mos-en] inserted on the en-inv double mutant chromosome described above, survives over all single en mutants tested, as well as over the double mutant Df(2R)enE. However, 12A-5R/Df(2R)enE flies have wings with severely disrupted posterior compartments and are sterile. Another rescue line, 10B-5, which contains an intact inv gene, also survives over Df(2R)enE. However, 10B-5/Df(2R)enE flies are fertile and have only minor defects in the posterior compartment.

en and inv are reported to be expressed in similar patterns (Gustavson et al., 1996), yet P[mos-en] in line 12A only rescued the en mutant phenotype in embryos. This suggested to us that 12A expression might differ from en expression, and that this might lead to larval lethality. We examined the expression of inv in wild-type embryos using a probe to the first exon of inv and compared it to mosquito en expression in 12A embryos and to Drosophila en expression in wild-type embryos (data not shown). We noted both differences and similarities. Mosquito en RNA in line 12A was evident earlier than inv, at about the same time as Drosophila en. inv expression was not evident until about 30 minutes later. This difference is probably due to a difference in transcript sizes: the inv transcript extends over about 25 kb and would take considerably longer to be transcribed than the 4.5 kb en mRNA and the 1.8 kb mosquito en mRNA. Like mosquito en in line 12A, we note that inv stripes in wild-type embryos seem to be narrower than Drosophila en stripes, and that sometimes gaps occur within inv stripes (data not shown). Thus, although similar, en and inv expression are not identical. We believe there may be other subtle differences in inv and en expression which explain why P[mos-en] in line 12A cannot rescue en mutants to viable adults. Likewise, the difference in inv and en expression patterns may explain why inv cannot substitute for en during Drosophila development.

We thank Elizabeth Gustavson and Tom Kornberg for sharing information on the sequence of en mutations prior to publication, for en mutant flies, and for interesting discussions; Charles Girdham and Pat O’Farrell for the anti-en antibody; Nipam Patel for the monoclonal antibody 4D09; Louis Miller and members of his laboratory for providing Anopheles gambiae embryos and adults; Pat Romans for A. gambiae DNA; Ward Odenwald for examining the nervous systems of 12A embryos; Jim Jaynes for examining the cuticles of homozygous 12A embryos; and Lesley Brown, Jim Jaynes, Ward Odenwald, Pete Mathers, Jim Kennison, Diane Mucci, and Philip Noguchi for comments on this manuscript. The mosquito en genomic clone was isolated by Rebecca Clayton. Finally, we thank Philip Noguchi for continuing to support our Drosophila work at CBER. This work was supported in part by a grant on gene therapy from the Cystic Fibrosis Foundation.