The expression of pannier and achaete-scute homologues in a mosquito suggests an ancient role of pannier as a selector gene in the regulation of the dorsal body pattern

During the past two or three decades, investigation into the genetic control of development has shown that, in the model organism, Drosophila melanogaster, the body is progressively subdivided into smaller and smaller domains, called compartments (Garcia-Bellido et al., 1973). Patterning within each domain then proceeds relatively independently. Compartments are derived from groups of cells, or polyclones, that are subsequently defined by lineage (Crick and Lawrence, 1975). The process of compartmentalisation and patterning within compartments is regulated by selector genes such as engrailed, apterous and the genes of the Bithorax Complex (BXC) (Blair, 1993; Diaz-Benjumea and Cohen, 1993; Garcia-Bellido, 1975; Lewis, 1978; Morata and Lawrence, 1975). The boundaries between compartments have been shown to be sources of signalling molecules that coordinate growth and patterning of each compartment (Lawrence and Struhl, 1996). Conservation of this developmental strategy is not clear since compartments in other animals have been difficult to demonstrate. Recently, however, evidence has been obtained for a new class of selector genes whose activity is not confined to lineage-based compartments (Calleja et al., 2000; Mann and Morata, 2000). They behave similarly to classical selector genes and generate morphological differences between different parts of the body plan (Mann and Morata, 2000). Two such genes are pannier (pnr) and the genes of the iroquois complex (iro-C; araucan/caupolican), that encode transcription factors of the GATA and homeobox-containing protein families respectively (Cavodeassi et al., 2001; Gomez-Skarmeta et al., 1996; Ramain et al., 1993).

pannier is expressed in a longitudinal, dorsal domain extending from the head to the end of the abdomen in both larvae and imagos and is involved in the subdivision of the dorsal component of each segment (Calleja et al., 2000; Maurel-Zaffran and Treisman, 2000). It acts in combination with engrailed and the genes of the BXC to specify the identity of the dorsal, medial domain (Calleja et al., 2000). The lateral domain is specified by the expression of the iro-C genes (Calleja et al., 2000; Diez del Corral et al., 1999; Kehl et al., 1998). The most obvious pattern elements of the notum are the large sensory bristles that arise in a stereotyped pattern as a result of the spatially regulated expression of the achaete-scute (ac-sc) genes (Cubas et al., 1991; Ghysen and Dambly-Chaudiere, 1988; Romani et al., 1989; Skeath and Carroll, 1991). Both Pnr and the iro-C gene products have been shown to activate transcription of ac-sc in Drosophila, and this regulatory function of the iro-C proteins appears to have been conserved in Xenopus (Garcia-Garcia et al., 1999; Gomez-Skarmeta et al., 1996; Gomez-Skarmeta et al., 1998; Haenlin et al., 1997).

There are about 60,000 species of true flies many of which display species-specific bristle patterns that differ from that of Drosophila (McAlpine, 1981; Simpson, 1999). Dipteran flies thus provide a convenient model group in which to investigate evolutionary changes in the regulation of expression of ac-sc by the selector genes of the iro-C and pnr. The more derived species of cyclorraphous Diptera, such as Drosophila, Ceratitis capitata and Calliphora vicina, display stereotyped bristle patterns. These result from the expression of ac-sc in discrete proneural clusters or stripes, corresponding to each bristle or bristle row (Cubas et al., 1991; Pistillo et al., 2002; Romani et al., 1989; Simpson et al., 1999; Skeath and Carroll, 1991; Sturtevant, 1970; Wülbeck and Simpson, 2000). In Drosophila this complex spatial expression relies on a number of cis-regulatory elements scattered throughout the ac-sc gene complex (ASC) (Gomez-Skarmeta et al., 1995). These are likely to be conserved in Ceratitis and Calliphora, together with the function of pnr, which is expressed in an identical medial dorsal domain in all three species (Pistillo et al., 2002; Wülbeck and Simpson, 2000).

The Nematocera comprises a group of basal Dipteran species in most of which the bristles are randomly positioned on the notum (McAlpine, 1981; Simpson et al., 1999). A few families, such as the Culicidae, do include species with a simple arrangement of bristles into two or three rows on the notum, most of the body being densely covered with sensory scales (Stone, 1981; McIver, 1975). Here we examine the expression patterns of pnr and an ac-sc homologue, Ag-ASH, on the notum of Anopheles gambiae (Culicidae), a vector of the malaria-causing parasite. We find that, on the medial notum, Ag-ASH is expressed in very broad domains coincident with domains of expression of Ag-pnr. This suggests that activation of Ag-ASH by Ag-Pnr has been conserved. Indeed expression of Ag-pnr in Drosophila mimics the effects of mis-expression of Dm-pnr, and causes the development of ectopic bristles. The coincident expression domains of Ag-pnr and Ag-ASH suggest that activation of Ag-ASH may not require the complex modular promoter characteristic of the ASC of Drosophila. We hypothesise that duplication of the ASC genes, acquisition of position-specific cis-regulatory sequences, and regulatory co-factors for Pnr, may only have been co-opted after the separation of the Nematocera and the Brachycera, some 200 million years ago, and may have allowed the evolution of stereotyped bristle patterns. Interestingly, all of the bristles on the medial notum of Anopheles appear to arise along the borders of the Ag-pnr (and Ag-ASH) expression domains. This indicates that pnr may specify the dorsal bristle pattern in both Drosophila and Anopheles, but in quite different ways.

Screening for homologues of the achaete-scute (ac-sc) and pannier (pnr) genes was performed under low stringency at 42°C with 20% formamide containing standard hybridisation buffer. Washes were carried out at 50°C with 2× SSC, 1% SDS. A genomic Anopheles library and a λZap cDNA library of non-infected fourth instar larvae (from Drs Larry Zwiebel and Fotis Kafatos, EMBL-Heidelberg) was plated and nylon replica filters (PALL, Biodyne A) screened with probes for ac-sc or pnr respectively, as described previously (Wülbeck and Simpson, 2000). Two overlapping genomic phages (AA2J1 and AA2J2) were isolated, and a 2.8 kb XhoI fragment containing the transcription unit of the entire ac-sc homologue, Ag-ASH, was subcloned in pBluescript SK+. To isolate an Ag-ASH cDNA, the fourth larval instar cDNA library was re-screened by using a probe specific to the Ag-ASH bHLH-encoding domain, which was generated using degenerated primers as described previously (Wülbeck and Simpson, 2000). Two phages were isolated and in vivo excised; one of them (pBS-AC3K1: 2116 bp) contained the entire protein encoding region. For pnr, five phages were isolated and in vivo excised but sequence analysis showed that only one of them (pBS-PAC3F1: 3285 bp) encoded a pnr homologue, Ag-pnr. Sequence analysis was performed as described previously (Wülbeck and Simpson, 2000) using the GCG programme. All sequences have been submitted to GenBank (Ag-ASH: AF395079, Ag-pnr: AF395080).

Mosquito larvae were kindly provided by members of the laboratory of Professor Fotis Kafatos, at the EMBL. Larvae were reared in humid chambers at room temperature and fed with cat food. Newly eclosed Anopheles adults were dehydrated and mounted in Euparal for microscopic analysis.

Digoxigenin-labelled RNA probes (DIG-UTP; Boehringer Mannheim) were generated using the standard protocol of Boehringer Mannheim. The resulting RNA was resuspended in 100 μl preHyb solution (50% formamide, 5× SSC, 0,1% Tween 20, pH 6.0). RNA was transcribed from linearized DNA templates: Ag-pnr pBS-PAC3F1 (T3 sense, T7 antisense), Ag-ASH pBS-AC3K1 (T3 sense, T7 antisense).

In situ hybridisation was performed (Wülbeck and Simpson, 2000) with some modifications. Incubation with proteinase K was for 5 minutes at room temperature and incubation with anti-digoxigenin alkaline phosphatase-coupled antibody (Boehringer Mannheim) was performed overnight at 4°C instead of for 2 hours at room temperature.

Larvae and pupae for RNA in situ hybridisation were dissected in ice cold PBS and fixed as described previously (Wülbeck and Simpson, 2000), then stored in 100% methanol at –20°C. For antibody staining Anopheles larvae and pupae were boiled for 5 minutes in PBS and the cuticle removed when possible. Staining was performed immediately afterwards using standard procedures and dilutions of 1/200 in 10% foetal calf serum (FCS) for the primary antibody (rabbit anti-horseradish peroxidase (HRP); Jackson) and 1/200 in 10% FCS for the secondary antibody (anti-rabbit coupled with HRP; Jackson). DAB staining was performed using standard protocols.

A full-length cDNA EcoRI-XbaI fragment of Ag-ASH (pBS-AC3K1) and a full-length cDNA SpeI-KpnI fragment of Ag-pnr (PAC3F1) were each subcloned into the corresponding restriction sites of the pUAST vector under the control of the Drosophila HSP70 minimal promoter. Germline transformants were obtained as described previously (Rubin and Spradling, 1982). Three independent lines were established. Expression of UAS-Ag-ASH was driven by GAL4-pnr^MD237 and that of UAS-Ag-pnr by Gal4-ap, C765, sca⁵³⁴, pnr^MD237, MD455 and MD410 (Brand and Perrimon, 1993; Calleja et al., 1996; Gorfinkiel et al., 1997; Garcia-Garcia et al., 1997). Standard procedures were used for X-gal staining. Flies were dehydrated and mounted in Euparal for microscopic analysis.

The ac-sc complex (ASC) of Drosophila comprises four genes: ac, sc, lethal of scute (l’sc) and asense (ase), which encode transcription factors of the basic helix-loop-helix (bHLH) family of proteins (Alonso and Cabrera, 1988; Gonzalez et al., 1989; Villares and Cabrera, 1987). Screening of a genomic library under moderately stringent conditions (see Materials and Methods) with a PCR-generated fragment containing the bHLH domain of the ac gene of Drosophila virilis uncovered two overlapping genomic phages encoding a single ac-sc homologue (Ag-ASH). Re-screening of a cDNA library of fourth instar larvae, with a PCR-generated probe from the Ag-ASH under moderately stringent conditions, allowed only the recovery of the same sequences.

Ag-ASH appears closest to Drosophila l’sc (l(1)sc – FlyBase). The complete protein sequence of Ag-ASH is compared with that of l’sc from Drosophila melanogaster and an ac-sc homologue from the butterfly Precis coenia, in Fig. 1A. Sequence analysis revealed that 81% of the amino acids in the bHLH domain are identical to those of the Drosophila l’sc protein. Outside of this functional domain, amino acid sequence conservation is low (ranging from 20-27% for the amino (N)-terminal portion to 25-38% for the carboxy (C)-terminal part). A single stretch of 15 conserved amino acids, which appears to be restricted to insect ac-sc proteins, can be seen at the C terminus (shaded blue box). The central tyrosine of this sequence has changed in the butterfly Precis coenia (Galant et al., 1998).

The pnr gene of Drosophila comprises two zinc fingers characteristic of the GATA family of transcription proteins, and a C-terminal domain bearing two α helices (Ramain et al., 1993). The 537 amino acid sequence of the Ag-pnr protein is shown in Fig. 1B, together with Pnr from Drosophila and Ceratitis capitata (Wülbeck and Simpson, 2000). It contains two zinc fingers that are very strongly conserved. The proteins are, however, quite divergent in the C-terminal domain. The proteins of Ceratitis and Anopheles carry a single α helix, in contrast to the two in Drosophila.

Ag-ASH displays strong proneural activity when expressed in Drosophila. We made use of the GAL4-UAS system and the driver pnr^MD237, to express Ag-ASH in the medial half of the notum. This leads to the formation of an excess of bristles in this region (Fig. 2A,B). These bristles are characteristic of the large bristles, or macrochaetes, of Drosophila.

In order to examine the effects of expression of Ag-pnr in Drosophila, three Drosophila lines carrying the UAS-Ag-pnr were established. Two of these exhibited mild phenotypes known to be associated with gain of function of Dm-pnr, such as a slight midline cleft and additional DC and SC bristles (Haenlin et al., 1997; Heitzler et al., 1996). Each of these lines was crossed to several Gal4 drivers (see Materials and Methods). Over-expression of either Dm-pnr or Ag-pnr in the medial notum where the endogenous Dm-pnr gene is expressed, using Gal4pnr, was without effect on the bristle pattern, other than a slight cleft and an occasional additional scutellar bristle (Calleja et al., 2000), not shown. Two of the lines, C765 and Gal4apterous, drive expression in the entire notum (Calleja et al., 1996; Gorfinkiel et al., 1997). The C765 driver resulted in flies of a uniform phenotype: a tuft of additional DC bristles situated laterally on the notum (Fig. 2C). This phenotype is very similar to that seen after ectopic expression of Dm-pnr in C765/UAS-Dm-pnr flies (Garcia-Garcia et al., 1999). The Gal4ap driver gave rise to flies with a range of phenotypes depending on the UAS-Ag-pnr line. These ranged from a loss of just one or both DC bristles with a deformed scutellum to loss of most of the notum. Flies from one of the lines were devoid of structures present on the lateral notum and developed only structures typical of the medial pattern that appeared normal apart from an excess of DC bristles (Fig. 2D). This phenotype is almost identical to that seen after ubiquitous expression of Dm-pnr in Gal4-ap/UAS-Dm-pnr flies (Calleja et al., 2000).

Many features of the life cycle of Anopheles are ancestral and characteristic of Nematocera (Clements, 1992). There are four larval instars after which the animal moults to a free-swimming pupa. The duration of larval development was variable under our laboratory conditions, but most animals pupated after about 12 days at room temperature. As in most Nematocera, the appendages develop from simple imaginal discs that are little more than invaginated pouches attached to the body wall (Clements, 1992). These can be seen at the second larval instar (Fig. 3A). Although the discs are enclosed in a peripodial membrane, their stalks do not close. The wing buds are situated laterally on either side of the larval mesothorax. By the fourth instar, the adult appendages have evaginated into a space outside the larval epidermis and lie flat against the body wall (Fig. 3C). The trunk of much of the adult body does not arise from imaginal discs. The abdomen of the larva, pupa and imago is made from the same epithelium that secretes successive cuticles at each moult (Clements, 1992). Thus, in Anopheles, the outline of the imaginal body is already present at pupation. Consequently the pupal period is short, lasting little more than 24 hours. This contrasts with cyclorraphous Schizophora, such as Drosophila, where it may last 5 days or more, during which time the larval body is destroyed and an entirely new adult body constructed from the imaginal discs and histoblasts.

We have investigated the origin of the dorsal mesonotum of Anopheles. It is not derived from the wing imaginal discs, but from a small epidermal thickening at the junction between the larval body wall and the wing pouch (Fig. 3B). Throughout the third and fourth larval instars, the cells of this region progressively grow across the dorsal part of the body just beneath the larval epidermis. The two edges eventually meet and close at the dorsal midline (Fig. 3C). Closure is always complete before the larval-pupal moult, but the rate of growth and the time of closure vary between individuals. The larval epidermis is gradually destroyed as the imaginal one expands (Fig. 3D). During larval stages, the epithelium of the future notum is very compact and the cells are tall and columnar. At pupation, the cells flatten out leading to an increase in surface area before secretion of the pupal cuticle.

The adult notum of Anopheles displays large sensory bristles as well as many small sensory scales (McIver, 1975; Stone, 1981) (Fig. 4A). The medial half of each heminotum bears two rows of bristles named the acrostichal (AC) and dorsocentral (DC) rows. In addition there is a small transverse row of prescutellar (PST) bristles. The lateral part of each heminotum bears a band of antealar bristles and the scutellum a row of scutellar (SC) bristles. The number of bristles in each row varies considerably between individuals (Simpson et al., 1999). Numbers and positions of scales are quite variable between individuals. However, scales do not cover the entire notum and are consistently found in specific regions: between the AC and DC bristle rows, around the positions of the PST and SC bristles and close to the lateral intercalary bristles (Fig. 4A). They are more or less absent from the area between the DC and antealar bristles. Scales are much smaller than bristles, each scale is composed of a socket and a short pedicel followed by a flattened blade (Fig. 4B). Both bristles and scales are innervated (Fig. 4C). It has been suggested, on the basis of expression of an ac-sc homologue, that the scales of butterflies are analogous to the bristles of flies (Galant et al., 1998).

The domains of expression of Ag-pnr and Ag-ASH were examined by means of in situ hybridisation. Both genes are expressed on the notum, but only Ag-ASH is expressed in the wing pouch and on the legs (not shown). We found that dorsal closure and patterning of the notal epithelium, proceed at different rates in different individuals, such that expression of Ag-pnr and Ag-ASH, as well as the appearance of sensory organ precursors, may start in the fourth larval instar or only after the pupal moult. Thus patterning of the notum could not be timed with respect to external larval morphology. We therefore looked at a large number of fourth instar larvae and pupae and staged them with respect to the expression domains of Ag-pnr, Ag-ASH and to HRP staining, that are seen after completion of dorsal closure. We were able to observe a clear progression of patterning events.

We have concentrated on the pattern in the medial notum where Ag-pnr and Ag-ASH are expressed in domains that appear to be identical. Expression is first evident in two longitudinal bands, one on either side of the dorsal midline that is itself devoid of expression (Fig. 5E). When viewed through the cuticle of fourth instar larvae, these domains appear to be quite broad (Fig. 5E), but after pupation and the subsequent cell shape changes of the epithelium, they appear as longer, thinner bands (Fig. 5A). These bands extend from the anterior border of the notum to almost the level of the future scutal-scutellar suture. They subsequently fade and three other expression domains appear in rapid succession: a small, posterior, triangle that straddles the dorsal midline (Fig. 5F) and that gradually transforms to a kidney-shaped domain (Fig. 5B,C,G) and a crescent-shaped domain along the future scutellum (Fig. 5D). Ag-ASH, but not Ag-pnr, is also expressed in another broad longitudinal domain on the lateral part of the notum (not shown).

As the expression domains fade, single, stained cells appear in their wake (Fig. 6A-C). These are also characterised by the expression of both Ag-pnr and Ag-ASH. We believe these correspond to the sensory organ precursors for the following reasons. The stained cells are always spaced apart from one another. They are of two sizes, large and small. The large cells are positioned at the sites of the future bristles; this is particularly evident for the AC and DC precursors whose numbers correspond roughly to the numbers of bristles found in imagos. The small, stained cells are only found in the areas corresponding to the expression domains of Ag-ASH (and Ag-pnr). These domains correlate with the areas covered by scales in the imago. No stained cells are found outside the domains of Ag-ASH expression, which are also devoid of scales and bristles in the imago. We also looked at neural differentiation with the anti-HRP antibody that labels neurons and axons (Jan and Jan, 1982). Staining was apparent in cells at similar positions to the presumed precursors and the imaginal sensory organs (Fig. 4C and Fig. 6D). Within any specific domain, sensory organ precursors appear to arise in a stochastic fashion, no particular order was apparent. Precursors of bristles and scales also appear to arise simultaneously, but in some preparations stained with anti-HRP, only bristle neurons are labelled, suggesting that neural differentiation may start earlier in bristles than in scales (Fig. 6D). Shortly after segregation of the precursors, both Ag-pnr and Ag-ASH stainings disappear.

The positions of bristles appear to coincide with the borders of the expression domains of Ag-pnr and Ag-ASH. We could not observe simultaneous staining of the broad domains of expression as well as the precursors, since the latter only appear as the domains are fading. However the positions of all bristle precursors correlate with the edges of the successive domains. The AC and DC bristles are situated along the medial and lateral borders of the first longitudinal band of expression. The posterior transverse AC precursors are positioned along the anterior edge of the triangle, the PST precursors along the anterior edge of the kidney-shaped domain and the SC precursors along the anterior edge of the scutellar crescent (Fig. 5).

Precursors of the sensory scales arise both on the borders and inside the Ag-ASH expression domains.

Cyclorraphous flies of the Schizophora, Calliphora vicina, Ceratitis capitata and Drosophila spp., possess three or four ASC genes (Alonso and Cabrera, 1988; Gonzalez et al., 1989; Pistillo et al., 2002; Villares and Cabrera, 1987; Wülbeck and Simpson, 2000). Each of the ASC genes of Drosophila is regulated independently and expressed in different, albeit overlapping domains of the developing nervous system (Martin-Bermudo et al., 1991; Romani et al., 1989; Ruiz-Gomez and Ghysen, 1993). Their products remain, however, largely functionally redundant (Balcells et al., 1988; Brand et al., 1993; Dominguez and Campuzano, 1993; Hinz et al., 1994; Rodriguez et al., 1990; Skeath and Doe, 1996). Anopheles belongs to the sub-order Nematocera, composed of basal species of Diptera that display a number of ancestral features (McAlpine, 1981). Our screening procedure allowed the isolation of a single Anopheles ASC homologue, Ag-ASH, but examination of the recently published genome of this species reveals the existence of an asense gene. Ag-ASH is closest to Drosophila l’sc, but may representative of an ancestral gene, which was present prior to the duplication events that gave rise to l’sc, sc and ac (Skaer et al., 2002). This may have taken place after separation of the Nematocera (including the mosquitoes) and Brachycera (including Drosophila and Ceratitis), two lineages that diverged about 200 million years ago. A single ASC homologue has been described in the butterfly Precis coenia (Galant et al., 1998).

When expressed in Drosophila, Ag-ASH has a conserved and strong, proneural function.

Several observations argue in favour of a role for Ag-Pnr in the activation of Ag-ASH. First, the two genes are expressed in what appear to be identical domains in the medial notum. Secondly, both genes are also expressed in sensory organ precursors. In the cyclorraphous flies examined to date, pnr is not expressed in bristle precursors (Pistillo et al., 2002; Ramain et al., 1993; Wülbeck and Simpson, 2000). Thirdly, expression of Ag-pnr is able to mimic the effects of mis-expression of Dm-pnr in Drosophila. Thus, when expressed in the lateral notum Ag-pnr elicits the development of ectopic DC bristles, strongly suggesting that it can activate the Drosophila ac-sc genes. Therefore we think it probable that regulation of ac-sc genes by Pnr has been conserved throughout the Diptera.

In Drosophila, pnr is expressed in a conserved broad medial domain but activates ac and sc in discrete proneural clusters within this domain (Cubas et al., 1991; Garcia-Garcia et al., 1999; Ramain et al., 1993; Romani et al., 1989; Skeath and Carroll, 1991) The ac-sc genes of Drosophila are organised into a complex containing multiple enhancer regions, each of which independently regulates expression in one or a small number of proneural clusters (Gomez-Skarmeta et al., 1995; Ruiz-Gomez and Modolell, 1987). In this species three proneural clusters arise in the domain of pnr expression and Pnr has been shown to directly activate ac-sc in the dorso-central cluster, through binding to a cis-regulatory sequence just upstream of ac (Garcia-Garcia et al., 1999; Haenlin et al., 1997). It is not entirely understood how the broad domain of Pnr is translated into the small clusters of ac-sc expression, but this is at least in part achieved through interaction of Pnr with regulatory co-factors (Cubadda et al., 1997; Haenlin et al., 1997; Ramain et al., 2000). The spatially complex expression of sc in Calliphora and Ceratitis suggests that the ASC genes of these species may also have modular promoters (Pistillo et al., 2002; Wülbeck and Simpson, 2000). Furthermore, the expression domain of pnr in these species is conserved with that of Drosophila (ibid).

In contrast, the regulatory interactions between the two genes appear to have diverged in Anopheles since Ag-ASH is expressed in all Ag-pnr-expressing cells. The common domains of expression suggest that Ag-Pnr may activate Ag-ASH in every cell in which it is expressed, in a simple straightforward fashion. This observation raises two possibilities. First, for the regulation of Ag-ASH, Ag-Pnr may not associate with the various co-factors known to modulate its activity in Drosophila. Second, in order to be activated in all Ag-pnr-expressing cells, Ag-ASH would not need to have a modular promoter structure like that of the Drosophila locus, and could have a less complex organisation. If so, the acquisition of position-specific enhancers may have occurred after the separation of Nematocera and Brachycera, at a time when further gene duplication events appear to have taken place (Skaer et al., 2002). In addition, modulation of Pnr activity through the use of different co-factors may have accompanied the acquisition of cis-regulatory enhancer sequences in the lineage leading to Drosophila.

Despite the inferred simple regulatory interaction between Ag-Pnr and Ag-ASH, it is remarkable that the effects of mis-expression of Ag-pnr in Drosophila are almost identical to those caused by mis-expression of Dm-pnr. For example, ectopic expression of either Dm-pnr or Ag-pnr on the lateral notum, causes the development of a tuft of ectopic dorso-central bristles. This is due to an expansion of the activity of the dorso-central enhancer element known to be regulated by Dm-Pnr (Garcia-Garcia et al., 1999). This result suggests that Ag-Pnr is able to recognise the relevant regulatory modules of the Drosophila ASC promoter which may indicate that these enhancers are derived from an ancestral regulatory sequence also present in Anopheles. Alternatively, a number of regulatory modules may in fact be present in the Anopheles promoter and govern expression in the various domains on the notum. Further understanding of the structure and regulation of Ag-ASH will require investigation of regulatory sequences from this organism. The ectopic expression assay also indicates that Ag-Pnr is probably able to associate with Drosophila co-factors such as U-shaped and Chip (Cubadda et al., 1997; Ramain et al., 2000). It has been shown that the N-terminal zinc finger of Dm-Pnr associates with U-shaped, while two C-terminal helical structures are components mediating association with Chip (Haenlin et al., 1997; Ramain et al., 2000). The two zinc fingers are strongly conserved in Ag-Pnr, and there is a single α helix. Thus Ag-Pnr appears to contain the relevant binding regions for these two factors. This complexity of the Ag-pnr protein may indicate association with endogenous co-factors, perhaps in a different tissue.

It has been demonstrated, that, in Drosophila, pnr and the iro-C genes are selector genes involved in the subdivision of the dorsal component of segments of the head, thorax and abdomen of the adult into medial and lateral domains (Calleja et al., 2000; Mann and Morata, 2000; Maurel-Zaffran and Treisman, 2000). While pnr regulates the pattern of the medial domain of the dorsal mesonotum, patterning of the lateral half is regulated by the iro-C genes (Gomez-Skarmeta et al., 1996; Calleja et al., 2000; Cavodeassi et al., 2001; Diez del Corral et al., 1999; Kehl et al., 1998). Thus, when either Dm-pnr or Ag-pnr is expressed from an early stage in the entire notum of Drosophila, only structures corresponding to the medial notum are formed, the lateral region fails to develop (Calleja et al., 2000). Ubiquitous expression specifies a single medial domain thought to include cells originally destined to form the lateral region (Calleja et al., 2000). In addition we find that Ag-pnr is expressed in the medial, but not the lateral, mesonotum of Anopheles, consistent with a conserved function in the medial domain. Thus the selector gene function of pnr may have been conserved. The function of proteins of other selector genes of Anopheles, such as engrailed, has been shown to be conserved (Whiteley and Kassis, 1997).

The precursors of the sensory scales on the notum of Anopheles are distributed in a random fashion within the domains of expression of Ag-pnr/Ag-ASH. In some respects the sensory scales resemble the small bristles or microchaetes of cyclorraphous Diptera, which are often randomly distributed although sometimes lined up into rows (McAlpine, 1981; Simpson et al., 1999). However, in the latter species they arise later than the large bristles or macrochaetes, from a second period of ac-sc expression, and are consequently positioned closer to one another than are the macrochaetes (Simpson et al., 1999; Wülbeck and Simpson, 2000; Pistillo et al., 2002). In contrast, the precursors of scales and bristles appear to arise simultaneously in Anopheles, which is consistent with the fact that they are equidistant from each other in the imago. In cyclorraphous flies, the macrochaete pattern is the result of spatially complex sc (ac) expression: one (or a very small number) of bristle(s) develops from each small cluster (or stripe) of sc (ac) expression. In Anopheles, however, the patterning mechanism is different: remarkably, the precursors of the bristles are exclusively positioned along the borders of the expression domains. Thus the positions of the rows of AC and DC bristles, as well as the PST and SC bristles, are coincident with the borders of the four domains of Ag-pnr/Ag-ASH expression. This suggests that the boundaries of Ag-ASH/Ag-pnr expression convey specific positional information causing neural precursors to develop into bristles rather than sensory scales.

Two observations in Drosophila may be relevant to this phenomenon. First, some of the macrochaete precursors arise from the edges of the corresponding proneural clusters of ac-sc expression, an observation that has been linked to distance from the source of the signalling molecules Wingless and Decapentaplegic (Cubas et al., 1991; Phillips et al., 1999). The expression pattern of these molecules in Anopheles is not yet known. Second, it has been demonstrated that the border between pnr-expressing and non-expressing cells does in fact display special properties. Cells of the medial domain manifest unique adhesive characteristics that prevent them from mixing with cells of the lateral domain (Calleja et al., 2000). So, as for compartment boundaries, this interface between cells expressing pnr and those expressing iro may be an important patterning boundary (Dahmann and Basler, 1999; Lawrence and Struhl, 1996; Mann and Morata, 2000). It has indeed been shown to be required for the growth and patterning of the Drosophila eye (Cavodeassi et al., 1999; Cavodeassi et al., 2000; McNeill et al., 1997; Yang et al., 1999). Interestingly, the five macrochaetes on the medial notum of Drosophila are pnr-dependent, and they are all positioned on the lateral border of the domain of pnr expression (Fig. 5). Experimentally contrived expression of ac-sc inside the pnr domain, however, results in the formation of ectopic macrochaetes, indicating that macrochaete formation in Drosophila, is not dependent on special properties at the border (Balcells et al., 1988; Cubadda et al., 1997; Haenlin et al., 1997; Rodriguez et al., 1990). Furthermore the prescutellar bristle of Ceratitis and the AC row of bristles in Calliphora, arise from sc-expressing cells situated inside the pnr expression domain (Pistillo et al., 2002; Wülbeck and Simpson, 2000).

Although the bristles on the notum of Anopheles are aligned into rows, the number and position of bristles within a row varies greatly between individuals, a feature that is thought to be ancestral (McAlpine, 1981; Simpson et al., 1999). Species of cyclorraphous Schizophora in contrast, have very defined rows in which the number and position of bristles varies little if at all. The stereotyped notal bristle patterns of species such as Drosophila are thought to be derived from an ancestral pattern of four longitudinal rows of bristles, still present in many extant species of Schizophora (Simpson et al., 1999; Pistillo et al., 2002). These include the AC and DC bristle rows that are in the medial domain of the notum. So, for example, the two DC bristles of Drosophila would be vestiges of the DC row. Whether the rows of bristles seen in some families of Nematocera such as the Culicidae, are in any way related by ancestry to the four rows of Schizophoran flies, is more difficult to assess. Nevertheless the DC row of Anopheles is positioned on the lateral border of the Ag-pnr expression domain, as in Ceratitis, Calliphora and Drosophila, which may indicate a common origin for this row. If so, this would mean that an ancestral pattern of bristle rows was already present in a common ancestor of the Brachycera and at least some families of Nematocera.

Our results indicate a conserved function for pnr in the regulation of the bristle pattern on the medial notum. This argues in favour of an ancient role for pnr as a selector gene specifying the dorsal medial pattern. The nature of the regulatory interactions between Pnr and its target genes ac-sc appears to have changed, however, over evolutionary time. We hypothesise that in Culicid mosquitoes, which have fewer ac-sc genes, the regulatory regions of this locus may not be organised in a modular fashion. Evolution of the stereotyped bristle patterns characteristic of species such as Drosophila and Ceratitis may have entailed the acquisition of a number of additional factors. These would include gene duplication within the ASC and the co-option of cis-regulatory sequences. Co-factors for Pnr, such as Ush and Chip, are also likely to have been co-opted for use in constructing the notal pattern at a later evolutionary stage, although our results suggest that Ag-Pnr has the requisite domains for association with these proteins (Cubadda et al., 1997; Haenlin et al., 1997; Ramain et al., 2000). In the lineage leading to Drosophila, these different levels of regulation might have been superimposed onto an ancestral patterning mechanism, similar to that of Anopheles, at different times in the 200 million years separating Drosophila from the Nematocera.

We acknowledge the financial support of the Institut National de la Santé et de la Recherche Médicale, the Centre National de la Recherche Scientifique, the Hôpital Universitaire de Strasbourg, as well as the Programme of the European Community (contract no. FMRX CT 96 0065), the Programme Génome du CNRS and the Wellcome Trust (grant number 29156). C. W. is the recipient of a fellowship from the DFG. We are very grateful to Professor Fotis Kafatos and the members of his laboratory for their generous supply of mosquitoes and for DNA libraries from Anopheles and for hosting C. Wülbeck as a visiting scientist at the outset of this work. We thank Nadia Masdecq for her help with sectioned material, Annie Bauer and the members of the sequencing facility of the IGBMC for technical help, Margit Pal, Daniela Pistillo and Claire Chapman for their generous help with transformation of Drosophila and for Drosophila crosses, and Lynn Riddiford, Jim Truman, Manuel Calleja and the members of our group for discussion and comments on the manuscript.