Evolution of the dorsal-ventral patterning network in the mosquito, Anopheles gambiae

The dorsal-ventral patterning of the Drosophila embryo is controlled by a well-defined gene regulation network that is deployed by Dorsal (reviewed by Moussian and Roth,2005), a sequence-specific transcription factor related to mammalian NF-kB (also known as Nfkb1)(Lenardo and Baltimore, 1989). The Dorsal protein is distributed in a broad nuclear gradient in precellular embryos. This transient gradient leads to stable circuits of cell differentiation that control gastrulation(Stathopoulos and Levine,2004), including the invagination and patterning of the mesoderm,and the establishment of diverse cell types within the ectoderm.

The Dorsal gradient regulates over 50 target genes in a concentration-dependent manner(Stathopoulos and Levine,2002). Approximately 40 of the genes encode sequence-specific transcription factors (TF) or components of signal transduction (ST) pathways that impinge on the activities of the TFs. Dorsal target enhancers have been identified for about half of these genes, and the DNA binding sites recognized by many of the TFs have been determined (reviewed by Stathopoulos and Levine,2005). This information has permitted the construction of a detailed gene network, or circuit diagram, containing nearly 200 functional interconnections among the 40 TF and ST Dorsal target genes(Levine and Davidson,2005).

It is our long-term goal to understand how changes in the Drosophila dorsal-ventral (DV) patterning network produce diverse gastrulation profiles in different insects. In the present study we compare dorsal-ventral patterning in Drosophila melanogaster and the malaria mosquito, Anopheles gambiae. Both insects are members of the same order, Diptera, but are highly divergent and last shared a common ancestor∼200 million years ago (Gailey et al.,2006). Genome turnover is so extensive that homologous enhancers do not display any vestige of sequence similarity. By contrast, sequence conservation is readily detected among extensively divergent vertebrates such as humans and pufferfish (Santini et al.,2003). Despite the turnover in the noncoding sequences of divergent insects, there is extensive conservation of the segmentation gene network, which serves to establish broadly similar body plans. For example,altered patterns in gap gene expression are balanced by compensatory changes in the regulation of downstream pair-rule genes(Goltsev et al., 2004).

Classical embryological studies revealed broad similarities in DV patterning among diverse Diptera (see Sander, 1975). However,notable differences were detected in the formation of the extraembryonic membranes (EMs). Specifically, higher dipterans such as D. melanogaster contain one EM, the amnioserosa(Demerec, 1950), whereas lower dipterans, such as mosquitoes, contain distinct amnion and serosa tissues(Christophers, 1960; Davis, 1967; Guichard, 1971; Idris, 1960; Ivanova-Kazas, 1949). Indeed,most insects contain separate tissues, suggesting that the formation of the single aminoserosa is a derived characteristic(Schmidt-Ott, 2000; Stauber et al., 1999). The analysis of segmentation gene expression in A. gambiae and specifically the repression of individual eve stripes in the presumptive serosa suggested early divergence in the DV patterning of the EMs of D. melanogaster and A. gambiae(Goltsev et al., 2004).

Here, we extend the previous analysis of segmentation to obtain a detailed picture of early dorsal-ventral patterning in A. gambiae. Particular efforts focus on the analysis of Dorsal target genes governing mesoderm invagination and the patterning of the ectoderm. Evidence is presented that the patterning of the ventral half of the embryo, the mesoderm and ventral neurogenic ectoderm, is highly conserved in A. gambiae and D. melanogaster. By contrast, the patterning of the dorsal ectoderm exhibits many differences.

The dorsal ectoderm of D. melanogaster produces just two cell types, dorsal epidermis and the amnioserosa. The latter tissue arises from a restricted region of the dorsal-most ectoderm, along the dorsal midline. In A. gambiae, the dorsal ectoderm is significantly expanded, and the dorsal midline is subdivided into distinct amnion and serosa lineages. The expansion of the dorsal ectoderm can be explained by a broadening in the domain of Dpp (BMP) signaling (reviewed by Podos and Ferguson, 1999) in the early A. gambiae embryo. Evidence is presented that this expansion results, in part, from the restricted expression of the Dpp inhibitor, Sog, within the presumptive mesoderm. In Drosophila, sogis expressed in a broad pattern that encompasses the entire neurogenic ectoderm (Francois et al.,1994). This broad sog pattern restricts Dpp signaling to the dorsal midline (Ashe and Levine,1999; Decotto and Ferguson,2001; Eldar et al.,2002; Holley et al.,1995; Marques et al.,1997; Mizutani et al.,2005; Shimmi et al.,2005), whereas the ventrally restricted sog pattern in A. gambiae appears to produce a broader domain of Dpp signaling. The broad sog pattern in Drosophila is driven by an intronic enhancer containing optimal Dorsal binding sites(Markstein et al., 2002). An analogous enhancer in A. gambiae contains low-affinity sites, and when expressed in transgenic D. melanogaster embryos it recapitulates restricted expression in the mesoderm, similar to the endogenous A. gambiae pattern. Thus, the interconversion of high- and low-affinity Dorsal binding sites appears to produce altered threshold responses to the Dorsal gradient. We discuss how subtle changes in the Dorsal patterning network can convert separate serosa and amnion tissues into a single tissue.

Anopheles gambiae Kisumu strain was reared at 26°C, 75%humidity, with a 12-hour light/dark cycle. Adults were maintained on a 10%sucrose solution and females were blood-fed on anesthetized hamsters. Drosophila melanogaster strain yw⁶⁷ was used for in situ hybridizations, as described previously (e.g. Stathopoulos et al.,2002). P-element-mediated transformation was performed using standard methods(Rubin and Spradling, 1982). Flies of the genotype P{Kr-zen-Gal4} were mated to those of the P{UAS-A.gzen}/+. The resulting embryos were of the genotype P{Kr-zen-Gal4}/+;UAS-A.g.zen/+; or P{Kr-zen-Gal4}/+;+/+. The embryos bearing P{UAS-A.g.zen} exhibit ectopic Race-expressing cells.

Mosquito DNA was derived from the Anopheles gambiae Kisumu strain. We partially resequenced the sog locus to correct for the gaps in the mosquito genome assembly (Fig. S3 in supplementary material). The A. gambiae sog enhancer fragments were amplified from genomic DNA with the following primers:

Fragment 1: ACCAGGTCGTGTGCAGCTCGCGTATGGTCTT,GGCGTGCGAGCTCTTGCGTCTCCTACGCAG;

Fragment 2: GAGAACCGGTAATGGTCTAGCCGCCAA, GCAACCCCAACAACAACTCTGTTCACA;

Fragment 3: TGGTAGCACTTCGCACATTCGAGTTAG, TCAGCATCGACGATGCAATACCATACG;

Fragment 4: GCCGGTACGTGGTAGAGTGGCAGAGTA, CTGACCAGACGGCAGACCACGGTAGAA;

Fragment 5: TCTGATATGTCTGGGACGGTGTGTTGT, CTGGATGTTCGCATCACGTCTTCCTCT.

PCR products were cloned into a [-42evelacZ]-pCaSpeR vector(Small et al., 1992). The exact coding sequence for the A. gambiae zen gene was determined by RACE using an A. gambiae embryonic cDNA Marathon library. The coding sequence was amplified with the following pair of primers:ATAAAGTTTCTGTTAAGCAACTGCAGTAA, CCAGATGTCGTAGTACCCATTATATGGTAA.

PCR products were cloned into the pUAST(Brand and Perrimon, 1993)expression vector. Constructs were introduced into the D. melanogaster germline by microinjection as described previously(Ip et al., 1992). Between three and nine independent transgenic lines were obtained for each construct.

Mosquito embryos were collected and fixed as described previously(Goltsev et al., 2004). Hybridization probes were prepared against specific A. gambiae genes identified by reciprocal BLAST analyses. The hybridization probes were generated by RT-PCR amplification from embryonic RNA. A 26 bp tail encoding the T7 RNA polymerase promoter (TAATACGACTCACTATAGGGAGA) was included on the 5′ side of the reverse primer. PCR products were purified with the Qiagen PCR purification kit and used directly as templates for in vitro transcription reactions. The following primer pairs were used to amplify each of the indicated A. gambiae segmentation genes. (The T7 promoter sequence is denoted by the symbol [T7].):

twi: CTTATACTGGACATTAGTGGAGCCGGTT,AAG[T7]GGAAGCTAGCCGGAGCGTCTGTATCTT;

sna: CCACACCTCGTTCAACTCGTACCTTTCGTC,AAG[T7]TGGCATGAAGCTGTCCTCCGAGATGTT;

sim: AGCGTCAATCATACGACTCACCACCTCGTA,AAG[T7]TAGAACATAGTTGACGCTAACGATACA;

vnd: CCGGTGCTGACCTGGTCGCCGCTGCTGTTT, AAG[T7]GG ACCGCCGTCAGCAGGTCGTGCGGTT;

brk: CCAGTTCAAGCTGCAGGTGCTCGACTCGTA, AAG[T7]TC CGGCTAATGTTGTACTTGGTCGCGA;

ind: TTCTAGTGGACTCGTTAATCAGTGATAAGC,AAG[T7]AGTGCGTACGATCTTCTGCTGATCGTT;

dpp: ACGCTAGTCGAGATAGAGAAGAACCTTCT,AAG[T7]CCGCAGCCAACGACCGTCATGTCCTGGTA;

tkv: CTGCTACTGCGAGGGCCACTGTCCGGGCAA,AAG[T7]GAGTCCGTCTCGAGCTTGACGAGCGTT;

zen: TCGCTGCTGACAGTTATATTGGTTCAACTA,AAG[T7]ATCATTATCGAGAGATGTGCTACAAGCCT;

hnt: CAGATGCAGGATGTGCCGCCCACGCCGGCC,AAG[T7]GGCGGTAGCTCAGCATCGCCCGACACCACGC;

tup: CGCTTATCCTTGTGCGTTGGATGCGGCGGTC,AAG[T7]CCATGTGCGAACCGATCGGAGGACCTGGCC;

Doc1: ACCGTCAGCAAATGTTGCAACGGATACCAG,AAG[T7]CGAGGAGGAGGTGTTGTTGCAGCCCATCTT;

ems: CTGGCGGCCCAGTTCCAGGCGGCCGCCCTT,AAG[T7]TCGGACAGTCGTCCATGTCGATGAACT;

hb: GGCTCGGACTGTGAGGATGGCTCGTACGAT,AAG[T7]CAGGTACGGGAACAGTGGCAGACTGCCGTT;

ttk: ATGGTGCAAACGAATCCGCTGCTCGGTACT,AAG[T7]CGCGAACGGACATCTCTGTGAGTGCTT;

sog: TGCCAGTTTGGCAAGACCATACGCGAGCTG,AAG[T7]CTTCTCGCACTTGTACTGCTGGTGGTCGCA;

tld: TGCTTGCGGAGGTCAGCTGGACACGCCGAA,AAG[T7]CTGATGTGGCTCAATATCGAACACATTGAA;

rho: CGGGTTCTTCGTCTACCACTCACTCACGTT,AAG[T7]ATACCTCTTCACTTTCCTCCTCTAGCCTCT.

Rabbit anti-pMad antibody was kindly provided by P. ten Dijke (Leiden University Medical Center, Leiden, Netherlands). D. melanogaster and A. gambiae embryos were fixed as described previously(Goltsev et al., 2004). Primary antibodies were used at a dilution of 1:200. Secondary anti-rabbit antibodies conjugated to alkaline phosphatase (Jackson ImmunoResearch) were used for staining.

Fly and mosquito embryos were fixed as for in situ hybridization. The embryos were subsequently post-fixed in 25% glutaraldehyde for 30 minutes,dehydrated and dried. Embryos were coated with gold-palladium and observed with a JOEL JSM 5800LV scanning electron microscope.

A Dorsal position weighted matrix(Papatsenko and Levine, 2005)was used to identify potential Dorsal binding clusters at the Anopheles sog locus. The recently developed ClusterDraw software was used for this analysis (Zinzen et al.,2006).

D. melanogaster and A. gambiae embryos exhibit distinct patterns of mesoderm invagination (Fig. 1A). Scanning electron photomicrographs of the ventral surface of the D. melanogaster embryo (top panel) show that the presumptive mesoderm constricts at the apical surface. Mesectodermal cells, and possibly some of the lateral mesoderm cells, have processes oriented towards the ventral furrow. It is conceivable that these protrusions contribute to mesodermal tube closure and mesectoderm intercalation(Leptin et al., 1992; Leptin and Grunewald, 1990). The presumptive mesoderm of A. gambiae also have apical constrictions(Fig. 1A, bottom panel),however, they lack the oriented protrusions mentioned above. There is a shallow groove at the midline in place of the deep furrow seen in D. melanogaster. These observations suggest that the mesoderm does not undergo the same type of coherent involution in A. gambiae as seen in D. melanogaster.

Mesoderm invagination has been studied by analyzing the expression of twist, an early determinant of mesoderm fate(Boulay et al., 1987; Thisse, 1987). This approach was used in staged A. gambiae embryos which were hybridized with a digoxigenin-labeled twist antisense RNA probe, and then mounted in plastic and sectioned (Fig. 1B). twist staining is restricted to the ventral-most 25%of the embryo circumference, as seen in D. melanogaster. There is transient apical constriction of the mesoderm plate(Fig. 1Bc), followed by the appearance of a shallow groove along the ventral midline. There is no organized involution of the mesoderm, but instead, individual mesoderm cells undergo progressive ingression during germband elongation(Fig. 1Bd-f). This ingression is similar to that seen in mutant D. melanogaster embryos lacking fog-concertina signaling. In these mutants, there is a severe reduction of the ventral furrow and mesoderm cells fail to invaginate(Costa et al., 1994; Dawes-Hoang et al., 2005). Nonetheless, many of the mutant embryos survive because of ingression of the mesoderm during elongation. Interestingly, the A. gambiae genome lacks a clear homologue of the fog gene (see Discussion).

A number of marker genes were analyzed to determine whether there have been significant changes in the DV patterning network responsible for the mesoderm and neurogenic ectoderm in flies and mosquitoes(Fig. 2). Despite the different modes of mesoderm invagination, the overall limits of the presumptive mesoderm are quite similar in flies and mosquitoes(Fig. 2A-D). In both cases, the twist and snail expression patterns are restricted to the ventral-most regions destined for later ingression during elongation. As in D. melanogaster, the snail pattern has somewhat sharper lateral borders than those seen for twist(Fig. 2C; compare with A). sim expression is restricted to single lines of cells immediately straddling the snail borders (Fig. 2E,F). These lines coincide with the ventral-most regions of the neurogenic ectoderm, and the cells will form specialized mesectodermal derivatives along the ventral midline of the nerve cord(Martin-Bermudo et al.,1995).

In D. melanogaster, intermediate and low levels of the Dorsal gradient lead to sequential patterns of vnd and indexpression, which pattern the medial and lateral portions of the future nerve cord (McDonald et al., 1998; Weiss et al., 1998). Similar sequential patterns are seen for the corresponding genes in A. gambiae (Fig. 2I,J; data not shown). The ind pattern has a segmental periodicity in A. gambiae (Fig. 2I), but is otherwise similar to the expression pattern seen in D. melanogaster(Fig. 2J).

The brinker gene encodes a transcriptional repressor that is a component of the Dpp (BMP) signaling pathway in the D. melanogasterembryo (Campbell and Tomlinson,1999; Jazwinska et al.,1999). It is activated in ventral and lateral regions of the neurogenic ectoderm, in a pattern similar to vnd(Fig. 2H). Once again, a comparable pattern is seen in A. gambiae(Fig. 2G). Overall, the preceding results suggest that the initial patterning of the mesoderm and neurogenic ectoderm depend on similar mechanisms in the fly and mosquito embryos. The only clear difference is the formation of a coherent ventral furrow and invaginated mesodermal tube in D. melanogaster.

There is a clear difference in the dorsal ectoderm of D. melanogaster and A. gambiae embryos. The A. gambiaeembryo is enclosed by the serosa, an external cuboidal layer of cells that forms an extraembryonic membrane (Fig. 3B,D,E,F; blue arrows in B and D, pseudo-colored blue in F). There is also a separate amnion that connects the embryo proper to the serosa (e.g. Fig. 3D,F - red arrow in D and pseudo-colored red in F) and therefore resides between the external serosa and the germband. The establishment of a double-layered extraembryonic envelope is a highly dynamic process, well described for a number of diverse insects(reviewed by Schmidt-Ott,2005) (see also van der Zee et al., 2005). The electron micrograph in Fig. 3A and the DIC image in Fig. 3D show the initial phases of germband elongation in the mosquito. At this stage the caudal regions of the germband begin to migrate beneath the serosa and the double-layered topology of the extraembryonic membrane is established. The EMs have not yet extended to ventral regions (Fig. 3A). Later, the amnion continues to migrate over the germband stretching the serosa around the embryo(Fig. 3B,E,F). Finally, the A. gambiae embryo becomes fully enclosed, whereby the amnion and serosa fuse along the ventral midline of the germband.

The D. melanogaster embryo contains a single amnioserosa arising from the dorsal midline (white arrow, Fig. 3C). The formation of separate amnion and serosa lineages is probably ancestral for insect embryos, since these are seen in a broad range of insects, including flour beetles, bees and grasshoppers(Dearden et al., 2000; Panfilio et al., 2006; Schmidt-Ott, 2000; Stauber et al., 1999; Stauber et al., 2002; van der Zee et al., 2005). Higher Dipterans, such as D. melanogaster, are somewhat unique in containing just a single amnioserosa.

A variety of dorsal patterning genes were examined in A. gambiaeembryos in an effort to determine the basis for the formation of distinct ectodermal derivatives. For example hindsight (hnt; also known as peb - Flybase) (Frank and Rushlow, 1996) is expressed along the dorsal midline of D. melanogaster embryos (Fig. 4B), while tailup (tup)(Thor and Thomas, 1997) is expressed in a broader pattern that encompasses both the presumptive amnioserosa and dorsolateral ectoderm (Fig. 4D). The hnt expression pattern seen in A. gambiae is similar to that detected in D. melanogaster, although there is a marked expansion in the dorsal-ventral limits of the presumptive extra-embryonic territory (Fig. 4A; prospective serosa is marked by red oval). By contrast, the tup pattern in A. gambiae is dramatically different from that seen in D. melanogaster - it is excluded from the prospective serosa and restricted to the future amnion(Fig. 4C).

The T-box genes Dorsocross1 (Doc1) and Doc2 are involved in amnioserosa development and expressed along the dorsal midline and in a transverse stripe near the cephalic furrow of gastrulating D. melanogaster embryos (Reim et al.,2003) (Fig. 4F). The Doc1 and Doc2 orthologues in A. gambiae exhibit restricted expression in the presumptive amnion(Fig. 4E,G; the white arrow in G indicates the amnion), similar to the tup pattern. The expression patterns of the two genes are identical but only Doc1 is shown. They are initially expressed in a broad dorsal domain (data not shown) but come to be repressed in the serosa. There is also a head stripe of expression comparable to the D. melanogaster pattern(Fig. 4E). Additional dorsal-ventral patterning genes are also expressed in a restricted pattern within the developing amnion (see Fig. S1 in supplementary material). Overall,the early expression patterns of tup, Doc1 and Doc2 (and additional patterning genes) foreshadow the subdivision of the dorsal ectoderm into separate serosa and amnion lineages in Anopheles.

In D. melanogaster, the patterning of the dorsal ectoderm depends on Dpp and Zen, along with a variety of genes encoding Dpp signaling components, such as the Thickveins (Tkv) receptor. Most of the corresponding genes are expressed in divergent patterns in A. gambiae embryos(Fig. 5). For example, dpp and tkv are initially expressed throughout the dorsal ectoderm (data not shown), but become excluded from the presumptive serosa and restricted to the amnion (Fig. 5A,C). By contrast, both genes have broad, nearly uniform expression patterns in the dorsal ectoderm of D. melanogaster embryos(Fig. 5B,D).

There is an equally dramatic change in the zen expression pattern. In A. gambiae, expression is restricted to the presumptive serosa territory, even at the earliest stages of development(Fig. 5E,G). By contrast, zen is initially expressed throughout the dorsal ectoderm of cellularizing embryos in D. melanogaster(Fig. 5F), and becomes restricted to the dorsal midline by the onset of gastrulation(Fig. 5H). Thus, the dpp/tkv and zen expression patterns are essentially complementary in A. gambiae embryos, but extensively overlap in Drosophila (see below).

The loss of dpp (Fig. 5A), tkv (Fig. 5C), Doc1 (Fig. 4E), Doc2 (not shown) and tup(Fig. 4C) expression in the presumptive serosa of A. gambiae embryos raises the possibility that zen activates the expression of one or more repressors in the serosa. It is unlikely that Zen itself is such a repressor since the expression of the A. gambiae zen gene in transgenic Drosophila embryos does not alter the normal development of the amnioserosa (see Fig. S2 in supplementary material).

Different segmentation genes were examined in an effort to identify putative serosa-specific repressors. For example, the gap gene hunchback (hb) is initially expressed in the anterior regions of A. gambiae embryos, in a similar pattern to that seen in D. melanogaster (Bender et al.,1988; Lehmann and Nusslein-Volhard, 1987), but by the onset of gastrulation a novel pattern arises within the presumptive serosa(Goltsev et al., 2004). hb expression has also been seen in the developing serosa of other insects, including a primitive fly (Clogmia) and the flour beetle, Tribolium (Stauber et al.,2002; Wolff et al.,1995).

Two additional segmentation genes behave like hb, empty spiracles(ems) and tramtrack (ttk)(Fig. 6A,C). ems is involved in head patterning in D. melanogaster(Dalton et al., 1989). Its expression is limited to a single stripe in anterior regions of cellularizing D. melanogaster embryos (Fig. 6B). Staining is seen in a comparable anterior region of A. gambiae embryos (Fig. 6A),but a second site of expression - not seen in Drosophila - is also detected in the presumptive serosa.

Ttk is a maternal repressor that helps establish the expression limits of several pair-rule stripes (Read et al.,1992). It is ubiquitously expressed throughout the early D. melanogaster embryo (Fig. 6D), but has a tightly localized expression pattern within the presumptive serosa of A. gambiae embryos(Fig. 6C). Thus, novel patterns of ems and ttk expression are consistent with the possibility that serosa-specific repressors help subdivide the dorsal ectoderm into separate serosa and amnion lineages in A. gambiae embryos (see Discussion).

The analysis of dorsal-ventral patterning genes identified two critical differences between the pre-gastrular fly and mosquito embryos. First, there are separate serosa and amnion lineages in A. gambiae, but just a single amnioserosa in D. melanogaster. Second, there is an expansion in the limits of the dorsal ectoderm in A. gambiae as compared with the D. melanogaster embryo. Localized repressors might help explain the former observation of separate lineages, but do not provide a basis for the expansion of the dorsal ectoderm.

In D. melanogaster, the limits of Dpp signaling are established by the repressor Brinker (Jazwinska et al.,1999) and the inhibitor Sog(Francois et al., 1994). Genetic studies suggest that Sog is the more critical determinant in early embryos. It is related to Chordin, which inhibits BMP signaling in vertebrates(Francois and Bier, 1995), and is expressed in broad lateral stripes encompassing the entire neurogenic ectoderm (Fig. 7B)(Markstein et al., 2002). The secreted Sog protein directly binds Dpp, and blocks its ability to interact with the Tkv receptor (e.g. Shimmi et al.,2005). However, Sog-Dpp complexes are proteolytically processed by the Tolloid (Tld) metalloprotease(Mullins, 1998), which is expressed throughout the dorsal ectoderm of early Drosophila embryos(Fig. 7G)(Marques et al., 1997). Tld helps ensure that high levels of the Dpp signal are released at the dorsal midline located far from the restricted source of the inhibitor Sog(Shimmi et al., 2005).

The expression patterns of the sog and tld genes in A. gambiae are very different from those seen in D. melanogaster (Fig. 7). sog expression is primarily detected in the ventral mesoderm,although low levels of sog transcripts might extend into the ventral-most regions of the neurogenic ectoderm(Fig. 7A,F). This pattern is more restricted across the dorsal-ventral axis than the D. melanogaster sog pattern (Fig. 7B). tld expression is restricted to lateral regions of A. gambiae embryos (Fig. 7C,H) and is excluded from the dorsal ectoderm, which is the principal site of expression in Drosophila(Fig. 7G). These significant changes in the sog and tld expression patterns might account, at least in part, for the expanded limits of Dpp signaling in the dorsal ectoderm of A. gambiae embryos (see Discussion).

Direct evidence for broader Dpp signaling was obtained using an antibody that detects phosphorylated Mad (pMad)(Persson et al., 1998), the activated form of Mad obtained upon induction of the Tkv receptor. In D. melanogaster pMad expression is restricted to the dorsal midline(Fig. 7I,J). This is the domain where Sog-Dpp complexes are processed and peak levels of Dpp interact with the receptor Tkv. The spatial limits of the sog expression pattern are decisive for this restricted domain of pMad activity. Just a twofold reduction in the levels of Sog (sog/+ heterozygotes) causes a significant expansion in pMad expression (Mizutani et al., 2005).

There is a marked expansion of the pMad expression domain in A. gambiae embryos as compared with Drosophila(Fig. 7D,E). The domain encompasses the entire presumptive serosa and extends into portions of the presumptive amnion. The dpp and tkv expression patterns are downregulated in the presumptive serosa(Fig. 5A,C), nonetheless, the pMad staining pattern clearly indicates that this is the site of peak Dpp signaling activity. The early expression of both dpp and tkvencompasses the entire dorsal ectoderm. It would appear that peak Dpp signaling is somehow maintained in the developing serosa even after the downregulation of dpp and tkv expression in this tissue (see Discussion). A similar scenario is seen in the Drosophila embryo, in that there is downregulation of both dpp and tkv expression along the dorsal midline of gastrulating embryos (e.g. Affolter et al., 1994).

To determine the basis for expanded Dpp signaling we identified and characterized a sog enhancer in A. gambiae. The D. melanogaster enhancer is located in the first intron of the sogtranscription unit (Fig. 8D). It is ∼300 bp in length and contains four evenly spaced, optimal Dorsal binding sites (Markstein et al.,2002). These sites permit activation of sog expression by low levels of the Dorsal gradient; however, closely linked Snail repressor sites inactivate the enhancer in the ventral mesoderm. A putative A. gambiae enhancer was identified by scanning the sog locus for potential clusters of Dorsal binding sites. The recently developed cluster-draw program was used for this purpose since it successfully identified a sim enhancer in the honeybee, Apis mellifera,which is even more divergent than Anopheles(Zinzen et al., 2006). The best putative Dorsal binding cluster was identified within the first intron of the A. gambiae sog locus (Fig. 8C). Several genomic DNA fragments were tested for enhancer activity, but only this cluster was found to activate gene expression in transgenic Drosophila embryos (summarized in Fig. 8D).

Two different genomic DNA fragments, 3.7 kb and 1.1 kb, that encompass the intronic binding cluster were tested in transgenic embryos (see Fig. 8D). Both fragments were attached to a lacZ reporter gene containing the core evepromoter from D. melanogaster, and both direct lacZexpression in the presumptive mesoderm(Fig. 8A,B; data not shown). They exhibit the same restricted dorsal-ventral limits of expression as that seen for the endogenous sog gene in A. gambiae, although the smaller fragment produces ventral stripes whereas the larger fragment directs a more uniform pattern (not shown). The change in the dorsal-ventral limits -broad expression in D. melanogaster and restricted expression in A. gambiae - might be due to the quality of individual Dorsal binding sites in the two enhancers (see Discussion).

A comprehensive analysis of dorsal-ventral patterning genes in the A. gambiae embryo reveals elements of conservation and divergence in the gastrulation network of D. melanogaster. There is broad conservation in the expression of regulatory genes responsible for the patterning of the mesoderm and neurogenic ectoderm, including sequential expression of sim,vnd and ind in the developing nerve cord. By contrast, there are extensive changes in the expression of regulatory genes that pattern the dorsal ectoderm. These changes foreshadow the subdivision of the dorsal ectoderm into separate serosa and amnion lineages in A. gambiae.

The major difference in the early patterning of the mesoderm in flies and mosquitoes concerns the manner in which mesoderm cells enter the blastocoel of gastrulating embryos. In D. melanogaster, there is a coherent invagination of the mesoderm through the ventral furrow, much like the movement of bottle cells through the blastocoel of Xenopus embryos(Keller, 1981). By contrast,there is no invagination of the mesoderm in A. gambiae. Instead, the mesoderm undergoes progressive ingression during germband elongation. This type of ingression is seen in D. melanogaster mutants lacking fog signaling (Costa et al.,1994; Dawes-Hoang et al.,2005). The A. gambiae genome lacks a clear homologue of fog, and it is therefore conceivable that fog represents an innovation of the higher Diptera that was only recently incorporated into the D. melanogaster dorsal-ventral patterning network.

D. melanogaster is somewhat unusual in having an amnioserosa,rather than separate serosa and amnion tissues as seen in most insects(Dearden et al., 2000; Panfilio et al., 2006; Stauber et al., 2002; van der Zee et al., 2005). In certain mosquitoes the serosa secretes an additional proteinaceous membrane that provides extra protection against desiccation(Harwood, 1958; Harwood and Horsfall, 1959). The changes in gene expression in the D. melanogaster and A. gambiae dorsal ectoderm provide a basis for understanding the evolutionary transition of two dorsal tissues in A. gambiae into a novel single tissue in higher dipterans.

The D. melanogaster amnioserosa expresses a variety of regulatory genes, including Doc1/2 and tup. The expression of most of these genes is restricted in the presumptive amnion of the A. gambiaeembryo. zen is the only dorsal patterning gene, among those tested,that exhibits restricted expression in the serosa. Several segmentation genes have a similar pattern, and one of these, ttk, encodes a known repressor. Ectopic expression of Ttk causes a variety of patterning defects in Drosophila embryos, including disruptions in head involution and germband elongation that might arise from alterations in the amnioserosa(Read et al., 1992). We propose that zen activates ttk in the serosa of A. gambiae embryos. The encoded repressor might subdivide the dorsal ectoderm into separate serosa and amnion tissues by inhibiting the expression of Doc1/2 and tup in the serosa. The loss of this putative zen-ttk regulatory linkage might be sufficient to allow Dpp signaling to activate tup and Doc1/2 throughout the dorsal ectoderm,thereby transforming separate serosa and amnion tissues into a single amnioserosa. According to this scenario, the loss of zen binding sites in ttk regulatory sequences might be responsible for the evolutionary transition of the amnioserosa (summarized in Fig. 9; see below).

The formation of separate amnion and serosa tissues is not the only distinguishing feature of A. gambiae embryos when compared with D. melanogaster. There is also a significant expansion in the overall limits of the dorsal ectoderm. This can be explained, in part, by distinct patterns of sog expression.

The broad expression limits of the Sog inhibitor are responsible for restricting Dpp/pMad signaling to the dorsal midline of the D. melanogaster embryo (summarized in Fig. 9A). This pattern depends on a highly sensitive response of the sog intronic enhancer to the lowest levels of the Dorsal gradient. The Dorsal binding sites in the sog enhancer are optimal sites, possessing perfect matches to the idealized position weighted matrix of Dorsal recognition sequences(Papatsenko and Levine, 2005). By contrast, the A.gambiae intronic sog enhancer contains low-quality Dorsal binding sites, similar to those seen in the regulatory sequences of genes activated by peak levels of the Dorsal gradient,such as twist. The binding sites in the D. melanogaster sogenhancer have an average score of ∼10. By contrast, the best sites in the A. gambiae sog enhancer have scores in the 6.5-7 range, typical of enhancers that mediate expression in the mesoderm in response to high levels of the Dorsal gradient (Fig. 9B). Although we did not explicitly test every potential regulatory sequence in the A. gambiae sog locus, none of the putative Dorsal binding clusters in the vicinity of the gene possess the quality required for activation by low levels of the Dorsal gradient in the neurogenic ectoderm. Thus, the narrow limits of sog expression in A. gambiae embryos can be explained by the occurrence of low-quality Dorsal binding sites, along with the loss of Snail repressor sites.

The altered sog expression pattern is probably not the sole basis for the expansion of the dorsal ectoderm. A. gambiae embryos also exhibit a significant change in the tld expression pattern. tld is expressed throughout the dorsal ectoderm in D. melanogaster, but restricted to the neurogenic ectoderm of A. gambiae. Tld cleaves inactive Tsg-Sog-Dpp complexes to produce peak Dpp signaling along the dorsal midline of Drosophila embryos(Fig. 9C). We propose that the altered tld pattern in combination with altered sog leads to two dorsolateral sources of the active Dpp ligand in mosquito embryos. The sum of these sources might produce a step-like distribution of pMad across dorsal regions of mosquito embryos (Fig. 9C). This broad plateau of pMad activity might be responsible for the observed expansion of the dorsal ectoderm territory, and the specification of the serosa.

In Drosophila, tld is regulated by a 5′ silencer element that prevents the gene from being expressed in ventral and lateral regions in response to high and low levels of the Dorsal gradient. This silencing activity is due to close linkage of Dorsal binding sites and recognition sequences for `co-repressor' proteins (e.g. Ratnaparkhi et al., 2006). Our preliminary studies suggest that Dorsal activates the A. gambiae tldgene, possibly by the loss of co-repressor binding sites in the 5′enhancer (Kirov et al.,1993).

We propose that there are at least two distinct threshold readouts of Dpp signaling in the dorsal ectoderm of A. gambiae embryos. Type 1 target genes, such as hb, ems, ttk and zen, are activated by high levels and thereby restricted to the presumptive serosa. Type 2 target genes,such as tup and Doc1/2, can be activated - in principle - by both high and low levels of Dpp signaling in the presumptive serosa and amnion. However, these target enhancers contain binding sites for one or more type 1 repressors expressed in the serosa. Our favorite candidate repressor is Ttk. Perhaps the type 2 tup enhancer contains optimal pMad activator sites as well as binding sites for the localized repressor Ttk, which keeps tup expression off in the serosa and restricted to the amnion (see diagram in Fig. 9C). As discussed earlier, the simple loss of ttk regulation by the Dpp signaling network might be sufficient to account for the evolutionary conversion of separate serosa and amnion tissues into a single amnioserosa. Localization of this single tissue within a restricted domain along the dorsal midline would arise from concomitant dorsal shifts in the sog and tld expression patterns.

We are grateful to Peter ten Dijke for generously providing the pMad antibody, and Dmitri Papatsenko for help with the computer analysis of sog enhancers. We also thank members of Levine lab for helpful discussions. We are deeply grateful to Lanzaro lab members and specifically to Claudio Menesis for maintaining the mosquito colony. This work was funded by grants from the NIH (GM46638 and GM34431).