Promoter Sequences of the Putative Anopheles gambiaeApyrase Confer Salivary Gland Expression in Drosophila melanogaster*

The saliva of blood-feeding arthropods contains an apyrase that facilitates hematophagy by inhibiting the ADP-induced aggregation of the host platelets. We report here the isolation of a salivary gland-specific cDNA encoding a secreted protein that likely represents the Anopheles gambiae apyrase. We describe also two additional members of the apyrase/5′-nucleotidase family. The cDNA corresponding to the AgApyL1 gene encodes a secreted protein that is closely related in sequence to the apyrase of the yellow fever mosquito, Aedes aegypti, and whose expression appears enriched in, but not restricted to, female salivary glands. The AgApyL2 gene was found searching anA. gambiae data base, and its expression is restricted to larval stages. We isolated the gene encoding the presumed A. gambiae apyrase (AgApy) and we tested its putative promoter for the tissue-specific expression of the LacZgene from Escherichia coli in transgenic Drosophila melanogaster. All the transgenic lines analyzed showed a weak but unambiguous staining of the adult glands, indicating that some of the salivary gland-specific transcriptional regulatory elements are conserved between the malaria mosquito and the fruit fly. The availability of salivary gland-specific promoters may be useful both for studies on vector-parasite interactions and, potentially, for the targeted tissue-specific expression of anti-parasite genes in the mosquito.

The salivary glands of blood-sucking arthropods secrete a complex array of specific factors with vasodilatory, anti-clotting, and anti-platelet activities that assist the mosquito dur-ing blood feeding. Since hematophagy has arisen independently several times in insects, even within the same order, a large variety of different molecules have evolved to accomplish the same or similar function (1)(2)(3)(4). Remarkably diverse substances act as vasodilators in the saliva of distinct arthropod species; they include prostaglandins in ticks, nitric oxide in the bugs Rhodnius prolixus and Cimex lectularius, peptides such as tachykinins or maxadilan in the mosquito Aedes aegypti and in the sandfly Lutzomya longipalpis, respectively, and the salivary peroxidase/catechol oxidase in the mosquito Anopheles albimanus (4 -6). Similarly, partly as a consequence of their different feeding habits, a multitude of different anticoagulants is found in different blood-eating species (7). In mosquitoes, thrombin and Factor Xa are the preferential targets within the blood coagulation cascade; anophelines produce anti-thrombin activities, whereas culicines secrete Factor Xa-directed anticoagulants (8). In contrast, inhibition of platelet aggregation seems to have been achieved in most hematophagous arthropods by the salivary apyrase (ATP-diphosphohydrolase, E.C. 3.6.1.5) (2,4). When vascular tissue is damaged, the disrupted cells release ATP and ADP at high concentrations into the extra cellular environment where ADP promotes platelet activation and aggregation. The activated platelets may in turn release in the medium their ADP-containing granules, recruiting additional platelets to the site of injury. The function of the apyrase, injected at the feeding site with the saliva, is to inhibit the ADP-induced platelet recruitment and aggregation by hydrolyzing the ADP to AMP and inorganic phosphate.
Molecular cloning and sequence analysis have revealed at least three classes of apyrases of different evolutionary origin. They are represented by the apyrases of the yellow fever mosquito A. aegypti (9,10), the intracellular parasite Toxoplasma gondii (11), and the bedbug C. lectularius (12). The T. gondii apyrase belongs to a large family of ecto-ATPases that are found in a wide variety of organisms and tissues ranging from plants (13) to humans (14), whose role is not yet well understood. The C. lectularius apyrase does not show sequence similarity to any previously characterized nucleotide binding enzyme and belongs to a novel type of ATPases (12). Finally the A. aegypti apyrase shows a high degree of sequence similarity to 5Ј-nucleotidases from different organisms (9).
Using the signal sequence trap technique (15), we previously identified two cDNAs expressed in the salivary glands of the malaria mosquito, Anopheles gambiae, showing similarity to the gene encoding the A. aegypti apyrase. Because of their tissue-specific pattern of expression we suggested that they could be derived from malaria mosquito apyrase and 5Ј-nucleotidase genes. We report here the isolation of the corresponding * This work was supported by European Union Grant ERBFM-RXCT960017 (to M. C and C. L.) and by a grant from the United Nations Developmental Program/World Bank/World Health Organization Special Program for Research and Training in Tropical Diseases (TDR) (to M. C.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The full-length cDNAs and their developmental expression profiles. Expression of tagged recombinant proteins in COS-7 cells shows that both proteins are secreted. Finally, we provide evidence that an 800 bp 1 fragment located at the 5Ј-end of the putative A. gambiae apyrase-coding region is able to drive specific expression of the Escherichia coli ␤-galactosidase reporter gene in the salivary glands of transgenic Drosophila melanogaster.

EXPERIMENTAL PROCEDURES
Mosquito Colony-The A. gambiae strain used in this study was the homokaryotypic GASUA reference strain (Xag, 2R, 2La, 3R, 3L). Individuals from different developmental stages were collected, frozen in liquid nitrogen, and stored at Ϫ80°C before nucleic acid isolation.
Libraries Screening and Sequence Analysis-If not otherwise specified, general nucleic acid manipulations were performed according to standard procedures (16,17). The A. gambiae thoracic cDNA library (15) and the genomic library (18) were screened by a gene amplification-based method (19) using the following gene-specific oligonucleotide primers: AgApy-F, 5Ј-AAAGTGCTGCTGCTAATC-3Ј; AgApy-R, 5Ј-AATACAGGTGTCACCTTCC-3Ј; AgApyL1-F, 5Ј-GGCAGAATGGCACTGGTACG-3Ј; AgApyL1-R, 5Ј-CACTCTTCAGCTGCTTGATC-3Ј. Signal peptide prediction analysis was performed by using the SIG-NALP program (20). Sequence comparison and data base searches were done by using the Wisconsin Package Version 9.1 (Genetics Computer Group, Madison, WI) and the BLAST program (21). Multiple alignments were obtained using the CLUSTAL W program (22) and the Java Multiple Alignment editor at the World Wide Web server of the European Bioinformatics Institute. The phenogram was obtained using the Neighbor joining option in PAUP* 4.0b2 (23) using the E. coli 5Јnucleotidase sequence as the out-group to root the tree. Tree topology was statistically tested by bootstrap analysis (2000 replicates).
RNA Purification and Expression Analysis-Five micrograms of total RNA was used for Northern analysis (16). The full-length cDNAs encoded by AgApy and AgApyL1 and a 590-bp fragment from the 3Ј-UTR of the A. gambiae actin gene (U02964: nucleotides 1707-2297) were used as probes (24).
Construction of myc-tagged Recombinant Clones-The myc epitope, EQKLISEEDL, was used to replace the peptides of identical length, SERSSKCKAA (amino acids 55-64) and NQKSSTCTNS (amino acids 52-61), respectively, in the AgApy and AgApyL1 proteins. The AgApymyc and the AgApyL1-myc were obtained by the overlap PCR amplification technique (26). The final amplification products were cloned into the pBKCMV vector (Stratagene), modified according to manufacturer's instructions to allow for higher expression levels in eukaryotic cells. The AgApyL1-myc-Crat construct contains the carboxyl terminus of the rat 5Ј-nucleotidase (amino acids 548 -576) in place of the corresponding AgApyL1 end (amino acids 549 -570). It was constructed from AgA-pyL1-myc by substitution of a BglII-XhoI restriction fragment with a PCR fragment containing the carboxyl-terminal domain of rat 5Ј-nucleotidase. All of the constructs were verified by sequencing before use.
Oligonucleotide primers used for the construction of the clones described above are available upon request.
Expression in COS-7 Cells and Western Blot Detection of Recombinant Proteins-COS-7 cells were cultured in DMEM supplemented with 10% heat-inactivated fetal bovine serum, 100 units/ml penicillin, and 100 g/ml streptomycin (DMEM complete) in a humidified incubator at 37°C with 5% CO 2 . Cells were transfected using the LipofectAMINE reagent (Life Technologies, Inc.) according to the manufacturer's instructions. Briefly, 1-3 ϫ 10 5 cells were seeded in 2 ml of DMEM complete in 35-mm wells and, after 24 -36 h, transfected for 6 h at 37°C using 2 g of plasmid DNA. Twenty-four hours after transfection the supernatant was removed, and the cells were carefully washed several times with fetal bovine serum-free DMEM to remove traces of serum and incubated in 2 ml of fetal bovine serum-free DMEM. After 48 to 72 h of incubation, the medium was removed, centrifuged twice at 1,000 ϫ g for 10 min and then once at 14,000 ϫ g for 5 min to remove detached cells and debris. After the addition of phenylmethylsulfonyl fluoride (40 g/ml), the supernatants were concentrated using Microcon YM-30 filter devices (Millipore) and stored at Ϫ20°C for Western blot analysis. Concentrated supernatants were subjected to SDS/polyacrylamide gel electrophoresis and transferred onto nitrocellulose filters (Schleicher & Schuell). myc-tagged recombinant proteins were stained with the mouse anti-c-myc-peroxidase monoclonal antibody (Roche Molecular Biochemicals) and detected using the ECL-plus system (Amersham Pharmacia Biotech).
Drosophila Transformation and Histochemical Stainings-The oligonucleotide primers AgApyPr-5ЈEcoRI (5Ј-CTAGGAATTCGCTTG-TAGGTGACGCTGTG-3Ј) and AgApyPr-3ЈBamHI (5Ј-CTAGCCTAG-GCACGCTTCGCAGATATTAC-3Ј), containing the EcoRI and BamHI restriction sites at their ends respectively, were used to amplify the 800-bp segment upstream of the AgApy gene. This segment was directionally cloned into the expression vector pCaSpeR-AUG-␤gal (27). The resulting pCaSpeR-Apy-␤gal was microinjected into yw D. melanogaster embryos (carrying a mutation in the yellow and white genes) along with an integration-defective helper plasmid as the source of P transposase. Several transformed individuals were obtained, and three independent homozygous lines were established through crosses with strains carrying appropriate balancer chromosomes. The lines, designated Apy5, Apy9, and Apy13, were analyzed by Southern blot hybridization and assayed for ␤-galactosidase activity. Apy5 and Apy13 contained a single insertion, whereas the Apy9 line carried a double insertion of the transposon.
Small openings were made in otherwise intact adult flies to allow the staining solution to enter the body cavity. Flies were assayed individually in 96-well plates and incubated at 37°C in 100 l of staining solution (50 mM sodium phosphate, pH 8.0, 2 mM potassium ferrocyanide, 2 mM potassium ferricyanide, 0.3% X-gal, 15% Ficoll-400). Several individuals of both sexes were analyzed for each line. Typically, staining started to appear after 5 to 6 h, but intense staining of the glands was only observed after overnight incubations. After incubation, flies were dissected in phosphate-buffered saline to extract the salivary glands. The staining pattern of each line was compared with that of the recipient strain yw and of the Lys␤-gal stock. The latter carries a P element insertion with ␤-galactosidase expression under the control of the salivary gland-specific lysozyme P promoter of D. melanogaster (28) and typically exhibits a strong staining in salivary glands after 1-2 h of incubation at 37°C. 2

RESULTS
The Full-length cDNA Corresponding to the cF3 Fragment May Encode the A. gambiae Apyrase-In a previous study, we identified two short cDNA fragments, cF3 and iC6, whose conceptually translated proteins showed similarity to the A. aegypti apyrase and to several members of the 5Ј-nucleotidase family (15). cF3 expression was found to be restricted to female salivary glands, with the corresponding transcripts mainly localized in the distal-lateral lobes. In contrast, iC6 expression, which was clearly enriched in female glands, could also be detected at a lower level in other tissues. These results were compatible with cF3 representing the A. gambiae salivary apyrase and iC6, a 5Ј-nucleotidase. To confirm these observations, we screened a thoracic cDNA library from A. gambiae adult females and isolated the corresponding full-length cDNAs. In the remaining part of the text we will keep the designation cF3 and iC6 to indicate the partial clones previously isolated by the signal sequence trap technique and we will refer to the corresponding full-length cDNAs as AgApy and AgApyL1, respectively.
Using cF3-specific oligonucleotide primers for a PCR-based screening we isolated a cDNA that was 2253 bp in length. Sequence analysis showed that it lacked 19 nucleotides at the 5Ј-end that were present in the cF3 clone (15). The 2272nucleotide-long mRNA that can be reconstructed from this two-clone contig is likely to represent the full-length transcription product and contains a 1671-nucleotide-long open reading frame and a 584-base 3Ј-UTR (Fig. 1A). The putative protein encoded by this mRNA is similar in size (557 amino acids), molecular mass (61.7 kDa), and isoelectric point (8.83) to the A. aegypti apyrase (9, 10) and contains five potential N-linked glycosylation sites. Prediction analysis showed the presence at the amino terminus of a cleavable signal peptide, whereas no regions with transmembrane properties could be detected, suggesting that this mRNA encodes a secreted protein. Sequence comparison of the conceptually translated protein to the A. aegypti apyrase showed an overall identity of 50.8% and a similarity of 60.8%, whereas identity and similarity to different members of the 5Ј-nucleotidase family were significantly lower (Table I). These observations along with the previously determined female salivary gland-specific expression provide further support for the notion that this mRNA could code for the A. gambiae salivary apyrase. However, we should point out here that we were unable to prove the apyrase activity of the AgApy gene product by using a myc-tagged recombinant version of the protein (see also "Discussion").
Because of our interest in the potential applications of upstream regulatory sequences determining salivary gland-specific gene expression in A. gambiae, we screened a genomic library and isolated a clone containing the entire region encoding this gene. The primary transcript and 800 bp of 5Ј end sequences as well as ϳ150 bp to the 3Ј end of the polyadenylation site are shown in Fig. 1A. The putative A. gambiae apyrase gene (AgApy) contains six exons separated by five small introns and, as outlined in Fig. 1B, it is comparable in its general organization with the A. aegypti apyrase. The position of introns I to IV is perfectly conserved in the two mosquito species, whereas the intron V of the A. gambiae gene clearly corresponds to intron VI of the A. aegypti apyrase. The other introns present in the A. aegypti gene are not found in A.
gambiae. An additional difference involves the 3Ј-UTR, which in A. gambiae is longer by almost 600 nucleotides, compared with the only 30-base-long 3Ј-UTR found in the A. aegypti apyrase gene.
Two Additional Apyrase-like Genes in A. gambiae-As previously reported, a second cDNA, iC6, showing similarity to the A. aegypti apyrase was isolated in the signal sequence trap screen (15). However, because its expression was not restricted to the salivary glands, it was thought to represent an A. gambiae 5Ј-nucleotidase. Champagne et al. (9) suggest that the A. aegypti apyrase evolved from a 5Ј-nucleotidase family member by gene duplication and divergent evolution. Because ecto-5Јnucleotidases are attached to the plasma membrane by glycosylphosphatidylinositol anchors, the evolution of the secreted apyrase proteins, adapted to blood-feeding, may have involved the loss of the hydrophobic carboxyl-terminal domain that includes this structure (29). With the aim of better understanding of the evolutionary relationship between these two proteins, we isolated the full-length cDNA corresponding to iC6, and we designated the corresponding gene AgApyL1 (A. gam-biae apyrase-like 1). The AgApyL1 cDNA is ϳ1.8 kilobases in length, with an open reading frame potentially encoding a protein of 570 amino acids and containing an amino-terminal signal peptide. Sequence comparison showed that the putative protein shared higher similarity with apyrases than with 5Јnucleotidases, and surprisingly, the degree of similarity was significantly higher to the apyrase of A. aegypti than to the putative A. gambiae apyrase (Table I). The conceptual translation products of AgApy and AgApyL1 can be easily aligned to several other members of the apyrase/5Ј-nucleotidase family (not shown). All these proteins show a common general structure, with an amino-terminal signal peptide of variable length and a high degree of conservation in the seven domains known to characterize enzymes having apyrase or 5Ј-nucleotidase activity (9,30). The six-amino acid sequence GKYVGR previously identified in the sixth domain of the A. aegypti apyrase as the putative nucleotide-binding site (9) is perfectly conserved in both the presumed A. gambiae apyrase and the apyrase-like-1 proteins. Fig. 2 shows the alignments of the carboxyl-terminal domains of the conceptually translated proteins AgApy and AgApyL1, the A. aegypti apyrase (9), the 5Ј-nucleotidases from rat, human, and from the cattle tick Boophilus microplus (29,31,32), and two additional dipteran members of the family, recently submitted to data bases: the 5Ј-nucleotidase from the sandfly L. longipalpis and the chrysoptin from Chrysops sp. The comparison of the carboxyl-terminal portions of these proteins shows that the 5Ј-nucleotidases from rat, human, and B. microplus contain an additional terminal domain that is highly hydrophobic and which is known to function as the signal for glycosylphosphatidylinositol anchoring to plasma membranes (29,32). This terminal portion is not present in AgApy, in the A. aegypti apyrase, and in the chrysoptin, whereas the L. longipalpis 5Ј-nucleotidase and the AgApyL1 protein contain shorter carboxyl-terminal regions of 5 and 13 amino acids, respectively. However, these regions do not show the characteristic hydrophobic profile exhibited by 5Ј-nucleotidases (not shown), suggesting that these proteins as well as the two mosquito apyrases and the chrysoptin also may be secreted. The relationships among these different members of the family are more strikingly represented in Fig. 3, where the tree obtained from the alignment of the entire peptide sequences is shown. Interestingly two different clusters can be clearly recognized; the first includes the A. aegypti apyrase, AgApy, AgApyL1, and the chrysoptin, whereas the remaining 5Ј-nucleotidases form a second group.
An additional member of this family of apyrase/5Ј-nucleotidase-like proteins was found by searching a data base that includes sequences from the ends of genomic clones of an A. gambiae bacterial artificial chromosome library. One of these clones, AK0AA011L15B1, contains sequences that showed similarity to apyrases and 5Ј-nucleotidases. The entry could potentially include an exon that is flanked by sequences matching the consensus for donor and acceptor splice sites and that has the potential to encode a 277-amino acid-long polypeptide. This putative partial protein showed similarity to the region encoded by the exons 3 to 5 of the AgApy cDNA, and it is most likely another representative of the apyrase/5Ј-nucleotidase family. We will refer to this additional member of the family as AgApyL2 (A. gambiae apyrase-like 2).
Developmental Expression of the Putative Apyrase and Apyrase-like Genes-We had previously shown that the AgApy gene is specifically expressed in the salivary glands of adult females, whereas AgApyL1 also was found to be expressed at lower levels in other female tissues and in males (15). This was confirmed by Northern analysis on total RNA from adult male and female mosquitoes (Fig. 4A), the only difference being that FIG. 1. A, sequence of a genomic fragment containing the AgApy gene (AJ237705). The nucleotide coding sequence is shown above the conceptual translation product of the corresponding cDNA. The putative TATA and CAAT boxes, the translation initiation (ATG), the putative signal peptide, and the polyadenylation signal are underlined. The arrow indicates the probable transcription start site. Introns are shown in lowercase letters, and the two invariable nucleotides of the donor and acceptor splice sites are in bold characters. Circles highlight potential N-linked glycosylation 5Ј-Nucleotidase Family Members from A. gambiae the AgApyL1 transcript, which is highly abundant in females, is not detectable in adult males. This is presumably the result of the lower sensitivity of this technique as compared with RT-PCR.

5Ј-Nucleotidase Family Members from A. gambiae
abundance increases shortly after blood-feeding, when the salivary glands are depleted of the corresponding protein. This pattern of expression would be consistent with the role of apyrase in blood-feeding. More complex is the expression pattern of AgApyL1, which is clearly expressed in females and at different larval stages but also at lower levels in males and in early pupae. Finally, the expression of the AgApyL2 gene is restricted to larval stages, suggesting that this member of the apyrase/5Ј-nucleotidase family may play some specific function during the larval stage, perhaps linked to nucleotide metabolism.
Expression of myc-tagged Recombinant cDNAs in COS-7 Cells-To ascertain whether the putative apyrase and the apyrase-like 1 proteins are indeed secreted, we tagged them with an epitope that would enable their immunological detection and followed the transient expression of the AgApy-myc and the AgApyL1-myc constructs in COS-7 cells. In addition, we also analyzed AgApyL1-myc-Crat, a construct derived from AgApyL1-myc, whose endogenous carboxyl-terminal domain was replaced by the carboxyl terminus of the rat 5Ј-nucleotidase (see boxed sequences at the carboxyl terminus in Fig. 2); this domain is known to be responsible for the glycosylphosphatidylinositol anchoring of this protein (31). After transfection and overexpression in COS-7 cells, the supernatants were concentrated and analyzed by Western blot for the presence of the recombinant proteins using an anti-c-myc monoclonal antibody. As shown in Fig. 5 (lanes 1 and 2), both the AgApy-myc and the AgApyL1-myc proteins were found in the supernatant of COS-7 cells. However, no proteins could be detected in the supernatant by the anti-c-myc antibody either in untransfected cells (lane 4) or in cells transfected with AgApyL1-myc-Crat (lane 3). The 13-amino acid difference in length between AgApy-myc and AgApyL1-myc can only partially account for the observed different relative mobilities of the two recombinant proteins, which may be due to post-translational modifications. Altogether, these results strongly suggest that both AgApy and AgApyL1 have the properties of secretory proteins.
Transformation of D. melanogaster with the AgApy Promoter-Because of our interest in salivary gland-specific promoters, which may be of use in future vector control campaign, we decided to test whether the 800 nucleotides located immediately upstream of the putative apyrase starting codon were able to drive the tissue-specific expression of a reporter gene in D. melanogaster. Using PCR, we inserted the 800-bp AgApy segment in front of the E. coli ␤-galactosidase gene in the transformation vector pCaSpeR-AUG-␤gal (27). The resulting construct pCaSpeR-Apy-␤gal, schematically represented in the upper part of Fig. 6, was used for transformation. Flies from different transgenic lines were histochemically stained for ␤-galactosidase activity, yielding essentially the same expression patterns. In all cases, weak staining was detectable in the thoracic region after 5 to 6 h of incubation at 37°C. Only after longer incubations (Ͼ16 h), a more intense color appeared that could be clearly ascribed to the staining of the glands. The staining was more intense in the Apy9 line that contains a double insertion of the transgene. The glands were stained both in adult males and females, with the blue color being especially dark in the terminal, convoluted portion of Drosophila salivary glands. No staining was detectable, even after extensive incubations, either in the larval glands or in the adult glands of the recipient strain yw. As can be seen in Fig. 6, A and B, a staining was also visible in the midgut after long incubation periods. However, staining of the midgut of similar intensity was also found both in the negative and in the positive control strains, i.e. the recipient strain yw and the Lys␤-gal strain, which contains a P element insertion carrying the ␤-galactosidase under control of the salivary gland-specific lysozyme P promoter of D. melanogaster. 2

DISCUSSION
We have isolated from A. gambiae a cDNA that shows strong similarity to the A. aegypti apyrase gene and to different members of the 5Ј-nucleotidase family. The mRNA is specifically expressed in the salivary glands of adult females, where it is produced mainly in the gland cells constituting the distallateral lobes, a region of the mosquito glands known to express genes whose function is related to blood-feeding (2,15,33). The deduced protein contains at the amino terminus a putative signal peptide and does not seem to carry any transmembrane region; these data imply that it is secreted. This interpretation is supported by the finding that a myc-tagged recombinant version of the protein is secreted in the cell culture medium when expressed in COS-7 cells. All these observations strongly suggest that the AgApy gene is likely to encode the A. gambiae apyrase and is responsible for the previously described antiplatelet activity ascribed to the salivary apyrase (34).
We have also identified another protein belonging to the apyrase/5Ј-nucleotidase family, and intriguingly, sequence comparison showed that it is significantly closer to the A. aegypti apyrase and to AgApy than to 5Ј-nucleotidases from different organisms. In the absence of additional information regarding its biochemical activity, we named the corresponding gene AgApyL1 (A. gambiae apyrase-like 1). As previously shown, the expression of AgApyL1 is enriched in female salivary glands (15), yet the corresponding transcript is detectable also in other female tissues, at different larval stages, and at a lower level in adult males. Such a pattern of expression would

5Ј-Nucleotidase Family Members from A. gambiae
be compatible with a possible 5Ј-nucleotidase function. Moreover, AgApyL1 is a secreted molecule, as suggested by the expression of the myc-tagged recombinant protein in cultured mammalian cells. These observations raise the possibility that AgApyL1 may have 5Ј-nucleotidase activity and, when secreted with the saliva and injected into the host skin, may play some role in blood-feeding. As was recently proposed for the salivary 5Ј-nucleotidase of the sandfly L. longipalpis (35,36), it may be involved in the production of adenosine from the ADP and ATP released from the injured host tissue. Adenosine is not only an antagonist of platelet recruitment, adhesion, and aggregation but also a potent vasoactive agent (14); therefore, the hemostatic action of apyrase, which results in the production of AMP, would be amplified by the activity of a salivary 5Ј-nucleotidase, capable of further converting the AMP to adenosine.
We should stress here that we tried to assay for apyrase and/or 5Ј-nucleotidase activity in concentrated cell supernatants containing, respectively, the AgApy-myc and the AgA-pyL1-myc recombinant proteins, but we could not detect any orthophosphate release according to the assay of Fiske and Subbarow (38). It is likely that the presence of the myc epitope interferes with the correct folding or with the activity of the recombinant proteins. We have proposed, based both on the RT-PCR expression analysis and on the RNA in situ hybridization to salivary glands (15), that AgApy and AgApyL1 encode, respectively, an apyrase and a 5Ј-nucleotidase. However, at this stage, we cannot rule out the possibility that they represent for example two apyrases or two secreted 5Ј-nucleotidases. The unambiguous assignment of the functions of these genes will require the purification of the corresponding proteins or their expression in other in vitro systems (i.e. baculovirus expression) that will preserve their enzymatic activity.
Along with the A. aegypti apyrase, AgApy would represent the second apyrase gene isolated from a mosquito species. If this is the case, then the observation that both enzymes are 5Ј-nucleotidase family members suggests that the emergence of the apyrase function adapted to blood-feeding may have originated by a gene duplication event that took place before the separation of Anophelinae and Culicinae from their common progenitor. This would imply a widespread occurrence of apyrases of the 5Ј-nucleotidase type in different mosquito species. Moreover, from our data it appears that culicines and anophelines used different copies of the duplicated genes as their salivary apyrase. Thus, AaApy and AgApyL1 are most likely orthologous, and AaApy and AgApy are paralogous. Both the exon-intron structure of the AaApy and AgApy genes and the difference in the length of their 3Ј-untranslated region would be in agreement with this hypothesis.
We have also revealed the existence of a third member of the family, AgApyL2, which was identified by searching an A. gambiae genomic data base; it exhibits a larval-specific expression profile. We do not know what function this gene may have during the larval stages and/or if there is any tissue-or organspecific expression in the larvae; however, it could play some role in connection to nucleotide metabolism. The existence of multiple genes related in sequence to 5Ј-nucleotidase has been demonstrated in other organisms (32), yet the significance of this redundancy remains to be clarified.
The putative apyrase gene represents the first salivary gland-specific gene isolated from the African malaria vector A. gambiae. The identification of control sequences capable of conferring salivary gland-specific expression as well as a more detailed understanding of the physiology of the glands are the necessary steps toward the development of malaria control strategies based on the genetic modification of the mosquito vector (33, 39 -41). Salivary glands represent a crucial target organ because malaria parasites can be transmitted to the vertebrate host with the saliva only after invading and traversing the salivary glands (40,42). The availability of specific promoters able to drive the expression of an "anti-parasite" gene in the glands may be of great help as part of a multi-step blocking strategy as soon as techniques for the introduction of foreign genes are available for A. gambiae. For these reasons we tested the promoter activity of the region located immediately upstream to the starting codon of the AgApy gene. We used transgenic D. melanogaster rather than A. aegypti because transformation of the yellow fever mosquito is a rather recent achievement and is a tedious technique. In comparison, the transformation of the fruit fly is well established, and it can take advantage of a wide variety of transformation tools (43)(44)(45). Moreover, promoter sequences both from chorion and silk gland genes of the distantly related insect Bombyx mori (46,47) and also from midgut-specific genes of A. gambiae (48) have been previously shown to be recognized and correctly expressed in D. melanogaster. Our observations show that some salivary gland-specific transcriptional regulatory elements are also con- served between A. gambiae and D. melanogaster. More specifically, control elements required to direct the correct stage-and tissue-specific expression patterns seem to be maintained. In contrast, the sex specificity of expression was not retained because we never observed any difference between male and female staining patterns. A similar result was found with the midgut-specific promoters of the trypsin genes Antryp1 and Antryp2 in transgenic D. melanogaster. These two genes are never expressed in male mosquitoes, but their promoters could drive tissue-specific LacZ expression both in male and in female D. melanogaster (48). It is likely that A. gambiae sex regulatory elements are either not recognized or absent from the DNA fragment that we used for D. melanogaster transformation. Moreover, the low level of ␤-galactosidase activity suggests the possibility that enhancer elements not included in the construct used for the fruit fly transformation may be important for high transcription rates. This possibility has also been suggested to explain the low levels of luciferase expression obtained after A. aegypti transformation with the endogenous apyrase promoter (37). An alternative possibility is that the 584-bp-long 3Ј-UTR of the AgApy gene is somehow important, maybe for the stability of the corresponding mRNA.
In conclusion we have shown that a short fragment of the A. gambiae putative apyrase promoter is able to drive ␤-galactosidase expression in the salivary glands of D. melanogaster. Larger fragments and/or promoters of other salivary gland genes may need to be tested to obtain a promoter able to drive a stronger and sex-specific expression in the malaria mosquito salivary glands. However, our observations reinforce previous results obtained with the trypsin promoters (48) and suggest that the fruit fly can be reliably used, at least in a preliminary stage, for A. gambiae promoter analysis, which may then be refined or confirmed by the A. aegypti transgenic technology.