Identification of a Tsetse Fly Salivary Protein with Dual Inhibitory Action on Human Platelet Aggregation

Background Tsetse flies (Glossina sp.), the African trypanosome vectors, rely on anti-hemostatic compounds for efficient blood feeding. Despite their medical importance, very few salivary proteins have been characterized and functionally annotated. Methodology/Principal Findings Here we report on the functional characterisation of a 5′nucleotidase-related (5′Nuc) saliva protein of the tsetse fly Glossina morsitans morsitans. This protein is encoded by a 1668 bp cDNA corresponding at the genomic level with a single-copy 4 kb gene that is exclusively transcribed in the tsetse salivary gland tissue. The encoded 5′Nuc protein is a soluble 65 kDa glycosylated compound of tsetse saliva with a dual anti-hemostatic action that relies on its combined apyrase activity and fibrinogen receptor (GPIIb/IIIa) antagonistic properties. Experimental evidence is based on the biochemical and functional characterization of recombinant protein and on the successful silencing of the 5′nuc translation in the salivary gland by RNA interference (RNAi). Refolding of a 5′Nuc/SUMO-fusion protein yielded an active apyrase enzyme with Km and Vmax values of 43±4 µM and 684±49 nmol Pi/min×mg for ATPase and 49±11 µM and 177±37 nmol Pi/min×mg for the ADPase activity. In addition, recombinant 5′Nuc was found to bind to GPIIb/IIIa with an apparent KD of 92±25 nM. Consistent with these features, 5′Nuc potently inhibited ADP-induced thrombocyte aggregation and even caused disaggregation of ADP-triggered human platelets. The importance of 5′Nuc for the tsetse fly hematophagy was further illustrated by specific RNAi that reduced the anti-thrombotic activities in saliva by approximately 50% resulting in a disturbed blood feeding process. Conclusions/Significance These data show that this 5′nucleotidase-related apyrase exhibits GPIIb/IIIa antagonistic properties and represents a key thromboregulatory compound of tsetse fly saliva.


Introduction
Efficient acquisition of a blood meal by hematophagous arthropods relies on a broad repertoire of physiologically active saliva components inoculated at the blood feeding site. These are primarily anti-hemostatic components that interfere with host responses such as vasoconstriction [1,2], primary hemostasis through the adherence and aggregation of thrombocytes [3,4] and a secondary hemostatic cascade mainly relying on serine proteases such as thrombin [5,6]. In the formation of the primary hemostatic plug, adenosine-59-triphosphate (ATP) and adenosine-59-diphosphate (ADP) released from injured cells and activated platelets, interact with purinergic P2 receptors (P2X 1 for ATP and P2Y 1 and P2Y 12 for ADP) and fulfil key roles in the activation and aggregation of platelets and strongly contribute to the amplification of the initial hemostatic response [reviewed in [7,8]]. ADP acts on the G protein-coupled P2Y 1 and P2Y 12 receptors resulting in thrombocyte shape change, aggregation (through activation of the fibrinogen receptor) and the production of thromboxane A 2 that has prothrombotic properties. ATP binds onto the cation-gated P2X 1 receptor that is present on thrombocytes resulting in cytoskeletal reorganisation and higher responsiveness in terms of aggregation and degranulation towards other platelet activating triggers such as collagen [reviewed in [7,8]]. To overcome these ATP-and ADP-related host responses, blood sucking insects have ATP/ADP-hydrolysing [ATP(D)ase] enzymes present in their salivary secretions [9]. These enzymes are often called apyrases (nucleosidetriphosphate diphosphohydrolases) and have been described in the saliva of a variety of blood feeding arthropods such as Cimex lectularius bed bugs [10], Ixodes dammini ticks [11], Aedes aegypti and Anopheles gambiae mosquitoes [12,13], Phlebotomus papatasi sand flies [14] and Triatoma infestans reduviids [15]. These apyrases were shown to mainly belong to two different genetic families: one group of apyrases belongs to the 59-nucleotidase gene family that has been described in several hematopaghous arthropod species [13,15], while a completely different type of apyrases has been found in bed bugs (Cimex sp.) and sand flies [10,14,16]. However, all these apyrases are cation-dependent enzymes that release inorganic phosphate (P i ) from ATP and ADP but not from AMP, thereby inhibiting the purinergic activation and subsequent aggregation of thrombocytes [9,11]. Beside their thromboregulatory role, apyrases present in the saliva of blood feeding arthropods have also been implicated in anti-inflammation by the degradation of ATP as agonist of the inflammatory purinergic receptors [17]. As such, salivary apyrases of hematophagous arthropods could inhibit several aspects of hemostasis and inflammation thereby promoting the blood feeding process and possibly facilitating pathogen transmission, a feature that we have described for the saliva of tsetse flies, blood feeding insects that transmits the protozoan agents of African trypanosomiasis [18].
In the tsetse fly, the presence of salivary ATP(D)ases has been suggested by the observation that ADP-induced thrombocyte aggregation is inhibited by an unidentified apyrase activity of .30kDa [4]. Strikingly, tsetse saliva could also disaggregate ADPtriggered platelets which was attributed by the authors to the apyrase enzymatic activity. In this study, we report on the identification of a 59nucleotidase-related protein in Glossina morsitans morsitans and provide molecular, biochemical and functional evidence that this is the major apyrase in tsetse saliva that is also able to bind to the fibrinogen receptor and inhibit and reverse ADP-induced platelet responses.

Results
In silico analysis of the 59Nuc cDNA Screening of the lgt11 salivary gland cDNA library with the 385 bp 59Nuc probe and subsequent sequencing of 5 positive clones identified a full-length 59nucleotidase encoding cDNA of 1976 bp. (Fig. 1A). In silico analysis revealed a 102-bp 59untranslated region followed by an open reading frame corresponding to 555 amino acid residues and a 185-bp 39-untranslated region that contains a polyadenylation signal (EMBL: AF384674).
Sequence analysis of the predicted translation product, designated as G. m. morsitans 59nucleotidase-related protein (59Nuc), identified the first 25 amino acids as a potential signal peptide leaving a mature protein of 530 amino acids with a calculated molecular weight of 59375 Da and an isoelectric point of 7.32. Figure 1. In silico analysis of the 59Nuc cDNA. (A) Alignment of the G. m. morsitans 59nucleotidase conceptual translation product with several members of the 59nucleotidase family. Homologous residues are shown on a shaded background, residues that were identified for E. coli 59Nuc to be important for catalysis, substrate and cofactor binding are indicated on an orange background. (B) Amino acids predicted to be involved in catalytic activity and co-factor and substrate binding of the E. coli 59nucleotidase and their homologous residues in tsetse 59Nuc (C). Structure prediction of tsetse 59Nuc relying on comparison of Hidden Markov Models, based on resolved structures of homologous 59nucleotidases. The predicted structure of tsetse 59Nuc and the resolved structures of Escherichia coli and Thermus thermophilus 59nucleotidases are depicted (respective PDBs: 1h05 and 2Z1a). doi:10.1371/journal.pone.0009671.g001 Four putative N-glycosylation sites (Asn 80 , Asn 173 , Asn 270 and Asn 440 ) and 2 O-glycosylation sites (Thr 256 and Thr 389 ) were identified. BLAST analysis revealed significant sequence similarities to a group of enzymes belonging to the 59-nucleotidase family (Table S1) but lacks a hydrophobic C-terminal domain that signals for glycosylphosphatidylinositol anchoring. High degrees of similarity were apparent within the seven domains known to characterize enzymes exhibiting 59nucleotidase activity [13] (Fig. 1A). Residues that were documented for E. coli 59nucleotidase to be crucial for co-factor and substrate binding and catalytic activity [19,20] are all present in the G. m. morsitans 59nucleotidaserelated protein (Fig. 1B). Structure modelling based on comparison of hidden Markov models, suggested strong structural similarity with resolved structures of the Escherichia coli and Thermus thermophilus 59nucleotidase (respective PDBs: 1hO5 and 2Z1a, Fig. 1C).

Identification of the 59Nuc gene and expression analysis
The 59Nuc structural gene of .4 kb including the 59 and the 39 UTRs was identified and revealed by Southern blot analysis to be present as a single copy in the tsetse fly genome. The gene was shown to contain 7 introns, including two with the size of approximately 1 kb ( Fig. 2A). Northern blot analysis using salivary gland RNA corroborated that the gene resulted in a single transcript of around 1.7 kb, corresponding to the identified 1668 bp coding sequence (data not shown).
RT-PCR analysis revealed that 59Nuc transcription occurs exclusively in the salivary glands ( Fig 2B). Western blot analysis with anti-59Nuc polyclonal rabbit serum resulted in a strong positive reaction around 65 kDa and a significantly weaker signal for the 125-140 kDa protein bands (previously identified as sgp3 that consists of a 59nucleotidase domain, EMBL:EF398273, [21]) in the tsetse fly saliva (Fig. 2C). The molecular weight of 65 kDa as compared to the predicted 59.4 kDa suggested the presence of glycosyl modifications, as demonstrated using a glycosylation detection reagent and a mobility shift of about 5 kDa after PNGase F treatment (data not shown). In the soluble extract of the foregut/proventriculus tissue, a weak but significantly positive signal around 65 kDa could be detected (Fig. 2C), suggesting overflow of salivary proteins into this part of the alimentary tract. In teneral (newly emerged, non-fed) flies and at different time points after blood feeding, the relative abundance of the 65 kDa 59Nuc protein in the total saliva protein pool remained at a similar constant level (data not shown).
Total tsetse fly saliva exerts ATP(D)ase activity Tsetse fly saliva exerts ATP(D)ase activity as illustrated in kinetic assays using ATP and ADP as substrates followed by determining the inorganic phosphate (P i ) release. No AMPase activity could be detected, illustrating that tsetse salivary nucleotidases exert apyrase activity. Beside ATP and ADP, other nucleoside triphoshate and diphosphates substrates were converted such as rATP.rGT-P = rUTP and rADP = rUDP.rCDP (data not shown). Separation of total tsetse saliva under native conditions followed by zymographic ATPase detection revealed activity confined to the upper four high molecular weight protein complexes (Fig. 3A) that are positive in a 59Nuc-specific western blot (data not shown). Analysis of the pH dependence revealed alkaline pH optima around 8.0 and 9.5 (Fig. 3B). This salivary apyrase activity is dependent on divalent ions as cofactors with a preference for Mg 2+ and Mn 2+ and is completely inhibited by the addition of EDTA (data not shown).
In vivo functional analysis of 59Nuc by RNA interference (RNAi) RNAi was applied for in vivo functional analysis of 59Nuc by intrathoracal injection of 15 mg dsRNA per fly. This resulted in a highly specific knockdown of the 65 kDa 59nucleotidase with 90% reduction at the mRNA level without silencing the homologous sgp3 gene (Fig. S1). A maximal translational silencing of 75% was obtained at day 12 post injection as illustrated by densitometric analysis after SDS-PAGE (Fig. 4A). Saliva from these RNAitreated and control dsRNA injected tsetse flies was compared in a P i -release based ATP(D)ase assay revealing an inhibition of the salivary ATPase activity by 49% and ADPase activity by 45% (Fig. 4B). As a result of silencing, feeding efficiencies on anesthetized mice of 59Nuc RNAi flies (64612 mg/s, n = 32) were significantly lower (p = 0.037, Fig. 4C) than those of control RNAi treated individuals (88612 mg/s, n = 41). Indeed, flies with reduced salivary apyrase activity generally acquired smaller blood meals (13.061.5 mg versus 16.861.7 mg, p = 0.061) and required slightly longer times to complete blood feeding (267628 s versus 248627 s, p = 0.200). Evaluation of the inhibition of ADP-induced thrombocyte aggregation by K serially diluted saliva samples from RNAi treated flies (5 -0.625 mg/ml, Fig. 4D), revealed an approximate 50% activity reduction upon 59Nuc silencing. Also the capacity of saliva to disaggregate ADP-triggered platelets was reduced to a similar degree by 59Nuc knockdown. To illustrate that the knockdown specifically affects the 59Nuc activity, the potent inhibitory activity of tsetse saliva to bovine thrombin remained unaffected after 59Nuc RNAi (data not shown).

Refolding of recombinant 59Nuc/SUMO fusion protein
Recombinant 59Nuc/SUMO fusion protein, purified from inclusion bodies using immobilized metal affinity chromatography, was refolded by rapid dilution in different non-denaturing buffers. Various pH conditions (pH 4.0 to 10) in combination with additives such as 0.5 M L-arginine, 10 mM b-mercaptoethanol and oxidized/reduced glutathion (0.1 mM/1 mM) were tested. Primary read-out for the resolubilisation and refolding efficiency were precipitation (O.D. 405nm) and ATPase activity (Fig. 5A). Recombinant 59Nuc could be readily solubilized in alkaline buffers from pH 8 onwards without the addition of L-arginine. Gain-ofactivity was obtained without the requirement of an oxidoshuffling system (Fig. 5A). Moreover, reducing agents such as bmercaptoethanol and dithiothreitol did not affect ATPase activity, illustrating that disulfide bridges are not required for activity of 59Nuc.

Enzymatic characteristics of refolded 59Nuc/SUMO fusion protein
Refolded monomeric 59Nuc exerted ATPase and ADPase but no AMPase activity, similar to what was observed for total tsetse fly saliva. Evaluation of the pH dependence illustrated a preference for alkaline conditions, with an optimum at pH 8.0 for both ATPase and ADPase activity (Fig. 5B). Different nucleoside triphosphate and diphoshate substrates could be converted such as rATP&dATP = dTTP.rGTP = rUTP and rADP = rUDP.rCDP (Fig. 5C). The temperature optimum for the ATP(D)ase activity was around 40uC but remained high in a broad range of suboptimal temperature conditions (20u-47uC) (Fig. 5D). Similar as for total saliva, ATPase and ADPase activities of the recombinant 59Nuc depended on divalent ions with a preference for Mn 2+ and Mg 2+ . To a lesser extent, Ca 2+ could also activate ATPase and ADPase activity, while Cu 2+ , Ni 2+ and Co 2+ support mainly ADPase activity (Fig. 5E). ATPase and ADPase experiments that were performed with total tsetse fly saliva (Fig. 6A) and refolded 59Nuc (Fig. 6B) suggested that the native and recombinant enzymes obeyed Michaelis-Menten kinetics, allowing the estimation of K m and V max from non-linear regression analyses. K m values were determined in three independent experiments, being 4364 mM for ATPase and 49611 mM for ADPase activity of recombinant 59Nuc. V max values were determined to be 684649 and 177637 nmol Pi/ min6mg for respectively ATPase and ADPase activities (Fig. 6C). Comparable K m (3364 mM for ATPase and 1963 mM for ADPase) and V max values (791627 and 239611 nmol Pi/ min6mg respectively) were obtained for total tsetse fly saliva. K cat /K m ratios confirmed a substrate preference of 59Nuc for ATP over ADP (Fig. 6C). Next, a range of ATPase/nucleotidase inhibitors was tested for their influence on the ATP(D)ase activity of 59Nuc. While 1 mM of the Na + /K + -ATPase inhibitor ouabain and alkaline phosphatase inhibitor levamisole had no influence on 59Nuc, 1 mM adenosine, AMP and AP5A as well as 1 mM vanadate and 10 mM NaN 3 had moderate inhibitory effects on enzymatic activity (Table S2). DEPC (2 mM), sodium fluoride (10 mM) and 4,49 diisothiocyanostylbene 2,29 disulfonic acid (DIDS, 100 mM) nearly completely abrogated substrate conversion while concanavalin A enhanced the activity as has been described for other 59 nucleotidases (Table S2).

Fibrinogen receptor (GPIIb/IIIa) binding properties of refolded 59Nuc/SUMO fusion protein
To evaluate the fibrinogen receptor antagonistic properties of 59Nuc/SUMO, its binding to solid-phase immobilized GIIb/IIIa was evaluated. Using both a SUMO-Tag specific and an anti-59Nuc IgG based detection, concentration-dependent 59Nuc binding onto the purified receptor could be demonstrated. Based on four experiments, an apparent K D of 92625 nM was calculated. To further illustrate the specificity of the binding, inhibition studies were performed using varying concentrations of purified human fibrinogen (Fig. 7). These illustrated a dosedependent inhibition of 59Nuc binding by fibrinogen, further confirming that the tsetse fly 59Nuc has GIIb/IIIa receptorspecific antagonistic properties.

Platelet inhibitory properties of refolded 59Nuc/SUMO fusion protein
Recombinant 59Nuc/SUMO protein was assessed for its activity on human platelets using an aggregation assay in microtiter plates. Similar as for total tsetse fly saliva, the recombinant protein inhibited ADP-induced platelet aggregation and even caused disaggregation of thrombocytes in a dosedependent manner (Fig. 8A). Pre-incubation of ADP with the recombinant protein completely abrogated the ADP-induced platelet response (Fig. 8A). Exposure of maximally ADPaggregated platelets to saliva or recombinant 59Nuc resulted in a disaggregation response that for 59Nuc was significant at the highest tested concentration of 20 mg/ml (Fig. 8B).

Discussion
Tsetse fly saliva is a complex mixture of proteins that is essential for the blood feeding process which requires the creation of a blood pool at the bite site and the maintenance of blood fluidity in the mouthparts, crop and anterior midgut. Previously, it was shown that an ADP-hydrolyzing activity, present in a crude salivary extract, can inhibit platelet aggregation (Mant & Parker, 1981).
In this report, we functionally characterized a secreted 65 kDa 59nucleotidase-related protein (59Nuc, EMBL: AF384674) as a major apyrase and fibrinogen receptor (GPIIb/IIIa) antagonist that is estimated to make up about 5% of the tsetse fly saliva protein pool [21]. Experimental evidence is based on the successful suppression of 59Nuc transcription in the tsetse salivary gland by RNA interference and on the production of functional recombinant protein with characteristics that are nearly identical to the native activity. This recombinant 59Nuc still had a strong tendency to aggregate in inclusion bodies despite the presence of a SUMOchaperone for better folding [22]. However, the advantage of purifying 59Nuc from these inclusion bodies was the possibility to obtain homogeneous preparations, as problems of proteolytic degradation and copurification with contaminants has precluded biochemical characterisation of several orthologs. Out of the different refolding conditions that were tested, it was sufficient to make a fast dilution of the protein into a non-denaturing alkaline buffer with pH.8.0, without the need of folding-assisting agents to obtain a good yield of active protein. Successful refolding in similar alkaline conditions has also been reported for rat NTPase ectodomains [23]. The active monomeric 59Nuc that we obtained by size exclusion chromatography after refolding was clearly an apyrase as it excluded AMP as substrate. Preferred substrates were rATP, rADP and rUDP while other nucleoside tri-and diphosphates were less potently converted. The enzyme had also a preference for ribonucleoside over deoxyribonucleoside triphos-phates as rATP was a much better substrate than dATP. These base and sugar preferences of 59Nuc provide the enzyme with the most relevant nucleotide specificities (for rATP and rADP as platelet triggers) in the context of anti-hemostasis. Co-factor studies using recombinant protein and total tsetse fly saliva identified Mn 2+ and Mg 2+ as most effective, while Ca 2+ could only induce moderate activity. This is similar to the triatomine 59nucleotidase related apyrase [15], while most other identified apyrases of blood feeding insects have a cofactor preference for Ca 2+ [10,14,24]. Cu 2+ and Ni 2+ were also found to activate the enzyme (primarily stimulating ADPase activity), similar as for some E. coli nucleotidases [25], whereas Cd 2+ and Zn 2+ could not induce the tsetse fly 59Nuc although these are suitable cofactors for the E. coli homologue [19]. Assumptions on the mechanism of substrate dephosphorylation as well as a prediction of the tsetse 59Nuc structure could be made based on previous studies on the homologous E. coli 59nucleotidase [19,20]. Structure prediction suggested a protein fold consisting of two domains, linked through an a-helix (see Fig. 1C). The E. coli homologue has the ability to shift between an open (inactive) and closed (active) conformation as a result from an interdomain rotation of 96u. At the interphase of the two domains, at the C-terminal end of 2 sandwiched babab-motifs, the dimetal catalytic site is located. All residues that were illustrated to coordinate the metal ions in E. coli 59Nuc are also present in the homologous Glossina m. morsitans 59Nuc (Asp 38 , His 40 , Asp 91 , Asn 123 , His 124 , His 249 ). Only the E. coli Gln 254 is substituted by a histidine (His 251 ), a replacement which is also found in e.g. human, Drosophila and Anopheles gambiae 59Nuc. The Asp-His dyad that is part of the catalytic core structure as well as most of the residues that constitute the substrate binding pocket are conserved in tsetse 59Nuc (respectively His 124 and Asp 127 and Arg 354 , Arg 358 , Arg 402 , Phe 421 and Phe 505 ). The crucial involvement of the bivalent cations and the histidine that might function as a general base to drive the nucleophilic attack by one of the coordinated water molecules was illustrated by the efficient inhibition by respectively EDTA and DEPC. Disulfide bridges and glycosylation were suggested to have little influence on the core structure and enzymatic function of other 59nucleotidase because the involved cystein residues are already in close proximity and glycosylated asparagine residues are located at the surface of the molecule [26]. Our data indeed show that reducing agents (b-mercaptoethanol and DTT) do not affect catalytic rates and that the non-glycosylated recombinant protein (expressed in E. coli) exerts potent activity. The fact that AMP does not serve as a substrate might suggest that AMP is not properly accommodated in the active site and that the a-phosphate group cannot be hydrolytically removed by the tsetse 59Nuc. Given that AMP as well as adenosine and AP5A had some inhibitory effects (20-50%) on 59nucleotidase activity at a 506 molar excess, these adenosine derivatives probably bind to the active site with low affinity, where the Phe 429 and Phe 498 residues could be responsible for stacking of the adenine ring [20]. Nevertheless, the functional result is that the saliva converts ATP and ADP released at the bite site not further than AMP, where AMP has been suggested to promote the blood feeding process by inducing vasodilatation [27]. As adenosine is not produced, the functional relevance of two tsetse fly salivary proteins with putative adenosine deaminase activity, tsetse salivary gland growth factors 1 and 2 (TSGF-1 and TSGF-2) [28], remains unclear.
Biochemical characterisation of the refolded recombinant 59Nuc/SUMO protein revealed K m values that were in the low micromolar range (43 mM and 49 mM) and maximal enzymatic rates of 684649 and 177637 nmol P i /min6mg or 0.68 and 0.18 U/mg for respectively ATPase and ADPase. Although limited biochemical information is available on apyrases of blood feeding arthropods, the measured K m values were comparable to those described for e.g. a recombinant mosquito (Aedes aegypti) apyrase [29] and the porcine hepatic canalicular ATP(D)ase [30]. With the definition of 1 unit corresponding to the amount of enzyme required to release 1 mmol of P i /min and the fact that tsetse flies inoculate approximately 4 mg saliva at the feeding site to obtain a 20 ml blood meal, this would correspond to 100 mU/ml  ATPase and 30 mU/ml ADPase under V max conditions (0.79 and 0.24 U/mg for saliva). This inoculated activity, which is confined to 4 high molecular weight (HMW) protein complexes as revealed by native gel separation and zymography, would theoretically be sufficient to convert local concentrations of 200 mM ATP and 70 mM ADP within an average feeding time of 4 minutes. The physiological advantage of these HMW complexes could be the centralization and potentation of several pharmacological activities similar as what has been described for snake venom (reviewed in [31]) and a reduction in diffusion rate from the blood pool where the tsetse fly feeds.
The exclusive expression of 59Nuc in the salivary gland tissue suggested that the activity is primarily required in the saliva inoculum at the bite site. In this context, the potency of tsetse saliva to counter ADP-induced hemostasis has been previously documented [4] and our thrombocyte aggregation assays have illustrated the potency of 59Nuc to inhibit and even reverse ADPinduced platelet responses, a feature which was also described for a 59nucleotidase related fibrinogen receptor (GPIIb/IIIa) antagonist in the saliva of Chrysops deerflies [32]. 59Nuc/SUMO binding studies onto purified human GPIIb/IIIa combined with fibrinogen inhibition studies have clearly illustrated that the tsetse fly 59Nuc is also a fibrinogen receptor antagonist, where the recombinant protein displayed an apparent K D of 92625 nM. This feature is likely causally linked to the platelet disaggregating potential of 59Nuc in tsetse fly saliva. The binding mode of 59Nuc and the homologous chrysoptin to GPIIb/IIIa remains to be elucidated, as both lack the characteristic RGD-motif that has been recurrently found in other fibrinogen receptor antagonists [32]. Moreover, these data illustrate that, unlike chrysoptin, the glycosylation is not absolutely required for GPIIb/IIIa binding of the tsetse fly 59Nuc.
In addition to influencing ADP-induced hemostasis, mediated through P2Y 1 and P2Y 12 receptor ligation and involving downstream fibrinogen-mediated reactions, substrate preference of 59Nuc for ATP suggested that interference with ATP-based host reactions plays an important role in the blood feeding process. It is known that ATP can bind to the P2X 1 receptor on platelets thereby amplifying their responsiveness to various activating stimuli. This feature has been illustrated to especially contribute to hemostasis in high shear rate conditions that might occur in the microvasculature at the blood feeding site [33]. Beside the involvement in thromboregulation, extracellular ATP was also shown to induce IL-1b and IL-18 and to stimulate phagosomelysosome fusion in macrophages [34,35], to provoke chemotaxis and distorted maturation of dendritic cells [36,37] and to amplify the chemotactic gradient that orients granulocyte migration [38]. In line with these properties of ATP and ADP, the presence of ectonucleotidase activities on the surface of various cell types has been shown to be important in thromboregulation [39] and the control of inflammatory reactions in the host [40,41]. As such, salivary apyrases of hematophagous arthropods might not only inhibit several aspects of hemostasis but also reduce inflammation at the bite site. Possibly, the documented anti-inflammatory properties of nucleotidases explain the salivary immunomodulatory action that we have previously described to promote trypanosome infection onset [18].
In addition to its activity at the blood feeding site, analysis of the apyrase activity in total saliva revealed that optimal activities were obtained in alkaline pH conditions that corresponded to the measured pHs in respectively saliva (pH 7.8) [21] and the proventricular tract (pH 9.5). Together with the observation that 59Nuc leaks into the proventriculus, this suggested that the activity is not restricted to the salivary gland and micro-environment of the bite site but that it might also act in the upper part of the alimentary tract. Our in vivo gene silencing experiments revealed that 59Nuc supports the blood feeding process as evidenced by slower feeding rates for RNAi-treated flies that generally obtained smaller blood meals and required slightly longer times to feed. This blood feeding phenotype resulted from a 50% knockdown of apyrase and anti-thrombotic activity despite the remaining 25% 59Nuc protein and the presence of another putative apyrase in tsetse fly saliva, the sgp3 protein that we have previously identified [21] and that remained unaffected by the RNAi. Detailed analysis of the salivary gland EST database ( [42], http://www.genedb. org/genedb/glossina/) further supports the strong dominance of the 59nuc that is represented by 25 ESTs in a total of 37 EST clusters putatively encoding 59nucleotidase-related proteins. Another cluster of 10 ESTs encodes Sgp3 that still remains to be characterized as a true apyrase, while another candidate, GM-544, is only represented by 2 ESTs that are moreover truncated at the 39 end. Collectively, based on the in vivo silencing, the biochemical analyses and platelet aggregation assays, we have calculated that the 65 kDa 59Nuc accounts for approximately 65% of the total ATP/ADP-hydrolyzing activity and the bulk of the anti-platelet aggregation/disaggregating activity in tsetse saliva and have uncovered its fibrinogen receptor antagonistic potential. These data clearly illustrate the important anti-hemostatic role of 59Nuc in the tsetse feeding biology and elucidation of its exact mode of binding to the fibrinogen receptor might be of significant biomedical interest.

Ethics statement
Blood sampling from 3 human volunteers was performed after obtaining a written informed consent. These individuals are the authors and co-author of this manuscript (GC, JVDA and KDR) that have been informed on the purpose of the sample collection that was approved by the Ethics Committee of the Institute of Tropical Medicine (ITM), Antwerp (Belgium).
Animal ethics approval for the tsetse fly feeding on mice and rabbits was also obtained from the Institutional Review Board of ITM.

Tsetse flies, tissue collection and saliva harvesting
Tsetse flies (Glossina m. morsitans, ITMA colony) were available from the insectaria at the Institute of Tropical Medicine Antwerp. Adult male flies were starved for 72 hours prior to dissection (at day 15 after emergence) of three different tissues: (i) the entire midgut (posterior + anterior), (ii) the proventriculus and foregut and (iii) the salivary glands. The tissues of 30 individuals were pooled into microtubes containing 60 ml sterile H 2 O and stored at 220uC until analysis. Saliva was harvested from the salivary gland tissue as described earlier [43], protein concentrations were determined by the BCA kit (Pierce Biotechnology) and aliquots stored at 220uC.

SDS-PAGE electrophoresis and protein electrotransfer
Tissue samples were thawed, homogenised with a Teflon pestle and run under reducing and denaturing conditions on 10% SDS PAGE. Proteins were stained with Coomassie Brilliant blue or by silversalts, or electrotransferred to nitrocellulose (Hybond C, Amersham) or PVDF membranes. Some protein bands were subjected to densitometric analyses as described earlier [44]. Detection of glycosyl modifications was achieved with the Gelcode glycoprotein staining kit (Pierce) and mobility shift assays were performed using native and PNGase F treated saliva.

N-terminal amino acid sequencing
The N-terminal amino acid sequences of selected salivary protein bands were determined by automated Edman degradation in an ABI 471-B sequencer, operated as recommended by the manufacturer.

cDNA library screening and clone isolation
A 385 base pair fragment, corresponding with the 59 part of the 59nucleotidase cDNA, was amplified using two primers that target respectively the 59 untranslated region (sense primer: 59-CCTAAATCCTTTCTTTAAC-39) and the region encoding the amino-terminal part of the protein (antisense primer: 59-CACGTGCCAGACCACCGATGC-39). To be used as a DNA probe for screening of the lgt11 salivary gland cDNA library, the PCR product was purified and labelled by random priming with dUTP-digoxigenin using the DIG High Prime DNA Labeling and Detection Starter Kit II (Roche). Approximately 20000 plaques were lifted with Nylon membrane (Roche) and hybridized with this specific probe. Five positive plaques were picked, confirmed by PCR and screened an additional two rounds with the same probe. Isolated positive plaques were amplified in the E.coli Y-1090 strain and phage DNA was purified using the Wizard Lambda Preps DNA Purification System (Promega).

cDNA sequence analysis and protein structure prediction
The obtained full-length cDNA was excised from the lgt11 arms with the Sfi and NotI restiction enzymes (Roche), ligated into pGEM-13Zf(+) vector (Promega) and transformed into Top 10 E. coli cells. Clones were sequenced using T7 and Sp6 primers as well as custom primers constructed from the internal sequence of the cDNA clone. Translated sequences were aligned using the CLUSTALW program and imported into GeneDoc (www.psc. edu/biomed/genedoc). Signal peptide prediction was based on the SPScan program (GCG software package). Homology detection and structure prediction was based on comparison of Hidden Markov Models (HHpred, http://toolkit.tuebingen.mpg.de/ hhpred). Generated PDB-files were used to produce 3Dfigures using Deepview/Swiss-PdbViewer (http://www.expasy. org/spdbv/).

59Nuc gene identification
Using G. m. morsitans salivary gland genomic DNA as a template, a PCR was performed using a sense primer in the 59UTR (59-CCTAAATCCTTTCTTTAAC-39) and a 39end primer (59-ATGAATAATCGTAATGCG-39) that span the entire coding sequence. The .4 kb amplification product was sequenced by using an additional number of internal primers. A universal GenomeWalker Kit (Clontech) was used to determine the 59 and 39 flanking sequences. A putative promotor sequence and the splicing sites were identified using software analysis tools (Neural Network Promotor prediction and Splice Site Prediction by Neural Network) that are available at the Berkeley Drosophila Genome project website (http://www.fruitfly.org/). The predicted splicing sites were confirmed by comparison with the cDNA sequence.

Recombinant 59nucleotidase purification, refolding and antibody production
The SUMO-tagged 59Nuc was purified on Ni-NTA (Sigma), eluted in 6 M urea with 0.5 M imidazole and refolded in pH conditions ranging from 4.0 to 10.0 [sodiumacetate (pH 4.0 to 5.4), MES (pH 5.6 to 6.6), HEPES (pH 6.8 to 7.8), Tris (pH 8.0 to 9.0) and piperazin (pH 9.0 to 10.0)] in the presence and absence of additives such as 0.5 M arginine, 1 mM reduced and 0.1 mM oxidized glutathion and 10 mM b-mercaptoethanol. Refolding efficiency was determined by quantifying precipitation (O.D. 405 nM) and ATPase activity. For the biochemical characterisation experiments, recombinant 59Nuc was refolded in 25 mM piperazin pH 10.0 supplemented with 250 mM NaCl, 5 mM KCl, 2 mM CaCl 2 and 2 mM MgCl 2 . Refolding was obtained by a fast injection and 1/20 dilution of the 59Nuc Ni-NTA elutes (at approximately 1 mg/ml) into the refolding buffer and incubation for 3 to 7 days at 4uC on a rocking platform at 70 rpm. Next, the 59Nuc protein was concentrated using centrifuge concentrators with polyethersulfone membranes and a 50 kDa molecular weight cut-off. Concentrates were run on a Superdex 200 gelfiltration column connected to an Ä kta explorer (GE Healtcare), allowing change of buffer to 1 mM piperazin pH 9.5, 100 mM NaCl and separation of active monomeric 59Nuc from aggregates and impurities.
59Nuc was also expressed with a T7-tag, affinity purified on a T7-tag antibody agarose column (Novagen) and used for raising rabbit polyclonal antiserum for Western blot analyses.

Southern, Northern and Western blot analyses and RT-(q)PCR
A P 32 labelled probe was prepared by PCR, using primers encompassing the entire 59nucleotidase coding sequence (1668 bp). The same probe was used for both Southern and Northern blot analysis. Northern blot was performed on salivary gland RNA from adult male flies. Southern blot analysis was performed on genomic DNA, from G. m. morsitans mixed males and females, digested by a range of restriction enzymes (PstI, XhoI, BamHI, BanII, SacI, EcoRI and HindIII). Western blot was used to assess expression in different tissues using the raised rabbit anti-59Nuc polyclonal serum (1/1000) and a peroxidase-conjugated anti-rabbit IgG (1/1000, Sigma). Conventional and quantitative RT-PCR analyses were performed using following primers: 59Nuc sense (59-CGGGTAATAAAGTTCTGGTCGTA-39), 59Nuc antisense (59-TTGGCAAGTCCACATTTGTTCTC-39) and primers that were used in a previous study [44]. The PCR cycles (35 for the conventional PCR, 45 for the qPCR) consisted of 1 min. denaturation at 94uC, 45 s. annealing at 54uC and 1 min. extension at 72uC. Gene expression was normalized using tubulin, actin and TAg5.

Nucleotidase activity measurements by detection of P i release
Zymographic detection of ATPase activity in total saliva was performed by separating the proteins at 4uC in native conditions on an 8% polyacrylamide gel using TAE (40 mM Tris, 5 mM sodium acetate, 1 mM EDTA) as running buffer and applying a voltage of 100V for 3 hours. ATPase activity was revealed by zymography as described elsewhere [10].
For pH dependence analyses, reactions were performed in 25 mM buffers (with 100 mM NaCl) ranging from 3.0 till 12.0 with pH differences of 0.2. The used buffers were sodiumacetate (pH 3.0 to 5.4), MES (pH 5.6 to 6.6), HEPES (pH 6.8 to 7.8), Tris (pH 8.0 to 9.0) and piperazin (pH 9.0 to 12.0). Reactions were performed in the presence of 1 mM of CaCl 2 and MgCl 2 . To determine the cofactor preference, activity was monitored in the presence of a range of divalent ions (Ca 2+ , Mg 2+ , Co 2+ , Cd 2+ , Cu 2+ , Zn 2+ and Ni 2+ ) at 1 mM final concentration. Temperature optima were determined between 20u and 60uC. For kinetic analyses, nucleotide concentrations ranged from 1000 to 0.78 mM.

Human platelet aggregation and disaggregation assays
Venous blood was obtained from healthy volunteers using Monovette coagulation tubes (Sarstedt) resulting in an anticoagulation with 10.6 mM citrate. Platelet-rich plasma (PRP) was obtained as the upper layer after 15 min centrifugation at 1506g at 22uC. Platelet-poor plasma (PPP), which served as the 100% transmission baseline, was prepared by pelleting the platelets at 15006g for 15 min. Aggregation of platelets in the presence or absence of total tsetse saliva (0.625-10 mg/ml) and the individual 59Nuc recombinant protein (10-20 mg/ml) in response to 5 and 10 mM ADP was evaluated at 37uC in a microtiter plate reader (MultiScan Ascent, Thermo). Reduction in optical density (increase in transmission) at 650 nm wavelength was monitored at 15 second intervals, with 60 rpm shaking between each reading, as a measure for platelet aggregation. For disaggregation experiments, human platelets were maximally aggregated with a 10 mM ADP trigger, followed by the addition of saliva (1.25-2.5 mg/ml) or recombinant 59Nuc/SUMO (10-20 mg/ml) and O.D. measurement at 650 nm.

In vivo RNA interference (RNAi)
For the in vivo functional analysis of the 59nucleotidase, the RNA interference (RNAi) method was applied as described previously [44]. To generate the template for 59Nuc-specific dsRNA (length: 496 bp) production by in vitro transcription (IT), a plasmid containing the full length 59Nuc coding sequence (pET17b:59Nuc) was used as PCR template in combination with following primers: 59Nuc(IT) sense (59-TAATACGACTCACTATAGGGGCAGA-CAGCTTGTACGACCA-39) and 59Nuc(IT) antisense (59-TAA-TACGACTCACTATAGGGTCATGAATTCGATCACGGAA-39). A control IT template was derived from pBlueScript SK(+) for the production of a control dsRNA as described earlier [44]. Purified IT templates were transcribed using the Megascript RNAi kit (Ambion), following the manufacturer's instructions. DsRNA was further purified as described previously and stored at 220uC until fly injection.
Tsetse flies, 48 h after the last blood meal, were briefly anaesthetised by cold shock and micro-injected intrathoracally with a single dose of 15mg dsRNA. For the evaluation of the transcriptional and translational silencing efficiency, 10 pairs of glands were isolated at eight and twelve days after dsRNA injection and saliva and RNA was purified followed by SDS-PAGE and RT-qPCR. At day 12 after dsRNA injection, feeding efficiencies (blood meal weights and feeding times) of individual flies on F1 (C57Bl/66Balb/c) mice were monitored within a maximum time of 10 minutes as described previously [43].