Six Anopheles species (An. gambiae, An. arabiensis, An. stephensi, An. funestus, An. albimanus and An. darlingi) were selected based on their significance as major vectors of human malaria in different parts of the world. Despite the availability of an An. albimanus transcriptome, protein sequences specifically matching secreted salivary gland proteins could not be identified in the whole body dataset 58 . This results from the hybrid nature of the transcriptome data from different tissues of adult female An. albimanus into a single transcriptome dataset, from which salivary gland-specific transcriptomic data have been excluded due to their low representation compared to other tissues. However, the sialome of four of the species (An. gambiae, An. stephensi, An. funestus and An. darlingi) has been characterised by high-throughput sialotranscriptomic studies 47 49 52 57 59 . These six Anopheles species were gathered in phylogenetically meaningful groups by analysing the degree of similarity of their cytochrome oxidase subunit II (COII) protein sequences (Figure 1A, 1B) 60 61 . Alignment of the six protein sequences shows 100% identity between An. gambiae and An. arabiensis (Figure 1B). The orthologous COII protein sequences from the others Anopheles species are more divergent with 97.1%, 95.2%, 97% and 96.6% identity for An. funestus, An. stephensi, An. albimanus and An. darlingi, respectively. A bootstrap consensus tree inferred from 10,000 replicates showed that the six Anopheles species are divided into the following two major groups based on their taxonomic classification: (i) a clade formed by An. gambiae, An. arabiensis, An. funestus and An. stephensi, which are all members of the Cellia subgenus; and (ii) a clade including An. albimanus and An. darlingi, which are two neotropical species belonging to the Nyssorhynchus subgenus. The large Cellia clade encompasses An. gambiae and An. arabiensis, which are two morphologically indistinguishable sibling species from the An. gambiae s.l. species complex. This phylogenetic analysis also indicated larger genetic distance between the An. gambiae s.l. (subgenus Cellia, Pyretophorus Series), An. funestus (subgenus Cellia Myzomyia Series) and An. stephensi (subgenus Cellia, Neocellia Series) species.

A total of 401 salivary gland protein sequences from the six Anopheles species were retrieved from public databases according to their annotation. Focusing on proteins potentially injected into the human host during mosquito blood feeding, protein sequences were sorted based on signal peptide predictions 64 65 . A total of 272 out of these 401 salivary proteins, heterogeneously distributed among the six Anopheles species (i.e., 71, 5, 44, 5, 117 and 30 protein sequences for An. gambiae, An. arabiensis, An. stephensi, An. albimanus, An. darlingi and An. funestus, respectively), were predicted to harbor a secretory-signal peptide following submission to SignalP server 3.0 and were thus retained for further analysis. A pairwise protein-protein sequence comparison was first performed on these selected SGPs using “BLAST 2 Sequences” 63 to find regions of local similarity between sequences from different Anopheles taxa. Briefly, secreted salivary protein sequences from each Anopheles species were gathered into six different databases in accordance with their species affiliation. Each protein from a database was then searched against all proteins from other database in a pairwise fashion. BLAST E-values were used as a parsing criterion in order to select best matches between two different salivary protein repertoires. In order to perform comparative analysis between salivary proteins repertoires, Raw BLAST score obtained from the match of a query protein sequence with a targeted protein sequence were divided against raw self-BLAST score from the match of the query protein sequence to itself to obtain normalised BLAST scores (or BLAST Score Ratio). Normalised BLAST scores range from 0 (no BLAST match) to 1 (perfect BLAST match between two salivary proteins) 66 67 . The average normalised BLAST scores and average percentage identity between local alignments that estimate the degree of homology between salivary protein sequences from each pair of Anopheles species are presented in Figure 1C and Additional file 1. Only protein sequences from each Anopheles species matching at least one other salivary protein in another species at 40% identity (q.v. Additional file 2) were considered in this analysis, representing 50, 5, 33, 5, 53 and 27 protein sequences for An. gambiae, An. arabiensis, An. stephensi, An. albimanus, An. darlingi and An. funestus, respectively. The average normalised BLAST score between salivary protein sequences from An. gambiae and An. arabiensis was 0.98 ± 0.01 (mean ± SD). Based on the same criteria, salivary protein sequences from both species from the Nyssorhynchus subgenus had normalised BLAST scores of 0.75 ± 0.17 and sequences from species belonging to the Cellia subgenus had normalised BLAST scores larger than 0.45. Lower normalised BLAST scores were observed when comparing salivary protein sequences from species belonging to Cellia with those of the Nyssorhynchus subgenera (all normalised BLAST scores were inferior to 0.40) (Figure 1C, Additional file 1). Thus, secreted salivary protein sequence similarities were the highest among closely related anopheline species and decreased with increasing phylogenetic distance (Figure 1A, B).

Hierarchical clustering of the salivary protein sequences was performed to determine paralogous (i.e., homologous intra-species protein derived from a gene duplication event) and orthologous (i.e., homologous inter-species protein derived from a speciation event) salivary proteins and their degrees of similarity among the six Anopheles species. Three clustering steps using the CD-HIT program 68 were sequentially performed at different similarity thresholds based on full-length sequences (≥ 90%, ≥ 70% and ≥ 40% identity). This agglomerative hierarchical clustering approach was used to maximise the quality of clustering 69 and produce a tree-like structure (Figure 2) to assess the level of homology among the proteins. Among the 272 secreted salivary proteins that were retrieved, the first clustering step (≥ 90% identity threshold) led to the determination of 162 (60%) non-redundant (NR) protein sequences (i.e., sequences that did not cluster with other sequences over a specified identity threshold) and 44 clusters composed of at least two protein sequences (mean number of proteins per cluster ± 95% confidence interval (CI) of the mean, 2.63 ± 0.36). These 44 clusters were almost exclusively composed of paralogous sequences with the exception of five An. arabiensis protein sequences, which all clustered with An. gambiae sequences (clusters 3, 4, 5, 21, 39) and cluster 43 composed of An. stephensi and An. funestus 6.3 kDa proteins (Figure 2, Additional file 2). The second clustering steps (≥ 70% identity threshold) identified 130 (48%) NR protein sequences and 49 clusters (2.94 ± 0.46). Among these 49 clusters, 19 consisted of orthologous protein sequences. The vast majority of salivary protein sequences from species belonging to the Nyssorhynchus subgenus did not cluster with those from the Cellia subgenus at this step (Figure 2A and Additional file 2). The last clustering steps (≥ 40% identity threshold) resulted in 73 (27%) NR protein sequences and 46 clusters (4.37 ± 1). Among these clusters, 36 were composed of orthologous protein sequences, half of which (18 out of 36) consisted of sequences belonging to both Nyssorhynchus and Cellia subgenera. These orthologous sequences belong to several protein families, including the apyrase/5’nucleotidase, antigen 5/gvag, GE-rich/30 kDa, long and short form D7, mucin/13.5 kDa, SG3, SG7, SG10 or hypothetical 6.2 kDa protein families. Among the NR sequences, 12 out of 19 (63%), 18 out of 38 (47%), 5 out of 11 (45%) and 1 out of 3 (33%) have no homologs in sialomes from other blood feeding arthropods, concerning An. gambiae, An. darlingi, An. stephensi and An. funestus, respectively. A majority of these species-specific secreted salivary proteins have low molecular weights (Additional file 2).

The salivary protein sequence repertoires available from public databases for both An. arabiensis and An. albimanus species are largely incomplete, and the vast majority of Anopheles protein sequences are inferred from transcriptomic or genomic sequence analyses. Thus, a proteomic analysis was performed to confirm the existence of predicted secreted proteins and evaluate the sequence diversity observed by in silico analysis at the protein level. Access to SGEs could only be achieved for four of the six Anopheles species by dissecting wild mosquitoes (An. arabiensis) or mosquitoes reared in laboratories (An. gambiae, An. stephensi and An. albimanus). Nonetheless, these four selected species are a representative sample of the Anopheles taxonomic diversity at the subgenus, species complex and species levels. SGPs of the four Anopheles species were separated by SDS-PAGE (Figure 3A). Despite slight quantitative variations (i.e., band intensities), superimposition of densitometric protein profiles indicated a high similarity between the An. gambiae and An. arabiensis species belonging to the An. gambiae s.l. species complex (Figure 3B). Conversely, SGP profiles of An. stephensi (Cellia sb.) and An. albimanus (Nyssorhynchus sb.) differed and were highly distinct from An. gambiae s.l. profiles at the qualitative (i.e., molecular weights of the bands) and quantitative (i.e., band intensities) levels.

To improve estimates of protein diversity, SGP repertoires from these four mosquito species were identified as previously described 54 70 . Briefly, each gel loading track was cut into several segments covering the entire protein profile and proteins from gel pieces were identified by MS/MS. As scarce protein sequences are available in protein database for some of the Anopheles species under study, the MS/MS spectra were searched against a non-redundant protein database including protein sequences from An. gambiae, An. arabiensis, An. stephensi and An. albimanus together. This strategy was implemented to identify homologous proteins among Anopheles species based on their peptides similarities. A total of 41, 44, 40 and 16 proteins were identified in SGEs from An. gambiae, An. arabiensis, An. stephensi and An. albimanus, respectively, representing a total of 77 unique proteins for all species (Additional file 3). Among these 77 unique proteins, 25, 27, 21 and 9 were identified as putative secreted proteins in the SGEs of An. gambiae, An. arabiensis, An. stephensi and An. albimanus, respectively, totalling 42 unique putative secreted proteins for all species (Additional file 4). Only 26, 2, 11 and 3 salivary proteins were identified in An. gambiae, An. arabiensis, An. stephensi and An. albimanus, respectively. The majority of these proteins (18/42) were identified in the An. gambiae strain PEST by homology. Protein members from the apyrase and 30 kDa/GE-rich/anti-platelet families were identified in all Anopheles species. A total of 18 secreted proteins were only identified in unique Anopheles species. The number of common proteins among the Anopheles species was highest among closely related anopheline species (i.e., 73% among An. gambiae and An. arabiensis) and decreased with increasing phylogenetic distance (i.e., 34% of proteins were identified in common among species from the Cellia subgenus and 12% of proteins were common among the four Anopheles species (Figure 4).

Protein sequence diversity observed in in silico and proteomic analyses among the four Anopheles species was also tested at the antigenic level. Using a pool of sera from 5 Senegalese individuals exposed mainly to An. gambiae s.l. and An. funestus 71 , a comparison of IgG antibody responses against SGEs from An. gambiae, An. arabiensis, An. stephensi and An. albimanus species was performed by immunoblot analysis. Several antigenic bands were observed in SGEs of all anopheline species, with a total of 7, 10, 6 and 3 antigenic bands detected in An. gambiae, An. arabiensis, An. stephensi and An. albimanus, respectively (Figure 5A). The pooled sera exhibited high reactivity with three antigenic bands at 40, 35 and 11 kDa in SGEs from An. gambiae s.l., and all other antigenic bands detected in SGEs from An. gambiae were also found in those from An. arabiensis. However, 3 antigenic bands at 26, 24 and 14 kDa were exclusively observed in An. arabiensis SGEs (Figure 5A and 5B). Three antigenic bands with molecular weights of 82, 30 and 11 kDa were recognised in SGEs from species belonging to the Cellia subgenus (i.e., An. stephensi and An. gambiae s.l.). However, the antigenic band with low molecular weight (11 kDa) was 7.5-fold less intense in An. stephensi compared to those of the same molecular weight in An. gambiae s.l. With the same molecular weight criteria, two antigenic bands (45 and 37 kDa) were only detected in the An. stephensi SGE antigenic profile. In An. albimanus SGEs, three antigenic bands were detected with molecular weights of 54, 47 and 34 kDa. Among them, the more intense antigenic band (34 kDa) was 2.8-fold less intense than its 35 kDa counterpart detected in An. gambiae s.l. SGEs. Interestingly, the 47 kDa band was uniquely observed in An. albimanus, and the 54 kDa antigenic band was faint but detected in all Anopheles antigenic profiles. All protein bands numbered in Figure 5C, corresponding to antigenic bands (Figure 5A and 5B), were submitted to MS analysis for identification. With the exception of protein bands numbered 1 (82 kDa, An. gambiae), 6 (30 kDa, An. gambiae), 8 (82 kDa, An. arabiensis), 13 (30 kDa, An. arabiensis), 14 (26 kDa, An. arabiensis), 15 (24 kDa, An. arabiensis),and 24 (54 kDa, An. albimanus) at least one protein was identified in all excised protein bands, resulting in the identification of 45 distinct proteins (17 housekeeping and 28 secreted proteins) according to their NCBI accession numbers (Additional file 4 and 5). As expected, several proteins could be identified in each excised band, and the same protein could also be identified in distinct excised bands from the same species as previously described 70 . Some proteins, such as the salivary apyrase [NCBI: gi|27372911] and anophensin [NCBI: gi|148189823] were identified in antigenic bands from all species belonging to the Cellia subgenus, including the non-African An. stephensi mosquito (Additional file 5). Orthologous proteins to An. gambiae s.l. antigens were identified in antigenic bands from the SGEs of An. albimanus, another non-African Anopheles mosquito (Figure 6). Interestingly, a GE-rich salivary gland protein [NCBI: gi|29501380] and a salivary gland protein [NCBI: gi|71389019] identified in An. stephensi and An. albimanus, respectively, shared 57.2% and 54.7% amino acid sequence identity with the anti-platelet protein [NCBI: gi|190576759] identified in the 30 kDa antigenic band from the An. gambiae s.l. complex (Additional file 6). These results highlight a potential link between protein sequence homology and the presence of cross-reactivity.

Vector species of the Anopheles genus throughout the world have different levels of phylogenetic relatedness 3 . Variation in the mitochondrial cytochrome oxidase subunit 2 (COII) sequence has been widely used to display the phylogenetic relationships and population structure of anopheline mosquitoes 73 74 . Molecular phylogenetic analysis using COII protein sequences from six selected malaria vectors (An. gambiae, An. arabiensis, An. stephensi, An. funestus, An. albimanus and An. darlingi) indicated that genetic distances between species were in agreement with their taxonomic classification. In addition, taking into account all of the secreted salivary proteins available from public databases for these six mosquito species, a pairwise protein-protein sequence analysis demonstrated that the proportion of salivary protein sequence homology decreases according to increasing phylogenetic distance between Anopheles species. Similar observations were reported for salivary proteins from phlebotomine sandflies 75 , underlining this diversity of salivary secreted proteins inside a hematophagous arthropod genus. Comparative analyses of sialotranscriptomes between Anopheles species highlighted that secreted salivary proteins are less conserved than housekeeping proteins 48 49 52 57 . Heterogeneous secreted salivary protein repertoires among Anopheles species are consistent with the existence of secreted salivary proteins occurring throughout the Anopheles genus and others limited to the subgenus or species level.

Despite the low number of sequences available for some Anopheles species, several salivary protein families were found in all Anopheles species tested using agglomerative hierarchical clustering. These protein families included the apyrase/5’nucleotidase, antigen 5/gvag, GE-rich/30 kDa, long and short form D7, mucin/13.5 kDa, SG3, SG7, SG10 or hypothetical 6.2 kDa protein families. At the protein level, the clustering of salivary secreted proteins identified by MS confirmed that some proteins, including proteins from the apyrase/5’nucleotidase and 30 kDa/GE-rich/anti-platelet families, are present in all the investigated Anopheles species. Several of these protein families were also found in the saliva of other hematophagous arthropods, which could result from convergent evolution or a common ancestry 32 76 . For instance, members of the apyrase/5’nucleotidase and antigen 5/gvag protein families are found in the salivary glands of diverse hematophagous insect and tick species across the Arthropoda phylum. Interestingly, members from the GE-rich/30 kDa protein family are exclusively found in salivary glands of both culicine and anopheline female mosquitoes 49 57 77 . Here, other protein families appeared to be exclusively found in saliva from anopheline mosquitoes, such as SG3, SG7 or hypothetical 6.2 kDa proteins, offering the opportunity to use these proteins as genus-specific immunological markers. However, cross-reactivity is likely to occur when antigenic proteins share more than 70% amino acid identity 78 . Most of these orthologous protein sequences belonging to the Anopheles genus mostly shared less than 70% identity, minimising the probability of characterising conserved epitopes inside the Anopheles genus 43 79 .

Although an antigenic band was commonly detected at 54 kDa in all Anopheles species, MS did not identify common or orthologous secreted protein in these antigenic bands that could explain the observed cross-reactivity. The low protein abundance in the corresponding gel bands and incomplete molecular sequencing data for some of the Anopheles species could explain this lack of identification. However, members of the GE-rich/30 kDa/anti-platelet family were identified in antigenic bands from the four Anopheles species, and are thus potential candidates to serve as pan-Anopheles genus markers of immunological exposure. Complementary experiments are required to evaluate the lack of cross-reactivity against salivary proteins from other mosquito species from the Culicidae family. Interestingly, in cases of human allergic reactions involving the production of IgE antibodies in response to mosquito bites, some salivary allergens from Aedes aegypti mosquito species have been characterized 34 80 . Among Aedes aegypti salivary allergens tested, recombinant forms of the 68 kDa salivary apyrase, the 37-kDa protein belonging to the D7 family and the 30 kDa salivary gland allergen were demonstrated to elicit an IgE responses in mosquito-allergic individuals. These data underline the antigenicity of some salivary proteins including 30 kDa family members, and the opportunity to use such proteins for the diagnosis and the desensitisation of mosquito allergic individuals.

Combinations of subgenus-specific immunological markers may be an alternative for assessing exposure to Anopheles bites. Moreover, such markers could be helpful in identifying the anopheline subgenus involved in malaria transmission in places where several vector subgenera with different behaviour and vectorial capacity are encountered within the same area, notably in Asia 81 82 . In silico analysis of secreted salivary protein sequences from the six Anopheles species tested in the present study indicated that secreted salivary proteins sharing between 70% and 90% amino acid sequence identity are mainly clustered according to their subgenus affiliation. Thus, several salivary secreted proteins, such as TRIO, gSG2, gSG6 proteins and numerous hypothetical proteins with low molecular weights (Hypothetical 4.2/13, 8.2, 10, 12, or 15/17 kDa proteins), clustered only in the Cellia subgenus. The same observation was made for the Nyssorhynchus subgenus (e.g., salivary peroxidase). Some of the secreted salivary proteins identified by MS were uniquely detected in the Anopheles species belonging to the Cellia subgenus, such as salivary apyrase, anophensin, D7 proteins or TRIO proteins. In silico analysis revealed that members of the gSG2 and gSG6 protein families are well conserved inside the Cellia subgenus with at least 67% and 77% identity, respectively, among An. gambiae, An. funestus and An. stephensi. Among these two protein families, An. gambiae gSG6, which is a reliable marker for exposure to An. gambiae 83 84 85 bites, was recently reported to be a good indicator for exposure to bites from three main African malaria vectors from Cellia sb. (i.e., An. gambiae, An. arabiensis and An. funestus). This cross-recognition could result from shared epitopes among these orthologous SG6 proteins 38 86 . According to their low molecular weights (approximately 11.7 and 13 kDa, respectively), gSG2 and gSG6 proteins should be contained in the large 11 kDa antigenic band detected in the Cellia subgenus. However, MS did not identify these last two proteins. Description of the salivary gland protein repertoire from An. gambiae performed by Kalume and colleagues identified gSG6 proteins only using a gel-free approach 87 . Thus, unsuccessful identification of these small proteins could be attributed to the combination of a high number of salivary proteins of this molecular weight and a low number of MS spectra generated by these small proteins, rendering this complex protein mixture unidentifiable by MS. The application of gel-free proteomic methods might increase salivary proteome coverage, especially concerning secreted protein with low molecular weights 88 . As orthologous secreted salivary proteins belonging to the same subgenus clustered only at important sequence identity levels, shared-epitopes are likely to occur, increasing the likelihood of observing cross-reactivity among these subgenus-specific proteins. Thus, identification of apyrase, anophensin and TRIO orthologs in antigenic bands from An. gambiae s.l. and An. stephensi SGEs highlight the potential of these salivary proteins to be Cellia subgenus-specific immunological markers.

The Anopheles subgenera encompass several groups of closely related species that are morphologically indistinguishable, such as the An. gambiae s.l. species complex, which includes at least 6 species 89 90 91 . In silico analysis revealed a high degree of homology (> 90%) among salivary protein sequences from each selected mosquito of this complex (i.e., An. gambiae and An. arabiensis). Moreover, protein profiles and protein repertoires were highly similar between these two closely related Anopheles species, although one was collected in the field (i.e., An. arabiensis) and the other came from continuous laboratory rearing (i.e., An. gambiae). These results point to the likelihood that salivary protein candidates from either of these species would be able to assess exposure to Anopheles mosquitoes pertaining to the An. gambiae s.l. species complex. Using pooled sera from individuals mainly exposed to An. gambiae, An. arabiensis and An. funestus 71 , most intense antigenic bands (40, 35 and 11 kDa) were revealed in SGEs from both species, suggesting that strong antibody responses against An. gambiae s.l. SGEs are elicited following exposure to these mosquitoes. Indeed, in a recent study conducted in the South of France, a positive association between mosquito exposure and the level of antibody response was reported and this immunological response appeared to be species-specific 92 . The major antigenic bands observed in An. stephensi and An. albimanus SGEs were about 4-fold less intense than the most intense antigenic bands detected in An. gambiae s.l. SGEs. Potential common antigenic salivary proteins were identified in major antigenic bands from An. gambiae and An. arabiensis salivary gland protein profiles, including anti-platelets, anophensin and proteins from the D7 and SG1 families. Even if members of these protein families are present throughout the Cellia subgenus, low amino acid sequence identities can occur between An. gambiae s.l. and other species, which could drastically affect antibody binding. Cross-reactivity usually implies a lower affinity for the cross-reactive antigen compared to the primary antigen 43 . The development of better quantitative methods, such as ELISA or Luminex®, combined with the production of salivary antigenic protein candidates may provide more distinct antibody responses to specific mosquito bites from the cross-reactivity attributed to partial shared-antigens.

It would be interesting to use immunological tools to assess individual exposure to a specific Anopheles species, even to closely related phylogenetically species, in areas harbouring several Anopheles species with distinct vector competences and behaviours. For instance, the An. gambiae s.l. complex encompasses species that often occur in sympatry but profoundly differ in their ability to transmit malaria parasites, in host feeding preferences, larval habitat requirements or responses to select vector control measures 90 93 . Thus, characterisation of Anopheles species-specific immunological markers may help to implement adapted vector-control strategies and assess their efficiency to prevent host/vector contact at the individual level. Additionally, the combination of different species-specific immunological markers using multiplex techniques, such as Luminex® 94 , may increase sensitivity 38 95 and specificity of the test. Such multiplexing of validated antigenic salivary proteins will distinguish exposure to the bite of malaria vectors from that of non-vector Anopheles. Species-specific immunological markers may provide a more detailed view of the history of exposure at an individual level in retrospective studies. Thus, these markers may be useful to determine the implication of different vector species in malaria epidemics. In silico analysis indicated that 73 (27%) salivary proteins have no orthologous proteins in other Anopheles species at the lowest amino acid sequence identity threshold tested (40% identity). In addition, MS only identified 18 secreted in unique Anopheles species. Some of these species-specific salivary proteins, including mainly proteins with low molecular weights, are not found in SGEs from other hematophagous arthropod species (Additional file 2). Thus, these can serve as potential immunological markers for the assessment of individual exposure to specific Anopheles species.

Although in silico analysis, protein patterns and protein repertoires indicated a low diversity of secreted salivary proteins between An. gambiae and An. arabiensis, the detection of three antigenic bands exclusively in An. arabiensis SGEs suggest that closely related species could be distinct at the antigenic level. Unfortunately, MS analysis could not successfully identify the corresponding proteins from the gel bands. To identify these antigenic bands, a two-dimensional immunoproteomic approach using a fluorescence-based method could be an alternative 96 . Although the challenge appears more important for closely related mosquito species, our data suggest that identification of species-specific immunological markers seems reasonably conceivable. Moreover, this hypothesis is supported by our recent work on the existence of species-specific serological responses against Ae. caspius SGEs 92 .

Sera from 5 individuals living in the Senegalese village of Dielmo (13°45’N, 16°25’W), sampled in March 1995 when malaria was holoendemic were used in this study 94 . These individuals were exposed to high levels of malaria transmission (about 200 infective bites/person/per year) and mosquito bites (human biting rate about 23.2), with An. gambiae (11%), An. arabiensis (56%) and An. funestus (33%) as principal vectors 71 . These individuals did not travel outside Senegal country in the twelve months prior to blood sampling. The protocol was approved by the Senegal National Ethics Committee (Dakar, Senegal). The informed consent of each participant was obtained at the beginning of the study, after a thorough explanation of its purpose.

The cytochrome oxidase II (COII) protein sequences from An. gambiae, An. arabiensis, An. stephensi, An. albimanus, An. funestus, An. darlingi and Aedes aegypti were retrieved from NCBInr database (May 10^th, 2011) and multiple sequence alignment was performed with Clustal W 1.7 multiple sequence alignment program 97 which is included in Molecular Evolutionary genetic Analysis 5 (MEGA 5) programs package 62 . The evolutionary history was inferred using the Neighbour-Joining method in MEGA 5 98 . The Ae. aegypti sequence was taken as an outgroup. The bootstrap consensus tree inferred from 10,000 replicates 99 is taken to represent the evolutionary history of the taxa analyzed. Branches corresponding to partitions reproduced in less than 50% bootstrap replicates are collapsed. The percentage of replicate trees in which the associated taxa clustered together in the bootstrap test (10,000 replicates) are shown next to the branches 99 . The evolutionary distances were computed using the Poisson correction method and are in the units of the number of amino acid substitutions per site. All positions containing gaps and missing data were eliminated. Evolutionary analyses were conducted in MEGA5 62 . Protein sequence alignments were generated in Geneious Pro 5.6.4 http://www.geneious.com/.

Uninfected 5-day-old adult females of the An. gambiae s.s., An. arabiensis, An. stephensi and An. albimanus species were used in this study. An. stephensi and An. albimanus species were reared at the Institut Pasteur (CEntre de Production et Infection des Anopheles, CEPIA, Paris). An. gambiae was reared at the Institut de Recherche pour le Développement (laboratoire de Lutte contre les Insectes Nuisibles, Montpellier). An. arabiensis species was collected on the field at the larvae stage at Dakar (Senegal) at the end of the rainy season in September 2008 101 and identified by polymerase chain reaction (PCR) 102 . After emergence, adult mosquitoes were reared at the Institut de Recherche pour le Développement (UR24, Dakar). All mosquitoes used in the experiments were maintained under identical standard conditions: 26°C and 60% humidity, took no blood meals and were maintained on a diet of 10% syrup solution. The salivary glands from adult mosquito females were dissected under a stereomicroscope at 4X magnification as previously described 92 . The salivary glands from each experiment were pooled by strains into a microcentrifuge tube on ice in phosphate-buffered saline (PBS) and then stored frozen at −20°C until needed.

Salivary glands were disrupted by ultrasonication (Vibracell 72412, Bioblock Scientific, Illkirch, France) for 5 min on ice at maximum amplitude. Salivary gland homogenates were then centrifuged for 15 min at 16,100 ×g 103 and protein concentration of the supernatant was determined in duplicate by the Lowry method (DC Protein assay Kit, Bio-Rad) according to the manufacturer’s instructions. Salivary gland proteins were then concentrated by precipitation with acetone (Sigma, St Louis, MI), and were suspended in a UTC buffer containing 8M urea (Sigma), 2M thiourea (Sigma), 4% (w/v) CHAPS (Sigma) and 30 mM Tris (Sigma), adjusted to pH 8.5 in order to obtain a protein concentration adjusted to 2.5μg/μL.

For each species, 20 μg of salivary gland proteins were separated onto a 12% SDS-PAGE in a PROTEAN II xi or Mini (BioRad, Hercules, USA). A broad range molecular weight marker (BioRad) was loaded on each gel. Gels were stained with Sypro Ruby (BioRad) according to the manufacturer’s protocol and digitalized using the Typhoon^TM Trio Image scanner (GE Healthcare, UK). Salivary glands densitometry profiles were analyzed using the ImageQuant^TM TL software (GE Healthcare, UK). Background subtraction was performed and the densitometry profiles were normalized to take into account global differences 54 .

For each species, 15 μg of SGPs were separated onto a 12% SDS-PAGE as described above. SGPs were minimally labelled with CyDye 5 according to the manufacturer’s protocol (GE Healthcare) prior electrophoresis, as described elsewhere 96 . Gels were then transferred to a nitrocellulose membrane (0.45-μm, Amersham Pharmacia, Saclay, France) by semidry blotting (0.8 mA per cm²) 96 . Blots were saturated 1 hour at room temperature with 5% w/v non-fat dried milk, and were carried out with human sera diluted at 1/50 in phosphate buffer saline (PBS) containing 0.1% v/v tween-20 with 5% w/v non-fat dried milk. After an overnight incubation, blots were incubated with anti-human Fcγ/IgG FITC conjugated antibody 1/1000 (Sigma, St Louis, MI). Immunoblots were directly digitalized using a Typhoon^TM Trio Image scanner (GE Healthcare) and densitometric analysis of IgG immune profiles was performed using ImageQuant^TM TL software (GE Healthcare), as previously described 54 .

Each loading tracks were excised covering the totality of the each lane as previously described 54 68 , using Shimadzu Biotech Xcise System (Champs sur Marne, France). Salivary protein identification was made in duplicate on two distinct gels. Protein bands were digested overnight at 37°C with sequencing-grade trypsin (12.5 μg/mL; Promega Madison, WI, USA) in 50 mM NH₄HCO₃ (Sigma). The resulting peptides were extracted with 25 mM NH₄HCO₃ for 15 min, dehydrated with acetonitrile (ACN) (Sigma), incubated with 5% acid formic (Sigma) for 15 min under agitation, then dehydrated with ACN, and finally completely dried using a SpeedVac. Samples were then stored at −20°C before analysis by mass spectrometry (MS).

For MS analysis, a LCQ DecaXPplus (ThermoFinnigan, San Jose, CA) ion trap was used. Nano-liquid separation of peptides was carried out using an Ettan MDLC chromatographic system (GE Healthcare) in high throughput configuration. Ten microliters of the digest were first trapped on a zorbax 300SB-C18 5 × 0.3 mm column and eluted at a flow rate of approximately 200 nl/min on a zorbax 300SB-C18, 3.5 μm, 150 × 0.075 mm by a linear gradient of eluant B (0.1% Formic acid, 84% ACN) in eluant A (1% Formic acid). The chromatographic system was piloted by the Unicorn 5.01 software (GE Healthcare). MS measurements were done on a LCQTM Deca XP Plus ion trap mass spectrometer (ThermoFinnigan) equipped with a LCQTM nanospray ionization source. A spray voltage of 1.8 kV was applied to the liquid junction via an in-union high voltage contact coupled to a silicaTip emitter (New Objective). Operation of the mass spectrometer was fully automated during the entire procedure using the Excalibur 1.3 data system (ThermoFinnigan). Continuous cycles of one full scan (m/z 500 to 1700) followed by three data-dependent MS/MS measurements at 35% normalized collision energy were done. MS/MS measurements were allowed for the three most intense precursor ions with a maximum rejection time limit of 1 min.

All MS/MS spectra from one species were gathered and sequence database searched using the Bioworks 3.1 (ThermoFinnigan) or Mascot Daemon v.2.2.2 (Matrix Science, London, UK). The MS/MS spectra were searched against the non-redundant protein database (NCBInr, NIH, Bethesda) of An. gambiae [7165], An. arabiensis [7173], An. stephensi [30069] and An. albimanus [7167] together (July 27th, 2009, 16677 sequences). The following search parameters were used: precursor-ion mass tolerance of 0.8 Da, fragment ion tolerance of 0.8 Da with methionine oxidation and cysteine carboxyamidomethylation specified as differential modifications, and a maximum of one missed cleavage site allowed. Scaffold (version Scaffold_3.6.2, Proteome Software Inc., Portland, OR) was used to validate MS/MS based peptide and protein identifications. Peptide identifications were accepted if they could be established at greater than 95.0% probability as specified by the Peptide Prophet algorithm 104 . Protein identifications were accepted if they could be established at greater than 95.0% probability and contained at least 1 identified peptide. Protein probabilities were assigned by the Protein Prophet algorithm 105 .