High-Resolution Proteomics of Aedes Aegypti Salivary Glands Infected With DENV2, ZIKV and CHIKV Reveal Virus-Specific and Broad Antiviral Factors


 Arboviruses such as dengue (DENV), Zika (ZIKV) and chikungunya (CHIKV) viruses infect close to half a billion people per year, and are transmitted through Aedes aegypti bites. Infection-induced changes in mosquito salivary glands (SG) influence transmission by inducing immunity, which restricts virus replication, and by altering saliva composition, which influences skin infection. Here, we profiled SG responses to DENV2, ZIKV and CHIKV infections by using high-resolution quantitative proteomics. We identified 218 proteins related to immunity, blood-feeding or cellular machinery. We observed that 58, 27 and 29 proteins were regulated by DENV2, ZIKV and CHIKV infections, respectively. While the regulation patterns were mostly virus-specific, we determined the function of four uncharacterized proteins that were upregulated by all three viruses. We revealed the anti-ZIKV function of gamma-interferon responsive lysosomal thiol-like (GILT-like), the anti-CHIKV function of adenosine deaminase (ADA), the pro-ZIKV function of salivary gland surface protein 1 (SGS1) and the antiviral function against all three viruses of an uncharacterized protein we called salivary gland broad-spectrum antiviral protein (SGBAP). The comprehensive description of SG responses to three global pathogenic viruses and the identification of new restriction factors improves our understanding of the molecular mechanisms influencing transmission.


Introduction
Arboviruses like dengue (DENV), Zika (ZIKV) and chikungunya (CHIKV) viruses are primarily spread through the bites of Aedes aegypti mosquitoes 1 . DENV and ZIKV belong to the Flavivirus genus (Flaviviridae family), while CHIKV is an Alphavirus (Togaviridae family). These viruses infect more than 400 million people yearly, mostly in tropical and sub-tropical countries where the environment is conducive to natural mosquito breeding 2 . In the absence of effective vaccine and curative drug 3 , mitigation of these mosquito-borne diseases relies on vector population control. Insecticide-based controls often fail to curb outbreaks partly due to insecticide resistance 4,5 . Alternatively, novel biological strategies are being developed to limit vector populations. Population suppression or replacement by releasing Wolbachia-infected mosquitoes is currently being evaluated in eld trials in several countries 6 . However, the ability to scale up the suppression strategy and the sustainability of the replacement strategy 7 remain to be validated. The production of pathogen-refractory mosquitoes by genetic modi cation offers another promising vector-targeted strategy 8, 9 . In this regard, identi cation of pro-and anti-viral factors in mosquitoes is a prerequisite.
While biting an infected host, female mosquitoes ingest viruses that infect the mosquito midgut before disseminating throughout their whole body nally reaching the salivary glands (SG), from where the virus is secreted during a subsequent bite. The period the virus takes from entering the mosquito midgut to nally exiting through the SG into mosquito saliva is de ned as extrinsic incubation period (EIP). EIP varies among virus species. EIP for aviviruses like DENV and ZIKV is estimated between 10-14 days 10,11 and for alphaviruses like CHIKV between two to nine days 12 . Aedes aegypti displays pathogen-responsive innate immune pathways throughout the above path of midgut, hemolymph and SG, which limit virus infection, and thus transmission 13,14 . Transcriptomics studies of SG showed that the induction of Toll and IMD pathways by DENV serotype 2 (DENV2) leads to the production of a cecropin-like antimicrobial peptide with anti-DENV2, anti-CHIKV and anti-Leishmania properties 15 . Using highthroughput RNA sequencing, we recently showed that DENV2, ZIKV and CHIKV infections trigger a broad antiviral response through the c-jun N terminal kinase (JNK) pathway, which activates complement and apoptotic effectors 16 . Alternatively, other factors unrelated to the canonical immune pathways have been implicated in the regulation of DENV2 infection in SG 17 . While multiple evidence indicates that mosquito SG response modulates infection, additional studies are required to characterize the response at the proteomic level, identify novel viral factors and determine how infection-regulated proteins in uence infection.
SG infection also alters the composition of saliva, thereby in uencing blood acquisition and skin infection [18][19][20] . Mosquito saliva contains a cocktail of biologically active molecules with functions in hemostasis, in ammation and immunity 21,22 . Among others, A. aegypti saliva contains a vasodilatory tachykinin decapeptide named sialokinin, a factor Xa-directed anticoagulant and an anti-platelet apyrase, all of which may facilitate blood acquisition by preventing clotting to maintain steady blood ow [23][24][25] .
Immune-modulators such as a secreted 387 kDa protein can supress cytokine release and proliferation of T and B cells in mouse splenocytes in vitro 26 . A venom allergen-1 protein detected in A. aegypti saliva was recently found to enhance DENV2 and ZIKV infection in skin cells by augmenting autophagy 27 .
Alternatively, salivary proteins can also inhibit skin infection. A 30 kDa collagen-binding protein called aegyptin 28 and a D7 protein 29 reduce DENV2 multiplication. Characterization of SG proteomic response to infection will inform about changes in saliva composition, which can affect transmission.
The earliest studies of A. aegypti SG proteome used uninfected mosquitoes and one-or two-dimensional gel electrophoresis (DGE) coupled with mass spectrometry (MS) to identify a few proteins 30,31 . Recently, using high-resolution MS, 1,208 proteins were detected in uninfected A. aegypti SGs 32 , although these included proteins identi ed by only one unique peptide. To our knowledge, only two studies for DENV2, one for CHIKV and none for ZIKV reported SG proteomic response to infection with low resolution MS [33][34][35] . Here, to bridge the knowledge gap in SG proteomic response to infections, we deployed high-resolution MS with isobaric tag for relative and absolute quantitation (i-TRAQ) on DENV2-, ZIKV-and CHIKV-infected and non-infected A. aegypti SGs. We identi ed 218 proteins using a custom protein database and described those with functions related to immunity, blood feeding, digestion, metabolism and ribosome, stress and mitochondria. DENV2 infection regulated 58 proteins, ZIKV infection 27, and CHIKV infection 29. While a majority of differentially expressed proteins (DEPs) were virus-speci c, we characterized the function in SG of the four proteins upregulated by all three viruses. We identi ed the anti-ZIKV function of gamma-interferon responsive lysosomal thiol-like (GILT-like), the anti-CHIKV function of an adenosine deaminase (ADA), the pro-ZIKV function of salivary gland surface protein 1 (SGS1) and the antiviral function against all three viruses of a protein we named salivary gland broad-spectrum antiviral protein (SGBAP).

Mosquito rearing
Aedes aegypti mosquitoes were collected in Singapore in 2010, and ever since reared in the insectary. Eggs were hatched in MilliQ water, larvae fed on a mixture of TetraMin sh akes (Tetra, Germany), yeast and liver powder (MP Biomedicals, France) and adults maintained on 10% sucrose (1st Base, Singapore).
Mosquitoes were maintained at 28°C and 50% relative humidity with a 12h:12h light: dark cycle.

Viruses
Dengue virus serotype 2 PVP110 was isolated from an EDEN cohort patient in Singapore in 2008 36 . Zika virus Paraiba_01/2015 was isolated from a febrile female in the state of Paraiba, Brazil in 2015 37 .
Chikungunya virus SGP011 was isolated from a patient at the National University Hospital in Singapore in 2008 38 . DENV2 and ZIKV isolates were propagated in C6/36 (CRL-1660) and CHIKV in Vero (CCL-81) cell lines. Virus stocks were titered with BHK-21 cell plaque assay as previously described 39 , aliquoted and stored at -80 °C.

Mosquito infection
Three-to-ve day old female mosquitoes were starved for 24 h and fed on an infectious blood meal containing 40% volume of washed erythrocytes from speci c pathogen free (SPF) pig's blood (Prestige BioResearch, Singapore), 5% 10 mM ATP (Sigma-Aldrich, USA), 5% human serum (Corning human AB serum, Thermo Fisher Scienti c, USA) and 50% virus solution in RPMI media (Gibco, Thermo Fisher Scienti c, USA), using Hemotek membrane feeder system (Discovery Workshops, UK). The virus titers in blood meals were 2 x 10 7 pfu/ml for DENV2, 6 x 10 6 pfu/ml for ZIKV, and 1.5 x 10 8 pfu/ml for CHIKV which resulted in 100% SG infection for each virus 16 . Bloodmeal titers were validated in plaque assay using BHK-21 cells. Control mosquitoes were fed with the same blood meal composition except for virus solution, which was replaced by RPMI media. Following oral feeding, fully engorged females were selected and kept in a cage with ad libitum access to a 10% sucrose solution in an incubation chamber with conditions similar to insect rearing.
For inoculation, mosquitoes were cold-anesthetized and intrathoracically injected with 0.5 pfu of either DENV2, ZIKV or CHIKV using Nanoject-II (Drummond scienti c company, USA). The same volume of RPMI media was injected as control. Virus inoculation was conducted four days post dsRNA injection.
Sample preparation and i-TRAQ labeling Mosquito SGs were dissected and collected in 1X phosphate buffer saline (pH 7.4, Cytiva HyClone, Thermo Fisher Scienti c, USA) at 14 days post oral infection (dpi) for DENV2 and ZIKV, and seven dpi for CHIKV. Ninety SGs were pooled together for each condition, and freeze-thawed twice. The samples were nally homogenized using mini beadbeater-96 (Biospec Products, USA) and centrifuged to collect the supernatant as salivary gland extract (SGE). The protein content of each sample was normalized based on their concentration as measured by Pierce BCA protein assay kit (Thermo Fisher Scienti c, USA). SGEs were denatured, alkylated, trypsin (Promega) digested, and labeled using i-TRAQ 8plex Protein quantitation kit (AB SCIEX, Singapore) following manufacture's protocol. Each condition was conducted in triplicate.

LC-MS/MS analysis
The 1st dimension of peptide separation was conducted using an Eksigent nanoLC Ultra and ChiPLCnano ex (USA) in TrapElute con guration. Subsequently, the samples were loaded on a 200 μm x 0.5 mm column and eluted on an analytical 75 μm x 15 cm column (ChromXP C18-CL, 3 μm). A gradient formed by mobile phase A (2% acetonitrile, 0.1% formic acid) and mobile phase B (98% acetonitrile, 0.1% formic acid) was used to separate 2 and 5 μl of the sample at a 0.3 μl/min ow rate. The following gradient elution was used for peptide separation: 0 to 5% of mobile phase B in 1 min, 5 to 12% of mobile phase B in 15 min, 12 to 30% of mobile phase B in 114 min, 30 to 90% of mobile phase B in 2 min, 90% for 7min, 90 to 5% in 3 min and nally held at 5% of mobile phase B for 13 min. The tandem MS analysis was Modi cation; Database for Aedes VB Search Effort: Thorough; and FDR Analysis: Yes. The MS/MS spectra were searched against a decoy database to estimate the false discovery rate (FDR) for peptide identi cation. The decoy database consisted of reversed protein sequences from the same custom protein database as mentioned earlier. Different modi cation states of the same peptide sequences were considered distinct by the software. Peptides with con dence score ≥ 95% were considered identi ed, and proteins with at least two unique identi ed peptides were quanti ed. Proteins were identi ed as upregulated when they had the i-TRAQ ratio above 1.5 (p-value < 0.05) and downregulated when they had the ratio below 0.67 (p-value < 0.05). Proteins with ratio from 1.5 to 0.67 were considered not regulated. Functional annotation of the proteins was done using Blast2go software with mosquito database from VB, and Diamond search algorithm with Drosophila melanogaster homologs from FlyBase 40,41 .

Phylogenetic analysis of SGBAP in different mosquito species
A phylogenetic tree for SGBAP and its homologs was inferred by using the maximum likelihood method and general time reversible model 42 . Twenty-one homologs in Aedes aegypti, Aedes albopictus and Culex quinquefasciatus were identi ed using paralogs and orthologs from VB. We did not nd homologs in other mosquito species or D. melanogaster. Trees for heuristic search were obtained by applying Neighbor-Join and BioNJ algorithms to a matrix of pairwise distances estimated using the maximum composite likelihood (MCL) approach on cDNA sequences. The topology with the superior log likelihood value (-15281.44) was selected. A discrete Gamma distribution was used to model evolutionary rates (5 categories (+G, parameter = 6.7916)) with some sites allowed to be evolutionarily invariable ([+I], 1.98% sites). Evolutionary analyses were conducted in MEGA X 43,44 . Salivary gland gene silencing using double stranded RNA Aedes aegypti salivary gland cDNA was used to amplify dsRNA targets with T7-tagged primers (Table  S2). The PCR products were in vitro transcribed using T7 Scribe kit (Cellscript, USA). dsRNA was annealed by heating to 95 °C and slow cooling in a thermocycler. Three to ve-day-old adult female mosquitoes were cold-anesthetized and intra-thoracically injected with 2 µg of dsRNA using Nanoject II (Drummond Scienti c Company, USA). The same quantity of dsRNA against the bacterial gene LacZ was injected as control (dsCtrl). Four days post dsRNA injection, gene depletion was validated in SGs by RT-qPCR.
Gene expression quanti cation using real-time quantitative polymerase chain reaction Total RNA was extracted from 10 SGs using E.Z.N.A. Total RNA kit I (Omega Bio-Tek, USA), DNAse treated using Turbo DNA-free kit (Thermo Fisher Scienti c, USA), and reverse transcribed using iScript cDNA synthesis kit (Bio-Rad, USA). Gene expression was quanti ed using qPCR with SensiFast Sybr no-rox kit (Bioline, USA) and gene speci c primers (Table S2). Actin expression was used for normalization. The reactions were performed using the following conditions: 95°C for 10 min, 40 cycles of 95°C for 5 s, 60°C for 20s and melting curve analysis. The 2 -DDCq method was used to calculate relative fold changes.
Quanti cation of viral genomic RNA (gRNA) copies using RT-qPCR At 8 days post inoculation, individual pairs of SGs were collected in 350 µl of TRK lysis buffer (E.Z.N.A Total RNA kit I, Omega Bio-Tek, USA) and homogenized with the mini beadbeater-96 (Biospec Products, USA) before RNA extraction with E.Z.N.A Total RNA kit I (Omega, Bio-Tek, USA). DENV2 gRNA copies were quanti ed by RT-qPCR using i-Taq universal probes one-step kit (Bio-Rad, USA), and ZIKV and CHIKV gRNA copies with i-Taq Universal SYBR green one-step kit (Bio-Rad, USA) with speci c primers (Table S2). CFX96 Touch Real-Time PCR Detection System (Bio-Rad, Singapore) was used for ampli cation with the following thermal pro le: 50°C for 10 min, 95°C for 1 min, 40 cycles at 95 °C for 10s, 60 °C for 15 s. A melt-curve analysis was added for the SYBR-based quanti cation.
To quantify gRNA copies, a standard curve for each qPCR target was generated. qPCR targets were ampli ed from viral cDNA with the qPCR primers and forward primer tagged with T7. RNA fragments were generated with T7-Scribe kit (Cell Script, USA). RNA target copies were estimated based on Nanodrop quanti cation and used to generate an absolute standard equation. Three standard dilutions per plate were then added to adjust for inter-plate variation.

Statistical analyses
Differences in gene expression and gRNA copies were tested with unpaired T-test (in Microsoft excel) after log2 and log10 transformation, respectively, to meet normality.

Results
High-resolution proteomics of salivary glands A total of 218 proteins with at least two unique peptides (95% con dence per peptide) were identi ed in uninfected A. aegypti SG (Table S1). To identify proteins, we used a custom protein database including annotated proteome database from VB and UniProt as well as de novo proteome data (unpublished) generated from our previous work on A. aegypti SG transcriptome collected at the same time point from the same mosquito colony 16 . This custom protein database is unique to our study conditions with the A. aegypti Singapore strain. We identi ed ve newly annotated proteins which included one putative-C-type lectin, one 18.6 kDa secreted protein, one uncharacterized (no homolog in Drosophila melanogaster) protein, one aggrecan core-like protein and one 34 kDa salivary protein. A large majority (175 out of 218) of the proteins that we detected were also found in the only other high-resolution SG proteomic analysis 32 (Fig. S1a). Discrepancies between the studies may stem from different starting protein amounts, mosquito colonies and mosquito age at collection. Signal peptides (SP), which directs proteins to the secretory pathways 45 , may indicate secretion into saliva. We detected 71 SP-proteins (32.57% of all proteins), among which 39 were previously found in A. aegypti saliva 27 (Fig. S1b). Of note, proteins could also be secreted by non-classical pathways 45,46 as exempli ed by SGS1 that do not have an SP but is detected in saliva 27 . Hereafter, proteins with a SP (including the secreted SGS1) were categorized as secretory proteins, whereas non-SP proteins were considered as cellular proteins.
Most of these (97 out of 103) did not have a SP (Fig. 1a, b). There were also seven proteins related to proteolysis (PROT), two related to polysaccharide digestion (DIG), seven related to transport (TRP), 30 related to diverse functions (DIV) and 51 had unknown functions (UNK) (Fig. 1a, 1b; Table S1). Interestingly, there were 11 proteins related to immunity, including three serpins (SRPNs), one serine protease, four C-type lectins (CTLs), two brinogen related protein (FREPs) and one lysozyme (LYS) (Table S1). We did not detect proteins related to signaling of the canonical immune pathways. Seven proteins were related to blood-feeding (BF) and included two D7 proteins, two apyrases, one ADA, a prosialokinin precursor and one odorant binding protein (OBP) (Fig. 1a, 1b; Table S1). All immunity-and BF-related proteins had a SP (Fig. 1a, 1b; Table S1).
Salivary gland proteome response to DENV2, ZIKV or CHIKV infections Owing to the different EIP for aviviruses and alphaviruses 10,12 , SG were dissected at 14 days post oral infection (dpi) for DENV2 and ZIKV, and at seven dpi for CHIKV. Blood inocula resulted in 100 % of infected SG at the collection time, as determined previously 16 . Controls for DENV2 and ZIKV, and for CHIKV were dissected at the corresponding times post uninfectious blood feeding. Using iTRAQ-based quantitative proteomics, we found 35, 17 or 16 upregulated, and 23, 10 or 13 downregulated proteins by DENV2, ZIKV or CHIKV infection, respectively (Fig. 2a, 2b; Fig. S2, Table S2). Detection of the corresponding viral proteins in the SG proteome further con rmed infection.
Among RSM-related proteins, protein disul de isomerase (PDI, AAEL002501) was commonly upregulated by both aviviral infections. However, another PDI (AAEL000641) was uniquely downregulated by DENV2 infection alone. Thioredoxin reductase (AAEL002886) and 3-ketoacyl-CoA thiolase (AAEL0010697) were commonly upregulated by DENV2 and CHIKV infections. Few other RSM-related proteins were uniquely regulated by all three infections (Table S2). The weak overlap between infections indicates an overall virus-speci c response at the protein level in SGs.

Functional evaluation of virus-induced proteins in SGs
To determine the function of the four upregulated proteins (i.e., SGBAP, SGS1, ADA and GILT-like) on DENV2, ZIKV or CHIKV infections, we depleted these proteins in SGs by RNAi-mediated gene silencing (silencing e ciency ranged from 48.7-73.9%; Fig. S3). DsRNA (dsCtrl) targeting the bacterial gene LacZ was injected as control. To study the impact of gene depletion in SG only, we bypassed the midgut barrier by infecting mosquitoes through intra-thoracic inoculation with an inoculum enabling an increase or a decrease in infection 16 . At 8 days post inoculation (dpin) with DENV2 and ZIKV, and 4 dpin with CHIKV, we quanti ed gRNA in SGs and calculated infection prevalence (de ned as percentage of infected SG) and infection intensity (measured as viral gRNA copies per infected SG). We used different mosquito batches to test the different genes, and because we observed that infection in control mosquitoes varied between batches (Fig. 3-5), infection outputs were compared within batches.
For all three viruses, infection prevalence was not altered by any gene silencing (Fig. 3-5). Of note, infection prevalence was 100% for ZIKV and CHIKV, thereby preventing observation of a pro-viral effect with this parameter. Interestingly, gene silencing altered infection intensity in a virus-speci c manner. DENV2 infection intensity was increased by SGBAP depletion (Fig. 3). ZIKV infection intensity increased upon SGBAP and GILT-like depletions, and decreased upon SGS1 depletion (Fig. 4). CHIKV infection intensity was higher when SGBAP and ADA were depleted (Fig. 5). By studying SG proteins with uncharacterized impact on infection, we identi ed the virus-speci c function of GILT-like, SGS1 and ADA, and the broad antiviral function of SGBAP (Table 1). Discussion SG response to infection regulates viral transmission (i) by modulating the mosquito antiviral response, which reduces virus amount in SG and saliva, and (ii) by altering the production of salivary components, which in uences skin infection. Despite the relevance of SG in transmission, there is a dearth of knowledge about its response to infection at the proteome level. Leveraging cutting-edge proteomics technology, this study bridges this knowledge gap by describing the SG response to DENV2, ZIKV and CHIKV infections in A. aegypti at the global proteome level. Using high-resolution MS, we identi ed 218 proteins expressed in SG with putative functions in immunity, blood-feeding and cellular machinery. Using isobaric-based quantitative proteomics, we detected 58 proteins that were regulated by DENV2 infection, 27 by ZIKV infection and 29 by CHIKV infection. While a majority of proteins were not commonly regulated by all three viruses, four proteins were signi cantly upregulated in SG by DENV2, ZIKV and CHIKV infections. Hypothesizing that their upregulation was related to an antiviral response, we separately tested their functions in SG. The results revealed the antiviral function of GILT-like against ZIKV, ADA against CHIKV and the proviral function of SGS1 for ZIKV. Most interestingly, we showed that SGBAP reduced DENV2, ZIKV and CHIKV infections in SG, thereby identifying a potential target to generate arbovirus-refractory mosquitoes.
We found that a large majority of the SG proteins related to immunity were regulated by the infections. For this, they bind trypsin-like targets through an arginine or lysine residue at P1 position 57 . Of note, the three SG SRPNs lack the characteristic inhibitory sequence and could therefore be non-inhibitory or act in a non-classical way as protease inhibitor 48,57 . The one LYS (LYSC9) expressed in SG was downregulated by both DENV2 and ZIKV infections. Its closest D. melanogaster homolog (i.e., LYSP) is speci cally expressed in SG 58 , while another LYSC in A. aegypti was upregulated in midgut by DENV2 infection 59 .
LYS have functions in both digestion and immunity 60 . Overall, we identi ed immunity-related proteins regulated by DENV2, ZIKV or CHIKV infections in SG. Determination of their roles in SG immune response will require functional characterization.
We also observed that SG infection in uenced the expression of proteins expectorated in saliva. OBP22 was previously found to be a ligand for fatty acids 70 and required for e cient biting 17 . Our data suggests that SG infection can modulate transmission by altering saliva composition. Moreover, proteins related to digestion, metabolism and redox are discussed in supplemental (S1 Text).
A large proportion of SG proteins that were regulated by the infections remains uncharacterized. In this study, we determined the function of four uncharacterized infection-responsive proteins in SG infection, i.e., ADA, SGS1, GILT-like and SGBAP (Table 1). Tissue expression analysis based on available A. aegypti transcriptome literature 49,71−74 showed that SGBAP and SGS1 are speci cally expressed in SG (Fig. S4).
ADA is highly expressed in SG but is also present at lower levels in female abdominal tips. GILT-like protein is expressed in a wide range of tissues. While all four proteins were upregulated by DENV2, ZIKV and CHIKV in SG in the current study, we observed that ADA and GILT-like had virus-speci c antiviral properties against CHIKV and ZIKV, respectively, that SGS1 had virus-speci c proviral function for ZIKV, and that SGBAP had broad-spectrum antiviral properties by inhibiting all three viruses. ADA is indirectly involved in immune regulation, as it degrades adenosine, which suppresses immune response 75 .
Accordingly, ADA enhances DENV2 infection in keratinocyte cells by inhibiting type I interferon response 75 . ADA levels in SG may thus regulate the balance between immune activation and repression. GILT-like was originally discovered as an interferon-inducible gene in mammals and subsequent characterization revealed its function in antigen presentation, bacterial infection and production of reactive oxygen species 76 , which provides a rational for its antiviral function. In mosquitoes, GILT-like interacts with Plasmodium parasites and limits the parasite motility in skin when expectorated during biting 77 . The SGS1 is secreted in saliva through a non-classical pathway 78 and is a potential receptor for cell entry of avian malaria sporozoites in A. aegypti SG 79 . Its proviral effect for ZIKV might be related to a similar mechanism.
Factors with broad antiviral properties are of particular interest in designing transmission blocking interventions. We revealed the antiviral function of SGBAP against two aviviruses and one alphavirus in SG of A. aegypti. SGBAP does not contain conserved functional domains as determined with NCBI conserved domain search, InterPro -EMBL-EBI or PROSITE-Expasy and has no homolog in D. melanogaster, making it di cult to speculate on its structure or mechanism of action. SGBAP is a small protein of 130 amino acid residues (mature protein) (we suggest a re-annotation of the gene in S1 text and Fig. S5) and is secreted in A. aegypti saliva 27 . An earlier transcriptomic study suggested it originated from a truncation of a gene from the 34 kDa protein family 48 . To determine whether SGBAP function is conserved in other arbovirus mosquito vectors, we built a phylogenetic tree (Fig. S6). Among all mosquito species, SGBAP showed 21 putative homologs in A. aegypti, Aedes albopictus and Culex quinquefasciatus. SGBAP sequence did not cluster with other genes, thereby not supporting the existence of SGBAP antiviral homologs. Similarly, the closest homolog protein (i.e., A. albopictus gene -AALF004420) had only 56% identity. Functional homology among SGBAP orthologues should be experimentally tested. In A. aegypti, SGBAP broad antiviral function warrants further studies to understand its mechanism in SG and its function in saliva, where it could inhibit virus propagation.
In conclusion, we expanded our understanding of SG response to DENV2, ZIKV and CHIKV infections by using high-resolution quantitative proteomics for the rst time in mosquito SG. We also identi ed new antiviral factors in SG, shedding new light on the antiviral response, which can be used to promote transmission blocking interventions.
Each graph combines results from the same mosquito batch.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.