Midgut transcriptomic responses to dengue and chikungunya viruses in the vectors Aedes albopictus and Aedes malayensis

Dengue (DENV) and chikungunya (CHIKV) viruses are among the most preponderant arboviruses. Although primarily transmitted through the bite of Aedes aegypti mosquitoes, Aedes albopictus and Aedes malayensis are competent vectors and have an impact on arbovirus epidemiology. Here, to fill the gap in our understanding of the molecular interactions between secondary vectors and arboviruses, we used transcriptomics to profile the whole-genome responses of A. albopictus to CHIKV and of A. malayensis to CHIKV and DENV at 1 and 4 days post-infection (dpi) in midguts. In A. albopictus, 1793 and 339 genes were significantly regulated by CHIKV at 1 and 4 dpi, respectively. In A. malayensis, 943 and 222 genes upon CHIKV infection, and 74 and 69 genes upon DENV infection were significantly regulated at 1 and 4 dpi, respectively. We reported 81 genes that were consistently differentially regulated in all the CHIKV-infected conditions, identifying a CHIKV-induced signature. We identified expressed immune genes in both mosquito species, using a de novo assembled midgut transcriptome for A. malayensis, and described the immune architectures. We found the JNK pathway activated in all conditions, generalizing its antiviral function to Aedines. Our comprehensive study provides insight into arbovirus transmission by multiple Aedes vectors.


Transcriptome regulations by CHIKV and DENV in A. albopictus and A. malayensis midguts at 1 and 4 dpi.
To determine the transcriptomic response to arboviral infections in midguts of secondary vectors, A. albopictus mosquitoes were orally infected with 10 7 pfu/ml of CHIKV and A. malayensis mosquitoes were orally infected with either the same concentration of CHIKV or 2 × 10 7 pfu/ml of DENV. Both blood inoculum are within the high-end of viremia measured in patients 48,49 and resulted in 100% infected mosquitoes 34 . Controls were fed non-infectious blood. Viruses and mosquitoes were sympatric as they were all collected in Singapore, and the CHIKV strain collected before 2010 did not possess the envelope mutation that enhances A. albopictus infection 50 . As Wolbachia can influence mosquito susceptibility to arboviruses 51 , we conducted a PCR detection assay and did not find the bacteria in our colonies (Fig. S1).
We then performed high-throughput RNA-sequencing on three replicates of 20 pooled midguts collected at 1 and 4 dpi for each virus and mosquito combination. We selected these two time points to cover the initial response to infection at 1 dpi and the established infection at 4 dpi. We obtained between 38 and 52 million reads per sample (Table S1). After quality filtering, we confirmed mosquito species and virus infection by mapping reads to species-specific cytochrome c oxidase (COI) genes and each viral genome, respectively (Table S1). We also quantified the percentage of viral reads and observed that CHIKV reads were more abundant than DENV and had plateaued at 1 dpi in both mosquito species, whereas DENV reads increased between 1 and 4 www.nature.com/scientificreports/ dpi (Fig. 1a). Since CHIKV and DENV have different patterns of infection kinetics 52,53 , this should be taken into account when interpreting gene regulations.
Using the most recent genome assembles for A. albopictus 54 (AalbFP1.0) and A. aegypti 55 (AaegL5.0) (A. malayensis genome has not been assembled), we mapped reads from A. albopictus and A. malayensis and calculated DEGs between infected samples and time-matched non-infected controls. To select the best genome approach and validate DEGs, we quantified expression levels of several genes at 4 dpi by RT-qPCR in midguts of separate mosquito batches for both species. For both A. albopictus and A. malayensis, DEGs obtained by mapping on the A. albopictus genome correlated best with qPCR data (Fig. S2 and S3) and were further analyzed. Of note, A. malayensis genes do not have IDs and hereafter, are identified from A. albopictus orthologs.
To identify similarities in the transcriptomes between our different conditions, we performed a principal component analysis (PCA) on normalized gene expression values (Fig. 1b). Interestingly, virus infection was not the main source of transcriptome variation, mosquito species and time post blood-feeding were responsible for the greatest variance. The PC1 explained 66% of the total variance and segregated the mosquito species, and the PC2 explained 20% of the total variance and separated the time of collection.
Upon CHIKV infection in A. albopictus, there were 1,793 DEGs at 1 dpi and 339 at 4 dpi (Table S2). Upon CHIKV infection in A. malayensis, there were 943 DEGs at 1 dpi and 222 at 4 dpi. Upon DENV infection in A. malayensis, there were 74 DEGs at 1 dpi and 69 at 4 dpi.
Among all the CHIKV infection conditions (i.e., both mosquito species at 1 and 4 dpi), we identified shared DEGs (Fig. 1c). First, we found 81 DEGs common to all conditions and that 79 were upregulated in all conditions (Table S2). The top eight most regulated were highly expressed (> 48 fold) and included two uncharacterized genes (AALFPA_051205, AALFPA_066142), one gene putatively involved in antiviral response (AAL-FPA_05206), one serine-rich protein kinase (AALFPA_068585), one deubiquitinase (AALFPA_074126), one cytochrome P450 (AALFPA_062825), one homologue of hunchback (AALFPA_070528) and one homolog of canoe (AALFPA_068948). Second, we looked at the effect of time post infection within mosquito species. We www.nature.com/scientificreports/ found that 223 DEGs were conserved in A. albopictus between 1 and 4 dpi (Table S2), representing 12% and 66% of all DEGs for each dpi, respectively. We observed that 153 DEGs were common in A. malayensis between 1 and 4 dpi, representing 16% and 69% of within-dpi DEGs, respectively. Third, we reported the shared DEGs among mosquito species at each time point. At 1 dpi, 530 DEGs were common between A. albopictus and A. malayensis, representing 29% and 56% of all DEGs for each species, respectively. At 4 dpi, 130 DEGs were common, representing 7.3% and 13.8% of all DEGs for A. albopictus and A. malayensis, respectively. Overall, the transcriptomic impact of CHIKV infection is relatively conserved across the two mosquito species and a majority of DEGs at 4 dpi were already altered at 1 dpi. We subsequently looked at the shared DEGs in A. malayensis infected with DENV or CHIKV at 1 and 4 dpi. Firstly, there was only one conserved DEG between all conditions (Fig. 1d) and it was the uncharacterized AAFLPA_064235, which was slightly upregulated in all conditions (Table S2). Secondly, we looked at the effect of time post infection. There were only two DEGs common between 1 and 4 dpi with DENV, which was less than between 1 and 4 dpi with CHIKV, as detailed above. Finally, we observed how the virus influenced the transcriptome within each time point. There were 28 DEGs common between DENV and CHIKV at 1 dpi, representing 38% and 3% of all DEGs for DENV and CHIKV, respectively. There were 19 shared DEGs between the two viruses at 4 dpi, representing 28% and 9% of DENV-and CHIKV-induced DEGs. Overall, in A. malayensis, DENV induced a different transcriptomic response at 1 and 4 dpi and the transcriptomic signature of DENV and CHIKV differed.
Functional annotations of genes responsive to CHIKV and DENV infections in A. albopictus and A. malayensis midguts. We annotated the DEGs into different functional categories based on characterized functions, homology, or protein domains. For both comparative purposes and because of the absence of an annotated A. malayensis genome, we relied on the orthologs in A. albopictus to describe virus-induced gene alterations in A. malayensis. Apoptotic-related genes were mostly regulated by CHIKV in A. albopictus at 1 dpi ( Fig. 2; Table S2) with the up-and downregulation of apoptosis-inducers such as CASP. Interestingly, the expression of an inhibitor of apoptosis (IAP; AALFPA_076435) was increased in all CHIKV conditions. Only one apoptotic gene was induced by DENV at 1 dpi in A. malayensis.
Several genes related to blood-feeding were regulated by CHIKV infection at the different time points and in both mosquito species but only one blood-feeding gene was regulated by DENV at 1 dpi in A. malayensis ( Fig. 2; Table S2). Particularly, one putative odorant binding (AALFPA_066142) was strongly (> 48 fold) induced in all CHIKV conditions and one gustatory receptor (AALFPA_047124) was similarly highly upregulated by CHIKV at both time points in A. malayensis. Genes related to cytoskeleton were mostly upregulated in all the CHIKV conditions. However, contrasting regulation of cytoskeleton genes were observed for DENV in A. malayensis, where they were downregulated 1 dpi, but upregulated at 4 dpi ( Fig. 2; Table S2). Genes related to digestion were mostly regulated by CHIKV at 1 dpi in both mosquito species and in lower quantity by CHIKV at 4 dpi and DENV at 1 dpi ( Fig. 2; Table S2). None of the digestion-related genes were regulated by DENV at 4 dpi. Of interest, two collagen genes and one chitinase gene were upregulated in all CHIKV conditions. A large proportion of genes with diverse functions were regulated in all conditions. Among the genes related to immunity, in CHIKV-infected A. albopictus, 36 DEGs were upregulated and 21 downregulated at 1 dpi, and 12 were upregulated at 4 dpi ( Fig. 2; Table S2). In CHIKV-infected A. malayensis, 30 DEGs were induced and five were reduced at 1 dpi, while expression of nine were increased at 4 dpi. In DENV-infected A. malayensis, there was one upregulated immune DEGs and five downregulated at 1 dpi, and www.nature.com/scientificreports/ four increased DEGs at 4 dpi. Among these DEGs, there were several PRRs. One GNBP and three PGRPs were downregulated by CHIKV at 1 dpi in A. albopictus. A variety of serine protease or serine protease inhibitor DEGs were regulated in either direction by CHIKV and by DENV in both mosquito species. For IMD pathway, one activator (sickie-like; AALFPA_052752) was inhibited and another one (hyperplastic discs; AALFPA_079063) was activated at 1 dpi in CHIKV-infected A. albopictus. Another activator (sickie-like; AALFPA_058168) was upregulated by CHIKV at 4 dpi in both mosquito species. For Toll pathway, two Toll-like genes (AALFPA_053631 and AALFPA_057415) were induced by CHIKV at 1 dpi in A. albopictus. For Jak/STAT pathway, three repressors (three Suppressor of cytokine signaling (SOCS)-like; AALFPA_051177; AALFPA_054706; AALFPA_056404) were activated at 1 dpi with CHIKV and at 4 dpi with DENV in A. malayensis, whereas one Bromodomain and WD repeat domain containing 3 gene (BRWD3, AALFPA_050507) was induced by CHIKV at 1 dpi in A. albopictus. For JNK pathway, two Kayak-like genes (AALFPA_079058 and AALFPA_079354) were upregulated by CHIKV at 1 dpi in both mosquito species and by DENV at 4 dpi in A. malayensis. At 1 dpi with CHIKV, two other JNK activators were induced in either mosquito species (A. albopictus: AALFPA_071724 and AALFPA_078630; A. malayensis: AALFPA_071724 and AALFPA_040889). In terms of immune effectors, two complement genes (AAL-FPA_045642 and C1q, AALFPA_055182) were increased by CHIKV at 1 dpi in A. albopictus, but no antimicrobial peptide (AMP) was directly regulated. Moreover, one helicase (AALFPA_080443), involved in RNAi, was induced at 1 dpi with CHIKV in A. albopictus. The same helicase plus another (AALFPA_066096) were upregulated at 1 dpi with CHIKV in A. malayensis. Overall, although we observed the regulations of components from several immune pathways, there was a consistent induction of JNK activators.
Immune regulations in A. albopictus midgut infected by CHIKV. Because the immune response chiefly determines vector competence 32,33 , we focused on immunity. We identified the immune genes transcribed in A. albopictus midguts in both infected and control conditions, and classified them based on the categories defined by Palatini et al. 54 (Table 1; Table S3). A majority of the immune genes identified in the A. albopictus genome were detected in our RNA-seq dataset (463 out of 664; not including JNK-related genes). Interestingly, most of the PRRs (GNBPs and PGRPs), all the components of the cytoplasmic signaling of the immune path- To identify patterns in the immune response to CHIKV infection at 1 and 4 dpi in A. albopictus, we clustered and plotted the expressions of the 50 most variable immune genes in a heatmap ( Fig. 2; Table S3). We detected 5 clusters. Cluster 1 contained only one C-type lysozyme that also belonged to the significant DEGs identified above and was highly upregulated (> 77 fold) at both time points. Cluster 2 was made of five genes only downregulated at 4 dpi and included two PRRs (one PGRP and one GNBP), two proteases involved in the regulation of the immune signaling (one CLIP and one SRPN), and one LYS. Cluster 3 consisted of 12 genes induced at 1 dpi and then reduced at 4 dpi. These included several genes implicated in regulating the immune activation, i.e., three SRPNs (two of which were significant DEGs), one SPZ and one TOLL. Their clustering with effectors such as one AMP (Attacin-B), one CTL and two FREPs, suggested their regulations through the above enzymes. Cluster 3 also included two SCRs involved in ROS regulation and one PPO that induces melanization. Inversely to cluster 3, cluster 4 was made of 14 genes downregulated at 1 dpi and upregulated at 4 dpi. Among the proteins that regulate immune signaling, there were one PGRP, one SPZ, one SRPN and one CLIP. Six immune effectors were also grouped in the same cluster and included three FREPs and three MLs. Three CASPs that trigger apoptosis were present in cluster 4, together with one SRRP. Finally, cluster 5 contained 18 genes only downregulated at 1 dpi. Genes involved in immune signaling regulation were two PGRPs, one TOLL, three SRPNs and two CLIPs. One ML effector was also present. Regulations of ROS was apparent due to the clustering of five HPXs, regulation of melanization due to two PPOs, and regulation of RNAi because of two SRRP aubergine homologs.

Immune regulations in A. malayensis midgut infected by CHIKV and DENV. In A. malayensis,
we undertook a similar immune-centered approach. However, to provide the most accurate description for this mosquito species that does not have an assembled genome, we de novo assembled transcripts before searching for protein homologs in A. albopictus transcriptome. We identified 162 immune proteins (Data S1 details the protein sequences) expressed in the midgut of A. malayensis and functionally categorized them as above 54 ( Table 1; Table S4). There were a lower number of expressed homologs for all immune categories than in A. albopictus, except for AMP. These lower numbers of transcripts may be due to the conservative parameters (see "Methods") we used for homolog identification. Nonetheless, for immune signaling, we found 12 PRRs (2 GNBPs and 10 PGRPs), 5 SPZs, 2 TOLLs, 9 components of the IMDPATH, 1 of the JAKSTAT, 11 of the JNK, 4 of the TOLLPATH and 2 REL homologs. Among enzymes that regulate the signaling, we detected 9 CLIPs and 8 SRPNs. For effectors, we noted 6 AMPs, 3 CTLs, 6 FREPs, 8 GALEs, 2 LYSs, 6 MLs and 1 TEP. Moreover, our results indicated a functional RNAi with 20 SRRPs. Autophagy, apoptosis, melanization and ROS regulation were also activated with 14 APHAGs, 13 apoptosis-related (3 CASPAs, 7 CASPs, 3 IAPs), 1 PPO, and 19 oxidative stress-related transcripts (3 CATs, 7 HPXs, 7 SCRs and 2 SODs), respectively.
We next clustered the expression of the 50 most variable immune genes upon CHIKV ( Fig. 4a; Table S5) and DENV ( Fig. 4b; Table S5) infection, separately. With CHIKV-infected samples, we identified four clusters. Cluster 1 contained ten genes only upregulated at 1 dpi. Among the components of the immune pathway signaling, there were two SRPNs, one JNK component, and one REL transcription factor. For effectors, one LYS, one FREP and one GALE were clustered. The detection of two CASPs and Ago2, a major SRRP, suggested apoptosis and RNAi activation, respectively. Cluster 2 was made of 16 genes induced at 1 dpi and then reduced at 4 dpi. For signaling regulation, there were one GNBP, three SPZs, one CLIP and one homolog of Kayak, the JNK transcription factor. Multiple effectors were in cluster 2 and included two AMPs, one TEP, one FREP, one LYS, one GALE and one ML. There were also one APHAG and two SCRs. Cluster 3 was composed of eight genes upregulated only at 4 dpi. For immune signaling, there were one GNBP, one TOLL and one CLIP. For effectors, there was only one FREP. Additionally, there were two ROS-related genes (one SOD and one SCR), one IAP and one APHAG. Cluster 4 consisted of 16 genes only downregulated at 1 dpi. Those related to immune signaling included one CTL, two SRPNs, one IMDPATH and Cactus, the negative regulator of TOLLPATH transcription. Among effectors, there were two MLs. One APHAG and three HPXs were also clustered. Interestingly, five SRRPs were grouped in cluster 4 and included Ago3, loquacious and two aubergine homologs.
With DENV-infected samples, we identified five clusters ( Fig. 4b; Table S5). Cluster 1 consisted of six genes moderately upregulated at 1 dpi and strongly upregulated at 4 dpi. Among signaling genes, there was only one JNK component. Effectors included one ML and one GALE. There were also one CASPA, one APHAG and one SCR. Cluster 2 included 11 genes only upregulated at 4 dpi. No immune signaling-related genes were in cluster 2 and there were one ML and one GALE effectors. There were three apoptosis-related genes (two IAPs and one CASP), two APHAGs, one SOD and three RNAi-related genes (SRRPs). Cluster 3 was made of three genes induced at 1 dpi and then reduced at 4 dpi. These included one AMP, one SPZ and one LYS. Cluster 4 consisted of seven genes only upregulated at 1 dpi. Immune signaling genes included one GNBP and one SPZ, while immune effectors were one TEP, three FREPs and one AMP. Eventually, cluster 5 was composed of 23 genes moderately upregulated at 1 dpi and moderately downregulated at 4 dpi. Immune signaling genes included two PGRPs, three SPZs, three CLIPs, two SRPNs and two IMDPATHs such as an IKK-β homolog. Effectors consisted www.nature.com/scientificreports/ of two AMPs, three MLs and one CTL. There were also one APHAG, two CASPs and two ROS-related genes (one CAT and one SCR).

Discussion
Public health threats from dengue and chikungunya are expanding geographically and in intensity 3,56 . While A. aegypti remains the main vector, other Aedes species are competent enough to trigger moderate-scale epidemics and sustain DENV and CHIKV transmissions 8,57 . Furthermore, the ongoing deployment of new interventions aimed at A. aegypti like the Wolbachia-mediated transmission reduction 58,59 will most probably promote the roles of these secondary vectors. Here, we set to prepare for the next steps of vector control by comprehensively describing the transcriptomic responses to CHIKV and DENV in two capable vectors, A. albopictus and A. malayensis 27 . Because infection onset in the mosquito midgut dictates vector competence, we analyzed midguts at early times post infection, i.e., 1 and 4 dpi. Our results profile DEGs in two mosquito species infected by two viruses at two time points. Next, we focused on immune-related genes and utilized a de novo transcriptome assembly in A. malayensis to identify immune genes for the first time in this species. Overall, our study reveals patterns of transcriptomic responses in two Aedes vectors and details the immune responses, shedding new light on shared and divergent mechanisms of arboviral transmission. The overall transcriptomic response was shaped by viral infections as well as time post infection and varied between the two mosquito species. Although we orally infected mosquitoes with a similar inoculum for both viruses, the transcriptomic response was greater upon CHIKV than DENV infection. Such difference may relate to the faster CHIKV replication kinetic 52 , which reached a plateau as early as 1 dpi in both mosquito species. However, in spite of the similar viral genomic RNA (gRNA) levels between the two time points, the number of CHIKV-induced DEGs did not correlate with gRNA copies across the time points and sharply decreased from 1 to 4 dpi. This suggests a return to homeostasis after quick establishment of viral replication factories. Interestingly, CHIKV-induced DEGs at 4 dpi were mostly already regulated at 1 dpi in both mosquito species. These conserved regulated genes may represent the cellular machinery required for CHIKV multiplication. In DENV-infected A. malayensis midguts, the transcriptomic response did not correlate with gRNA quantities, which grew significantly between 1 and 4 dpi whereas DEG numbers were similar between the time points. In contrary to CHIKV infection, there was a very limited number of shared DEGs between 1 and 4 dpi in DENVinfected A. malayensis. There was also little overlap between DENV-and CHIKV-induced DEGs in this mosquito species. These transcriptomic differences document divergences in how DENV and CHIKV establish infection in mosquito midguts.
Strikingly, we identified a transcriptomic signature of CHIKV infection. There were 81 DEGs regulated in the same direction upon CHIKV infection in the two mosquito species (comparison was made using homologs from both species) and at both time points ( Fig. 1; Table S2). While these common DEGs were related to multiple functional groups, we discuss the top eight most regulated ones as they were highly induced in all conditions. AALFPA_052026 and its A. malayensis homolog have a D. melanogaster homolog (CG8492) that reduces viral infection 60 . When we detailed the immune response, this gene did not cluster with other immune genes in A. albopictus (Fig. 3). However, in A. malayensis, it responded similarly to a Relish homolog (Fig. 4a), suggesting a potential regulation through one of the immune pathways that induces Relish. A deubiquitinase was part of the CHIKV transcriptomic signature and may have influenced the immune response and cellular homeostasis, given the function of ubiquitination in both processes 61,62 . CHIKV infection consistently induced a serine-rich protein kinase (SRPK), which may influence gene expression through chromatin remodeling 63 . A cytochrome P450 monooxygenase was upregulated and could have metabolized xenobiotic substances 64,65 . Interestingly, two highly expressed genes of the CHIKV signature are related to cellular development, i.e., a homolog of hunchback, a transcription factor that regulates cellular development 66,67 , and a homolog of canoe, which maintains adherens junction between cells. Both genes may be required for the cell multiplication that occurs in infected mosquito midguts to compensate for damaged cells 68 . Together with two other uncharacterized homologs, these top induced DEGs shared across all our CHIKV conditions deserve further functional studies.
Given the importance of midgut immunity in determining vector competence 30,32,33 , we detailed the immune response in both mosquito species. In A. albopictus, most of the components of the IMD, TOLL, JAK/STAT and JNK pathways were expressed, indicating that the corresponding pathways are functional in midguts. In contrary, a lower proportion of enzymes involved in signaling regulation were detected and this suggests a midgutspecific regulation of immunity as different CLIP or SRPN interact with different immune components 69,70 . Apoptosis, autophagy, RNAi, ROS and melanization were also functional in midguts. In A. malayensis, smaller proportions of the immune pathway components were detected in midguts. While this may be related to the technical limitations associated with de novo assembly such as high error rates and short assembled contigs 71 , each canonical pathway was still represented by several key members. For TOLL and IMD pathways, the two downstream transcription factors, homologs of D. melanogaster Dorsal and Relish, respectively, were expressed in A. malayensis midguts (Table S4). The negative regulators for Dorsal and Relish, namely Caspar and Cactus respectively, were also found. Similarly, the downstream transcription activators for JNK and JAK/STAT, i.e., Kayak and STAT respectively, were transcribed, confirming the conservation of these pathways across insects 72 and indicating that the canonical immune pathways function in A. malayensis midguts.
Activation of the TOLL, IMD, JAK/STAT and JNK pathways occurs through transcriptional regulation of their components in response to infection 34,43,73,74 . Interestingly, the JNK pathway appeared to be induced in both mosquito species and upon both viral infections. Two Kayak-like homologs were significantly upregulated (i.e., DEG) at 1 dpi with CHIKV in both mosquito species and at 4 dpi with DENV in A. malayensis. In A. aegypti salivary glands, Kayak was previously shown to be induced by both CHIKV and DENV and to be co-regulated with Basket, one of the upstream components of the JNK pathway 34 . Supporting the activation of JNK, we also www.nature.com/scientificreports/ observed an induction of the Basket homolog (among the 50 most regulated immune genes) in A. malayensis at 1 dpi with CHIKV and 4 dpi with DENV. JNK activation clears viral infection by activating apoptosis and the complement system 34 . Together with the JNK components, we observed co-regulations of multiple CASPs and www.nature.com/scientificreports/ TEPs. Although our transcriptomic data indicate an activation of the JNK pathway, other immune pathways appeared induced as well. For instance, the induction of a Relish homolog together with the reduction of its inhibitor Cactus among the most regulated immune genes in A. malayensis at 1 dpi with CHIKV suggest the activation of the Toll pathway. The induction of Ikk-β in DENV-infected A. malayensis at 1 dpi show activation of the IMD pathway 75 . However, while components of several immune pathways were regulated, the activation of the JNK pathway was consistent in both A. albopictus and A. malayensis upon infection by CHIKV and DENV.
In conclusion, we conducted a high-throughput transcriptomic analysis in two Aedes vectors with DENV and CHIKV at two early time points in midguts. Our dataset provides a multidimensional picture of the transcriptomic and more specifically immune regulations in these conditions. In A. albopictus, we identified different gene regulation clusters, which may represent different immune pathway responses to CHIKV infection in midguts. In A. malayensis, we provided the first description of the immune response to arboviral infections. This knowledge is required to broaden our understanding of arboviral transmission by Aedes vectors.

Material and methods
Mosquito rearing. Aedes albopictus and A. malayensis mosquito colonies were established in 2010 and 2014, respectively, from eggs collected using oviposition traps in the parks of Singapore 27 . Since then, these colonies were reared in the insectary where eggs were hatched in MilliQ water, larvae fed on a mixture of fish food (TetraMin fish flakes), yeast and liver powder (MP Biomedicals) and adults maintained on 10% sucrose and fed pig's blood twice weekly. Mosquito colonies were maintained at 28 °C and 50% relative humidity with a 12 h:12 h light: dark cycle.

Mosquito infection.
Three-to-five day-old female mosquitoes were starved for 24 h before they were fed on an infectious blood meal containing 40% volume of washed erythrocytes from specific pathogen free (SPF) pig's blood (PWG Genetics), 5% 10 mM ATP (Thermo Scientific), 5% human serum (Sigma) and 50% virus solution in RPMI media (Gibco), using Hemotek membrane feeder system (Discovery Workshops) covered with porcine intestine membrane (sausage casing). The virus titers in blood meals were 2 × 10 7 pfu/ml for DENV and 1 × 10 7 pfu/ml for CHIKV. Blood titers were validated by 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. Following oral feeding, the 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. Validation of RNA-seq by real-time quantitative PCR. Aedes albopictus and A. malayensis mosquitoes from different batches than the ones used for RNA-seq were orally infected with DENV and CHIKV. At four dpi, midguts from eight mosquitoes were dissected and pooled in triplicates. Total RNA was extracted from the midgut samples using a E.Z.N.A. Total RNA kit I (OMEGA), DNAse treated using a Turbo DNA-free kit (Thermo Fisher Scientific), and reverse transcribed using iScript cDNA synthesis kit (Biorad). Expression for 11 genes was quantified with qPCR using the SensiFast Sybr no-rox kit (Bioline) and primers designed based on A. albopictus genome, as detailed in Table S6. Actin expression was used for normalization. Reactions were performed with the following cycle conditions: an initial 95 °C for 10 min, followed by 40 cycles of 95 °C for 5 s, 60 °C for 20 s and ending with a melting curve analysis. The delta delta method was used to calculate relative fold changes. Pearson correlations were calculated between qPCR-based and DEG log2 fold changes with excel. www.nature.com/scientificreports/ mine log2 Fold-changes between infected and control conditions and DEGs with the same criteria as detailed above.

De novo identification of immune genes in
To confidently annotate all A. malayensis immune transcripts, the protein products of A. albopictus immune genes identified by Palatini et al. 54 and an additional 13 JNK pathway genes were used as a database to identify orthologues from the translated A. malayensis de novo assembled transcriptome. This was done with Diamond Blastp 92 and a strict search criteria of an e-value 1.0E −10 or less, followed by a second search against all annotated A. albopictus proteins to confirm identifications. From this identified set, only transcripts that could be fully translated with no ambiguities, covered at least 90% of database subject sequences, and shared less than 98% identity with each other were considered confident A. malayensis immune genes. These were also searched against the FlyBase 85 D. melanogaster peptide database for identification of orthologues.

Data availability
Sequenced reads generated during this study are available under NCBI accessions: SRR14621613-SRR14621644, BioProject PRJNA731987.