Transcriptome of the Aedes aegypti Mosquito in Response to Human Complement Proteins

Aedes aegypti is the primary mosquito vector of several human arboviruses, including the dengue virus (DENV). Vector control is the principal intervention to decrease the transmission of these viruses. The characterization of molecules involved in the mosquito physiological responses to blood-feeding may help identify novel targets useful in designing effective control strategies. In this study, we evaluated the in vivo effect of feeding adult female mosquitoes with human red blood cells reconstituted with either heat-inactivated (IB) or normal plasma (NB). The RNA-seq based transcript expression of IB and NB mosquitoes was compared against sugar-fed (SF) mosquitoes. In in vitro experiments, we treated Aag2 cells with a recombinant version of complement proteins (hC3 or hC5a) and compared transcript expression to untreated control cells after 24 h. The transcript expression analysis revealed that human complement proteins modulate approximately 2300 transcripts involved in multiple biological functions, including immunity. We also found 161 upregulated and 168 downregulated transcripts differentially expressed when human complement protein C3 (hC3) and human complement protein C5a (hC5a) treated cells were compared to the control untreated cells. We conclude that active human complement induces significant changes to the transcriptome of Ae. aegypti mosquitoes, which may influence the physiology of these arthropods.


Introduction
Aedes mosquitoes, especially Ae. aegypti and Aedes albopictus are responsible for major arboviral epidemics of yellow fever (YFV), dengue (DENV), chikungunya (CHIKV), and Zika (ZIKV) around the world [1,2]. Mosquitoes acquire these viruses from a vertebrate host following the ingestion of an infected blood meal [3]. Once inside the mosquito, viruses must overcome several barriers before they can be transmitted back to a vertebrate host [4]. The first barrier is the midgut, where the pathogen would need to reach permissive cells before it is attacked by the mosquito digestive enzymes and active factors contained in the blood meal that can cause lysis of the virions [4,5]. Surviving pathogens must could be used to find novel ways of decreasing mosquito survival and inhibit virus transmission. Our rationale to focus on mosquito proteins that are activated against human complement is that these proteins could represent alternative candidates for control interventions, such as transmission-blocking vaccines (TBVs), in contrast to using viral proteins as targets because these may be more prone to mutation due to the immunological pressure exerted by both vertebrate and invertebrate hosts.

Sequencing and Read Mapping
To understand the molecular interactions between human complement and Ae. aegypti (Rockefeller strain), RNA-sequencing was used to explore the changes in the mosquito transcriptome. More than 135 million (135,411,293), Illumina 50 bp single-end directional reads (strand-specific) were sequenced across all samples (Table S1). Each in vitro test was represented by three libraries with an average of 9.13 million reads per library. The in vivo tests were represented by two libraries with an average of 8.87 million reads per library. The results obtained from the mapping of processed reads (131,210,742) against Ae. aegypti Liverpool Aedes Genome Working Group (LVP_AGWG) AaegL5.2 showed a higher percentage of uniquely mapped reads using the Spliced Transcripts Alignment to a Reference (STAR), showing between 6.71 to 7.72% more mapped reads than using Hisat2. Thus, these were used for expression analysis.

Transcript Differential Expression in NB and IB Fed Mosquitoes (In Vivo Experiments)
RNA-sequencing was conducted to explore the transcriptome changes in the Ae. aegypti abdominal region in response to oral ingestion of human red blood cells reconstituted with either autologous heat-inactivated plasma (IB) or normal plasma (NB) at 3 h post-feeding. RNA-seq analyses showed that 70.1% (19,866 out of 28,353) of annotated protein-coding transcripts were expressed in the mosquitos. A total of 3369 out of 28,353 of annotated protein-coding transcripts (equivalent to 11.9% of transcripts) were found differentially expressed (Fold change (FC) >4, p < 0.05) when comparing NB and sugar-fed (SF) control females. In this comparison, we found 1452 up-and 1917 downregulated transcripts.
Furthermore, 3270 out of 28,353 transcripts were found differentially expressed in SF when compared against IB (1496 upregulated and 1774 downregulated) ( Table S2). The comparison of NB vs. IB showed 1065 transcripts up-and 1285 downregulated differentially expressed in NB (Table S2). Further analysis of transcripts found differentially expressed in both NB and IB when compared to sugar-fed mosquitoes revealed 733 upregulated and 795 downregulated transcripts found in common.
The comparison between hC5a-treated cells and the untreated control cells revealed that 3.5% (981/28,353) of the transcripts had significant differential expression (504 upregulated and 477 downregulated) (Table S2). The upregulated transcripts were enriched in ion binding, phosphotransferase activity, and nucleotide-binding functions. In contrast, the downregulated transcripts were associated with functions such as actin-binding, cytoskeletal protein binding, guanosine-5'-triphosphate (GTP)ase regulator activity, and regulation of catalytic activity.
The comparison between hC5a-treated cells and the untreated control cells revealed that 3.5% (981/28,353) of the transcripts had significant differential expression (504 upregulated and 477 downregulated) ( Table S2). The upregulated transcripts were enriched in ion binding, phosphotransferase activity, and nucleotide-binding functions. In contrast, the downregulated transcripts were associated with functions such as actin-binding, cytoskeletal protein binding, guanosine-5'-triphosphate (GTP)ase regulator activity, and regulation of catalytic activity.
A total of 1.2% (329/28,353) differentially expressed transcripts, 161 up-and 168 downregulated, were in common between hC3 and hC5a treatments. The upregulated transcripts presented enrichment in multiple molecular functions like binding, regulation, and oxidoreductase activity. The GSEA of transcript differentially expressed in cells treated with either hC3 or hC5a compared to untreated cells is shown in Figure 2.
In the in vitro experiments, the Aag2 cells treated with hC3 and hC5a also showed a change in the expression of IRTs (Table S3). The cells treated with hC3 compared to control cells presented a change in the expression of 13 transcripts (6 up-and 7 downregulated), including the upregulation of caspase (AAEL014348-RF), Class B Scavenger Receptors (i.e., AAEL000234-RL, AAEL000256-RC),

Shared Transcripts between the In Vivo and the In Vitro Assays
The comparison between transcripts found differentially expressed in live mosquitoes (in vivo assays) and those found in Aag2 cells (in vitro assays) shows 44 transcripts (21 up-and 23 downregulated) in common for both experiments ( Figure 5, Table 1).

Shared Transcripts between the In Vivo and the In Vitro Assays
The comparison between transcripts found differentially expressed in live mosquitoes (in vivo assays) and those found in Aag2 cells (in vitro assays) shows 44 transcripts (21 up-and 23 downregulated) in common for both experiments ( Figure 5, Table 1).

Discussion
As demonstrated in previous studies, our current study confirms that human blood ingestion induces significant mosquito transcriptome changes. Moreover, our study highlights how a blood meal containing active complement (NB group) leads to important changes in the transcriptome profile of Ae. aegypti females compared to those fed with sugar (SF group). A significant number of these differentially expressed genes are linked to digestive processes, maturation of eggs, suppression of the response to stimuli, and even activation of the immune response [28]. Since complement activation may be detrimental to the midgut tissue, Ae. aegypti females have developed a serine proteases mechanism to prevent complement-induced damage. Serine proteases cleave and inactivate human complement proteins and catalases that reduce the oxidative stress caused by the hemoglobin heme group, extending the time of mosquito survival by days [29]. Serine-type endopeptidases, such as trypsin, are essential for blood digestion and may be associated with DENV replication in the mosquito midgut [5]. Furthermore, it has been demonstrated that late-phase trypsin (5G1) is associated with a significant reduction in DENV2 replication in mosquito midgut [5]. Another study also showed that treating an infectious blood meal with a trypsin inhibitor reduced DENV-2 midgut titers and delayed viral dissemination [30]. Our study revealed that several serine-type endopeptidases upregulated are the most abundant transcripts after feeding with human red blood cells reconstituted with autologous plasma. In contrast, several serine-type protease inhibitors were found downregulated, which is consistent with previous studies suggesting that these enzymes are critical in blood meal digestion.
Since a blood meal has significant reproductive effects compared with a sugar meal (which is the primary source of energy), using whole abdomens instead of midguts allows evaluating broad transcriptome changes at the digestive and reproductive tracts level. Using all the abdominal tissues (ovaries, Malpighian tubules, fat body, and hemocytes), we obtained a comprehensive perspective on the effect of human complement over the mosquito physiology. Our comparison between SF and NB fed mosquitoes showed a higher overall number of upregulated transcripts than those found between the IB vs. SF groups. Our results suggest that the presence of the active complement induces a different type of response in the mosquitoes, especially when we see that more than 3000 transcripts are differentially expressed in both groups, but only half of those transcripts are shared between treatments.
The comparison between NB and IB fed mosquitoes showed that a significant number of upregulated genes in the NB group are associated with iron metabolism and oxidative stress. Previous studies have shown that these functions are important in defense against pathogens and protecting the arthropod against blood-derived factors, such as heme [31]. It is not surprising to find that transcripts related to iron metabolism were upregulated in NB fed mosquitoes since this is the way blood is up taken by mosquitoes. Regarding the oxidative stress in the midgut, previous studies suggest that the presence of heme significantly decreases reagent oxygen species as a protection mechanism induced by the heme overload product of hemoglobin digestion in the mosquito midgut [32]. Nevertheless, a recent study showed that the presence of heme in midgut induces significant changes in the expression of genes associated with energy metabolism and antioxidant activities, as is also shown in our study [31].
Heat-inactivation of plasma is mainly performed to denature complement proteins blocking the activation of all three complement pathways and avoiding complement-mediated signaling in adjacent cells. Other studies have demonstrated that heat treatment of serum reduces phagocytosis and chemotactic signaling in human immune cells [33]. Early research on the effect of temperature on serum reported the formation of heat-labile and heat-stable anticomplementary activity or ACA. The studies revealed that while heat-labile ACA can be completely inactivated through serum heating at 56 • C for 30 min, the heat-stable ACA increases progressively with continued heating [34,35], and it has been attributed to immunoglobulin aggregates [21]. In the current study, the transcripts highly represented as downregulated in the NB group compared to the IB inactivated fed mosquitoes were mainly associated with carbohydrate, lipid, and nitrogenous compounds metabolism, suggesting that heat inactivation of the plasma may decrease signals in these midgut-associated pathways. It is thought that feeding the mosquitoes with heat-inactivated plasma does not produce the same physiological signals as when the mosquitoes ingest an intact blood meal. Further studies are needed to investigate whether thermolabile ACA or other by-products of heat-inactivation induced the changes in transcripts observed in the present study.
One of the most noteworthy transcripts differences induced by active complement was the upregulation of AAEL009681-RB (Rhomboid-like protein) and AAEL029038-RA (Cecropin) with a fold increase higher than 5. Rhomboid proteins are associated with mitochondria homeostasis and cell signaling [36,37]. In the case of the AMP cecropin, approximately ten genes have been reported in Ae. aegypti mosquitoes [38], and different cecropins can be induced by the microbiota or the presence of pathogens [39][40][41]. Our results suggest that human complement plays a role in the modulation of cecropin transcription. A previous study using live Ae. aegypti mosquitoes (in vivo) suggested that activation of complement-related proteins in this species induce the production of AMPs, including cecropins [28]. Other in vivo studies also confirm the importance of Toll-pathway induced cecropins in controlling DENV replication in the mosquito Ae. aegypti [42,43]. Significant downregulation of cecropins upon infection with DENV [36] could represent a potential viral defense mechanism. Moreover, the increased expression of three defensins (AAEL003841-RB, AAEL027792-RA, and AAEL003857-RA) and three cecropins (AAEL029038-RA, AAEL029041-RA, and AAEL029047-RA) in the NB group compared to IB, may indicate that human complement proteins directly influence the immune response of Ae. aegypti and corroborate our previous findings suggesting that activation of human complement impacts the synthesis of AMPs [8].
We also aimed to evaluate the direct effect of two specific human complement proteins, hC3 and hC5a, on mosquito cells. First, hC3 protein is one of the most abundant human complement factors in plasma [44]. The hC3 can be physically adsorbed to tissue surfaces, and the binding to membranes in its fluid form as C3(H2O) may trigger responses in human tissue [45]. Thus, our current study suggests that in the event of C3 interaction with the midgut cells, it may also induce significant transcriptome changes. We observed more than 300 transcripts upregulated and more than 400 transcripts downregulated with an enrichment mainly in functions such as phosphotransferase and oxidoreductase activities.
hC5a has shown an effect on DENV replication in our previous studies [8]. This molecule interacts with human cells through two possible receptors C5aR1 (C5a receptor 1 (CD88)) and C5aR2 (C5a receptor 2 (C5L2)). These receptors are mainly found in bone tissue and immune cells [46,47], where the association is associated with bone metabolism [48], induction of oxidative stress, and modulation of cytokine production in immune cells [49]. Our results show that transcripts modified by treatment with human hC5a are mainly associated with an increase in ion binding, phosphotransferase activity, and nucleotide-binding functions. We also observed decreased functions of actin-binding, cytoskeletal protein binding, GTPase regulator activity, and regulation of catalytic activity. Moreover, our results suggest that hC5a modified a larger number of transcripts than hC3, while only 329 transcripts were communally found modified in Aag2 cells after treatment with both complement proteins. In this regard, it is important to clarify that although both hC3 and hC5a were used in the in vitro experiments at the same concentration that was used in our previous study (1 µg/mL), the molarity in the solution was higher for hC5a (0.09 µM) than for hC3 (0.005 µM). A recent study showed that in patients with severe DENV infection, there is a dysregulation of the complement activity leading to a significant decrease in hC3 and an increase in the anaphylatoxins hC3a, hC4a, and hC5a [26]. Based on our previous studies and the shared IRT between the complement proteins, we hypothesize that mosquitoes feeding on DENV infected patients with severe clinical presentations (leading to higher complement activation) may present less infection and dissemination of the virus than those feeding on viremic asymptomatic individuals. In fact, previous studies suggest that asymptomatic DENV carriers may be more infectious to mosquitoes than those with symptoms at a given viremia level [27] that may be associated with the impact of human complement on the transcriptome of Ae. aegypti mosquitoes. Interestingly, we also observed that the commonly upregulated transcripts are associated with oxidoreductase activity and other enzymatic activities involved in mosquito metabolic processes, such as ion binding and phosphotransferase activity, suggesting a direct effect of human complement proteins in mosquito physiology.
Recent studies showed that the physical presence of blood (infected or uninfected) upon ingestion causes significant midgut distention. In the case of an infectious blood meal, this distention may allow the exit of viral particles into the hemolymph. Additionally, the expression of molecules involved in the degradation/remodeling of the midgut extracellular matrix during the blood-feeding may impact the scape of viruses [50,51]. We speculate that human complement proteins may interact with mosquito cells in two ways, through receptors in the surface of midgut cells before forming the peritrophic matrix ( Figure 6A) or by escaping through the distended midgut cell layer during feeding ( Figure 6B) where hC3 and hC5a molecules in the hemolymph could also interact with hemocytes. A previous study revealed that Aag2 cells are a hemocyte-like cell line and represent an appropriate in vitro model for the study of immune responses of Ae. aegypti mosquitoes [52], whereby we used it to evaluate the potential effect of complement proteins on mosquito immune responses. We believe that these interactions are responsible for the transcript changes observed in this study. Although no specific receptors for any human complement proteins have been identified in mosquito cells so far, these molecules may start transcription changes by binding proteins on the mosquito cells surface or by internalization via pores or endocytosis. In humans, clathrin-dependent endocytosis of the GPCR receptor is activated upon interaction with hC5a [53] and internalization of hC5a and leads to inflammation and oxidative burst [54]. Other studies suggest that hC3 can be internalized via an unknown mechanism, and its internal cleavage and production of hC3a leads to inflammatory responses [55,56]. Binding of hC3 can direct the intracellular route of the cargo modulating immune responses [57]. Mosquitoes may have evolved different routes to dispose of harmful components in the human blood (as noted above) in addition to a mechanism in which the human blood can be an immune buster protecting the arthropod from harmful pathogens.
The current study demonstrates the effect of human complement proteins on mosquito physiology; however, we acknowledge that it has limitations. First, we only measured transcripts at one point in time for both experiments, at 3 h for the in vivo, and at 24 h for the in vitro experiments. In addition, the in vivo experiment was only performed in duplicates. We are planning to expand these experiments and measure additional time points in the in vivo experiments (6, 12, 24, and 48 h) in addition to compare transcripts in DENV infected mosquitoes vs. those in non-infected to determine changes in the transcripts in relation to infection status. We believe that the findings of the current study support previous studies demonstrating the impact of human blood components in the physiology and immune response of arthropod vectors and highlights the importance of using normal non-inactivated human blood component to accurately evaluate vector competence. Furthermore, Ae. aegypti has a large number of transcripts involved in immune system processes; however, there is currently no curated list of these transcripts, to the best of our knowledge. Therefore, the proposed list established in this work of IRTs can serve as a frame of reference to establish a complete and refined list of immune-related genes of Ae. aegypti that can be used by the research community. In conclusion, our study shows that active complement induces significant changes in the transcriptome of Ae. aegypti mosquitoes with an important number of modulated genes involved in immune responses.
vitro model for the study of immune responses of Ae. aegypti mosquitoes [52], whereby we used it to evaluate the potential effect of complement proteins on mosquito immune responses. We believe that these interactions are responsible for the transcript changes observed in this study. Although no specific receptors for any human complement proteins have been identified in mosquito cells so far, these molecules may start transcription changes by binding proteins on the mosquito cells surface or by internalization via pores or endocytosis. In humans, clathrin-dependent endocytosis of the GPCR receptor is activated upon interaction with hC5a [53] and internalization of hC5a and leads to inflammation and oxidative burst [54]. Other studies suggest that hC3 can be internalized via an unknown mechanism, and its internal cleavage and production of hC3a leads to inflammatory responses [55,56]. Binding of hC3 can direct the intracellular route of the cargo modulating immune responses [57]. Mosquitoes may have evolved different routes to dispose of harmful components in the human blood (as noted above) in addition to a mechanism in which the human blood can be an immune buster protecting the arthropod from harmful pathogens. The current study demonstrates the effect of human complement proteins on mosquito physiology; however, we acknowledge that it has limitations. First, we only measured transcripts at one point in time for both experiments, at 3 h for the in vivo, and at 24 h for the in vitro experiments. In addition, the in vivo experiment was only performed in duplicates. We are planning to expand

Blood Source
The Institutional Review Board (IRB) approval for collecting human blood from healthy volunteers was granted by the University of South Carolina IRB (IRB# Pro00045351, approval date: 25 August 2015 Human whole peripheral blood was collected in ethylenediaminetetraacetic acid (EDTA) and processed for use in the experiments. Briefly, plasma was separated from whole human blood by centrifugation, and the red blood cells (RBCs) were washed three times in 1 × PBS and kept at 4 • C until use. For the in vivo experiments, the inactivation of plasma was accomplished by heating the plasma for 30 min at 56 • C. Inactivation of all complement pathways was verified using the Complement System Screen kit (Euro Diagnostica, Malmo, Sweden) according to the manufacturer's instructions.

Mosquito Blood-Feeding
To measure the in vivo effect of human complement inactivation on mosquito physiology, female adult mosquitoes were fed with either heat-inactivated blood plasma (500 µL inactivated plasma + 500 µL of homologous packed RBC), or normal non-inactivated plasma (500 µL normal plasma + 500 µL homologous packed RBC). Mosquitoes were fed for 30 min at room temperature using 1 mL of blood mixture in a Hemotek feeder maintained at 37 • C. Engorged females were sorted in different cages and held under standard conditions. Age-matched sugar-fed mosquitoes were used as controls. Three hours (3 h) post-feeding, whole abdomens were dissected and transferred to 1.5-mL tubes in pools of 10 abdomens. The tissue was homogenized in RLT buffer (Qiagen, CA, USA) with Basal Medium Eagle (BME, Thermo Scientific, Waltham, MA, USA) as the lysis buffer.

Aag2 Cells Exposure to Human Complement Protein hC3 and hC5a
To measure the in vitro effect of the specific human complement proteins C5a and C3, we used the Aag2 Ae. aegypti cell line. Cells were seeded in a 24-well cell culture plate 24 h before the experiment. Cells were treated with 1 µg/mL of either recombinant hC3 (Abcam) or hC5a (R&D systems) dissolved in complete media [8]. Control cells contained complete media only. After 24 h incubation at 28 • C, cells were harvested using the lysis buffer described above.

Mosquito RNA Isolation and Sequencing
Mosquito RNA was extracted using a Quick-RNA miniprep kit (Zymo Research, Irvine, CA, USA). Each in vivo experiment was conducted in duplicates for the feedings with sugar (SF), heat-inactivated (IB) blood, and normal blood (NB) (n = 6). Each in vitro experiment was conducted in triplicates for Aag2 control untreated cells, cells exposed to hC3, and cells exposed to hC5a (n = 9), for a total of 15 samples sequenced. A total of 3 ng of RNA per sample was sent for sequencing to LC Sciences (Houston, TX, USA), where the sample QC, library preparation (with enrichment for mRNA using poly(A) selection, single-molecule clonal amplification), and Illumina sequencing (San Diego, CA, USA) was performed.
The package NOIseq v2.31.0 [63] was used to perform differential expression analysis in the R software environment (https://www.r-project.org/), implementing the non-parametric approach NOIseqBio that improves the handling of biological variability specific to each gene, and is very successful in controlling the high false discovery rate (FDR) in experiments with biological replicates. The count filter was used to remove transcripts with Counts per Million (CPM) <1 in the samples, which avoids noise from lowly expressed transcripts. The normalization method used was the Trimmed Mean of M-values (TMM) approach. To identify those differentially expressed (DE) between comparisons, the q-value cut-off 0.95 was implemented, and only protein-coding transcripts with a fold change ≥4 (Log2FC ≥2) were analyzed.
The gene description and gene ontology (GO) terms of DE transcripts were obtained from VectorBase.org using the BiomaRt package v.2.44.1 [64,65]. The Gene Set Enrichment Analysis (GSEA) was performed using g: Profiler (https://biit.cs.ut.ee/gprofiler/gost). The transcripts related to immune response (IRTs), which include immunity and other processes, such as small regulatory RNAs, apoptosis, and autophagy, were identified considering the list of 477 transcripts obtained by in silico comparative genomic analyses and manual annotation that have established or putative associations with defense mechanisms published by Bonizzoni et al. [66]. Because this list of transcripts was generated with an older genome assembly (AaegL1.2), only 243 of the 477 transcripts were found in the current assembly (AaegL5.2). With VectorBase.org, it was possible to convert some of the old IDs to the new ones; others have been lost. To look for new IRTs in the current assembly, the following functional annotations were used: GO:0002376 (immune system process), GO:0006955 (immune response), GO:0045087 (innate immune response), GO:0006959 (humoral immune response), GO:0006952 (defense response), GO:0006915 (apoptotic process), GO:0042981 (regulation of apoptotic process), GO:0008592 (regulation of Toll signaling pathway), GO:0008236 (serine-type peptidase activity), GO:0004252 (serine-type endopeptidase activity), GO:0004867 (serine-type endopeptidase inhibitor activity), GO:0008234 (cysteine-type peptidase activity), and GO:0004869 (cysteine-type endopeptidase inhibitor activity). The updated and curated list of IRTs contains 883 transcripts (Table S3).
The GSEA plots were generated using the ggplot2 package v3.3.0 [67,68]. The heatmaps, clustered on expression profiles, were created to visualize changes in the experiment's profiles between transcripts. The z-score, shown as the scale key on each figure's side, represents the normalized gene expression measurements, where 0 has no difference from the mean. To illustrate the transcripts' expression, red color represents positive fold change (FC), indicating higher expression in the corresponding transcripts, and blue color represents negative FC, indicating decreased expression. The heatmaps were generated using the Pheatmap package v1.0.12 (https://cran.r-project.org/web/pac kages/pheatmap/index.html). The Venn diagrams represented the number of differentially expressed transcripts when comparing the different treatments (FC ≥ 4 and q-value > 0.95) and were generated using the Venneuler package v1.1-0 (http://www.rforge.net/venneuler/index.html).