Modulation of Anopheles stephensi Gene Expression by Nitroquine, an Antimalarial Drug against Plasmodium yoelii Infection in the Mosquito

Background Antimalarial drugs may impact mosquito’s defense against Plasmodium parasites. Our previous study showed nitroquine significantly reduced infection of Anopheles stephensi by Plasmodium yoelii, but the underlying mechanism remains unclear. In order to understand how transmission capacity of An. stephensi was affected by nitroquine, we explored the transcriptome of adult females after different treatments, examined changes in gene expression profiles, and identified transcripts affected by the drug and parasite. Methodology/Principal Findings We extended massively parallel sequencing and data analysis (including gene discovery, expression profiling, and function prediction) to An. stephensi before and after Plasmodium infection with or without nitroquine treatment. Using numbers of reads assembled into specific contigs to calculate relative abundances (RAs), we categorized the assembled contigs into four groups according to the differences in RA values infection induced, infection suppressed, drug induced, and drug suppressed. We found both nitroquine in the blood meal and Plasmodium infection altered transcription of mosquito genes implicated in diverse processes, including pathogen recognition, signal transduction, prophenoloxidase activation, cytoskeleton assembling, cell adhesion, and oxidative stress. The differential gene expression may have promoted certain defense responses of An. stephensi against the parasite and decreased its infectivity. Conclusions/Significance Our study indicated that nitroquine may regulate several immune mechanisms at the level of gene transcription in the mosquito against Plasmodium infection. This highlights the need for better understanding of antimalarial drug’s impact on parasite survival and transmission. In addition, our data largely enriched the existing sequence information of An. stephensi, an epidemiologically important vector species.


Introduction
Malaria parasites go through major transitions of differentiation as they cross tissue barriers of the mosquito with dramatic number changes [1][2]: between gametocytes and ookinetes, between ookinetes and mature oocysts, and between midgut and salivary gland sporozoites [3]. The most critical bottleneck of Plasmodium development occurs during ookinete invasion of the midgut epithelium, prior to oocyst development on the basal lamina [4]. Digestive enzymes and mosquito defense proteins are partly responsible for the parasite loss at this stage [5][6]. The latter includes pathogen recognition proteins, serine proteases, phenoloxidases, antimicrobial peptides, and others.
External factors may also affect mosquito susceptibility to parasite infection, including ingested antibodies, subsequent blood feeding, and antimalarial drugs in blood meals. Among them, effects of antimalarial drugs on infectivity of parasites in mosquitoes are well documented. Chloroquine mainly targets Plasmodium at the erythrocytic stage. Nitroquine (CI-679 or 2,4diamino-6-[(3,4-dichlorobenzyl)nitros-amino]quinazoline) caused nuclear and cytoplasmic damage in asexual forms of P. berghei and P. cynomolgi [7]. While there was no further report on this compound from Western countries, safety and efficacy of nitroquine were tested in rodents, chicken, primates, and humans [8]. Nitroquine is highly effective against P. yoelii, P. gallinaceum, and P. cynomolgi at the erythrocytic and exoerythrocytic stages. It interferes with structure and function of the cytoplasm and nucleus of P. yoelii exoerythrocytic forms [9]. The action mechanism of nitroquine involves inhibition of Plasmodium DNA and protein synthesis [10][11]. After further pharmacological and toxicological tests, nitroquine was used in clinical trials on malaria patients after 1973. In nine provinces and fifteen regions, this drug successfully cured 11,407 patients in the field tests [8]. Nitroquine and chloroquine were equally effective against P. falciparum. Chloroquine ingestion by mosquito at the time of blood feeding associates with an increase in parasite numbers in the insect host [12]. In contrast, ingested nitroquine leads to a decrease in Plasmodium number and infectivity [13]. Using the model system of Plasmodium yoelii and Anopheles stephensi, we demonstrated that nitroquine induced transcription of a few mosquito genes encoding pattern recognition receptors, signal transducers, cell adhesion molecules, and oxidative stress proteins.
An. stephensi is a major malaria vector in the Indian subcontinent [14]. Rapid development and urbanization in this region has led to increase in the mosquito population resulting in frequent malaria epidemics [15]. While recent malaria outbreaks occurred at a higher frequency, mortality became considerably lower. For example, during 2003, only 1006 of the reported 1.78 million cases in India caused deaths but reasons for the mortality reduction are unclear [16]. As an important disease vector, An. stephensi has not yet been intensively investigated at the molecular level, which hinders the elucidation of mechanisms for various physiological processes such as how immunity is possibly affected by nitroquine.
Although generation and analysis of cDNA clones from midgut tissue of adult female An. stephensi yielded useful sequences [17], the acquired information is limited by the method used and additional experiments are needed to quantify their mRNA levels in different stages or tissues. Recently, a high throughput method is established to efficiently discover genes along with their expression profiles using next-generation sequencing technology [18]. Without resorting to a reference genome and thereby directly uncover process-related gene expression, the study revealed over 103 differentially regulated defense genes in Manduca sexta.
To better understand the impact of antimalarial drugs on mosquito's defense against Plasmodium parasites, we designed a study to discover genes with significantly altered mRNA levels in naïve versus P. yoelii-infected adult females of An. stephensi fed on nitroquine-treated or -untreated mice by the RNA-Seq approach. We largely increased the throughput of sequencing by adopting the Illumina technology and studied effects of the infection and drug on mosquito gene expression.

Ethics Statement
This study was carried out in strict accordance with the recommendations in the Guide for the Ethics Committee of Third Military Medical University in China. The protocols involving mice were approved by the committee and performed under anesthesia to minimize suffering. Sodium pentobarbital was used to induce anesthesia.
Mosquitoes Rearing, Nitroquine Treatment, and P. yoelii Infection An. stephensi (Hor strain) mosquitoes were raised at 24uC, 75% humidity under a 12:12 light-dark cycle and maintained on a 5% sucrose solution during adult stage. Female BALB/c mice were inoculated intraperitoneally with 200 ml of infected blood containing about 1610 7 P. yoelii (By265-GFP) parasitized erythrocytes obtained from a donor mouse with about 10% parasitemia. The course of infection was followed by examining Giemsastained blood smears prepared from tail blood samples collected at different times after inoculation. When the gametocitaemia reached 1%, a curative dose of nitroquine (12 mg/kg) was administered intragastrically to the mice. At 4 h after the drug treatment, gametocytaemia and parasite exagellation were confirmed. The female mosquitoes were fed on anesthetized mice for 2 h and collected at 24 h after blood feeding. Four-day-old adult females (25 per group) were allowed to blood fed on one of the following four groups: uninfected BALB/c mice treated with nitroquine (12 mg/kg) (UD, uninfected mice treated with nitroquine) for 4 h, P. yoelii-infected mice treated with the drug (12 mg/kg) (ID, infected mice treated with nitroquine) for 4 h, uninfected mice treated with buffer without nitroquine (UB, uninfected mice treated with buffer) for 4 h, or P. yoelii-infected mice treated with buffer only (IB, infected mice treated with buffer) for 4 h. Unfed mosquitoes were removed from the groups. The oocyst number in midgut was measured at 10 days post-blood feeding, and each assay was done with at least 25 mosquitoes, and the data represent three independent experiments. Difference of infection intensity between groups was analyzed by paired sample t-test (Prism 6.01, GraphPad Software, Inc.).

RNA Extraction, Library Construction and Sequencing, and Read Assembling
Each of the four groups of female mosquitoes was collected 24 h after blood feeding, washed in ice-cold 95% ethanol to remove cuticle lipids, and rinsed in ice-cold water. All carcasses from the same group were combined, frozen in liquid nitrogen, and ground into fine powder for RNA extraction using TRIzol reagent (Invitrogen). mRNA was separately purified from the total RNA samples (1.0 mg each) by binding to cellulose in Mag-Bind mRNA Enrichment Kit (Omega Bio-Tek). For cDNA synthesis, mRNA was reverse transcribed with MMLV-RT (Promega) in the presence of oligo(dT) 15 primer. Paired-end libraries were constructed from the four groups (UD, ID, UB, and IB) and sequenced at Macrogen Inc (Korea), following Illumina specifications. The four cDNA pools were sequenced on one lane for 101 cycles from both ends on Illumina GA-IIx HiSeq2000.

De Novo Assembly of Transcriptomes
Filters were applied to remove low quality reads with .33% N's (indetermination), .33% A's from the 59 end (or T's from the 39 end) suggestive of poly-A tail, or .34% nucleotides with low Phred quality scores (,20 i.e. 1% error), according to Crawford et al [19]. After adaptor trimming, Velvet [20] was employed to assemble the remaining reads in each of the four libraries at different hash lengths (k: 29,35,41). All contigs from the three exploratory assemblies were summarized by clustering using CD-HIT [21] to generate four datasets of UD, ID, UB, and IB. The default threshold of 90% was used as identity cutoff. In addition, all retained reads after filtration in the four libraries were assembled at four hash lengths (k: 21,35,49,59). All contigs in the four exploratory sets were further assembled into one dataset designated ''UIDB'' (k: 39). Note that this ''assembly of assemblies'' may contain some misassembled contigs. The UIDB contigs were used as queries to search all insect sequences deposited at GenBank (http://ncbi.nlm.nih.gov/) and An. stephensi EST sequences [17] using BLASTX and BLASTN at a cutoff Evalue of 1610 25 . For discovering process-related genes by quantifying their mRNA level changes, numbers of the UD, ID, UB, and IB reads assembled into each UIDB contig were extracted from the Velvet output and tabulated using Microsoft Excel.

Read Normalization and Ratio Calculation
Read normalization and ratio calculation were performed as described by Zhang et al [18]. Briefly, based on frequencies of commonly used standards in each of the four libraries (e.g. number of rpS3 reads in UD 4 number of total reads in UD), a set of six ribosomal protein genes were selected as internal standards, which had high total read numbers (.10,000) and low coefficients of variation (i.e. SD/mean ,10%) in their frequencies. The sums of their read numbers for specific libraries, or library normalization factors (LNFs), which already reflected the differences in library sizes, were directly used to calibrate other read numbers in the corresponding libraries. For a specific contig in UIDB, its relative abundance (RA) in libraries X and Y is defined as: RA x/y = (actual read # in library X 4 LNF x )/(actual read # in library Y 4 LNF y ). In case read # in library Y is zero, adjusted read number (ARN) is calculated as: ARN x = actual read # in library X 6 LNF y /LNF x . Some of the contigs in UIDB, whose RAs are above a threshold, are categorized as infection induced (UPi: RA ID/UD or RA IB/UB .2), infection suppressed (DNi: RA ID/UD or RA IB/UB , 0.5), nitroquine induced (UPn: RA ID/IB or RA UD/UB .2), and nitroquine suppressed (DNn: RA ID/IB or RA UD/UB ,0.5). These contigs were used as queries to search insect sequences including An. stephensi ESTs, as described in Section 2.4.

Quantitative Real-time PCR Analysis
Quantitative real-time PCR was performed to confirm the RNA-Seq expression profiling by using the same total RNA samples. Eight immunity-related genes were selected for validation of the expression data, including TEP1, APL1C, PGRP, FBN8, Eater, C-type lectin, CLIPC7, and SP (Fig. S1). Two cDNA samples from a similar experiment were examined by quantitative real-time PCR to further verify the RNA-Seq data. In that experiment, the female mosquitoes were fed on P. yoelii-infected mice or uninfected mice for 2 h and collected for RNA isolation at 24 h after blood feeding. Specific primers and SYBR Premix EX Taq (TaKaRa, Japan) was used in real-time PCR analysis on an ECO TM Real-Time PCR System (Illumina, USA). The reaction mixture (15 ml total volume) contained 7.5 ml 26reaction buffer, 0.45 ml primers, 6.05 ml ddH 2 O and 1 ml cDNA template. The ribosomal protein S7 (rpS7) mRNA (GenBank AF539918) was used as an internal control for data normalization with the forward and reverse primers (Fig. S1).

Identification of Differentially Regulated Genes
To study the effect of P. yoelii infection (I) and nitroquine (D for drug) on An. stephensi gene transcription, we isolated mRNAs from adult female mosquitoes in the treatment (I and D) and control (U for uninfected, B for buffer) groups. Using oligo-dT primer annealing to the 39 end of mRNA molecules, we generated four cDNA libraries: UD, ID, UB, and IB. To increase the read number and sequence coverage, we sequenced the libraries on an Illumina GA-IIx HiSeq2000. After removing ,2% of the reads flagged as low quality or low complexity, we obtained a total of 25,505,495 reads from UD, 25,582,317 from ID, 22,561,525 from UB, and 21,145,138 from IB (Table 1). We established 'exploratory' assemblies of the paired end reads in each library at three hash lengths.
To quantify changes in gene expression, we also assembled reads from all the four groups into one set of contigs (UIDB) ( We extracted numbers of UD, ID, UB, and IB reads assembled into each UIDB contig. As read numbers directly correlated with library sizes and, therefore, had to be normalized first against control genes. Examination of the frequencies of commonly used internal standards in each of the four libraries resulted in a short list of six genes with low coefficients of variation (,10%) and high total read numbers (.10,000). We used the sums of their read numbers as library normalization factors (LNFs) to calibrate read numbers and calculate relative abundances (RAs). Based on the RA values, 3,046 or 9% of the 32,648 contigs in UIDB were categorized into four groups: infection induced, infection suppressed, nitroquine induced, and nitroquine suppressed. The total reads for each of the 3,046 contigs is greater than 1,000, so that low abundant transcripts (and, therefore, less reliable RA values) are eliminated from our study. To validate the RAs, we performed quantitative real-time PCR on eight genes using the same total RNA samples and found the fold changes were in good agreement in all these groups (Fig. S1). Similar results were obtained using cDNA samples from a different experiment. In the nitroquine induced and suppressed groups, we have found a total of 848 contigs. Among them, 550 contigs (65%) do not have BLAST hits and 111 are related to immunity, oxidative stress, detoxification, cytoskeleton assembling, or cell adhesion. In comparison, there are 2,198 contigs in the infection induced and suppressed groups: 1,806 (82%) have no BLAST hit and 109 are related to the above five processes. The low hit rates may stem from the cDNA synthesis and sequencing methods used.
The Illumina sequencing technology greatly increases read numbers and hence accuracy of RA values, but this comes at a cost. Short read length complicates their assembling into contigs with long coding regions that are highly desirable for species lacking sequenced genomes. Using random primers should alleviate this problem to certain extent, since they favorably bind to GC-rich coding sequences. In contrary, the oligo-(dT) primer causes a bias for contigs containing 39 untranslated region, which contributes to the low BLAST hit rates of 18-35%. When the An. stephensi genome is published, we will overlay all our contigs to the gene models and expand the transcriptome search not only to the groups studied here but also to other processes affected by the drug or infection.

Sequence Analysis and Function Prediction of Nitroquine-induced Genes
We discovered 356 UIDB contigs whose RA ID/IB or RA UD/UB values were higher than 2. As anticipated, some of these contigs encoded polypeptides either similar to defense proteins identified in An. gambiae, Aedes aegypti, Culex quinquefasciatus and other insects (e.g. Drosophila melanogaster) (Table 3), or related to proteins previously not known to participate in immune responses, or having no significant sequence similarity to known proteins. In the following sections, we describe them in the order of their putative functions.
Key components of the protease cascades include clip-domain SPs/SPHs (CLIPs), which take part in several defense mechanisms in insects and crustaceans such as the activation of signaling pathways leading to the synthesis of antimicrobial peptides [32], hemolymph coagulation [33], and melanization [34]. We identified a nitroquine-induced homolog of A. gambiae CLIPC7 (contig 2706). So far, An. gambiae CLIPB14 and CLIPB15 are found to be responsive to bacterial or Plasmodium infection: B14 showed persistent up-regulation in Plasmodium-infected mosquitoes while B15 showed transient up-regulation during midgut invasion [35][36]. Signal transduction mediated by extracellular SP pathways is down regulated by serpins, which are irreversible, suicide inhibitors that covalently bind to the active site Ser of their target SPs [37]. We have identified one nitroquine-induced serpin (contig 3479).
2.3. Intracellular signal transducers. Intracellular immune signaling pathways transmit the alarm signal originated from pathogen-associated PRRs to effector genes. We found two ankyrin repeat protein (ANK) contigs (16368, 11526). Ankyrin repeat is a 33-residue motif often occurring in tandem arrays that cooperatively fold into structures for molecular recognition via protein interactions. An. gambiae REL2 (a Relish homolog containing an ANK domain) regulates expression of antimicrobial peptide genes cec1, cec3, and gam1 [38]. Spectrins are components of G-protein coupled receptor (GPCR) and synaptic multiprotein complexes involved in cell cycle by regulating expression of membrane receptors [39]. Invertebrates have a small repertoire of spectrin genes. We found aspectrin (contig 16565), G-protein components (contigs 12518 and 4877), as well as arrestins (contigs 517 and 3493). Arrestins mediate cellular processes via interactions  with secondary signal transduction cascades by means of recruiting and activating mitogen-activated protein kinase (MAPK) and other effectors [40][41]. Other intracellular proteins possibly involved in signal transduction or modulation include a GTP/ GDP exchange factor. 2.4. Oxidative stress and detoxification proteins. Several genes for oxidative stress and detoxification were up-regulated by nitroquine, including sorbitol dehydrogenase (contig 57082, RA ID/IB : 2.99, RA UD/UB : 2.73), 14-3-3 epsilon protein (contig 12429), three cytochrome P450 contigs (30351, 10946, 24943).

2.5.
Cytoskeleton and cell adhesion molecules. Cytoskeletal dynamics and remodeling are considered as key factors affecting P. berghei invasion of the mosquito midgut [12]. Nitroquine affected the expression of some cytoskeletal and adhesion proteins. A highly conserved, transforming growth factor (TGF) -b inducible matrix protein (contig 6349, RA ID/IB : 6.18) may be a bifunctional linker between individual matrix components and resident cells [42]. Six innexin genes in Ae. aegypti are closely related to their Drosophila homologs, suggesting critical roles of gap junctions in diverse cellular and tissue functions [43]. We identified an innexin gene (contig 7935, RA ID/IB : 5.28, RA UD/UB : 2.44) in An. stephensi, which was highly induced after the drug treatment. A putative formin (contig 33424, RA ID/IB : 2.35, RA UD/UB : 7.83) may govern microtubule and microfilament dynamics by functioning as an effector of r small GTP-binding proteins during cell adhesion, cytokinesis, polarization, and morphogenesis [44]. Titin (contigs 484, 14623, and 14852), the third most abundant protein in muscle (after actin and myosin), is responsible for much of the myofibril elasticity in various organisms. Titin molecules are associated with each thick filament in skeletal muscle, where they interact with myosin [45]. Ten myosin contigs were up-regulated (Table 3). Gliotactin (contig 2492), first identified in the tricellular junction, is necessary for the development of tricellular junction and septate junctions [46]. Cadherin (contig 5828, RA UD/UB : 6.50) is a cell surface glycoprotein that may act as a receptor for envelop proteins of dengue or West Nile viruses when they enter mosquito cells [47]. Two kakapo protein contigs (2452, 302) encode a giant cytoskeleton protein that has multiple isoforms with characteristics of the spectrin and plakin superfamilies. Previously characterized short isoforms are similar to spectrin and dystrophin with an actinbinding domain followed by spectrin repeats [48].
In the previous study [13], at five days post-feeding on P. yoelii infected mice, mosquitoes were continually fed on 0.1% nitroquine in sucrose solution prior to collection at 7, 9, and 13 days post infection for morphology, PCR, and PO activity tests. Under those conditions, we observed melanized oocysts and an increase in PPO mRNA levels in the nitroquine-treated mosquitoes. Here, we administered the drug to P. yoelii-infected mice at a dosage similar to that for people in malaria endemic areas, but did not observe melanized ookinetes and detected a decrease in PPO mRNA levels. A likely reason for the discrepancy is the low dosage and no drug administration after the blood meal, which caused a milder immune response that hampered but did not completely block the parasite development. There was no clear difference in gametocyte The total read numbers, if lower than 300, are shown in italic and bold so that their RAs need to be interpreted with caution. *:adjusted read number (ARN). The footnote also applies for Tables 4-6. doi:10.1371/journal.pone.0089473.t003 Table 4. A list of 60 nitroquine-suppressed contigs for pathogen recognition, signal transduction, effector, oxidative stress, detoxification, cytoskeleton, and cell adhesion proteins. number or morphology between the ID and IB groups after treatment (data not shown). Nonetheless, we detected significant decrease in median oocyst count at 10 days post-blood feeding (Fig. S2), indicating that pre-oocyst stages (including those in mouse) were affected by low dosage of nitroquine but stayed viable for a period of time. After all, nitroquine did cause complex changes in gene expression.

Genes Induced by Plasmodium Infection
4.1. Pattern recognition receptors. While mRNA levels are up-or down-regulated after nitroquine treatment, we also examined how parasite infection may affect transcription (Tables 5  and 6) and identified overlaps with the contigs in Tables 3 and 4. Two FREP contigs (54250, RA ID/UD : 5.99; 9243, RA IB/UB : 5.20) and one Eater (contig 9561) were up-regulated in infected mosquitoes but down-regulated in drug-treated ones. We found one C-type lectin (contig 4534), one complement 4b-like protein (contig 4650), and one TEP (contig 13138). The A. gambiae genome contains 19 TEP genes [49]. TEP1 is a typical one characterized in some detail [53]. It is unclear whether all mosquito TEPs are engaged in similar functions. An. gambiae TEP3 and TEP4 are up-regulated upon bacterial challenge or parasite infection [54]. Two LRR protein contigs (2407 and 4502) had RA IB/UB of 2.07 and 2.38, respectively.

4.2.
Other up-regulated immunity-related genes. Similarly, nine PPO contigs (7548, 10561, 24062, 18477, 29061, 24063, 39231, 3605, 3163) were up-regulated in infected mosquitoes (Table 5). There are over 20 PPOs identified in Ae. aegypti, An. gambiae, An. stephensi, An. culicifacies and Ar. subalbatus, with multiple PPOs in each species [55]. We also found three SP contigs were induced by parasite infection (contigs 479, 20619, 2323). Several intracellular proteins possibly involved in signal transduction/modulation. These include two GTP/GDP exchange factors, two ANK proteins and one MAPK. One defensin (contig 22126, RA ID/UD : 3.09) was up-regulated in P. yoelii-infected mosquitoes, which was down-regulated after nitroquine treatment. 4.3. Oxidative stress and detoxification proteins. Ten oxidative stress-responsive and detoxification genes were upregulated: 7 cytochrome oxidase contigs (37606, 10946, 12511, 12532, 3114, 10811, 14692), a peroxidase (contig 18603), an oxidase (contig 4356), an esterase (contig 10978) and a 14-3-3 epsilon protein (contig 12429). Reactive oxidative species (ROS) can cause cell damage and, hence, are rapidly inactivated in cells by oxygen scavengers and reduction reactions. The inactivation is facilitated by enzymes such as peroxidases and cytochrome c [56]. It is intriguing that transcription of these enzymes was downregulated after nitroquine treatment but up-regulated in P. yoeliiinfected mosquitoes, indicating that nitroquine may suppress the production of these enzymes. Insufficient reduction of ROS may cause damage in parasite by reacting with its lipids, proteins, and nucleic acids. The imbalance of free radical-generating andscavenging mechanisms then led to loss of homeostasis and Plasmodium death. The involvement of ROS in mosquito immunity against bacteria and Plasmodium was investigated in the malaria vector An. gambiae [57]. The production of ROS is thought to contribute to the refractory phenotype, as experimentally elevating ROS in susceptible mosquitoes makes them more refractory [58]. Similarly, a variety of synthetic insecticides are known to suppress the activity of key reducing enzymes [59][60].
The high oocyst number in midgut (Fig. S2) and the RAs of infection-affected genes suggest that P. yoelii (By265) and An. stephensi (Hor) form a highly compatible vector-parasite pair in which the host immune system is no longer effective. Jaramillo-Gutierrez et al. [64] showed that silencing several genes involved in oxidative (OXR1 and GSTT1) or immune (LRIM1 and CTL4) responses had no effect on P. yoelii (17XNL) infection of An. stephensi (Nijmegen Sda500). In this study, nitroquine induced immune responses of An. stephensi (Hor) by unknown mechanism. In nitroquine-treated mosquitoes, mRNA levels of TEP1, APL1C (contig 26579), and other LRR proteins (contigs 38607, 695, and 51750) increased (Table 3). The Plasmodium infection induced FREPs, TEP2, and other PRRs (Table 5) but suppressed APL1C (contig 26579) and the LRR protein (contig 38607) ( Table 6). While these observations are interesting, we cannot conclude whether this fluctuation of different PRR mRNA levels is responsible for the reduction of P. yoelii infectivity in nitroquinetreated mosquitoes. Functional data from RNAi experiments may inform us more on the roles of these PRRs in governing the fate of P. yoelii in the mosquito. We are also cautioned by the opposite effects of chloroquine and nitroquine on mosquito response to Plasmodium infection, as well as the differences caused by sampling time and method of drug administration. It is desirable to assess whether chloroquine, nitroquine, and other antimalarial drugs have similar effects on the mosquito response at malaria endemic regions.

Conclusion
Our study shows nitroquine has a significant impact on transcript abundances of certain mosquito genes implicated in defense against Plasmodium, which encode pathogen recognition receptors, signal transducers/modulators, and cytoskeleton-, adhesion-, and oxidative stress-related proteins. Such changes may stimulate the innate immune system of An. stephensi to fight P. yoelii infection. Even though we do not understand the mechanisms of this drug effect, its impact on the mosquito is evident from the present and previous studies. Together, the results suggest nitroquine induces the production of LRIM1/APL1C/TEP1 complex and down-regulation of enzymes involved in ROS reduction, enhances the attack on parasites, and decreases the Plasmodium infectivity. Mechanisms for the infectivity decrease may be distinct, when the time or method of nitroquine administration is different. Figure S1 Confirmation of the RNA-Seq expression profiles by quantitative real-time PCR using the same RNA samples (ID, UD, UB, and IB) (A) and two different RNA samples from a similar test (B).

Supporting Information
(DOC) Figure S2 Comparison of oocyst counts in the infected mosquitoes fed on nitroquine-and buffer-treated mice. (DOC)