Assessing Aedes aegypti candidate genes during viral infection and Wolbachia‐mediated pathogen blocking

Abstract One approach to control dengue virus transmission is the symbiont Wolbachia, which limits viral infection in mosquitoes. Despite plans for its widespread use in Aedes aegypti, Wolbachia's mode of action remains poorly understood. Many studies suggest that the mechanism is likely multifaceted, involving aspects of immunity, cellular stress and nutritional competition. A previous study from our group used artificial selection to identify a new mosquito candidate gene related to viral blocking; alpha‐mannosidase‐2a (alpha‐Mann‐2a) with a predicted role in protein glycosylation. Protein glycosylation pathways tend to be involved in complex host–viral interactions; however, the function of alpha‐mannosidases has not been described in mosquito–virus interactions. We examined alpha‐Mann‐2a expression in response to virus and Wolbachia infections and whether reduced gene expression, caused by RNA interference, affected viral loads. We show that dengue virus (DENV) infection affects the expression of alpha‐Mann‐2a in a tissue‐ and time‐dependent manner, whereas Wolbachia infection had no effect. In the midgut, DENV prevalence increased following knockdown of alpha‐Mann‐2a expression in Wolbachia‐free mosquitoes, suggesting that alpha‐Mann‐2a interferes with infection. Expression knockdown had the same effect on the togavirus chikungunya virus, indicating that alpha‐Mann‐2a may have broad antivirus effects in the midgut. Interestingly, we were unable to knockdown the expression in Wolbachia‐infected mosquitoes. We also provide evidence that alpha‐Mann‐2a may affect the transcriptional level of another gene predicted to be involved in viral blocking and cell adhesion; cadherin87a. These data support the hypothesis that glycosylation and adhesion pathways may broadly be involved in viral infection in Ae. aegypti.


INTRODUCTION
The mosquito Aedes aegypti plays an extensive role in transmitting the following viruses: dengue (DENV), Zika (ZIKV), chikungunya (CHIKV) and yellow fever, which have serious health consequences in humans globally (Gould et al., 2017). Without effective, licensed vaccines, prevention of arboviral diseases is focused on mosquito control via habitat reduction and the use of insecticides. Such interventions are difficult to implement in regions of impoverishment, urbanization and poor sanitation (Fauci & Morens, 2016). Furthermore, insecticide resistance in mosquitoes has hampered these efforts and assisted in the increase in dengue cases (Dusfour et al., 2019). It is estimated that yearly 390 million become infected with dengue, with 96 million experiencing pathological symptoms, and efforts to create a vaccine have faced difficulties due to the circulation of four different dengue serotypes across the world (Izmirly et al., 2020).
One alternative means of disease control is through the use of the bacterium Wolbachia, which is being released globally to decrease virus transmission to humans through population 'suppression' and population 'replacement' (Ferreira et al., 2020). Wolbachia are endosymbiotic bacteria that naturally infect arthropods and are maternally transmitted to offspring (Zug & Hammerstein, 2012). Wolbachia species native to a range of other insects have been introduced into the naturally Wolbachia-free Ae. aegypti where they have formed stably inherited infections (McMeniman et al., 2009;Xi et al., 2005). 'Population suppression', involving the release of Wolbachia-infected males whose offspring are not viable when mated with wild females, has led to reductions in mosquito populations where trialled in the United States and China (Crawford et al., 2020;Zheng et al., 2019).
'Population replacement' is another control approach and is achieved by releasing female Ae. aegypti that pass on Wolbachia to their offspring, which then spreads through the population. Female mosquitoes infected with Wolbachia harbour reduced dengue viral loads, exhibit longer extrinsic incubation periods or fail to transmit virus in the saliva all together (Moreira et al., 2009;Ye et al., 2015). Field trials of replacement strategies have shown some efficacy in reducing the incidence of dengue fever in humans (Carrington et al., 2018;Nazni et al., 2019;Utarini et al., 2021). Wolbachia also limits the replication of other arboviruses in mosquitoes, including CHIKV, ZIKV, yellow fever and West Nile (WNV), suggesting a fundamental shared mechanism that could be useful for many diseases (Joubert & O'Neill, 2017;Kambris et al., 2009;Moreira et al., 2009;van den Hurk et al., 2012).
The potential use of Wolbachia population replacement for biocontrol by pathogen blocking is immense. While the mechanisms of blocking are still poorly understood, there is a growing consensus that it is multifaceted (Lindsey et al., 2018). Understanding how mosquitovirus-Wolbachia genetic interactions dictate pathogen blocking is necessary to assure the efficacy of the biocontrol agent across a global landscape of mosquito and viral genetic diversity. In addition, understanding the mechanism will be key to developing strategies to counter the evolution of resistance against Wolbachia or its effects (Joubert et al., 2016), either in the mosquito or the virus. Our own work has shown that we can rapidly artificially select for differences in DENV blocking strength in mosquitoes collected from a single field release site (Ford et al., 2019), demonstrating the importance of mosquito genetic diversity in the trait's expression. Genome-wide association (GWA) in selected lines exhibiting stronger and weaker blocking revealed a set of candidate mosquito genes that underpin blocking phenotypes. One of the top candidate genes was alpha-mannosidase-2a (alpha-Mann-2a). Alpha-mannosidase proteins cleave mannose residues on glycoproteins as they move through the secretory pathway in the endoplasmic reticulum (ER) and Golgi, creating the substrate for the next step in a complex process of N-linked glycosylation (Stanley, 2011). In the ER, alpha-mannosidases also recognize misfolded proteins and facilitate degradation through the ER-associated degradation pathway (Hosokawa et al., 2001). The predicted ortholog to alpha-Mann-2a in Drosophila, alpha-Man-II-b, has also been demonstrated to have deglycosylation activity, though its role is unknown, and its increased variation compared with other Drosophila alpha-mannosidases suggests a deviation in its function (Rosenbaum et al., 2014).
Alpha-mannosidases have not directly been associated with DENV infection; however, the expression of an alpha-mannosidase-1 positively correlated with Hepatitis B virus during infection (Jiang et al., 2016), and a hantavirus protein colocalizes with mammalian alpha-mannosidase-2 protein during infection (Ravkov & Compans, 2001). Another possibility is that alpha-Mann-2a may affect the glycosylation of viral proteins rather than those of host origin. Glycoproteins are a major component of viral surface proteins and are important for CHIKV and DENV attachment to cells, and DENV secretes non-structural glycoproteins (Banerjee & Mukhopadhyay, 2016). With no functional evidence of the role of this gene in mosquitoes, further investigation during infection is required. Since the key single nucleotide polymorphisms (SNPs) in alpha-Mann-2a identified in our GWA were located in the non-coding regions (Ford et al., 2019), we sought to manipulate the expression of this gene to see the most effect. We therefore used RNA interference (RNAi) to assess whether reduced expression of alpha-Mann-2a had any impact on the DENV and CHIKV infection loads in mosquitoes. This gene may offer a potential target for genetic modification in the effort to create virus refractory mosquitoes.

The effect of DENV and Wolbachia infection on alpha-Mann-2a expression
We hypothesized that infection may regulate the expression of alpha-Mann-2a and so we measured the expression in both Wolbachia-free (WÀ) and wAlbB mosquitoes (W+) that consumed either DENVinfected (D+) or DENV-free blood (DÀ). The midgut, salivary gland and remaining tissues (remainder) were collected at 3, 7 and 14 days post-feeding for each treatment group. Expression of the candidate gene was normalized to the housekeeping gene Rps6. The expression levels of alpha-Mann-2a were regulated by tissue, time and DENV infection status, but Wolbachia had no effect (mixed effects model: tissue, time and infection p < 0.0001, Wolbachia p = 0.44, Table S1).
Overall expression was lowest in the midgut and highest in the salivary glands, where on average expression levels at 7-and 14 days post-feeding on blood without DENV were at least 100-fold higher than in the other tissues ( Figure 1). In the midgut, the impact of DENV was limited to 14 days post-feeding, where expression in both wild-type and Wolbachia-infected mosquitoes decreased more than four-fold in the presence of DENV when compared with uninfected blood fed mosquitoes (post hoc Tukey time Â infection: p < 0.0001).
Salivary gland gene expression trends were similar to that of the midgut, with DENV-infected mosquito lines having 2.5-to 3-fold lower alpha-Mann-2a expression at 7 and 14 days post-feeding, compared with uninfected blood fed mosquitoes (post hoc Tukey time Â infection: p = 0.001 and p < 0.0001, respectively). In contrast, in the remainder tissue, alpha-Mann-2a expression increased in the presence of DENV at 7 days post-feeding and remained elevated over 12-fold at 14 days post-feeding for both WÀ and W+ mosquitoes (post hoc Tukey time Â infection: p = 0.008 and p = 0.007, respectively).  (Salazar et al., 2007). The Wolbachia-mediated blocking significantly reduced DENV-infected individuals as expected ( Figure S1); however, knockdown was not effective in any of the tissues of Wolbachiainfected mosquitoes (Figure 2a-c). We also checked the Wolbachia densities across treatment groups and found that RNAi had no effect on wAlbB densities ( Figure S2). We did see effective knockdown in the midgut of wild-type mosquitoes infected with DENV 7 dpi, with 90% of the control levels of expression suppressed (Figure 2e). In the carcass the transcript levels surprisingly increased compared to controls ( Figure 2f).

Alpha-mann
Knockdown of alpha-Mann-2a significantly increased the intensity (viral load) of DENV infection in the midgut at 7 dpi ( Figure 3a, Mann-Whitney test p < 0.01). Of those mosquitoes, the DENV prevalence (percent infected) was 60% higher in alpha-Mann-2a knockdown midguts, and prevalence also increased by 33% at 10 dpi in the midgut F I G U R E 3 Dengue virus (DENV) infection during RNA interference knockdown of alpha-Mann-2a. DENV intensity and prevalence were detected by absolute qRT-PCR in midgut and carcass samples at 7 and 10 dpi. (a) Infection intensity (genome copy number) over time. Lines mark the median, circles represent DENV quantities in individual dsGFP (yellow) or dsAlpha (magenta) tissue samples, and whiskers depict the 95% confidence intervals. Mann-Whitney test, **p < 0.01. (b) Prevalence (presence or absence) of DENV. Bars contain the percentage of mosquitoes uninfected (green) and DENV infected (blue). Binary logistical regression for Midgut: p < 0.0001, likelihood ratio test for treatment, ****p < 0.0001. Binary logistical regression for carcass: p = 0.48. N = 8-15 7 and 10 dpi. This is important for potential downstream effects on DENV dissemination and invasion of the salivary glands.

alpha-Mann-2a knockdown during CHIKV infection
Next, we were interested in comparing the potential functional role of alpha-Mann-2a in the presence of an alphavirus vs. a flavivirus. We used the same dsRNA for knockdown as above and infected mosquitoes with CHIKV. Knowing that CHIKV infection proceeds faster, we took time points at what would be biologically comparable in terms of viral dissemination.
In wild-type mosquitoes at 10 dpi, the intensity of CHIKV infection was significantly higher in the alpha-Mann-2a knockdown mosquito midguts (Figure 4a). This was largely driven by the increase in the number of infected mosquitoes in this group (Figure 4b), and similar to the effects observed during DENV infection ( Figure 3). By 10 dpi in control midguts, the prevalence of CHIKV was 81%, and alpha-Mann-2a knockdown increased CHIKV prevalence to 100% (likelihood ratio test, p = 0.03). Though similar to the effects observed during DENV infection, comparatively the trend was not as strong.
The overall prevalence and intensity of CHIKV infection were higher than in DENV control mosquitoes and likely due to the higher titre that was harvested for the bloodmeal. The expression of alpha-Mann-2a in dsAlpha mosquitoes was reduced by 75% prior to infection; however, by 5 dpi expression was not significantly lower ( Figure S3). This recovery of expression could be another reason the effects are not as strong. As with DENV infection, no effects were observed in the carcass ( Figure S4). Overall, the data supports that alpha-Mann-2a limits viral prevalence in the midgut during both DENV and CHIKV infections.

alpha-Mann-2a and cadherin87a interactions
Since the conventional role of alpha-mannosidase-2 is to help convert a high mannose glycosylated structure to a complex mannose structure, we hypothesized that the role of alpha-Mann-2a may function through its enzymatic activity on another protein. For multiple reasons we looked into cadherin87a (AAEL023845), the next top candidate gene from our previous study (Ford et al., 2019). In addition to both genes containing genetic variation that associated with the strength of DENV blocking, some cadherins require glycans in order to mediate cell adhesion (Lommel et al., 2013). Therefore, it is plausible there could be functional interactions with alpha-Mann-2a. Second, there are data from mammalian systems showing that decreasing the expression of an alpha-mannosidase gene also changed the expression of an E-cadherin (Tian et al., 2008). Finally, these genes are close in proximity to each other, just under 1.5 Mb apart. Cadherins are a large and diverse family of proteins that all have extracellular cadherin repeat domains and many roles including cell-cell adhesion, cell polarity and intracellular signalling (Halbleib & Nelson, 2006 (Mateo et al., 2015).
Though a cadherin protein is yet to be identified as a viral receptor in any organism, multiple mosquito cadherins have been identified that bind DENV proteins (Colpitts et al., 2011;Muñoz et al., 2013). We were interested to see if knockdown of alpha-Mann-2a affected expression of cadherin87a.
We checked the levels of cadherin87a in our dsAlpha treated mosquitoes. When alpha-Mann-2a expression decreased by more than 90% at 7 dpi in the midguts of dsAlpha mosquitoes (Figure 2e), We then went back to our initial expression experiment and checked cadherin expression in these same tissues.  1 and 6).

DISCUSSION
Current control approaches releasing Wolbachia-mosquitoes aim to decrease mosquito susceptibility to viral infection or to reduce the number of wild mosquitoes (Ferreira et al., 2020). Both strategies require large-scale mosquito rearing facilities for establishment and long-term monitoring for success (Ritchie & Staunton, 2019). These high investment approaches are proving to have high returns with respect to effectively crashing mosquito populations and for reducing the incidence of dengue fever in humans via pathogen blocking (Carrington et al., 2018;Nazni et al., 2019;Zheng et al., 2019).
Insights into how pathogen blocking operates at the molecular level may provide strategies and solutions for maintaining the success of Wolbachia in the face of evolutionary or ecological hinderances. Additionally, independently targeting the mosquito pathways acted upon by Wolbachia could lead to GM technology that is more accessible and easier to deploy in mosquitoes than symbiont-based methods.
A range of mechanisms has been proposed to explain Wolbachiamediated virus blocking. The growing consensus is that the effect is likely multifaceted (Lindsey et al., 2018). There is some evidence that mosquito immunity (Pan et al., 2012;Terradas et al., 2017), cellular stress (Wong et al., 2015) and competition for space or nutritional resources (Caragata et al., 2013(Caragata et al., , 2016 as well as other factors could play a role. We recently sought to identify the basis of viral blocking using an unbiased approach by artificially selecting upon the natural variation observed in this phenotype (Ford et al., 2019). By identifying SNPs that differentiated low-and high-blocking selected lines, we identified a suite of new candidate loci associated with these phenotypes. Few of the genes we identified containing key SNPs recapitulated previously proposed hypotheses, although there was some evidence of selection driving variation in oxidative stress genes (Ford et al., 2019). The two top candidate genes, surprisingly, pointed to potential impacts on protein glycosylation and or cell adhesion.
In this study, we sought to obtain functional data on the role of alpha-Mann-2a, in virus replication in the mosquito by using RNAi. We in the foregut (Chen et al., 2020). This gene's distinct expression dynamics and localization underline the diverse function of genes in this family. Further work into the spatial and temporal dynamics of alpha-Mann-2a and cadherin87a will inform on their roles during infection. In particular, the DENV-induced increase in alpha-Mann-2a in the tissues of the body, which are opposite of the decreases in expression observed in the midgut and salivary glands, could be reflective of tissue differences or differences in pathogen interactions. Furthermore, these differences may have contributed to the inability to capture an observed knockdown in the carcass. The carcass also includes hemocytes, fat body, the heart, aorta, muscles and additional tissues in which alpha-Mann-2a may be expressed, and the time course of knockdown could proceed at a different rate.
In keeping with our data on alpha-Mann-2a, we propose two potential modes of action for its involvement in viral infection. The first hypothesis is that alpha-Mann-2a enzymatic activity acts upon viral proteins. The second is that alpha-Mann-2a enzymatic activity acts upon mosquito proteins that alter infection. In support of the first hypothesis, the presence of N-linked glycosylation on flavivirus proteins directly impacts binding to factors on the cell's surface (Carbaugh & Lazear, 2020). N-linked glycosylation can be further classified as high mannose, which is the first step in the stepwise process to further becoming classified as complex mannose (Yap et al., 2017).
Alpha-mannosidase-2 is required in the first step to convert a high mannose glycosylated structure to a complex mannose structure.
Insects and mammals have different glycosylation pathways, and therefore glycosylation patterns are different in these two hosts The presence of N-linked glycosylation at a specific position on the WNV Envelope protein reduced viral infectivity in Aedes albopictus C6/36 mosquito cell lines, and this effect was stronger in mosquito cells compared with mammalian (Hanna et al., 2005). In human cells, WNV interactions with the C-type lectin viral attachment factor DC-SIGN have been shown to be dependent on the presence of high mannose residues on envelope proteins (Carbaugh & Lazear, 2020).
Though this receptor is not present in insects, other C-type lectins promote DENV, Japanese encephalitis virus and WNV infection (Caragata et al., 2019). The studies mentioned above have compared non-glycosylated DENV to glycosylated or comparatively studied mosquito versus mammalian derived virus; however, it is unknown if altering a high mannose glycan to a complex mannose, as would be facilitated by alpha-mannosidase, would change infectivity in mosquitoes. Lastly, it is also possible that high mannose glycosylation of viral proteins results in an evasion of immune responses as has been shown for both flaviviruses and alphaviruses in mammalian cells (Rogers & Heise, 2009), and therefore alpha-Mann-2a activity could result in increasing virus vulnerability to immune responses.
Our second hypothesis is that alpha-Mann-2a activity may alter the glycosylation of mosquito proteins. This could occur through altering the glycosylation of proteins involved in the viral response or that facilitate infection. Furthermore, the enzymatic activity of alpha-Mann-2a could lead to protein turnover, as inhibitors of alpha-mannosidase lead to decreased degradation of glycoproteins (Segal & Winkler, 1984). In summary, alpha-Mann-2a may destabilize the activity or recycling of a protein that facilitates viral infection.
The Changes in mosquito cell integrity during infection have not been described; however, in humans haemorrhagic fever during DENV infection associates with decreases in vascular cadherins involved in membrane permeability (Dewi et al., 2008;Kanlaya et al., 2009).
Interestingly, alpha-Mann-2a gene knockdown by RNAi only occurred in the absence of Wolbachia. This is intriguing since we hypothesize that alpha-Mann-2a decreases viral infection. We show that DENV changes alpha-Mann-2a expression but the presence of Wolbachia had little effect. This is not completely surprising since the variation that correlates with pathogen blocking in this gene occurs within introns (Ford et al., 2019). This suggests that Wolbachia may affect splicing and not expression. Recent work in D. melanogaster identified alternative splicing of genes in the presence of Wolbachia and Sindbis virus (Bhattacharya et al., 2017). So further assessment of the role of these genes in pathogen blocking will be done in combination with transcriptomics looking at splicing variation. Wolbachia could also potentially alter the rate of protein turnover, which would indirectly affect transcription and the ability to knockdown a transcript. This is supported by the possibility of overlap in physical space, as alpha-mannosidase proteins associate with the ER and Golgi, and Wolbachia physically interacts with the ER and alters its subcellular organization (Fattouh et al., 2019).

Mosquito rearing
Ae. aegypti mosquitoes previously infected with the wAlbB strain of Wolbachia were backcrossed for seven generations into a wild line, AFM, isolated from Merida Mexico in 2017. All mosquitoes were reared and maintained at 27 C, 60% relative humidity and a 12 h:12 h light/dark photoperiod. Larvae were reared in deionized water and fed fish food (TetraMin) ad libitum. Prior to eclosion, the pupal stage was transferred in small cups of water to a cage, and emerging adults were fed 10% sucrose.

Virus cultivation and mosquito blood feeding
The dengue virus serotype 2 (DENV-2) strain JAM 1409 (Bennett et al., 2002) was propagated in A. albopictus C6/36 cells, as previously described (Terradas et al., 2017). with a sausage casing, warmed to 37 C with water from a water bath.
After 2 h, the mosquitoes were anaesthetized on ice and unfed or partially fed mosquitoes were removed. All CHIKV work was carried out in the Pennsylvania State Pell BSL-3 Laboratory.

Dissections
Prior to dissection, mosquitoes were anaesthetized with triethylamine (Sigma-Aldrich) via exposure to a moistened 47 mm circular filter paper. The head was removed with tweezers, followed by removal of the salivary glands from the thorax, and the midgut from the abdomen using insect pins in phosphate-buffered saline (PBS). The remaining thorax and abdominal tissues are referred to as the remainder, and in cases where the salivary glands were left intact, we refer to as the carcass.

RNA and DNA extractions
Individual tissues from single mosquitoes were placed in 300 μl of

Gene and Wolbachia quantification
All quantitative polymerase chain reactions (qPCR) were carried out on a LightCycler 480 Real-Time machine (Roche). For gene quantification, mRNA was used in a one-step reverse transcriptase-qPCR (qRT-PCR) reaction. Using qScript One-step SYBR Green qRT-PCR (Quantabio) according to the manufacturer's protocol, the synthesis of cDNA occurred at 50 C for 5 min followed by Taq inactivation at 95 C for 2 min. Amplification and quantification cycled 45 times at 95 C for 3 s and 60 C for 30 s, and the final cycle was followed by a melting curve analysis. Gene-specific primers were used at a final concentration of 0.2 μM for amplification of target genes (Table S2), and the insect gene 40S Ribosomal protein S6 (Rps6) was used as a housekeeping gene (Pan et al., 2012). For wAlBb quantification by qPCR, DNA samples were amplified using PerfeCTa SYBR Green SuperMix (Quantabio) according to the manufacturer's protocol and genomic primers specific to wAlBb and the mosquito 40S Ribosomal protein S17, and thermocycling conditions as previously published (Dutra et al., 2021). Gene expression and Wolbachia levels were calculated by the 2 ÀΔCt method (Livak & Schmittgen, 2001).

Virus quantification
DENV was quantified using TaqMan Fast Virus 1-step Master Mix (Thermo Fisher Scientific) for qRT-PCR in 10 μl reaction volumes with DENV specific primers and probes (Table S2) (Ye et al., 2014). The following protocol was used: reverse transcription at 50 C for 5 min, followed by 95 C for 20 s, and amplification cycling at 95 C for 3 s and 60 C for 30 s. A standard reference curve of known quantities of a DENV-2 genomic fragment was used for absolute qRT-PCR. The DENV-2 genomic fragment was inserted into a plasmid and transformed into Escherichia coli as previously described (Ye et al., 2014).
The linearized and purified fragment was serially diluted ranging from 10 7 , 10 6 , 10 5 , 10 4 , 10 3 , 10 2 , and 10 copies and were used to create a standard curve of DENV amplification. The standard curve was run in duplicate on each 96-well plate, and the limits of detection were set at 10 2 copies. CHIKV was quantified using the same reagents and protocols as DENV above, with CHIKV specific primers and probes (Table S2) (Rückert et al., 2017). A 140 bp fragment was amplified and cloned into the PGEM-T vector system (Promega, Madison, WI), and the plasmid was transformed into E. coli. Plasmid was extracted by phenol-chloroform, linearized by restriction enzyme digest with PstI (NEB, Ipswich, MA) and a standard curve of 10 8 , 10 7 , 10 6 , 10 5 , 10 4 , 10 3 , 10 2 , 10, 5, 1 was serial diluted based on the copy number obtained using a Qubit fluorometer (Thermo Fisher Scientific). The limits of detection of CHIKV were set at 10 copies.
Focus forming unit assay Viral titres were quantified by FFU in Vero cells (Brustolin et al., 2018;Payne et al., 2006). Cells were plated and grown for 24 h to confluency in 96-well plates at 37 C with 5% CO 2 . The next day, cells were prepared by washing in RPMI medium (described above) without FBS. Viral supernatant or homogenized mosquito bodies were serially diluted 10-fold in medium at concentrations from 10 À1 to 10 À6 and added to cells in 30 μl volumes. Plates were incubated for 2 h at 27 C.

RNA-interference gene silencing
The Ae. aegypti gene alpha-Mannosidase-2a (AAEL004389) was targeted and silenced by RNAi. We designed many primer sets to target different regions of alpha-Mann-2a within exons 4-9. Knowing that this gene likely has a high amount of variation, and experiencing multiple primer sets that did not amplify cDNA, we designed the primers in areas where there were low numbers of SNPs based on genomic sequencing (Ford et al., 2019). Additionally appending the T7 promoter sequence for dsRNA synthesis requires a specific nucleotide sequence. To achieve the best binding, we aligned alpha-Mann-2a with the predicted ortholog in Anopheles gambiae (AGAP0004032) and chose conserved GGG sequences between these two species for the forward primer T7 binding location and then used the same approach to design the reverse primers. We also designed our dsRNAs to be over 400 bp in length and therefore would be processed into many siRNAs. Gene-specific dsRNA was synthesized using the MEGAscript T7 transcription kit (Invitrogen) as previously described (Sigle & Hillyer, 2018). Briefly, templates were amplified from mosquito cDNA using PCR and primers appended with the T7 promoter sequence (Table S2). PCR products were gel purified using the Pur-eLink Quick Gel Extraction kit (Invitrogen). Approximately 1 μg of template was used to synthesize dsRNA following the manufacturer's protocol. The dsRNA was purified by phenol/chloroform precipitation and resuspended in sterile PBS. Adult mosquitoes at 4 days posteclosion were anaesthetized on ice and 69 nl of dsRNA (300 ng final concentration) in PBS was injected into the hemocoel through the thoracic anepisternal cleft. Only two out of five designed dsRNAs made it through our stringent requirements and both targeted exon 9. We proceeded with the dsAlpha used in this study based on preliminary KD efficiencies being the greatest. As a control, dsGFP was injected in the same cohort of mosquitoes (Pan et al., 2012). At 3 days post-RNAi, mosquitoes were offered an infectious bloodmeal as described above. Following infection, the midgut was separated from the carcass and each tissue was individually placed in TriReagent as above and stored at À80 C until extracted.
Data were tested for normality using the Kolmogorov-Smirnov test.
All significant differences in gene expression are two-fold or greater changes in expression. Infection prevalence was analysed for each tissue using a binary logistical regression model with the presence or absence of infection as the response and with the variables of treatment (RNAi: dsGFP vs. dsAlpha) and time (dpi) as predicting effects.
The significance of the effects of time and treatment on prevalence were compared using a likelihood ratio test. The Mann-Whitney test was used for t-test comparisons of non-parametric data. Analyses were performed using PRISM 6 Software (GraphPad, La Jolla, CA), with the exception that General linear models and logistical regression were performed in JMP PRO 15 (SAS, Cary, NC). Effects were deemed significantly different at p < 0.05.

SUPPORTING INFORMATION
Additional supporting information may be found in the online version of the article at the publisher's website.