In order to over-express TEP1 in mosquitoes in the endogenous source tissue, we first sought to determine which cells express TEP1 mRNA and attempted to perform RNA in situ hybridization in both larva and adult mosquitoes. However, the fat body tissue did not withstand the procedure and was lost from the analysis. Thus, we constructed a transgenic reporter line expressing GFP under control of the TEP1 promoter. For this, we cloned a 3.1 kb genomic DNA fragment located 5’ of the TEP1s gene from the A. gambiae G3 strain, ending at the TEP1s initiator ATG codon. Transfection assays in Drosophila S2 cells confirmed that this cloned TEP1s 5’ region is responsive to bacterial challenge (see below), as is the endogenous TEP1 gene [20]. Investigation of the expression pattern of GFP in the transgenic mosquitoes revealed strong expression in the fat body tissue at larval (Fig 1a and 1b), pupal (Fig 1c and 1d) and adult stages (Figs 1n and 2). In addition, intense GFP expression was detected in cells of the midgut proventriculus (cardia) of adult mosquitoes, both males and females, regardless of blood feeding (Fig 1i and 1j, and S1a and S1b Fig). Other tissues showing some GFP expression included the oviduct, and more variably, the Malpighian tubules (S1c Fig). GFP was expressed in fat body loosely attached to the testis, but was not observed in the male reproductive organs themselves.

As a control for fat body and hemocyte expression we used the Lipophorin (Lp) promoter-driven RFP reporter line constitutively expressing RFP in the fat body and the hemocyte-specific PP06-RFP line [34]. Comparing the expression of TEP1-GFP to the Lp-RFP pattern shows similar expression in the same cells in dissected carcass tissues (Fig 2a). Mosquitoes carrying both reporter genes showed perfect GFP/RFP overlap. Upon crossing TEP1-GFP mosquitoes to the hemocyte reporter PPO6-RFP line, the observed RFP-positive hemocytes were never positive for GFP, indicating that the TEP1 reporter is not strongly expressed in the hemocytes that express PPO6 (Fig 1e–1h). Although the reporter mosquitoes did not display obvious GFP-positive hemocytes, we do not rule out a possible expression of endogenous TEP1 in some hemocytes, as the cloned regulatory sequences may lack distant hemocyte-specific enhancers.

To identify the expression pattern of other genes known to act in the TEP1 pathway, we also constructed transgenic reporter lines expressing GFP under the control of the promoters of the leucine-rich proteins LRIM1 and APL1C that associate with TEP1 in the hemolymph. Both reporter lines showed a GFP expression pattern similar to TEP1, namely expression in the fat body at all developmental stages (Fig 1k–1n). However, neither LRIM1-GFP nor APL1C-GFP exhibited GFP expression in the proventriculus. Of note, all these reporters were created using the same docking line, and are thus subjected to the same potential positional effects. Many other fluorescent reporter constructs inserted at the same locus exhibit different, promoter-specific patterns of expression [34], indicating that the TEP1, LRIM1 and APL1C promoters must contain sequences determining fat body-specific expression.

In mosquitoes, fat body cells are challenging to manipulate. Dissection of mosquitoes in buffer results in immediate loss of large parts of the fat body. The fat body cells that remain attached to the carcass detach easily during fixation and disappear during permeabilization in buffers containing detergents. To verify the expression of TEP1 in the fat body cells that remain attached to dissected mosquito carcasses, we performed an antibody staining of fixed samples using a modified protocol (see materials and methods). Indeed, careful examination showed that some fat body cells that express the fat body specific protein Lipophorin (Lp) also stain positive for TEP1 (Fig 2b). In mosquitoes where TEP1 was silenced by RNAi, this fat body TEP1 staining was markedly reduced (Fig 2c), indicating that the signal is TEP1-specific.

Larvae and adults of the transgenic TEP1-GFP reporter line express GFP at strikingly variable levels (S1d and S1e Fig). We followed the levels of GFP during larval development (S2 Fig) and found an increase in GFP levels in time, most fourth instar larvae having reached a maximum level of GFP expression, though striking differences from larva to larva persist until pupation. Newly eclosed adult mosquitoes also exhibit variable levels of GFP expression (S1e Fig), probably reflecting a high variability in TEP1 expression that contrasts with the highly reproducible expression level of the Lp-RFP reporter. We hypothesized that the variability in GFP expression reflected TEP1 induction by immune pathways [35] upon encounter with various microorganisms, which may differ from larva to larva. We tested whether there is a correlation between the level of GFP in adult mosquitoes and intensity of P. berghei infection. Adult mosquitoes were scored for GFP expression prior to infectious blood feeding and grouped according to weak or strong reporter expression level. Mosquitoes were infected and oocysts were counted 7 days post infection. S1 Table summarizes the result of 5 independent experiments, showing no significant differences in oocysts loads between the high and low GFP expressing mosquitoes in 4 out of 5 experiments.

Given the localization of GFP expression under control of the TEP1 promoter, we decided to express the TEP1 transgene in adult fat body tissue. To avoid unnaturally high TEP1 expression throughout development, we used the blood-meal inducible Vg promoter. Vg is induced at high levels after blood feeding [36] peaking 24h after blood meal, at the time when P. berghei ookinetes reach the basal lamina and encounter circulating TEP1. This experimental design aimed at achieving high levels of TEP1 simultaneously to Plasmodium infection.

Since the TEP1r allele is associated with refractoriness to P. berghei infection and shown to be more potent in Plasmodium killing, we based our construct on the TEP1r allele. TEP1r cDNA was cloned under the Vg promoter and injected into the TEP1s-containing X1 docking line [34], to create the Vg-TEP1r,3xP3-RFP transgenic line (henceforth referred to as Vg-TEP1r). Analysis of TEP1r mRNA in the transgenic mosquitoes (S3a Fig) showed blood meal induction of TEP1r similar to induction of Vg [37]. Like Vg mRNA, TEP1r mRNA levels monitored with TEP1r-specific primers peaked 24h after blood meal and dropped gradually in the following days. However, Western blot analysis of the transgenically expressed TEP1r protein in the TEP1s background did not reveal a notable increase in TEP1 protein levels in the transgenic mosquitoes 24h after blood feeding (S3b Fig), suggesting tight post-transcriptional control of hemolymph TEP1 protein levels.

In order to verify expression from the Vg-TEP1r transgene, we introduced it into two different TEP1 null mutant lines [27]. The TEP1Δct2 mutant has a nonsense mutation causing a premature stop codon and loss of the entire C-terminus. The TEP1ΔT mutant has a deletion of a threonine located 16 amino acids before the thioester cysteine. Both mutants lack detectable TEP1 protein in the hemolymph at 24h after blood feeding, but low levels of mutated TEP1 become detectable in old TEP1ΔT mosquitoes only. Samples from TEP1Δct2; Vg-TEP1r mosquitoes analyzed by polyclonal anti TEP1 antibodies showed the presence of both forms of TEP1, namely TEP1-full and TEP1-cut in the hemolymph expressed from the transgene (Fig 3a). In contrast, hemolymph samples from TEP1ΔT; Vg-TEP1r showed only the TEP1-full form (Fig 3b), suggesting that traces of the mutant TEP1ΔT protein inhibit maturation of wild-type TEP1 expressed from the transgene. In addition, TEP1ΔT (but not TEP1Δct2) mosquitoes show reduced levels of LRIM1 and APL1C (Fig 3a–3c and 3e). Unlike LRIM1 and APL1C depletion by RNAi silencing [21], this decrease does not result in massive TEP1-cut deposition on tissue (Fig 3d). Thus, traces of the mutant TEP1ΔT protein appear to downregulate both the endogenous LRIM1/APL1C complex and the cleavage of transgenic TEP1-full.

It has been previously demonstrated that injection of bacteria into mosquitoes promotes TEP1 maturation and activates TEP1 transcription [20]. Furthermore, shortly after bacteria injection, TEP1-cut binds to bacterial surfaces and acts as a convertase to recruit the full-length protein [38,39]. The recruited TEP1-full is then cleaved close to bacterial surfaces and increases bacterial opsonization, thus marking bacteria for phagocytosis by hemocytes [40]. We tested whether injection of bacteria induced formation of the TEP1-cut form in the TEP1ΔT;Vg-TEP1r mosquitoes (Fig 3e). While the positive control TEP1r mosquitoes showed an increase in both TEP1-full and TEP1-cut after bacterial injection, transgenic TEP1-full in the TEP1ΔT;Vg-TEP1r mosquitoes’ hemolymph still failed to be converted to TEP1-cut.

In order to verify that the TEP1r transgene was functional, we infected TEP1 mutant mosquitoes complemented with the Vg-TEP1r transgene. The TEP1Δct2;Vg-TEP1r mosquitoes resemble the L3-5 line in that they express only the TEP1r allele. However, an important difference from the L3-5 line is that TEP1r is expressed exclusively in the fat body and only after a blood meal. In the TEP1ΔT;Vg-TEP1r line, only the full form of TEP1 is detectable in the hemolymph, which offers an opportunity to examine the activity of the TEP1-full form in the absence of detectable TEP1-cut.

P. berghei infection of TEP1Δct2;Vg-TEP1r was strikingly reduced compared to mutant controls, with some live oocysts and mostly melanized parasites (Fig 4a), indicating that the introduced TEP1r is fully functional and sufficient to kill parasites. Infection prevalence (S2 Table) in the transgenic is also significantly lower than in control Vg-GFP mosquitoes that express TEP1s only. P. berghei infection of TEP1ΔT;Vg-TEP1r, in which only the full form of TEP1 is detected, also resulted in a significant reduction in live oocyst numbers and in the concomitant appearance of high numbers of melanized parasites in some midguts compared to the non-rescued mutant (Fig 4b). High levels of melanization are reminiscent of the L3-5 infection phenotype, where complete refractoriness is achieved by TEP1r-mediated parasite killing followed by parasite melanization [19,33]. However, contrary to L3-5 that are completely refractory to the parasite, the rescued mutant lines which derive from the susceptible G3 background are not fully refractory to P. berghei infection, with about 70% of infected midguts containing live oocysts 7 days after infection (S2 Table) and infection intensities comparable to wild-type control G3 mosquitoes. In fact, expression of TEP1r in mutant backgrounds rescued the mutant at least to the level of the wild-type TEP1s-carrying G3 line in terms of prevalence and intensity, and further conferred a gain-of-function phenotype (melanization) (Fig 4a). In the TEP1ΔT;Vg-TEP1r line, it is striking that transgenic TEP1r appears to be functional in spite of its impaired maturation into the TEP1-cut form. These results imply either that TEP1 cleavage may be dispensable for some of its functions, or that low levels of TEP1-cut that we could not detect in the hemolymph (Fig 3b and 3e) or on parasites (Fig 5) may be sufficient for function. The formation of TEP1-cut at foreign surfaces is dependent on a convertase complex controlled by SPCLIP1 [39]. To test if maturation of TEP1-full into TEP1-cut is required for the observed phenotype, we silenced SPCLIP1 by dsRNA injection in TEP1ΔT;Vg-TEP1r mosquitoes (S1 File). This fully abolished melanization and tended to increase parasite numbers, suggesting that some TEP1r cleavage by the convertase complex, though undetectable, must be required for activity of the TEP1r transgene.

Importantly, transgenic expression of TEP1r exclusively in the fat body using the Vg promoter rescued the TEP1 loss-of-function mutations (Fig 4a and 4b), showing that TEP1 produced in the fat body and secreted into the hemolymph is sufficient for parasite killing.

We tested whether expressing TEP1r in a wild type background where TEP1s is present could also reduce parasite loads after P. berghei infection. Oocyst counts revealed no difference in live oocyst numbers between the parental TEP1s line and the TEP1s;Vg-TEP1r transgenic line (Fig 4c, left panel). However, the number of melanized parasites was significantly elevated in these mosquitoes (Fig 4c, right panel). We verified that the increase in melanized parasites was attributable to the TEP1r transgene using allele specific RNAi. When transgenic TEP1r, but not endogenous TEP1s, was knocked down in the TEP1s;Vg-TEP1r transgenic line, the number of melanized parasites reverted to a level similar to control dsLacZ-injected mosquitoes (Fig 4d).

In the TEP1s;Vg-TEP1r line, transgenic TEP1r is induced to a peak level about 24 hours after a blood meal, possibly too late to efficiently fight parasite invasion. To induce TEP1 expression prior to the infectious blood meal and attempt to increase TEP1 concentration in the hemolymph, we offered a non-infectious blood meal and infected mosquitoes three days later, before the Vg promoter had returned to its off state. The results of such an experiment (S4 Fig) show no difference in infection levels between mosquitoes that were pre-induced by blood meal and naïve mosquitoes.

We verified that transgenic TEP1r binds to parasites in the TEP1Δct2;Vg-TEP1r line (S5 Fig) and asked whether this was also the case in TEP1ΔT;Vg-TEP1r mosquitoes in which formation of the mature TEP1-cut form is inhibited. To this end, we performed immuno-histochemical staining of TEP1ΔT;Vg-TEP1r midguts 24h after infection using TEP1 specific antibodies (Fig 5). Results show that in the presence of TEP1ΔT, transgenic TEP1r protein is not detectable on ookinete surfaces, despite its observed, SPCLIP1 dependent, anti-parasitic and melanization activity.

We tested whether the fat body-expressed, full length TEP1r in the TEP1ΔT;Vg-TEP1r line can enter hemocytes after bacterial injection. To this end we examined TEP1 staining in carcasses of blood fed mosquitoes, injected with heat killed E. coli, 24h after blood meal (Fig 6). Results show that in blood-fed TEP1ΔT;Vg-TEP1r mosquitoes, TEP1r artificially induced in the fat body by the blood meal accumulated in the hemocytes only when mosquitoes were injected with E. coli. Control wild type mosquitoes showed that TEP1 accumulation in hemocytes depended on bacterial injection, regardless of blood feeding. The TEP1ΔT mutant line, used here to control for the specificity of antibodies for TEP1, showed no TEP1 staining in hemocytes under any condition. These results show that TEP1 detection in hemocytes can result from uptake of what is most likely TEP1-labeled bacteria.

Since transgenic expression of TEP1r in a wild-type TEP1s background failed to produce refractory mosquitoes, we asked whether artificially elevating the levels of the endogenous TEP1s allele may lead to increased immunity against Plasmodium. To this aim, we stimulated expression of the endogenous TEP1 gene using synthetic transcription factors designed to bind specific promoter sequences.

Transcription Activator Like Effectors (TALEs) are transcription-activating proteins derived from Xanthomonas bacteria, containing a DNA-binding domain composed of nucleotide-binding repeats that can be arranged in the desired order to bind chosen DNA sequences [41,42,43]. In order to trigger over-expression of the endogenous TEP1 gene, we designed TALE proteins that bind specific DNA sequences in the promoter region of TEP1. To find the best potential binding position for a TALE, we first analyzed the 3 kb promoter region of TEP1 (S6 Fig). Using a luciferase reporter assay in Drosophila S2 cells and nested promoter deletions, we found that a 250 bp fragment of the promoter was sufficient to induce reporter activity. The minimal promoter sequence, harboring a potential NF-κB binding site, responded to the addition of bacteria to the culture medium by increased luciferase activity (S7 Fig). Mutating this NF-κB binding site abolished the reporter response to bacterial challenge (S8 Fig). This suggests that the NF-κB pathway of S2 cells can activate the mosquito TEP1 promoter, in agreement with the known activation of native TEP1 transcription by the mosquito NF-κB pathway [35]. We used this minimal promoter to test several transcriptional activation domains for reporter activation by TALEs, namely those of yeast Gal4, Drosophila Heat Shock Factor 1 and Herpes virus VP16, and found that the VP16 activation domain fused to a TALE DNA-binding domain leads to highest reporter induction (S9 Fig). The VP16 domain was thus chosen as the activation domain for all further designed TALEs. TALEs without an activation domain did not increase TEP1 expression and served as negative controls.

We constructed five TALEs that bind at different positions along the TEP1 minimal promoter sequence (Fig 7a and S6 Fig) and tested their effect on reporter transcription in S2 cells (Fig 7b). Results show that TAL₃ binding the closest to the transcription start site had no effect on reporter activity, while the other TALEs had a modest effect, increasing activity by about 1.5–2 fold. However, TAL₆, binding the furthest from the transcriptional start site and just upstream of a sequence resembling a TATA box, triggered a strong, 20 fold increase in reporter activity. TAL₆ was thus chosen to create transgenic mosquitoes with the aim of over-expressing endogenous TEP1.

In plants, natural and artificial TALEs induce target gene expression from alternative, TALE-specific transcription start sites [44,45]. We investigated whether TALEs triggered TEP1 expression through an alternative transcription site when transfected with the luciferase reporter in S2 cells. PCR products from the 5' RACE assay were sequenced and revealed that unlike in plants, both TAL₀ and TAL₆ preserved the normal transcriptional start site of the mRNA which remained located 43 bp upstream of the ATG (S6 Fig).

We generated several transgenic lines that express TALEs designed to bind and activate the TEP1 promoter. Our first choice of promoter for the expression of the synthetic transcription factor itself was the inducible fat body specific Vg promoter, in order to increase endogenous TEP1 levels after a blood meal, concomitant with Plasmodium parasite invasion. TAL₆ fused to a VP16 (TAL₆^vp16) activation domain was chosen due to its high activity in S2 cells. A control transgenic line was created with the same TAL₆ lacking an activation domain (TAL₆^ΔAD). Analysis of mRNA after blood feeding revealed that in both transgenic lines TALEs are induced by a blood meal, more efficiently in the case of Vg-TAL₆^vp16 (Fig 7c). Furthermore, TAL₆^vp16, but not TAL₆^ΔAD, induced a two-fold elevation in TEP1 mRNA levels. This elevation in the endogenous levels of TEP1 mRNA is modest compared to the 20-fold average elevation of the reporter in the cell line.

Since the Vg promoter triggers a pulse of transcription after a blood meal, TALEs may activate TEP1 too transiently to achieve a significant effect. Therefore, we also tested if a constitutive fat-body promoter would durably induce high levels of endogenous TEP1. For this, TAL₆^vp16 and TAL₀^vp16 were placed under the control of the Lp promoter, constitutively expressed in the fat body in larvae, pupae and adults. We included TAL₀^vp16 in addition to TAL₆^vp16 despite its modest effect in S2 cells (~1.5 fold increase, Fig 7b) because of its unique binding position on the NF-κB binding site. TAL₀^vp16 was designed to bind and activate not only the promoter of TEP1, but also similar sequences found in the promoters of APL1C and LRIM1, two proteins that are known to bind TEP1 and contribute to its activity (See Methods).

Analysis of TEP1 mRNA in transgenic mosquitoes that constitutively express TAL₆^vp16 and/or TAL₀^vp16 (Fig 7d) revealed a maximum of 2-fold induction of endogenous TEP1 mRNA in mosquitoes that expressed both TALES simultaneously (progeny of a cross between the two lines). Western blot analysis of TEP1 in these mosquitoes (Fig 7e) confirmed a 3–4 fold increase in TEP1 protein levels circulating in the hemolymph (Fig 7f). We thus proceeded to test whether this elevation in TEP1 levels in the hemolymph had any effect on parasite loads after Plasmodium infection.

Infection of the blood meal inducible Vg-TAL₆^vp16 transgenic line did not result in any significant change in parasite loads (Fig 7g). Likewise, analysis of parasite loads after infection of the constitutively expressing Lp-TAL₆^vp16 and Lp-TAL₀^vp16 lines did not show any significant difference in oocyst numbers between transgenic and control lines (Fig 7h).

TEP1 is a secreted protein that circulates in the hemolymph of mosquitoes at all developmental stages starting from larval stages [20]. Previous observations suggested that TEP1 is produced by hemocytes and secreted from these cells as a full-length protein which is then cleaved in the hemolymph [20]. Evidence to support this model included immuno-histochemical staining of mosquito tissues that showed TEP1 accumulation inside hemocytes, particularly after bacterial challenge and 24-48h after infection with P. berghei [22,35,46]. However, these results may also be explained by uptake of TEP1 protein, or TEP1-labeled particles, by hemocytes. In addition, analysis of RNA expressed by hemocytes using microarrays has shown that TEP1 transcript is present in these cells, but not enriched compared to whole female tissues, and is not further induced by infection [47]. TEP1 did not emerge as a hemocyte transcript in the hemocyte transcriptomics study of Pinto et al. [48]. Of note, it is technically challenging to acquire a pure population of hemocytes without contamination by fat body cells that detach into the hemolymph during sample collection [49]. Here we obtain evidence that TEP1, LRIM1 and APL1C are expressed primarily in the fat body. This is similar to the expression pattern of many insect anti-microbial peptides [50,51] and to that of some Drosophila TEP family member genes as determined by in situ hybridization [52]. In mammals, complement proteins phylogenetically and functionally homologous to TEP1 are expressed and secreted from the liver ([53],reviewed in [54]). Interestingly, some of the functions of the liver are performed by the fat body in insects.

We crossed the TEP1-GFP reporter mosquitoes to our hemocyte-specific PPO6-RFP line, and observed that the DsRed-positive hemocytes were GFP negative. We note that this result does not rule out that sub-populations of hemocytes not expressing the PPO6 promoter may also express TEP1; fine transcriptomic analyses will be required to quantitatively determine the contribution of each tissue to the global pool of circulating TEP1.

Our transgenic reporter lines also reveal that TEP1, but not APL1 and LRIM1, is also expressed in the proventriculus of adult mosquitoes. Interestingly, the proventriculus was found to be a site of immune gene expression in A. gambiae [55] and in other insects such as tsetse flies [56]. In A. gambiae, mRNA transcripts of antimicrobial peptides Defensin 1, Cecropins, Gambicin, as well as TEP1, are enriched in the proventriculus [55]. TEP1 is also expressed in the neighboring gastric caeca during larval stages [57]. Future studies may clarify whether proventriculus-expressed TEP1 has a function that bypasses the requirement for LRM1 and APL1-C, perhaps in the lumen of the digestive tract where TEP1 could interact with the gut microbiota.

Variation in GFP levels among individual mosquitoes led us to test whether GFP fluorescence intensity correlates with infection outcome. We found no such correlation and concluded that GFP expression either did not reflect TEP1 levels in the mosquitoes at the time of P. berghei infection, or that TEP1 level variations do not correlate with resistance to the parasite. The lack of correlation between GFP levels and immune status may be due to the remarkable stability of GFP protein, which persists throughout mosquito life in the fat body after a blood feeding-induced pulse of expression in the Vg-GFP line [34] compared to the highly dynamic turnover of TEP1 protein, which disappears by 24h after RNAi silencing [37].

Lysis of ookinetes at the basal lamina is associated with binding of TEP1 directly to the ookinete surface [19]. However, TEP1 does not induce direct lysis of bound pathogens, and the molecular mechanism of killing remains unknown. It has been previously shown that TEP1-cut is the major form that binds ookinetes [21,22]. Binding of circulating TEP1-cut to pathogen surface triggers the recruitment of a convertase complex that requires the serine protease homologue SPCLIP1 and that cleaves circulating TEP1-full leading to further accumulation of TEP1-cut at the pathogen surface [38,39]. We expressed the TEP1r allele in two different TEP1 mutant backgrounds: TEP1ΔT carries a single threonine deletion that appears to destabilize the protein, as mutant TEP1ΔT protein becomes faintly detectable only in aged mosquitoes [27]. TEP1Δct2 lacks the C-terminal third of the protein and is likely to be a complete null. In these two mutant backgrounds, transgenic TEP1r produced in the fat body and secreted into the hemolymph was sufficient to rescue parasite killing, and to trigger parasite melanization. Interestingly, in the TEP1ΔT background, but not in the TEP1Δct2 background, maturation of transgenic TEP1r-full to TEP1-cut appeared to be inhibited as we never detected circulating TEP1-cut in the hemolymph, even after bacterial injection. This suggests that very low amounts of the non-functional TEP1ΔT protein inhibit maturation of TEP1-full, perhaps by blocking the enzyme responsible for TEP1 cleavage in the hemolymph. Possibly related to this, LRIM1/APL1C levels are reduced in the TEP1ΔT, but not in the TEP1Δct2 mutant. In the TEP1ΔT;VgTEP1r mosquitoes, TEP1 binding is not detectable by antibody staining on ookinete surfaces, consistent with the lack of detectable circulating TEP1-cut form. Yet, both mutants rescued with TEP1r showed an increase in parasite melanization. We initially hypothesized that contrary to current views, processing of TEP1-full to TEP1-cut by the SPCLIP1-dependent convertase and accumulation of TEP1-cut at the parasite surface may be dispensable for parasite melanization in some contexts. However, disabling the TEP1 convertase complex through silencing SPCLIP1 in TEP1ΔT;VgTEP1r mosquitoes abolishes melanization and, to some extent, increases parasite numbers. Therefore, conversion of TEP1-full to TEP1-cut seems to be a prerequisite for parasite killing and melanization even in TEP1ΔT;VgTEP1r mosquitoes, but a large accumulation of TEP1-cut at the surface of parasites may not be necessary to trigger killing and melanization.

Another function of TEP1 in the hemolymph is the opsonization of bacteria [20,39,58,59], marking them for phagocytosis by hemocytes. Shortly after the injection of bacteria, TEP1 protein can be detected inside the hemocytes that are attached to dissected mosquito carcasses. The appearance of TEP1 inside these cells may be due to induced expression of TEP1 within hemocytes, presumably aimed at replenishing TEP1 levels in the hemolymph, or due to phagocytosis of TEP1-covered bacteria by hemocytes. We observed that in a TEP1 null mutant background, TEP1r expressed from a transgene in the fat-body could enter hemocytes. Our results suggest that TEP1 detected in hemocytes after bacterial challenge comes at least in part from uptake of circulating TEP1 protein.

TEP1r could be a candidate gene for creating malaria resistant mosquitoes by driving the gene into the wild mosquito population, for example using CRISPR-Cas9 gene drive [14,15]. This should not be achieved by replacement of endogenous TEP1s alleles with the more active TEP1r, as a few failed gene conversion events could result in Cas9-induced TEP1s loss-of-function mutations that would generate hyper-susceptible mosquitoes. Instead, the anti-parasitic factor should be driven into an independent neutral genomic locus. We thus tested whether over-expression of TEP1r in wild type background affects infection outcome. We found that transgenic TEP1r is indeed expressed in the presence of endogenous TEP1s, but fails to enhance parasite killing, while still able to stimulate melanization. Using allele specific knock-down we showed that the allelic difference between TEP1r and TEP1s is sufficient to explain the different intensities of the melanization reaction. In addition, silencing endogenous TEP1s in the same mosquitoes restored the ability of transgenic TEP1r to kill parasites and further elevated melanization (Fig 4d), as observed when transgenic TEP1r was expressed in a mutant background. This indicates that although less efficient than TEP1r for parasite killing, endogenous TEP1s protein is able to outcompete transgenic TEP1r for the parasite killing function. This is reminiscent of the apparent recessivity of TEP1r in TEP1s x TEP1r genetic crosses [25]. Therefore, simple transgenic expression of TEP1r in a wild type background where TEP1s is the most frequent allele would not be sufficient to render a mosquito population refractory to Plasmodium infection.

Since transgenic expression of TEP1r in a wild-type TEP1s background failed to produce refractory mosquitoes, we asked whether artificially elevating the levels of the endogenous TEP1s allele may lead to increased immunity against Plasmodium. To this aim, we stimulated expression of the endogenous TEP1 gene using synthetic transcription factors designed to bind specific sites in the TEP1 promoter.

We designed a luciferase reporter for TEP1 promoter activity and tested it in Drosophila S2 cells. The promoter contained an NF-κB binding site, which mediated elevation of reporter activity when S2 cells were treated with heat-killed bacteria. Thus, the cloned TEP1 promoter fragment is activated by the S2 cell NF-κB pathway, consistent with TEP1 activation by the mosquito NF-κB pathway [35].

Once we determined the optimal TALE sequence for binding TEP1 promoter, we created several transgenic mosquitoes lines that expressed TALEs either under an inducible promoter (Vg) or a constitutive fat body promoter (Lp). In all cases, the increase in endogenous TEP1 expression was very modest both at mRNA and protein levels, indicating that cell culture assays are poor predictors of TALE activity in vivo. TEP1 mRNA levels may be tightly regulated in vivo, or may depend on regulatory features that were absent or inactive in the S2 cells assays.

In wild type mosquitoes, some parasites are bound by TEP1 and killed, while others evade TEP1 binding. If the quantity of circulating TEP1 were the rate-limiting factor in parasite binding, we expected that even a modest 2-fold increase in TEP1 expression would yield increased resistance to Plasmodium. However, infection assays with mosquitoes that expressed TALEs and showed a 2 to 3-fold elevation in TEP1 levels did not show any increase in resistance. This is similar to the lack of increased parasite killing in mosquitoes expressing transgenic TEP1r in addition to endogenous TEP1s, and suggests that the survival of some parasites is not due to a lack of TEP1. The absence of effect of the TALE-mediated over-expression of TEP1s contrasts with the elevated melanization obtained by transgenic expression of TEP1r in mosquitoes carrying endogenous TEP1s, again confirming that enhancing melanization is an allelic property of TEP1r.

To conclude, our results indicate that simply augmenting the level of a given anti-parasitic factor may not be sufficient to achieve greater resistance levels, and biotechnological approaches aiming at rendering the mosquitoes resistant to malaria will need to integrate the full complexity of molecular interactions between immune factors.

A. gambiae mosquitoes were maintained in standard conditions (28°C, 75–80% humidity, 12-hr/12-hr light/dark cycle). Larvae were raised in deionized water and fed finely ground TetraMin fish food. Adults were fed on 10% sucrose ad libitum and females were blood-fed on anaesthetized mice. For infection, mosquitoes were fed on CD1 mice infected with the P. berghei GFP-con 259cl2 clone (ANKA strain) that constitutively expresses GFP [60]. Mice infected with Plasmodium berghei were sacrificed before they developed malaria symptoms by cervical dislocation. Mice used to blood feed and infect mosquitoes were anesthetized with a mix of xylazine and ketamine. Infected mosquitoes were maintained at 21°C, unfed mosquitoes were removed 24h post feeding. Seven days after infection, live midguts were mounted on microscope slides in PBS, photographed under fluorescent light, and parasites were counted on the pictures. Statistical significance of differences in parasite counts was evaluated with a Mann-Whitney non-parametric test.

Transgenic mosquitoes were created in the A. gambiae X1 attP docking line using the pDSA R/T/G/Y/P and pattB-RfB2 transgenesis vectors as described [34]. The following plasmids were constructed by Goldengate cloning to generate transgenic mosquitoes: pDSAP-LRIM1-GFP, pDSAT-APL1-GFP. In the case of the TEP1-GFP,3xP3-RFP reporter and of the Vg-TEP1r,3xP3-YFP constructs, plasmids were assembled using Gateway cloning into pattB-RfB2 [34]. Vg-TALE and Lp-TALE constructs were assembled by Goldengate cloning in pDSAR, pDSAT or pDSAG as described for TALEs [42]. Plasmid sequences, as well as promoter and TALE modules are provided in supplementary file S1 Text. In the case of TAL₀, we exploited the TAL repeat “NS” di-residue binding ambiguous bases to allow simultaneous targeting of the TEP1, LRIM1 and APL1C promoters by TAL₀ (target sites indicated in S4 Table). Transgenic lines PPO6-RFP,3xP3-YFP and Lp-RFP,3xP3-GFP and Vg-GFP,3xP3-RFP were previously described [34]. Promoters and genes were amplified from genomic DNA or cDNA by PCR with primers containing appropriate sites for Goldengate or Gateway cloning. All constructs were verified by sequencing.

To introduce the TEP1r,3xP3-YFP transgene in the background of the TEP1^ΔT and TEP1^Δct2 mutations [27], the transgenic line was backcrossed three times to the mutant line. After each backcross, YFP-positive progeny larvae were purified with the COPAS flow cytometer [61] to retain the transgene. After the last backcross, one leg from 48 male and 48 female mosquitoes kept alive in individual tubes was analyzed by PCR as described [27] using the Phire Direct Tissue PCR kit (Thermofisher). Mosquitoes showing homozygosity for the TEP1 mutation were pooled to establish a homozygous mutant and heterozygous transgenic line. For experiments, COPAS sorting of this line was used to generate cultures containing 50% non-transgenic mutant control and 50% transgenic mutant mosquitoes. In some cases, non-mutant control mosquitoes with RFP eye fluorescence (such as Vg-GFP,3xP3-RFP) were added to the same culture. On the day of dissection, transgenic and non-transgenic mosquitoes were separated under a fluorescence microscope based on the eye fluorescence of the transgenesis markers. This process ensured identical exposures of all mosquito genotypes to the same growth and environmental conditions.

Allele specific knockdown of TEP1 was performed by injection of dsRNA targeted specifically against TEP1r or TEP1s as previously described [25]. dsRNA was injected into the thorax of one day old female mosquitoes, three days prior to infection. Allele specific knockdown was verified by qRT-PCR. SPCLIP1 RNAi knockdown was performed as described [39][62].

Mosquitoes were dissected in PBS, fixed in paraformaldehyde for 20 min, permeabilized in PBS containing 0.1% triton (PBST) for 10 min, blocked in PBST containing 1%BSA (PBSTB) for one hour. Mosquito tissues were incubated with primary antibody (rabbit anti TEP, mouse anti Lipophorin) for at least two hours, washed 3 times and incubated with fluorescent secondary antibody (Alexa488, or Cy3) in PBSTB containing 1μg/μl DAPI. Tissues were washed three times in PBST and mounted on a slide with a drop of Aqua/poly mount (Polysciences). For fat body analysis, mosquitoes were dissected and treated as previously described [63]. Images were collected under a Zeiss LSM 710 confocal microscope.

RNA for RT PCR was extracted from groups of 5–10 mosquitoes. Mosquitoes were collected on ice into 300 μl RNAzol RT (MRC), homogenized with a Precellys homogenizer, with 2 pulses of 20 seconds. RNA was purified on DirectZol RNA columns (Zymo Research) following manufacturer's instructions. RNA was treated with RNAse-free DNAse I (ThermoFisher). Reverse transcription was performed using iScript cDNA Synthesis Kit (Bio-Rad) using 500 μg of RNA. SYBR Green PCR Master Mix, (Applied Biosystems) qPCR reactions were set up on 96 well plates with three technical and three biological repetitions and were run on 7500 Real-time PCR machine (Applied Biosystems).

Mosquito hemolymph from 10 mosquitoes was collected by proboscis clipping directly into 20 μl of Laemmli buffer. Samples were heated at 90°C for 2–5 minutes and 2 to 8 μl (depending on the experiment) were loaded on polyacrylamide gels. Rabbit polyclonal antibodies against TEP1, APL1C, LRIM1, PPO2 and Vg were used. Secondary HRP conjugated antibodies were used at 1:15000 dilution. Images were collected on Fusion FX7 scanner (Vilber Lourmat).

S2 cells were grown at 23°C in Schneider's medium (Biowest) supplemented with 10% FCS (Invitrogen) and 1% Penicillin/Streptomycin solution (Sigma). Cells were seeded in 24 well plates (Starsted) at 50,000 cells/well one day prior to transfection. Constructs expressing both firefly luciferase and Renilla luciferase were used as reporters. TEP1 promoter was cloned upstream of firefly luciferase. Renilla luciferase under the constitutive actin5c promoter was used as transfection control. Cells were transfected using Effectene (Qiagen) according to manufacturer's instructions. Cells were collected three days after transfection for luciferase activity analysis. Luciferase assay was performed using the Dual-Luciferase Reporter Assay System (Promega) according to manufacturer's instructions. Luciferase activity is depicted as Firefly activity divided by Renilla activity, averaged on three wells transfected with a set of given plasmids.

RNA from S2 cells was extracted using RNAzol RT (MRC) and purified on DirectZol RNA columns (Zymo Research). 5' RACE was performed using First Choice RLM-Race kit (Ambion) according to manufacturer's instructions. Briefly, ligation of specific primers to 5' and 3' ends of the RNA allows the amplification of a PCR product that corresponds to the transcription start site of gene of interest. The resulting PCR product was cloned into CloneJet PCR Cloning Kit (ThermoFisher Scientific) followed by sequencing of several clones.