Solution structure, glycan specificity and of phenol oxidase inhibitory activity of Anopheles C-type lectins CTL4 and CTLMA2

Malaria, the world’s most devastating parasitic disease, is transmitted between humans by mosquitoes of the Anopheles genus. An. gambiae is the principal malaria vector in Sub-Saharan Africa. The C-type lectins CTL4 and CTLMA2 cooperatively influence Plasmodium infection in the malaria vector Anopheles. Here we report the purification and biochemical characterization of CTL4 and CTLMA2 from An. gambiae and An. albimanus. CTL4 and CTLMA2 are known to form a disulfide-bridged heterodimer via an N-terminal tri-cysteine CXCXC motif. We demonstrate in vitro that CTL4 and CTLMA2 intermolecular disulfide formation is promiscuous within this motif. Furthermore, CTL4 and CTLMA2 form higher oligomeric states at physiological pH. Both lectins bind specific sugars, including glycosaminoglycan motifs with β1-3/β1-4 linkages between glucose, galactose and their respective hexosamines. Small-angle x-ray scattering data supports a compact heterodimer between the CTL domains. Recombinant CTL4/CTLMA2 is found to function in vivo, reversing the enhancement of phenol oxidase activity in dsCTL4-treated mosquitoes. We propose these molecular features underline a common function for CTL4/CTLMA2 in mosquitoes, with species and strain-specific variation in degrees of activity in response to Plasmodium infection.

motif. This suggests that CTLMA2 arose in a common ancestor of Anopheles and Aedes whereas CTL4 may be Anopheles-specific.
It has been shown that An. gambiae CTL4/CTLMA2 is stabilized by an intermolecular disulfide between cysteines of the N-terminal CXCXC motif; mutation of these three cysteines to alanine abrogates interchain disulfide formation 20 . To further refine the location of the interchain disulfide, we constructed nine mutants containing a single cysteine in the N-terminal CXCXC motif of CTL4 and CTLMA2, co-expressed the proteins in Sf9 cells, and performed Western Blotting to detect CTL4 and CTLMA2. Intermolecular disulfide formation was inefficient but evident on non-reducing SDS-PAGE in all cases (Fig. 2c).
This suggests that N-terminal intermolecular disulfide formation between CTL4 and CTLMA2 is promiscuous, rather than involving a specific cysteine residue from each protein. If intermolecular disulfide formation is promiscuous, CTL4/CTLMA2 heterodimers could in principle form multivalent disulfide-bridged oligomers. However, no disulfide-linked oligomers are detected on NR-SDS-PAGE for An. gambiae CTL4/CTLMA2 (Fig. 2a). Similarly, while CTL4 and CTLMA2 can form disulfide-bridged homodimers in vitro 20 , the purified product is overwhelmingly (>95%) heterodimer. Some minor bands (~10%) are observed on NR-SDS-PAGE for An. albimanus CTL4/CTLMA2 (Fig. S1b) that may represent homodimer or oligomer formation.
Both An. gambiae CTL4/CTLMA2 and An. albimanus CTL4/CTLMA2 are polydisperse in solution. Higher-order non-covalent oligomerization is evident as a shoulder of the main peak in SEC with increasing concentration, and is more pronounced for An. albimanus CTL4/CTLMA2 (Fig. S1). We confirmed the existence of oligomeric species by sedimentation-velocity analytical ultracentrifugation (AUC) (Fig. 2d). At pH 7.5 a series of species of decreasing intensity is observed in the c(s) distribution: s 1 = 3.1 s, s 2 = 4.4 s, s 3 = 5.9 s. Oligomerization is independent of Ca 2+ for A. gambiae CTL4/CTLMA2 but increases substantially with Ca 2+ for A. albimanus CTL4/CTLMA2 (Fig. S1b), and is not correlated with pH according to dynamic light scattering (DLS) (Fig. 2e).
Calcium binding. As CTLs, both CTL4 and CTLMA2 may bind calcium. However, the canonical Ca 2+ binding residues in CTL4 are mutated, suggesting it may be a CTLD but not a C-type CRD. Hence, we measured the calcium binding affinity of An. gambiae CTL4, CTLMA2 and the CTL4/CTLMA2 heterodimer of An. gambiae Unrooted phylogenetic trees for CTL4 (blue), CTLMA2 (red) in ten Anopheles species including An. gambiae (purple) and An. albimanus (cyan). The N-terminal portion of both multi-sequence alignments is shown with conserved cysteine residues highlighted yellow, other conserved residues highlighted in gray. The CXCXC motif is conserved in all sequences except those truncated at the N-terminus. and An. albimanus by isothermal titration calorimetry (ITC). Under equivalent conditions, binding was observed for CTLMA2 and CTL4/CTLMA2, but not for CTL4 ( Fig. 3a-c). Binding constants and thermodynamic parameters were calculated from the results of three independent experiments (Table 1). CTLMA2 Ca 2+ binding is well fit by a single site model with K D = 173 ± 27 μM, ΔH = 12 ± 2 kcal/mol. The affinity of An. gambiae CTL4/CTLMA2 for calcium is ~40 × higher than CTLMA2 with K D = 4.9 ± 0.5 μM, ΔH = −23 ± 4 kcal/mol. An. albimanus CTL4/ CTLMA2 (Fig. 3d) has a similar affinity for calcium with K D = 2.82 μM, ΔH = −12.1 kcal/mol. However, calcium binding to both An. gambiae and An. albimanus CTL4/CTLMA2 was sub-stoichiometric (N = 0.36-0.50).
Glycan binding. CTL4 and CTLMA2 belong to the lineage of myeloid CTLs -including macrophage mannose receptor (MMR) and DC-SIGN -that form a conserved family of immune receptors in metazoans 15,25 . Among CTLs with known structure, CTLMA2 has 30% sequence identity to the carbohydrate recognition domain (CRD) of mouse scavenger receptor (SCRL, PDB ID 2OX9) 26 and porcine surfactant protein D (SP-D, PDB ID 4DN8) 27 . CTLMA2 conserves residues associated with Ca 2+ binding in the glycan binding loop, and the canonical EPN motif associated with D-mannose selectivity (Fig. 4a). In contrast, CTL4 lacks all residues associated with Ca 2+ binding, consistent with the fact that no binding was observed by ITC. The only CTL of known structure with considerable similarity to CTL4 is factor IX/X binding protein (X-bp) from the venom of the Chinese moccasin Deinagkistrodon acutus (1IOD). X-bp is a modified CTLD in which the glycan binding domain is replaced by a long loop that mediates dimerization to generate the Factor IX/X binding site. Hence, although some insect CTLs do not require calcium for glycan binding, it is unclear if CTL4 should bind glycans at all.
In order to define their lectin activity, we analyzed CTL4, CTLMA2 and the CTL4/CTLMA2 heterodimer on glycan arrays that display 367 unique glycan structures 28 . These studies demonstrated binding to a range of glycans ( Table 2). The monomers of CTL4 and CTLMA2 demonstrated binding to only four and six glycans, respectively, whereas the heterodimer bound 18 different glycans. There is appreciable difference between ligands Data is fitted to a spherical particle whose radius reflects the size distribution in solution, no trend is apparent in the pH range 6-9.5 for An. gambiae or An. albimanus CTL4/ CTLMA2 in either 1 mM EDTA or 10 mM CaCl 2 . Result from one of two independent experiments. (2019) 9:15191 | https://doi.org/10.1038/s41598-019-51353-z www.nature.com/scientificreports www.nature.com/scientificreports/  www.nature.com/scientificreports www.nature.com/scientificreports/ bound by the individual monomers and the heterodimer; 3/4 (75%) of glycans recognized by CTL4 and 2/6 (33%) of glycans recognized by CTLMA2 were not recognized by CTL4/CTLMA2.
To confirm the glycan array results and to determine binding preferences for the CTLs, SPR analysis was performed (Table 3). In almost all interactions the glycan array and SPR were in agreement for the presence of interactions with the SPR indicating the glycan array had four false negative results: CTLMA2 monomer with H-antigen, chondroitin sulfate and chondroitin-6-sulfate, and CTL4 monomer with chondroitin-6-sulfate. However, all of the false negative results showed binding on the array with the CTL4/CTLMA2 heterodimer. www.nature.com/scientificreports www.nature.com/scientificreports/ The CTLs did not recognize mannose-containing glycans, with the exception of mannose-6-phosphate by CTL4/CTLMA2, despite the canonical EPN motif present in CTLMA2. Rather, the structures recognized are generally glycosaminoglycan (GAG) motifs comprising β1-3/β1-4 linkages between glucose (Glc), galactose (Gal) and their respective hexosamines GlcNac and GalNac. The Galβ1-4Glc linkage was present in 12/23 (52%) of glycans recognized, including 6/23 (26%) containing Galβ1-4GlcNac and 4/23 (17%) containing the keratan motif Galβ1-4GlcNac β1-3Gal or GlcNac β1-3Galβ1-4Glc.

Structural analysis.
To further probe the structure of CTL4 and CTLMA2, we generated structural models of CTL4 and CTLMA2 (Fig. 4b) using MODELLER 29 with additional manual editing. Both CTL4 and CTLMA2 have an extended loop (loop 1) following the second helix of the CTLD (Fig. 4a) with a high density of complementary charged residues; basic residues for CTL4 and acidic residues for CTLMA2. The complementary electrostatics of these loop 1 residues and their proximity to the N-terminal CXCXC motif suggest this is a potential protein/protein interface within the heterodimer (Fig. 4b), for which a hypothetical model was generated with a single disulfide bond in the CXCXC motif. However, the glycan/Ca 2+ binding loops are a second potential interface, as observed for the two chains of D. acutus X-bp. The alternate hypothesis is that the two CTL domains are independent of one another, with flexible linkers joining them via the intermolecular hypothesis.
To test this hypothesis, we analyzed the solution structure of CTL4/CTLMA2 by small angle x-ray scattering (SAXS). Experiments were conducted in 0.5 M NaCl, 20 mM CHES pH 9.0, 0.5 mM CaCl 2 , and 1% glycerol to minimize interparticle interactions. In these conditions, the protein displayed a linear relation between extrapolated intensity at zero angle and concentration (I 0 vs. c) up to a concentration of 3.1 mg/ml. The buffer-subtracted curve of intensity I vs. q (Fig. 4c) was submitted to SAXSMoW2 which yields R G = 23.0 Å (I 0 = 0.61) and molecular weight MW = 39 kDa, only 11% greater than the expected heterodimer MW of 35 kDa 30 . However, the calculated Guinier plot is based on only seven data points; fitting over an extended range (Fig. 4c, inset) yields R G = 24.5 Å (I 0 = 0.64), while fitting the pairwise distribution function P(r) (Fig. 4d) yields an R G = 25.4 Å (I 0 = 0.65). The P(r) distribution fit by DAMMIF 31 with 20 ab initio bead models with an average normalized structural discrepancy NSD = 1.0 ± 0.2. The main body of the ab initio models are similar in shape to the expected CTL4/CTLMA2 heterodimer.
Additional density extends from the main body of the bead models that can reflect either the N-terminal sequence of both proteins including the CXCXC motif, or a minor population of CTL4/CTLMA2 with a higher radius gyration. To test this hypothesis we performed multi-state modeling of the SAXS profile with the program MULTIFOXS 32 . We generated a complete model for CTL4-6xHis/CTLMA2 with an N-terminal coiled-coil terminated by an intermolecular disulfide between CTL4 C39 and CTLMA2 C34. MULTIFOXS generated an ensemble of 10,000 variants of the model to compare to the experimental scattering curve. The only flexible residues were CTL4 40-45 and 178-183 (6xHis) and CTLMA2 35-39, with an N-terminal coiled-coil serving as a rigid body connecting the two chains. The best one-state model fit the data with χ 2 = 1.13 and had R G = 23.7 Å. The minimum χ 2 = 1.07 was achieved with a 3-state model (Fig. 4e), in which 80% of the scattering is contributed by two compact models with R G = 22.0 Å And R G = 23.7 Å. This data is consistent with formation of a compact heterodimer between CTL4 and CTLMA2. CTL4/CTLMA2 inhibit phenol oxidase activation in response to E. coli. CTL4 and CTLMA2 function as inhibitors of the mosquito melanization response to infection. It was previously reported that CTL4 knockdown did not lead to increased phenol oxidase (PO) activity in response to infection with a mixture of E. coli and S. aureus 20 . The same study however, found that dsCTL4 and dsCTLMA2 mosquitoes were specifically susceptible to Gram-negative bacteria. Hence, we re-examined the effect of CTL4 and CTLMA2 knockdown on hemolymph PO activity in response to only E. coli infection (Fig. 5a). At 4 h post-infection with E. coli, PO activity was significantly enhanced for dsCTL4 (p = 0.02) and dsCTLMA2 (p = 0.004) mosquitoes compared to dsLacZ controls (Fig. 5b). The average knockdown efficiency for CTL4 and CTLMA2 was 93 ± 3% and 89 ± 3%, respectively, with >80% in any single experiment.
Melanization of Plasmodium ookinetes upon CTL4/CTLMA2 silencing is dependent on LRIM1 17,22 , TEP1 33 and SPCLIP1 33 . Since these are all elements of the TEP1 complement-like immune response, we reasoned that enhanced PO activity in the absence of CTL4 should be TEP1-dependent. Accordingly, we compared the enhancement of PO activity in dsCTL4 mosquitoes with co-administration of dsTEP1 to co-administration of dsLacZ (Fig. 5c). The average knockdown efficiency for TEP1 was 82 ± 9% and >80% in 5/6 experiments (64% in one experiment). Indeed, there was no significant enhancement of PO activity in dsCTL4 mosquitoes when TEP1 was also silenced. This confirms that melanization in the absence of CTL4/CTLMA2 is TEP1-dependent.
Co-administration of recombinant CTL4/CTLMA2 with E. coli significantly reversed the enhancement of PO activity in dsCTL4 mosquitoes compared to BSA (p = 0.007, Fig. 5d). These results demonstrate that CTL4/ CTLMA2 is directly involved as a negative regulator of PO activity. In dsLacZ mosquitoes however, CTL4/ CTLMA2 did not suppress E. coli-induced PO activity compared to bovine serum albumin (BSA). Also, E. coli-induced PO activity in dsCTL4 mosquitoes co-injected with recombinant CTL4/CTLMA2 (dsCTL4 + CTLs) was higher than that of dsLacZ mosquitoes co-injected with BSA (dsLacZ + BSA). Hence, injecting recombinant CTL4/CTLMA2 does not completely rescue the loss of endogenous protein.

Discussion
Here we describe the biochemical characterization of two C-type lectins, CTL4 and CTLMA2, that play key roles as negative regulators of the melanization cascade in the malaria vector An. gambiae. This is relevant to malaria transmission as knockdown of CTL4 and CTLMA2 results in reduced susceptibility to P. berghei and, with comparable infection levels, P. falciparum. The effect of CTL4 and CTLMA2 on Plasmodium is species-dependent, Scientific RepoRtS | (2019) 9:15191 | https://doi.org/10.1038/s41598-019-51353-z www.nature.com/scientificreports www.nature.com/scientificreports/ suggesting divergent evolution of the trait during speciation within the Anopheles genus. Here we focus on the common molecular features of these proteins conserved throughout Anopheles.
Heterodimerization of CTL4 and CTLMA2 is mediated by promiscuous intermolecular disulfide bonding of an N-terminal CXCXC motif conserved throughout Anopheles and, for CTLMA2, the Aedes genus. Aside from Zn-binding sites in DNA/RNA binding proteins, the CXCXC motif is uncommon in proteins of known structure. At time of publication the motif is present in only 74 representative protein-only structures at <4.0 Å resolution in the Protein Databank. The sequences are generally linear with disulfide bonds (if any) to different strands. Angepoietin-1 has an internal disulfide bond between the first and second cysteines of a CXCXC motif, with the backbone carbonyls serving as ligands to a calcium binding site. This is not observed for CTL4/CTLMA2 since we only observe a single Ca 2+ binding site by ITC corresponding to the CTLMA2 Ca 2+ /glycan loop. www.nature.com/scientificreports www.nature.com/scientificreports/ Steric hindrance tends to limit the formation of multiple disulfide between adjacent strands of closely spaced cysteine residues. However, intramolecular disulfides involving a CXC motif are rare but known within proteins. When CTL4 and CTLMA2 were restricted to a single cysteine within the CXCXC motif, neither protein had a reproducible preference for a single cysteine comparing multiple independent experiments. However, it may still be that a specific intermolecular disulfide is preferred in the context of the wild-type heterodimer produced in vivo.
Oligomerization of C-type CRDs is a common feature of CTLs in the immune system 13,14 . The collectins MBP, SP-A, and SP-D form trimeric structures mediated by an N-terminal cysteine-rich region followed by a collagen-like domain. DC-SIGN tetramerizes on cell surfaces via an extended neck region [34][35][36] . We see no evidence of higher order disulfide-mediated oligomerization of the CTL4/CTLMA2 heterodimer. However, there is clear evidence that CTL4 and CTLMA2 form non-covalent higher-order oligomers by SEC and AUC, especially so for An. albimanus CTL4/CTLMA2 in high calcium concentrations (10 mM). The constant ratio of sedimentation coefficients measured by AUC s 2 /s 1 = s 3 /s 2 = 1.4 is consistent with serial (1, 2, 4, …) oligomerization but no details of the structure or mechanism of this association is yet determined.
The Ca 2+ /glycan binding loop is mutated in CTL4, which does not bind Ca 2+ but may play a role in oligomerization as observed for the dimerization loop of X-bp. For CTLMA2 Ca 2+ affinity is enhanced in the presence of CTL4 suggesting a physical interaction either directly or allosterically alters its structure. The CTL4/ CTLMA2 Ca 2+ binding curve is best fit by substoichiometric binding. This could results from residual calcium either incompletely removed by EDTA or reintroduced before the ITC measurement. Or, the Ca 2+ binding site in CTLMA2 could be partially occluded by misfolding, presence of another divalent metal, oligomerization, interaction with CTL4 or perhaps bound glycans carried over during purification. The observed polydispersity of the recombinant heterodimer may contribute to these experimental factors.
CTL4 and CTLMA2 display several properties analogous to those of collectins. They are oligomeric serum proteins with a protective phenotype vs. Gram negative bacteria. Yet CTLMA2 displays the binding attributes of a selectin. The second calcium binding site in MBP, SP-D, and DC-SIGN is disrupted by mutation of conserved Asn/Asp residues (MBP D188, D194) to His and Arg (CTLMA2 H132, R144). CTLMA2 also binds Lewis A / Lewis X antigens as do selectins; the similar affinity of CTL4/CTLMA2 for Lewis A /Lewis X suggests this interaction solely involves CTLMA2. CTLMA2 binding of sialylated and sulfated Lewis structures likely involves direct ligation of fucose to calcium, as illustrated by the selectin-like mutant of MBP 37 and DC-SIGN 38,39 .
Binding of polymeric GAGs by CTL4 and CTLMA2 may be relevant to their physiological role as inhibitors of melanization. Insect connective tissues are rich in acidic and neutral GAGs (a.k.a. mucopolysaccharides, mucins), which serve to cement hemocytes together to encapsulate parasites and foreign bodies in the hemocoel 40 . CTL4/ CTLMA2 may line connective tissues in mosquitoes, inhibiting self-melanization in response to injury or infection. Such a function is analogous to association of vertebrate Factor H with sialylated surfaces, which serves to inhibit complement activation. Chondroitin and heparin sulfate are present in anopheline mosquitoes and are utilized by Plasmodium parasites during invasion 41,42 . If so, we speculate that pathogens, perhaps Plasmodium, could recruit glycosaminoglycans or CTL4/CTLMA2 directly to their surface as a means of inhibiting melanization by the host.
A number of questions remain unanswered. Does CTL4/CTLMA2 interact with specific proteins as well as glycans, and if so are such interactions glycan-dependent? Does CTL4/CTLMA2 inhibit melanization even after PO is activated, or does it simply inhibit the proteolytic activation of PO or an upstream factor? Is the binding of protein cofactors dependent on their conformation or proteolytic activation? The ability to reverse enhanced PO activity in dsCTL4 mosquitoes with co-administration of CTL4/CTLMA2 allows these questions to be addressed by site-specific alterations to the recombinant protein. Elucidating the mechanism of CTL4/CTLMA2 in the melanization response to infection can ultimately address its role in the natural susceptibility or refractoriness of different Anopheles species to Plasmodium infection.

Methods
Protein expression and purification. Full-length An. gambiae CTL4 (AGAP005335) and CTLMA2 (AGAP005334), An. albimanus CTL4 (AALB014534) and CTLMA2 (AALB005905) were obtained by total gene synthesis and subcloned into pFastbac1 with C-terminal 6 × His tag. For heterodimeric CTL4/CTLMA2 the genes were cloned into pFastbac-Dual with a C-terminal 6 × His tag on CTL4. Additional constructs were cloned into pFB-GP67-Hta with a TEV-cleavable 6 × His tag. www.nature.com/scientificreports www.nature.com/scientificreports/ Western blotting. Sf9 cells were infected at MOI of 0.1 and conditioned media collected 96 hpi. For non-reducing SDS-PAGE DTT was excluded from the 6 × loading buffer; samples were not heated prior to electrophoresis on 4-20% mini protean TGX precast gels (BioRad). Gels were transferred to Odyssey nitrocellulose membrane (LI-COR) at 30 V overnight in Towbin buffer with 10% methanol; but 1% SDS in the electrophoresis buffer was necessary for efficient transfer of the heterodimer in non-reducing conditions. Western blotting was performed with anti-6 × His mouse mAb (Clontech), with secondary IRDye 800CW goat anti-mouse (LI-COR) for. An. gambiae CTL4.
For detection of CTLMA2, full length His-tagged CTLMA2 was expressed in sf9 cells and purified from conditioned media by Co (Talon) affinity chromatography. The protein was used for inoculation into a rat for production of polyclonal antibodys by Cocalico Biologicals (Stevens, PA). The resulting antiserum was affinity purified using recombinant CTLMA2 immobilized on Aminolink resin (Thermo Fisher). Western blotting was performed with affinity purified anti-CTLMA2, with secondary IRDye 680LT goat anti-rat (LI-COR).
Analytical ultracentrifugation (AUC). Sedimentation velocity AUC was performed on a Beckman XL-I ultracentrifuge with absorbance optics set at 280 nm. First An. gambiae CTL4/CTLMA2 was analyzed under same conditions as for SEC (Fig. 2d). To test effect of calcium (Fig. S1b) Analysis of glycan binding by glycan array. Glycan arrays were produced and the glycan library used for screening as previously described 28 . Results were derived from a single batch of each protein, values are the average of four technical replicates. Glycan arrays were probed with 1 µg of 6 × His-tagged CTL4, CTLMA2 and CTL4/CTLMA2 in PBS containing 1 mM CaCl 2 and 1 mM MgCl 2 . A mouse anti-His primary antibody was added (1:1 molar ratio with protein) followed by rabbit anti-mouse Alexa647 (1:0.5 ratio with primary), and goat anti-rabbit Alexa647 (1:0.5 ratio with secondary). The final volume applied to the array was 300 µl incubated for 15 min. Arrays were washed 3 times in PBS containing 1 mM CaCl 2 and 1 mM MgCl 2 . Arrays were scanned on an Innoscan 1100AL using the 632 nm laser. Binding was considered to be positive when values were 3 standard deviations above the background of the array.

Analysis of glycan binding by surface plasmon resonance (SPR). SPR was performed on a Biacore
T200 using a CM5 senor chip. Flow cell one was left blank with ethanolamine only blocking the NHS activated carboxydextran. The proteins were immobilized onto flow cell 2-4 (CTL4, CTLMA2, CTL4/MA2 respectively). Glycans were flowed from 160 nM to 100 µM across a 1:5 dilution. A 15 s enhancement injection of PBS containing 1 mM CaCl 2 and 1 mM MgCl 2 preceded the sample injection and a 60 second regeneration injection of 10 mM Tris 1 mM EDTA followed the sample injection. Analysis was performed using the Biacore T200 evaluation software using the surface bound menu; Affinity; Steady state affinity. Glycan were run in triplicate with glycans run in single replicates with the program repeated three times.
Small-Angle X-ray scattering. SAXS data was collected in-house on a BioSAXS-2000 (Rigaku Corp.) with a DECTRIS PILATUS 100 K detector. Samples were prepared as a dilution series from 1-4.5 mg/ml in 0.5 M NaCl, 20 mM HEPES pH 9.0, 0.5 mM CaCl 2 , 1% glycerol. Data acquisition was 30 min collected as 5 min exposures to ensure no measurable radiation-induced changes within the aquisition period. Following conversion to I vs. q curves, primary data analysis using the ATSAS software package 20,44 . Buffer subtraction and Guinier analysis was performed with PRIMUS, P(r) calculation with GNOM. Construction of ab initio bead models were derived for each structure was performed with DAMMIF using D max = 80 Å, P1 symmetry.
Mosquito rearing and maintenance. The An. gambiae G3 strain was obtained through BEI Resources, NIAID, NIH: An. gambiae, Strain G3, MRA-112, contributed by Mark Q. Benedict. Mosquitoes were reared on (2019) 9:15191 | https://doi.org/10.1038/s41598-019-51353-z www.nature.com/scientificreports www.nature.com/scientificreports/ a 12 hr light/dark cycle at 28 °C and 75% relative humidity. Larvae were provided fish flakes (Tetramin) and dry cat food (Friskies) until pupation. Adults were maintained on 10% sucrose and fed sheep blood (HemoStat Laboratories, #SBH100) for egg production. Experiments were performed with 2-3 d old adult females from independent generations. Gene silencing by RNAi. T7 promoter-tagged templates for dsRNA synthesis were generated from a clone in plasmid pIB (LacZ), a clone in plasmid pIEx-10 (CTLMA2), and from mosquito cDNA (CTL4 and TEP1) using the iProof High-Fidelity PCR Kit (Bio-Rad, #1725331) and purified with the GeneJET PCR Purification Kit (Thermo Fisher Scientific, #K0701) according to the manufacturer provided instructions. Double stranded RNA reactions were performed using the HiScribe T7 High Yield RNA Synthesis Kit (NEB, #E2040S) and purified with the GeneJET RNA Purification Kit (Thermo Fisher Scientific, #K0731) according to the manufacturer's instructions. Double stranded RNA was reconstituted in ultrapure water to 3 μg/ul for microinjection. For single and double gene knockdown experiments, mosquitoes were injected with 69 or 138 nl dsRNA, respectively. Experiments were performed 3-4 d after dsRNA injection. Gene silencing efficiencies were performed as described previously 19 with the following modifications: cDNA was synthesized from 1 μg total RNA using the iScript cDNA Synthesis Kit (Bio-Rad, #1708891) following the manufacturer's instructions and qRT-PCR was performed with a QuantStudio 6 Flex Real-Time PCR System (Applied Biosystems) using PerfeCTa SYBR Green SuperMix, Low ROX (Quanta BioSciences, Inc. #95056-500). T7 template and qPCR primer pair sequences are provided as Supporting Information (Table S1).
Phenol oxidase (PO) activity assay. PO activity was measured in mosquito cohorts 4 h after the injection of E. coli strain DH10B or a mixture of E. coli and protein. Mid-log phase E. coli was rinsed and resuspended in PBS to OD 0.8 (Average dose: 124,063 CFU/μl). For PO assays involving the co-administration of E. coli and protein, bovine serum albumin (BSA, Sigma #B2518) or recombinant An. gambiae CTL4/CTLMA2 was combined with E. coli to a final concentration of 2.5 µg/µl just prior to injection. Mosquito injections, hemolymph collection, protein quantification, and the PO activity assay were performed as described previously 20 with the following modifications: (1) For PO assays following the injection of E. coli or a mixture of E. coli and protein, PO activity was assessed using 4-5 µg of hemolymph protein or the total hemolymph protein obtained from 100 mosquitoes, respectively. (2) Absorbance at 492 nm was recorded every 10 mins for 1 hour in a Molecular Devices SpectraMax 190 plate reader.

Data availability
The small-angle x-ray scattering dataset generated analysed in the current study is available in the SASBDB repository as entry SASDFL4, https://www.sasbdb.org/data/SASDFL4.