Introduction

Malaria is a deadly mosquito-borne disease that caused an estimated 263 million cases and 597,000 deaths in 2023, with the majority occurring in sub-Saharan Africa1. Frontline control measures, such as pyrethroid-treated bed nets, are increasingly threatened by widespread and rising pyrethroid resistance in major vectors like Anopheles gambiae1,2. This escalating resistance crisis underscores the urgent need for alternative, resistance-informed vector control strategies1,3. One such alternative is pirimiphos-methyl (PM), an organophosphate (OP) insecticide used in Indoor Residual Spraying (IRS) formulations such as Actellic® 300CS, which has gained widespread deployment due to its prolonged residual activity and historically low resistance levels in Anopheles populations4,5,6. However, emerging reports from West Africa indicate increasing resistance to PM. The development of such resistance is likely accelerated by the extensive genetic diversity within An. gambiae populations, which facilitates rapid adaptation to insecticide pressure7,8. Early identification of resistance mechanisms, combined with proactive surveillance, is therefore essential to contain the spread of resistance and sustain the efficacy of current interventions.

PM is a pro-insecticide requiring bio-activation by cytochrome P450 enzymes, monooxygenases (P450s), a large detoxification enzyme family, into its toxic metabolite, pirimiphos-methyl oxon (PMO). PMO irreversibly inhibits acetylcholinesterase (AChE), encoded by the Ace1 gene, an enzyme essential for neurotransmission9,10,11. Resistance arises through multiple mechanisms, including mutations and duplications in Ace1 and the overexpression of detoxifying enzymes8,12,13,14,15,16,17,18,19,20,21,22. P450s detoxify organophosphates through oxidative metabolism. Carboxylesterases (CCEs) either hydrolyze ester bonds to inactivate the insecticide or sequester organophosphates through phosphorylation of the enzyme active site, thereby neutralizing their activity.

In An. gambiae, members of the CCE family (containing greater than 50 genes) have been strongly associated with resistance to organophosphates and carbamates, and in some cases, pyrethroids. Understanding the specific enzymatic interactions involved in resistance to PM and PMO is critical for developing effective resistance management strategies.17,23,24,25.

Conventional strategies for elucidating resistance mechanisms predominantly rely on transcriptomic and genomic profiling of resistant mosquito populations to identify candidate genes, followed by downstream functional validation26,27,28. While informative, these approaches are intrinsically retrospective, offering mechanistic insights only after resistance has become established. To address this, there is a critical need for more proactive functionally driven methodologies capable of detecting early resistance liabilities in phenotypically susceptible populations before they become widespread. In this study, we developed and implemented a chemoproteomics approach leveraging activity-based protein profiling (ABPP) to functionally map resistance-associated enzymes in An. gambiae before the onset of detectable resistance phenotypes (Fig. 1). Specifically, we employed fluorophosphonate (FP) probes that covalently label active serine hydrolases (SHs), enabling competitive displacement by PM and its oxidized metabolite PMO. This enabled direct profiling of functional insecticide targets and enzyme activity in the native PM-susceptible proteome, revealing early resistance drivers. Among the enzymes identified, we prioritized the carboxylesterase Coeae6g to investigate its role in PM resistance, as its function beyond known association with pyrethroid/carbamate resistance was unclear29. Corroborating our ABPP findings, subsequent biochemical validation and over-expression in transgenic mosquito strains confirmed that Coeae6g confers resistance to PM and mediates cross-resistance to malathion and additional insecticide classes. Together, these findings highlight ABPP as a robust and predictive platform that complements genomic approaches. By enabling early detection of resistance mechanisms, ABPP strengthens the foundation for targeted, evidence-based vector control strategies in malaria-endemic regions.

Fig. 1: Chemical proteomics workflow to identify serine hydrolase targets of the organophosphate insecticide pirimiphos-methyl in An. gambiae.

a Homogenates from insecticide-susceptible An. gambiae (Kisumu strain) are prepared and centrifuged to obtain lysates. Samples are preincubated with the active organophosphate metabolite pirimiphos-methyl oxon (PMO), followed by treatment with a desthiobiotin-conjugated fluorophosphonate (FP) probe. FP-labelled active and heat-denatured homogenates serve as positive and negative controls, respectively. b SH labelling mechanism: FP probes target the conserved catalytic triad (Ser-His-Asp) of serine hydrolases (SHs), irreversibly binding to the nucleophilic serine residue. FP-labelled proteins from PMO-treated and control groups are enriched, purified, and on-bead digested for downstream peptide analysis. Proteins are analyzed by liquid chromatography-tandem mass spectrometry (LC-MS/MS) for identification and quantification. The resulting PMO-specific protein targets are validated by comparison of intensity signals in PMO-treated samples versus FP treatment and heat desaturated controls. c Chemical structures of FP probes and d organophosphate inhibitors, including pirimiphos-methyl (PM) and pirimiphos-methyl oxon (PMO). Created with BioRender (Ismail, H. (2025); https://BioRender.com/y61m555) and chemical structures of probes and inhibitors were created using ChemOffice version 22.0.

Results

Cross validation of FP probes for profiling An. gambiae serine hydrolase activity

We employed a dual-probe ABPP strategy to profile serine hydrolase (SH) activity in the insecticide-susceptible An. gambiae Kisumu strain (Fig. 2). A TAMRA-tagged fluorophosphonate (T-FP) probe enabled in-gel fluorescence detection, while a desthiobiotin-FP probe facilitated affinity enrichment and MS-based target identification. T-FP labelling revealed strong, activity-dependent signals in mosquito homogenates, which were abolished to minimal background level by heat denaturation and absent in DMSO controls (Fig. 2a). Prominent bands at ~62–70 kDa, consistent with esterases, were markedly reduced upon pre-incubation with PM or its bioactive metabolite PMO at both 10 and 100 µM, indicating shared target engagement. Although differences between the two concentrations were subtle, dose–response analysis confirmed PMO’s higher potency (IC₅₀ ~0.83 µM) relative to PM (IC₅₀ ~10.55 µM; Supplementary Fig. 1). Notably, PMO also diminished a distinct ~90 kDa band not affected by PM, suggesting unique target binding. The desthiobiotin-FP probe effectively competed with T-FP at both 1× and 5× concentrations, validating its use for proteomic enrichment, although minor differences in ~28–38 kDa bands were observed. These findings demonstrate the robustness and complementarity of both probes for SH profiling in An. gambiae.

Fig. 2: Proteome-wide mapping of fluorophosphonate-reactive targets in An. gambiae.

a Fluorescent gel image showing TAMRA-FP (tetramethylrhodamine fluorophosphonate) probe labelling of An. gambiae adult female lysates under various conditions. Lanes 1–12 represent treatments with DMSO (control), a non-fluorescent control probe desthiobiotin-conjugated FP probe (2 or 10 μM), competitive inhibitors (PM or PMO at 10 or 100 μM), or heat-denatured lysates subsequently labelled with 2 μM TAMRA-FP at 30 °C. Mosquito homogenate preparations and subsequent ABPP analyses were independently repeated twice with similar results (Source data file 1); one representative fluorescence image is shown. Red brackets highlight proteins affected by PMO pretreatment, while red arrowheads indicate proteins band where labelling was specifically inhibited by PMO but not PM. Blue arrowheads mark proteins unaffected by pretreatment with the desthiobiotin-conjugated FP probe. Controls include DMSO-treated sample (lane 1) and heat-denatured homogenates (Δ, lane 2), confirming the FP probe’s specificity for active homogenates. Molecular weight markers (kDa) are shown on the left. Coomassie blue staining (bottom panel) confirms equal protein loading across all samples. b Volcano plot showing differential protein enrichment in FP-treated samples relative to boiled controls. Each protein is represented by a dot, with the x-axis showing the log₂ fold change (FC) and the y-axis displaying the adjusted p-value. Serine hydrolases (SHs) are highlighted in red. Proteins enriched (log₂ fold change > 1, p < 0.05, two-sided t-test) are shown in black, negatively enriched proteins in yellow. c Scatterplot comparing mean abundance of proteins in FP-treated versus heat-denatured (Heat_FP) controls, showing a strong enrichment of SHs in active (non-boiled) samples. d Donut plot summarizing the proportion of FP-enriched proteins identified as serine hydrolases (24.5%, 73/298). e Protein domain enrichment analysis (InterPro) of the top 20 enriched serine hydrolases grouped by conserved functional domains to identify enzyme families that are disproportionately represented. Carboxylesterases, phospholipases, and acetylcholinesterases were the most represented, enzyme classes known organophosphate targets. The right panel lists top SHs by gene ID and domain type, providing functional context. The right panel lists the corresponding top-ranking SHs by gene ID and domain type, providing a functional context for the identified targets.

Proteomic profiling reveals diverse serine hydrolase activity in An. gambiae

To profile active serine hydrolases (SHs) in An. gambiae, we performed affinity enrichment using a fluorophosphonate (FP) probe, followed by LC-MS/MS and label-free quantification (LFQ) using a data-dependent acquisition approach. Streptavidin-enriched proteomes were analyzed under three conditions: active labelling (FP), heat-inactivated control (Heat_FP), and competitive inhibition (PMO_FP). LFQ analysis identified 380 proteins across conditions (Supplementary Data S9). Quality control metrics confirmed high reproducibility and condition-specific profiles (Supplementary Fig. 2). Comparing active samples to heat-inactivated controls revealed 298 proteins significantly enriched by the FP probe (adjusted p ≤ 0.05, log2 fold change (FC) ≥ 1), confirming activity-dependent labelling (Fig. 2b, c). Manual annotation based on functional keywords (e.g., hydrolase activity (KW-0378), exopeptidase, endopeptidase)30, identified 73 of these enriched proteins as SHs, representing 24.5% of the FP-labelled proteome (Fig. 2d and Supplementary Data 9). Network analysis of 283 significantly enriched proteins highlighted a densely interconnected cluster associated with cellular amide metabolism (Supplementary Fig. 3a). Furthermore, InterPro analysis identified the Alpha/Beta hydrolase fold (IPR029058), common feature of insecticide-metabolizing enzymes, in 32 of the positively enriched proteins (Fig. 2e). Crucially, known OPs targets including AChE1/29,16, various CCEs17,20,21,22, lysophospholipases31,32,33,34, and fatty acid synthases31,32,33,34, were identified, demonstrating the utility of FP probe for functionally relevant SHs to PM within the An. gambiae proteome.

Competitive ABPP identifies pirimiphos-methyl targets including key carboxylesterases

Having established the active SH profile, we next specifically analysed the competitive inhibition arm of the experiment to identify direct molecular targets of PM insecticide. We performed competitive ABPP by comparing FP probe labelling in proteomes pre-incubated with the bioactive form of the insecticide, PMO, versus those without pre-incubation (active samples). This analysis identified 23 proteins, including 18 putative SHs, with significantly reduced FP labelling upon PMO treatment (log₂FC ≥ 1, adjusted p ≤ 0.05; Fig. 3a). These targets were generally less abundant than other probe-reactive proteins, ranging from low to moderate expression levels (Supplementary Fig. 4). Notably, the established toxicity targets AChE1 and AChE2 were strongly inhibited9. PMO also targeted eight CCEs; four of these-Coeae5g, Coebe2o, Coe09916, and Coeae6g-showed the most profound inhibition, approaching background levels (Fig. 3b). Network analysis illustrated in Supplementary Fig. 3b, contextualized these interactions, revealing several inhibited enzymes, including AChE1/2 and the CCEs Coeae5g, Coeae6g, and Coebe2o, as central hubs with predicted links to diverse insecticides and metabolites. Among these, Coeae6g protein warranted further investigation, as its role in OPs resistance remained unclear, particularly given that field populations carrying Ace1 mutations and overexpressing Coeae6g gene remain susceptible to malathion29. We therefore prioritized Coeae6g orthologs from An. coluzzii and An. gambiae for functional validation to assess their contribution to PM resistance and potential cross-resistance to malathion, carbamate, and pyrethroid insecticides.

Fig. 3: Identification of potential targets of pirimiphos-methyl oxon (PMO) toxicity in An. gambiae using ABPP.

a The volcano plot illustrates changes in protein abundance between An. gambiae samples treated with fluorophosphonate (FP) and those exposed to PMO. Each protein is represented by a dot, with the x-axis showing the log₂ fold change (FC) and the y-axis displaying the adjusted p-value. Proteins with significantly reduced (Pos: positive enrichment) labelling upon PMO exposure (log₂ fold change > 1, p < 0.05, two-sided t-test) are highlighted in red, indicating potential direct targets of PMO toxicity. b Among the 23 proteins inhibited by PMO, 23 were mapped to VectorBase software (release 68), and 18 of these, classified as SHs, are displayed in the dot plots. The dot plots illustrate the log₂ intensity values for these 18 SHs across three conditions: FP treatment, PMO FP (PMO treatment following FP labelling), and Heat FP (heat-inactivated sample with FP labelling). Log₂ intensity reflects the abundance of probe-labelled (i.e., active) protein detected by mass spectrometry; a decrease in signal after PMO treatment indicates target engagement and inhibition. Each dot represents an individual replicate, with colours indicating the number of replicates (n = 3). Statistically significant differences between conditions are marked by asterisks (*p < 0.05, **p < 0.01, ***p < 0.001, two-sided t-test). The SHs identified in the dot plots are shown in red, with esterases denoted by an asterisk (*). Other SHs are labelled with their respective protein names, manually retrieved from VectorBase, and displayed in each box. Proteins with longer names are abbreviated using their initials, such as Adcp-12 (abhydrolase domain-containing protein 12) and VCPLP (vitellogenic carboxypeptidase-like protein).

Coeae6g is localized in the cytoplasm

Computational tools (TargetP-2.0, Phobius, MitoFates, iPSORT), predicted a mitochondrial localization for Coeae6g protein. To investigate this, we performed crude subcellular fractionation from thoracic tissues of female An. gambiae mosquitoes. The tissue was selected as flight muscles, involved in energy intensive processes, are located there, thus total mitochondrial mass is expected to be higher, compared to other tissues35,36,37. Isolated fractions were analysed by Western blot using antibodies against Coeae6g enzyme (Supplementary methods), a mitochondrial marker (anti-ATP5A), and a cytosolic marker (anti-alpha-tubulin). Coeae6g signals were much stronger in the cytosolic fraction, contradicting computational predictions and indicating localization in a cytoplasmic compartment (Fig. 4). Control antibodies gave the expected pattern, albeit with low cross-contamination of mitochondria in the cytosolic fraction as picked up by the anti-ATP5A antibody. This cytoplasmic localization aligns with its role in insecticide detoxification, consistent with xenobiotic metabolism occurring mainly outside of mitochondria38,39,40.

Fig. 4: Western blot analysis of cytosolic and mitochondrial protein fractions of mosquito thoraxes indicating Coeae6g sub-cellular localization.

ATP5A and alpha-tubulin are used as mitochondrial and cytosolic markers, respectively. Approximately 30 μg of total protein deriving from the same experiment were loaded from each fraction, in two separates, simultaneously processed, 10% SDS gels, and immuno-blotted against the three protein targets; a single nitrocellulose membrane was stained initially against Coeae6g and, after mild stripping using a low pH glycine solution and re-blocking, it was re-blotted against alpha-tubulin. Mitochondria isolation experiments and subsequent western blot analyses were independently repeated twice, yielding similar findings.

Inhibition kinetics of recombinant Coeae6g with different insecticides

The protein sequences of Coeae6g from An. gambiae and An. coluzzii are highly homologous (93.8% identity), carrying nine amino acid substitutions, of which two are non-conservative (Supplementary Fig. 5). These mutations do not lie in the conserved catalytic triad. We chose the An. coluzzii Coeae6g to express in Sf9 insect cells, using the baculovirus system. Successful expression was confirmed by Western blot analysis (Supplementary Fig. 6). Resuspended cell pellets expressing Coeae6g enzyme displayed 8-12-fold increased esterase activity towards the model substrates α-naphthyl acetate (α-NA), β- naphthyl acetate (β-NA) and p-nitrophenyl acetate (p-NPA), compared to YFP-expressing and uninfected Sf9 cells (Supplementary Table 1). The kinetic parameters for α- NA and β- NA were determined as: Km 68.49 μΜ (95% CI: 57.19–81.73) and Vmax 307.7 nmol min−1 mg total protein−1 (95% CI: 297.3–318.3); and Km 369.4 μΜ (95% CI: 306.0–446.5) and Vmax 253.0 nmol min−1 mg total protein−1 (95% CI: 238.2–269.3), respectively (Supplementary Table 1).

Activity inhibition assays towards α- NA, having shown higher affinity to Coeae6g than β-NA, were then performed to quantify the interaction of a number of insecticides (and their activated forms) with recombinant Coeae6g (Table 1). PMO displayed the strongest inhibition (IC**50 of 0.03 μΜ (95% CI: 0.024–0.031)), followed by malaoxon (0.24 μΜ (95% CI: 0.23–0.25)). No inhibition was detectable by their pro-insecticide forms, PM and malathion. Two carbamate insecticides were shown to act as lower affinity inhibitors, bendiocarb had an IC**50 value of 7.27 μΜ (95% CI: 6.43–8.19) and propoxur an IC**50 of 3.52 μΜ (95% CI: 3.06–4.05). Lastly, permethrin exhibited a very low inhibitory effect (IC**50 129.7 μΜ; 95% CI: 127.5–131.9), while no inhibition was observed by deltamethrin.

In vivo functional validation of the role of Coeae6g gene in insecticide resistance

Establishment of UAS-Coeae6g An. gambiae responder lines

The functional expression of Coeae6g gene in An. gambiae was performed in two independent labs using the GAL4 system. The coding sequence of the gene was amplified from two strains: an insecticide-susceptible An. gambiae (Kisumu) and an insecticide-susceptible An. coluzzii (NGousso). Responder lines carrying the An. gambiae (UAS-AngCoeae6g) or An. coluzzii (UAS-AncCoeae6g) Coeae6g coding regions under UAS control and marked with YFP (under the 3 × P3 promoter), were established by integrase-mediated transgene exchange into the genomic sites of docking lines Ubi-GAL441 and A11 (previously described in Lynd et al.)42, respectively, as depicted in Supplementary Fig. 7.

Coeae6g is highly expressed in transgenic An. gambiae lines

The UAS-Coeae6g responder lines were crossed with Ubi-GAL4, a driver/docking line expressing GAL4 under the An. gambiae polyubiquitin promoter, that directs widespread tissue expression. Coeae6g transcription levels were assessed in the progeny and compared to the native expression levels in the respective docking lines by qPCR (Supplementary methods). As shown in Supplementary Fig. 8a similar levels of overexpression were observed for the An. gambiae (71-fold) and An. coluzzii (81-fold) Coeae6g genes compared to controls (p < 0.01). An. coluzzii and An. gambiae Coeae6g protein abundance was additionally verified by Western blot analysis in whole transgenic female mosquito homogenates (Supplementary methods; Supplementary Fig. 8b, c), whereas total carboxylesterase activity was assayed in whole extracts from An. gambiae Coeae6g overexpressing mosquitoes and shown to be significantly higher against both standard esterase substrates, p-NPA (2.9-fold, p < 0.0001) and β-NA (2.3-fold, p < 0.0001) than control parental mosquitoes (Supplementary methods; Supplementary Table 2).

Coeae6g over-expressing An. gambiae are resistant to PM

Initially standard WHO tube diagnostic bioassays43 were used to assess the impact of Coeae6g over-expression on resistance against PM. PM resistance was observed in both transgenic lines (An. gambiae gene 67% mortality and An. coluzzii gene 79% mortality) (Fig. 5). To further quantify the level of PM resistance conferred by Coeae6g over-expression, we performed dose dependent or time dependent assays to calculate relative resistance ratios (RRs) for the An. gambiae and An. coluzzii gene over-expressing progeny, respectively (Table 2). Very similar RRs for PM exposure of 8.6 and 7.6 were observed for the two lines. Further work was performed with the AncCoeae6g line to investigate cross-resistance to other insecticides.

Fig. 5: Estimation of the effect of multi-tissue Coeae6g up-regulation on insecticide sensitivity, using standard WHO bioassays.

Insecticide sensitivity was assessed in WHO discriminating doses upon 1 h exposure, comparatively between a GAL4/UAS-An. gambiae (Ang)- Coeae6g and the control docking line Ubi-GAL4; and b GAL4/UAS-An. coluzzii (Anc)- Coeae6g and the control docking line A11. Individual values from n = 3 to 7 biological replicates (20–25 randomly picked females, from different batches) are shown and bars represent SEM n. Asterisk (*) denotes statistical significance between mortality rates of the two lines, estimated by Welch’s t-test (two-tailed), as: a p = 0.027; and b p = 0.046. Statistically non-significant differences are not indicated. The red dotted line corresponds to the 90% mortality WHO threshold indicating resistance. PM Pirimiphos methyl.

An. coluzzii Coeae6g over-expressing mosquitoes show cross-resistance to carbamates and the pyrethroid permethrin

One hour exposure caused 95–100% mortality for GAL4/UAS-AncCoeae6g and the control line for bendiocarb, propoxur, malathion and deltamethrin, while confirmed resistance was shown against permethrin (81.8% mortality) (Fig. 5b). Although Coeae6g overexpression did not confer resistance to the WHO discriminating dose and exposure time for the insecticides bendiocarb, propoxur and malathion, we tested the response of mosquitoes at shorter exposure times and observed clear differences between the GAL4/UAS-AncCoeae6g and the control line. To quantify the level of cross-resistance we performed time response assays and estimated the resistance ratios (RR50), which ranged between 2.2 and 4.6-fold (Table 3). For permethrin, where resistance was observed to the discriminating dose, a RR of 3 was observed. No resistance occurred against deltamethrin (exposure for 10 and 15 min resulted in 100% mortality for both strains, three replicates each).

Discussion

This study establishes a chemoproteomic framework for interrogating insecticide–protein interactions in insecticide-susceptible malaria vectors, advancing resistance biology beyond the retrospective scope of conventional genomic and transcriptomic approaches. While sequencing-based methods remain indispensable for cataloguing known resistance alleles, they often fail to capture dynamic and functionally relevant changes at the protein level, particularly those mediated by post-transcriptional regulation or enzyme activity modulation44,45,46. By directly probing the active proteome of a susceptible An. gambiae population with ABPP, we provide a forward-looking strategy to uncover mechanistic liabilities to new chemistries before resistance emerges in the field.

Our previous work demonstrated that ABPP using pyrethroid-mimetic probes could capture pyrethroid-metabolizing enzymes in rat liver microsomes and selectively label recombinant mosquito P450s47. In the present study, this approach was extended to intact mosquitoes, applying FP-based ABPP directly to An. gambiae to map active serine hydrolases within the native proteome. Using fluorophosphonate probes in an insecticide-susceptible strain, we systematically mapped catalytically active SH enzymes and identified enzymes with direct interactions with PM. This expands the application of ABPP beyond its established roles in pharmacology and toxicology30,31,48,49,50,51,52 into vector biology and resistance surveillance.

In-gel analyses of the insecticide-susceptible An. gambiae Kisumu strain confirmed selective and efficient labelling of active proteome, consistent with previous FP-based applications30,50,53, and validated the probe’s specificity in intact mosquitoes. Building on this validation, downstream pulldown proteomics resolved a baseline inventory of 73 catalytically active SHs, representing ~19% of the predicted repertoire. Because FP probes selectively label only catalytically active and accessible enzymes, shaped by tissue distribution, expression levels, and active-site conformation, this dataset provides a functional view of the enzyme landscape of which some are relevant to PM toxicity and resistance, such as AChEs and CCEs30,49. Competitive profiling with PMO narrowed SHs landscape to 18 proteins, including canonical OP insecticides targets such as AChE1/29,16, various CCEs17,20,21,22, lysophospholipases31,32,33,34, and fatty acid synthases31,32,[33](#ref-CR33 “Moriconi, D. E., Dulbecc