Contemporary status of insecticide resistance in the major Aedes vectors of arboviruses infecting humans

Both Aedes aegytpi and Ae. albopictus are major vectors of 5 important arboviruses (namely chikungunya virus, dengue virus, Rift Valley fever virus, yellow fever virus, and Zika virus), making these mosquitoes an important factor in the worldwide burden of infectious disease. Vector control using insecticides coupled with larval source reduction is critical to control the transmission of these viruses to humans but is threatened by the emergence of insecticide resistance. Here, we review the available evidence for the geographical distribution of insecticide resistance in these 2 major vectors worldwide and map the data collated for the 4 main classes of neurotoxic insecticide (carbamates, organochlorines, organophosphates, and pyrethroids). Emerging resistance to all 4 of these insecticide classes has been detected in the Americas, Africa, and Asia. Target-site mutations and increased insecticide detoxification have both been linked to resistance in Ae. aegypti and Ae. albopictus but more work is required to further elucidate metabolic mechanisms and develop robust diagnostic assays. Geographical distributions are provided for the mechanisms that have been shown to be important to date. Estimating insecticide resistance in unsampled locations is hampered by a lack of standardisation in the diagnostic tools used and by a lack of data in a number of regions for both resistance phenotypes and genotypes. The need for increased sampling using standard methods is critical to tackle the issue of emerging insecticide resistance threatening human health. Specifically, diagnostic doses and well-characterised susceptible strains are needed for the full range of insecticides used to control Ae. aegypti and Ae. albopictus to standardise measurement of the resistant phenotype, and calibrated diagnostic assays are needed for the major mechanisms of resistance.


Introduction
Arboviruses cause severe disease and death in humans with over 4 million disability-adjusted life years worldwide attributed to mosquito-borne viruses in 2013 [1,2]. Furthermore, a marked increase in mortality and morbidity associated with dengue virus infections has been noted over the last decade in contrast to a drop in other major infectious diseases [3]. Each arbovirus is transmitted by multiple species and the relative importance of each species is determined by their vectorial capacity and their geographical distribution. Both Aedes aegypti and Ae. albopictus mosquitoes are competent vectors of 5 important arboviruses (chikungunya virus, dengue virus, Rift Valley fever virus, yellow fever virus, and Zika virus) and both are widely distributed, making these species an important factor in the worldwide burden of infectious disease. There are currently no registered vaccines for 3 of these arboviral diseases, and although the Dengvaxia vaccine for dengue has now been registered in several countries, it is not yet in widespread use. Yellow fever is the exception, with a long-standing vaccine that is used worldwide; however, funds for routine vaccination are limited [4]. In addition, prophylaxis options are not available for these diseases, meaning transmission is not interrupted during the human infection phase. Vector control by insecticides coupled with larval source reduction is therefore absolutely critical in the prevention and control of these arboviral diseases.
Control of Aedes vectors is currently primarily based on insecticides and community engagement for habitat management. The use of alternatives such as Wolbachia infection, sterile insect techniques, and genetic manipulation is promising but they have only been tested in restricted locations worldwide. The long lead time needed for new control measures means that current insecticide-based approaches will continue to play a central role for many years to come. Control of Aedes vectors using insecticides, primarily through larviciding or space spraying of pyrethroids and organophosphates, is fraught with complications including high cost, slow operational response, low community buy-in, and ineffective timing of application [5][6][7]. A further major concern is the rapid spread of insecticide resistance with the potential to reduce efficacy of current insecticide-based control strategies [8], but there is a lack of accessible data documenting the prevalence and mechanisms of insecticide resistance at specific geographical locations. These data are important to guide national programmes in the choice of the most effective compounds to use in each resistance setting. In this context, the Worldwide Insecticide Resistance Network (WIN), supported by the World Health Organization (WHO), was established to track insecticide resistance in the vectors of the arboviruses at a global scale and to evaluate the potential for deployment of alternative vector-control interventions [9]. Here, we review the current knowledge of insecticide resistance in Ae. aegypti and Ae. albopictus worldwide and assess underlying resistance mechanisms. Evidence on the global distribution of resistance to the major classes of neurotoxic insecticide is assessed and the tools available to monitor resistance phenotypes and genotypes are reviewed.

Database of bioassay records linked to geographical locations
A previously compiled database that searched the published literature up to October 2008 was provided by its authors [8]. We then searched the published literature from January 2009 to August 2016 using the Web of Knowledge database of published articles and the search terms "resistance" and "aedes". Abstracts were scanned and full-text copies obtained for articles of potential relevance. Additional unpublished datasets were provided by members of the WIN. Data were extracted from all articles reporting either mortality or the lethal concentration required to kill half of the sample (LC 50 values) from bioassays of Ae. aegypti and/or Ae. albopictus mosquito populations using 1 or more of the 4 major classes of neurotoxic insecticide (carbamates, organochlorines, organophosphates, and pyrethroids). Only data linked to a field collection at a defined site(s) with no more than 4 generations in the laboratory were included. Bioassays that used a synergist were also excluded. A PRISMA flow diagram of the full process and details of how these data were then mapped are provided in S1 File.

Review of resistance mechanisms
Published studies of insecticide-resistance mechanisms were sought from Web of Knowledge using 3 sets of search terms: "Aedes" and ("pyrethroid" or "DDT") and "resistance"; "Aedes" and ("organophosphate", "temephos", or "carbamate") and "resistance"; and "Aedes" and ("kdr", "P450", "monooxygenase", "glutathione", or "esterase"). Because our focus was primarily to include studies capable of identifying specific genes and molecular markers associated with insecticide-resistance mechanisms, which can lead to diagnostic assays, studies solely investigating resistance mechanisms using biochemical assays or synergist tests were not reviewed unless complemented by more specific approaches. In studies of metabolic resistance, the number of genes identified as overexpressed in resistant populations or strains is highly variable, which in part relates to differences in the filtering strategies used to identify significance, and we have focused wherever possible on the genes or gene families for which replication was evident.

Results and discussion
The geographical distribution of the insecticide-resistance phenotype A database of 6,888 bioassay records was compiled including 1,190 records from the 2008 review [8] and 5,698 new records. All data derived from published articles and those unpublished data for which we received permission to share are provided in S2 File, and the total number of data records for each class of insecticide resistance is given in Table 1.
The geographical distributions of records of resistance/susceptibility to pyrethroids and organophosphates in Ae. aegypti and Ae. albopictus are shown in Fig 1. Good data coverage is evident for Ae. aegypti populations in much of the Americas and Asia but is lacking for Australasia and large parts of Africa. Less data is available for resistance to the main neurotoxic insecticides in Ae. albopictus with the exception of Southeast Asia. The amount of data available from the last decade is much greater than the amount available from the preceding decades, as shown in Fig B in S3 File. The data points shown on these maps, however, represent both susceptibility and dose-response bioassays performed on adults and larvae. To provide a more consistent measure of resistance, data for the most commonly used insecticide and test type were plotted and Fig 2 provides the results from bioassays that challenged adult Ae. aegypti with the most commonly tested pyrethroid, deltamethrin. These maps indicate that prevalence of resistance to deltamethrin is consistently high in Brazil and French Guiana and lower in the few locations in West Africa that have been sampled. Insecticide resistance appears to be highly variable or patchy in Southeast Asia. This may indicate that insecticide resistance is spatially heterogenous at a fine resolution but could also be the result of variation in test protocols, variation in collection dates, or small sample sizes. Apparent trends need to be treated with caution because these data do not come from a single study with a systematic sampling design and consistent methods. The dataset is highly biased, clustered in both time and space, and contains values generated using different methods.
For Ae. aegypti, the most commonly tested organophosphate was temephos using larval diagnostic dose bioassays, and Fig 3 shows the distribution of results from these assays in each region for which data were available. No assumptions have been made about the levels of resistance that equate to "susceptible", "possible resistance", "moderate resistance", "resistant", or "highly resistant", so the ratios obtained using the Rockefeller strain for comparison were classified into quartiles. The levels of resistance to temephos are higher in Brazil, French Guiana, and the Caribbean and lower at the few locations in West Africa that have been sampled. As seen for the pyrethroids, temephos resistance appears to be highly variable in southern and Southeast Asia. These apparent trends need to be assessed in the context of a number of potential confounding factors, as noted above.
For the other insecticide classes reviewed, resistance to the organochlorines is consistently high in Ae. aegypti populations whereas resistance to the carbamates is more variable, although evidence for resistance to the carbamates has been reported in Asia, Africa, and Latin America  Evidence for resistance to all 4 major classes of neurotoxic insecticides has been reported in Ae. albopictus populations from Southeast Asia, and resistance to the organophosphates has also been recorded in the Americas (Fig 4).

Gaps in the insecticide-resistance phenotype data
A comparison of the respective distributions of Ae. aegypti and Ae. albopictus with the full insecticide-resistance datasets available for each species over the last decade (2006-2015) highlights those locations lacking contemporary data (Fig 1). For Ae. aegypti, there are very large gaps in the data available in Africa and far southeastern Asia/Australasia, in contrast to insecticide resistance in malaria vector species where data are available across much of Africa [10,11]. Fewer data are available for Ae. albopictus and there are substantial data gaps on all continents. The maps in Fig 1 show the geographical spread of the data but combine multiple years and therefore mask any gaps in the temporal data coverage. The only region with good annual data coverage in the dataset presented here is Brazil (S2 File), although similar insecticide-resistance monitoring datasets may be held by other nations. In order to understand spatial heterogeneity and temporal trends, future analyses require good data coverage in both space and time and, ideally, standard measures of insecticide resistance. As shown in Table 1, there is currently considerable variation in the methods used to measure insecticide resistance in these 2 species and, in addition, Table A in S1 File shows that a diversity of strains with varying susceptibility have been used to calculate resistance ratios.

Mechanisms of insecticide resistance
Before considering the geographical distribution of the most influential mechanisms of resistance, specific mechanisms need to be identified and validated. Mechanisms of resistance to insecticides in mosquitoes can include reduced penetration of insecticides into the insect, nonsynonymous mutations affecting the proteins targeted by insecticides (target-site mutations), or enhanced enzymatic biodegradation or sequestration (metabolic resistance) [12,13]. Target-site mutations and metabolic resistance are thought to be the 2 main resistance mechanisms in Aedes mosquitoes. Reduced insecticide penetration is mainly caused by modifications of the insect cuticle, and though such changes have been identified in Anopheles species [14,15], this remains poorly understood as a resistance mechanism in mosquitoes, and its importance in Aedes species is yet to be confirmed.  Target-site resistance The proteins targeted by chemical insecticides play a key role in the functioning of mosquito nervous systems. Any nonsynonymous mutation in these proteins can potentially decrease the affinity of insecticides to their target protein and contribute to resistance. Target-site mutations affecting all chemical insecticide classes are widespread in insects and are often well-conserved across taxa, permitting meaningful cross-species searches for mutations [16,17].
Organophosphates and carbamates (acetylcholinesterase mutations). Resistance to acetylcholinesterase (AChE) agonists is common in Aedes (Figs 3 and 4  Pyrethroids and DDT (voltage-gated sodium channel mutations). Mutations in the voltage-gated sodium channel (VGSC), sometimes referred to as knockdown resistance (kdr) mutations, are very common in Ae. aegypti, with 10 mutations at 8 codon positions in VGSC domains II-IV, which have been found to comprise 15 haplotypes to date (Table A in S4 File). These mutations vary in frequency, geographical spread, and effects on resistance (Fig 5). The most widespread mutation in Ae. aegypti, 1534C, confers resistance to permethrin and deltamethrin when in combination with other mutations and is also linked with DDT resistance [28,29]. It is found in both Ae. aegypti aegypti and Ae. aegypti formosus and occurs across 3 continents, with sequence data from the nearby short intron suggesting a single origin followed by worldwide spread [30]. In contrast, the V1016I and V1016G mutations have distinct geographical distributions, with 1016G found in Asia (spanning as far as Saudi Arabia) and V1016I in the Americas (although with recent detection at low frequency in Ghana) [30]  (Table A in S4 File). Phenotypic effects of the 2 1016 mutations also differ. Prior to the discovery of F1534C, 1016I was thought to be a major cause of pyrethroid resistance [31, 32] but recent analyses, which also screened the F1534C position, found that the 1016I/1534C haplotype more strongly associates with pyrethroid resistance [33] and is increasing in frequency and range, whereas the 1016I/1534F haplotype is very rare [34,35]. Thus, 1016I appears to enhance the pyrethroid-resistance phenotype conferred by 1534C rather than being directly causal, and at present in the Americas, only the F1534C and I1011M mutations appear capable of directly causing resistance [36]. The Asian variant V1016G can cause resistance alone but is more potent when in combination with another mutation, 989P, with which it commonly cooccurs (Table A in S4 File), and especially in a triple mutant 989P/1016G/1534C haplotype, which can engender extreme resistance [37]. Fortunately, 989P/1016G and 1534C usually occur on alternate chromosomes, but the triple mutant haplotype has been detected at low frequency in at least 3 distinct Asian locations (Table A in S4 File), likely created by independent recombination events. Careful monitoring of the occurrence and frequency of these 3 mutations in Asia is a high priority.
Although most commonly associated with positions 989, 1016, and 1534, other, less frequent mutations also occur in combinations: D1763Y with V1016G, G923V with I1011M, and T1520I with F1534C, and to date, their effects on resistance are unclear. Individually, VGSC mutations typically appear to be recessive in Aedes [38][39][40] but this can depend on their combination; for example, a S989P/V1016G/F1534C heterozygote genotype confers moderate resistance to deltamethrin and may present an important step in the evolution of high-level resistance or a genotype providing benefit across different pyrethroid classes [41].
In Ae. albopictus, only 4 VGSC mutations have been detected to date, affecting 2 codons at positions 1532 and 1534 (Table A in S4 File). The I1532T variant has only been detected in Italy to date but the 3 mutations at codon 1534 (F to C, L, or S) have broad distributions spanning multiple continents, which might reflect the rapid invasive spread of the species rather than multiple mutant origins. Only the F1534S variant has been demonstrated to be moderately associated with resistance to DDT and pyrethroids [42,43]. Nucleotides at codons 989 and 1016 in the Ae. albopictus Folshan reference strain (VectorBase) are identical to the Ae. aegypti wild types, suggesting no intrinsic barrier to the evolution of elevated target-site resistance. Nevertheless, current evidence suggests that VGSC mutation is less common and, where present, may also link less strongly with pyrethroid resistance in Ae. albopictus.

Metabolic resistance
Metabolic resistance is caused by elevated activity via overexpression or conformational change of enzymes that are involved in the processes of insecticide metabolism, sequestration, and excretion. Detoxification pathways may be complex and whilst the primary metabolic enzymes involved belong to large gene families (cytochrome P450 monooxygenases [P450s], glutathione S-transferases [GSTs], and carboxy/cholinesterases [CCEs]), other families are likely to be involved [13,44]. Metabolic resistance is very common in mosquitoes and has been reported against all insecticides used in public health as well as agricultural pesticides that might be used for vector control in the future [44,45].
Cyclodienes (gamma aminobutyric acid receptor mutation). Despite withdrawal of the organochlorine insecticides targeting the gamma aminobutyric acid (GABA) receptor decades ago, the resistance to dieldrin (Rdl) mutation, A302S, has been detected in a cyclodiene-resistant strain of Ae. aegypti [46] and in wild populations of Ae. albopictus from La Reunion and Java [47,48]. The persistence of the A302S mutation at high frequency might be selected by long-term environmental persistence of organochlorine pesticides or negligible fitness costs [49]. However, intriguingly, a recent electrophysiological study raised the possibility that there might be some contemporary insecticidal selection, with evidence for antagonism of the GABA receptor by both imidacloprid (a neonicotinoid) and deltamethrin, for which it may be a secondary target [50].
Pyrethroids. Cytochrome P450s are involved in metabolism and detoxification of a wide range of compounds [51] and their overexpression is frequently associated with pyrethroid resistance in mosquitoes, although other detoxification enzymes are also implicated [12]. Multiple P450 genes, especially members of the CYP6 and CYP9 subfamilies, have been linked to resistance via overexpression in transcriptomic studies of insecticide-resistant versus susceptible strains (Fig A in S4 File). Whilst this suggests that a diversity of P450 genes may be involved, strains could be resistant to multiple insecticides or differ in other ways from susceptible colonies. Among the most consistently expressed genes, 4 (CYP9J10, CYP6BB2, CYP9J26, CYP6J28) have been proven to metabolise pyrethroids [52,53] or to confer pyrethroid resistance when expressed transgenically in Drosophila [54]. In transcriptomic studies linking more closely with pyrethroid-resistant phenotypes, whilst there is substantial variation in results among strains within and among studies, both CYP6BB2 and CYP9J28 are detected as significant across studies ( Table 2). The geographical distribution of results for functionallyvalidated genes is broader than is generally the case for VGSC mutations, with key candidate genes found to be significantly overexpressed across continents (Fig 6). In the sole study of Ae. albopictus, P450 genes again appear important, and although none of those overexpressed have been tested for metabolic activity, CYP6P12 conferred pyrethroid resistance in transgenic Drosophila melanogaster [55]. Whilst the majority of work has focused on gene expression, targeted sequencing of 760 candidate genes identified 55 nonsynonymous variants potentially linked to deltamethrin resistance in 10 P450 genes [56]. Studies of the association of expression of specific alleles of these genes with pyrethroid resistance in field populations are an important next step because the broad distributions of overexpression may have different underlying genetic causes, which also warrants studies testing allele specificity of metabolism. Interestingly, CYP6 and CYP9 genes were also the most commonly duplicated P450s, suggesting that copy number variation (CNV) plays an important role in differential expression phenotypes, and thus that design of informative DNA-based diagnostic assays could be possible.
Organophosphates. Quantitative trait locus mapping suggests the molecular underpinning of temephos resistance in Ae. aegypti is extremely multivariate [57]. This is supported by the diversity of genes highlighted in transcriptomic studies linking to temephos resistance, with multiple P450s, CCEs, and GSTs identified as overexpressed, although there is apparently greater consistency in the specific genes involved than for pyrethroid resistance ( Table 2). This consistency also extends across species, with some common genes overexpressed in temephosresistant Ae. aegypti and Ae. albopictus. Whilst the mode of action is unclear for most of these genes, CCEs catalyse the hydrolysis of ester bonds and have been frequently linked to both organophosphate and carbamate resistance, either through insecticide sequestration or direct metabolism [58]. In both Ae. aegypti and Ae. albopictus, the alpha-esterases CCEAE3A and CCEAE6A can be considered key candidates for temephos resistance, with up-regulation at least partially due to gene amplification. This is further supported by genetic crosses in Ae. albopictus that demonstrated a strong association between temephos survival and gene copy number [18,59]. In both species, CCEAE3A can strongly sequester and slowly metabolize the temephos oxon and is a widespread resistance mechanism (Fig 6) present in American and Asian populations of Ae. aegypti and in Ae. albopictus populations from Europe and the United States of America [60]. Alternative gene families contributing to resistance. Although the mode of action of P450s in pyrethroid metabolism is better understood, other enzyme families may be involved. For example, CCEs can hydrolyse pyrethroids, with subsequent metabolism of the secondary metabolites by P450s [61]. Other detoxification genes are also implicated by ribonucleic acid interference (RNAi) knockdown in Ae. aegypti, notably GSTE2 and GSTE7 in deltamethrin resistance [62]. Epsilon GSTs are more commonly associated with resistance to DDT through direct dechlorination and their mode of action in pyrethroid-resistant Ae. aegypti resistance is currently unclear but might involve a peroxidase-activity protection from oxidative stress [63]. Gene families beyond the 3 major superfamilies are also likely to be important. For example, UDP-glycosyltransferases (UGTs), which can conjugate a glycosyl group to electrophilic substrates, are repeatedly detected as overexpressed in Aedes strains resistant to both pyrethroids and temephos ( Table 2) and have been found amplified and affected by point mutations [56]. The activity of UGTs against hydrolysed pyrethroids has been demonstrated in rats [64] and activity-based probe fishing identified multiple UGTs interacting with pyrethroids [65]. Other enzymes potentially involved in insecticide resistance include adenosine triphosphate (ATP)binding cassette (ABC) transporters, sulfotransferases, aldehyde oxidases, alcohol dehydrogenases, and short-chain dehydrogenases ( Table 2). Further work is required to elucidate the role of each in the detoxification pathway.

Resistance to alternative insecticides currently being deployed
Bacillus thuringiensis var. israeliensis (Bti) formulations and insect growth regulators (IGRs) are important alternatives to the neurotoxic insecticides for larviciding, in light of emerging widespread resistance to the latter. It has been proposed that Bti is resilient to the emergence of resistance due to multiple modes of toxicity; however, the persistence of some Bti toxins in the environment may increase the selection pressure for the sequential development of resistance [66]. Several mechanisms of Bti resistance have been described [67] but studies have focused on laboratory-selected strains. Few field populations have been tested, although 14 Ae. aegypti populations from Brazil in 2011 [68], 1 from Martinique [69], and 1 from Malaysia in 2008 [70] were all found to be susceptible. Both Ae. aegypti and Ae. albopictus from 6 populations in Cameroon and Gabon were tested in 2007 and found to be susceptible to Bti [71]. In contrast, there is some evidence of emerging resistance to IGRs in field populations. Recently, Ae. aegypti populations from 12 municipalities in Brazil were challenged with the IGR diflubenzuron and found to be susceptible [72] and susceptibility to pyriproxyfen and methoprene was also found in Mayotte [73], whereas a population from Martinique showed tolerance of diflubenzuron and pyriproxyfen [69], and 2 populations from the US showed some tolerance to pyriproxyfen and methoprene [74]. Furthermore, Ae. aegypti and Ae. albopictus populations from 12 states in Malaysia displayed variable levels of resistance to 5 IGRs [75].

Conclusions
Since the geographical distribution of pyrethroid resistance was reviewed in 2010 [8], the data available has increased by more than 5-fold. The geographical spread of the data has also improved, most noticeably in West Africa, yet coverage remains extremely patchy and there is a pressing need for studies in Africa and Australasia for Ae. aegypti and in most areas for Ae. albopictus.
In order to provide a comprehensive assessment of the geographical distribution of insecticide resistance in these vectors, standard bioassay protocols need to be implemented using well-characterised susceptible reference strains. Monitoring and surveillance at a large number of locations then needs to follow. Bioassay surveillance data should also be used in conjunction with data from well-characterised markers for specific mechanisms of resistance that have been shown to be important.
Despite a growing number of studies on the mechanisms of resistance to pyrethroids and temephos, in particular, many knowledge gaps remain. Presence of VGSC mutations is unknown across large areas, especially in Africa and Australasia, which may be key foci for spreading resistance. The complexity of metabolic resistance precludes a full understanding of detoxification pathways in the near future but some key candidate genes and gene families have been identified and can be prioritised for screening or assay development. Judging the relative importance of metabolic and target-site mechanisms is difficult because of the qualitatively different nature of the 2 mechanisms. VGSC mutations affect a single gene and are easily typed, whereas metabolic resistance is much more complex with considerable potential for selection of different genes of similar function in different genetic backgrounds or environments and with variation in the specificity of the causal variants involved. Development of DNA markers for key metabolic genes will facilitate screening alongside VGSC mutations in field populations whilst controlled crosses and genome editing in the laboratory should be used to segregate and study the independent effects of each mechanism.
In addition to the standardisation of methods, development of diagnostics, and integration of phenotypic and genotypic data to improve monitoring, it is important that the resulting data are shared. Research data need to reach decision-making bodies, and health agency data held by individual countries need to be shared to address a critical global issue that is not restricted by national boundaries. Regular updates to online databases needs to be a key goal coupled with appropriate training of end users. A number of databases already exist for vector data, namely VectorBase, managed by the University of Notre Dame and Imperial College London, the Malaria Atlas Project, managed by the University of Oxford, and IR Mapper, managed by Vestergaard and the Kenya Medical Research Institute [76][77][78]. Incorporating the current dataset into existing repositories will make it easier to find and help ensure longerterm access to the data. We have started this process here by releasing 6,562 bioassay records, each linked to a defined time and place of mosquito collection and accompanied by all available information on how the bioassay was conducted.

Key learning points
• There is abundant evidence for resistance to the 2 insecticide classes most widely used to control Ae. aegypti and Ae. albopictus, pyrethroids and organophosphates, from the Americas and Asia, and some evidence for arising pyrethroid resistance in the parts of West Africa that have now been sampled.
• Evidence for the frequency and level of insecticide resistance at different times and places comes from a wide array of study methodologies and there is a critical need for internationally agreed standards coupled with surveillance in order to understand and map trends in resistance. A sensible starting point is the WHO's test procedures for insecticide-resistance monitoring in malaria vector mosquitoes, which would need to be expanded to incorporate the range of insecticides used in the control of Aedes vectors.
• Insecticide resistance in Aedes mosquitoes is mainly due to target-site mutations and increased detoxification; however, further work is needed to identify specific mutations and genes involved in metabolic resistance and to develop robust, phenotype-calibrated diagnostics.
Supporting information S1 File. Additional details on the bioassay data processing and mapping.