Research

Identifying genomic changes associated with

insecticide resistance in the dengue mosquito
Aedes aegypti by deep targeted sequencing
Frederic Faucon,1,2,3 Isabelle Dusfour,4 Thierry Gaude,1,2,3 Vincent Navratil,5
Frederic Boyer,1,2,3 Fabrice Chandre,6 Patcharawan Sirisopa,7,8
Kanutcharee Thanispong,9 Waraporn Juntarajumnong,7,8 Rodolphe Poupardin,10
Theeraphap Chareonviriyaphap,7,8 Romain Girod,4 Vincent Corbel,6,7,8
Stephane Reynaud,1,2,3 and Jean-Philippe David1,2,3
1
Laboratoire d’Ecologie Alpine (LECA), CNRS, UMR 5553, 38041 Grenoble Cedex 9, France; 2Université Grenoble–Alpes, 38041
Grenoble Cedex 9, France; 3Environmental and Systems Biology (BEeSy), Université Grenoble–Alpes, 38041 Grenoble Cedex 9,
France; 4Unité d’Entomologie Médicale, Institut Pasteur de la Guyane, 97306 Cayenne Cedex, France; 5Pôle Rhône Alpes de
Bioinformatique, Université Lyon 1, 69100 Villeurbanne, France; 6Institut de Recherche pour le Développement (IRD), Maladies
Infectieuses et Vecteurs, Ecologie, Génétique, Evolution et Contrôle (IRD 224-CNRS 5290 UM1-UM2), 34394 Montpellier Cedex 5,
France; 7Department of Entomology, Faculty of Agriculture, Kasetsart University, Lat Yao Chatuchak Bangkok 10900, Thailand;
8
Center for Advanced Studies for Agriculture and Food, Kasetsart University Institute for Advanced Studies, Kasetsart University,
Bangkok 10900, Thailand (CASAF, NRU-KU, Thailand); 9Bureau of Vector Borne Diseases, Department of Disease Control, Ministry
of Public Health, Mueang, Nonthaburi 11000, Thailand; 10Vector Biology Group, Liverpool School of Tropical Medicine, L35QA
Liverpool, United Kingdom

The capacity of mosquitoes to resist insecticides threatens the control of diseases such as dengue and malaria. Until alternative
control tools are implemented, characterizing resistance mechanisms is crucial for managing resistance in natural popula-
tions. Insecticide biodegradation by detoxification enzymes is a common resistance mechanism; however, the genomic chang-
es underlying this mechanism have rarely been identified, precluding individual resistance genotyping. In particular, the role
of copy number variations (CNVs) and polymorphisms of detoxification enzymes have never been investigated at the ge-
nome level, although they can represent robust markers of metabolic resistance. In this context, we combined target enrich-
ment with high-throughput sequencing for conducting the first comprehensive screening of gene amplifications and
polymorphisms associated with insecticide resistance in mosquitoes. More than 760 candidate genes were captured and
deep sequenced in several populations of the dengue mosquito Ae. aegypti displaying distinct genetic backgrounds and con-
trasted resistance levels to the insecticide deltamethrin. CNV analysis identified 41 gene amplifications associated with resis-
tance, most affecting cytochrome P450s overtranscribed in resistant populations. Polymorphism analysis detected more than
30,000 variants and strong selection footprints in specific genomic regions. Combining Bayesian and allele frequency filter-
ing approaches identified 55 nonsynonymous variants strongly associated with resistance. Both CNVs and polymorphisms
were conserved within regions but differed across continents, confirming that genomic changes underlying metabolic resis-
tance to insecticides are not universal. By identifying novel DNA markers of insecticide resistance, this study opens the way
for tracking down metabolic changes developed by mosquitoes to resist insecticides within and among populations.
[Supplemental material is available for this article.]

Mosquitoes are vectors of numerous human diseases, representing by resistance mechanisms developed by mosquitoes. Insecticide
a major threat for public health worldwide (Lounibos 2002). resistance is widespread in Ae. aegypti and affects most insecticides
Dengue and Chikungunya viruses are both transmitted by the used for vector control (Ranson et al. 2010). Resistance to pyre-
mosquito Aedes aegypti and represent a burden in more than 100 throid insecticides, the primary insecticide family used against
countries putting more than 2.5 billion people at risk (WHO adult mosquitoes, is particularly worrying in the context of the
2009, 2014). Since the Second World War, chemical insecticides re-emergence of dengue and other arboviruses worldwide (Bhatt
have been massively used for controlling vector populations and et al. 2013). Although attempts are made to develop new insecti-
reducing disease transmission, but their efficacy is now threatened cides or alternative mosquito control strategies (Scholte et al.
2004; Lacey 2007; Hoffmann et al. 2011; Walker et al. 2011;
Harris et al. 2012), their large-scale implementation in tropical
Corresponding author: jean-philippe.david@ujf-grenoble.fr
Article published online before print. Article, supplemental material, and publi- © 2015 Faucon et al. This article, published in Genome Research, is available
cation date are at http://www.genome.org/cgi/doi/10.1101/gr.189225.115. under a Creative Commons License (Attribution-NonCommercial 4.0 Inter-
Freely available online through the Genome Research Open Access option. national), as described at http://creativecommons.org/licenses/by-nc/4.0/.

25:1347–1359 Published by Cold Spring Harbor Laboratory Press; ISSN 1088-9051/15; www.genome.org Genome Research 1347
www.genome.org
Faucon et al.

regions will require at least a decade. Until this, characterizing mo- elements known to favor duplication events (Nene et al. 2007),
lecular mechanisms underlying resistance is crucial for tracking conducting a comprehensive screening of gene amplifications as-
down resistance alleles and improving resistance management sociated with insecticide resistance may allow identifying reliable
strategies (Corbel et al. 2013). DNA markers of metabolic resistance. Besides, most genes linked
Resistance to pyrethroids can be the consequence of various to metabolic resistance were identified based on their differential
mechanisms, such as nonsynonymous mutations affecting the expression while the importance of nonsynonymous changes
voltage-gated sodium channel targeted by these insecticides, i.e., affecting detoxification enzymes have clearly been neglected by
knockdown resistance (kdr) mutations, a lower insecticide penetra- molecular screenings. As for gene amplifications, these polymor-
tion, its sequestration, or its biodegradation (metabolic resistance) phisms are of high interest for monitoring the dynamics of in-
(Hemingway et al. 2004; Li et al. 2007). Kdr mutations and meta- secticide resistance mechanisms among and within mosquito
bolic resistance are known as the two primary resistance mecha- populations.
nisms in mosquitoes. Several kdr mutations have been identified In this context, the present study aimed at combining the tar-
in Ae. aegypti, and the association between the V1016G/I and get enrichment technology with high-throughput sequencing
the F1534C mutations and pyrethroid resistance has been con- for identifying gene amplifications and polymorphisms linked
firmed (Brengues et al. 2003; Saavedra-Rodriguez et al. 2007; to pyrethroid resistance in the dengue vector, Ae. aegypti. More
Yanola et al. 2011). Monitoring the frequency of these mutations than 760 genes, including those encoding for detoxification en-
in field populations is possible through simple DNA-based diag- zymes, cuticle proteins, ATP-binding cassette (ABC) transporters,
nostic assays and provides key data for resistance management neurotransmitter receptors, and voltage-gated channels, were cap-
strategies. tured and deep sequenced in several susceptible and pyrethroid-
Metabolic resistance is far less understood, although this type resistant populations from three geographical regions (Asia,
of resistance is frequent and often accounts for a significant part South America, and laboratory populations). After phenotyping
of the resistance phenotype (Li et al. 2007; Nkya et al. 2013). populations for resistance to the pyrethroid deltamethrin and
Such a resistance mechanism is caused by an increased activity segregating most resistant individuals, gene amplifications and
of detoxification enzymes. These detoxification enzymes include polymorphisms associated with deltamethrin resistance were
cytochrome P450 monooxygenases (P450s or CYPs for genes), car- identified by deep targeted sequencing of large pools of individu-
boxy/cholinesterases (CCEs), glutathione S-transferases (GSTs), als. Gene amplifications were validated on individual mosquitoes,
and UDP-glycosyl-transferases (UDPGTs), although other families and their impact on transcription levels were validated. Polymor-
can be involved (Hemingway et al. 2004; David et al. 2013). Their phism data were first used to identify genomic regions under selec-
high diversity (roughly 300 genes in Ae. aegypti) and the complex- tion through a hierarchical Bayesian approach taking into account
ity of insecticide biodegradation pathways make the identification population structure. Then, nonsynonymous variants strongly
of those conferring resistance challenging. Theoretically, metabol- linked to deltamethrin resistance were identified by combining
ic resistance can be the consequence of an increased expression those identified by the Bayesian approach and those displaying al-
of one or multiple detoxification enzymes capable of metaboliz- lele frequencies mostly associated with resistance. Quantitative
ing the insecticide and/or the selection of variants showing a PCR assays were used for cross validating polymorphism data
higher insecticide metabolism rate due to conformational modifi- and studying the association between gene amplifications and
cations. Because overexpression is frequently associated with over- known target-site mutations. Results are discussed in regard to
transcription, most candidate genes were identified based on their known insecticide resistance mechanisms and the identification
differential transcription in resistant populations compared to sus- of novel DNA markers of insecticide resistance.
ceptible counterparts using transcriptomics (for review, see David
et al. 2013). Although subjected to inherent difficulties associated
with gene expression studies (e.g., RNA handling and degradation, Results
uncontrolled variation across time, tissues, and populations),
these approaches identified several detoxification enzymes over- Deltamethrin resistance levels
transcribed in resistant populations, some of them being later val- Bioassays on adult females revealed a broad range of resistance to
idated as pyrethroid metabolizers (for review, see Vontas et al. deltamethrin across populations (Table 1). Exposing mosquitoes
2012; David et al. 2013). Although these approaches proved their for 1 h to various doses of insecticide showed that the lethal
value for identifying metabolic resistance genes, they failed to pin- dose killing 50% of individuals (LD50) varied up to 750-fold be-
point the genomic changes associated to their overexpression. In tween susceptible populations and the most resistant populations
addition, because a limited quantity of mRNA can be extracted from French Guiana. Resistant populations from Thailand showed
from a single mosquito, these studies were conducted at the popu- an intermediate resistance level (about 250-fold), and the laborato-
lation level without estimating the frequency of resistant alleles ry selected population showed a slight resistance (approximately
within populations. A constitutive overtranscription can be the fivefold). These different resistance levels were confirmed by
consequence of an up-regulation controlled by cis/trans genomic WHO diagnostic assays, in which the time necessary to knock
regulatory elements or an increased gene copy number (Li et al. down 50% of individuals (KDT50) varied from 11 min for suscepti-
2007). Although gene amplifications have been associated with ble populations to >8 h for the most resistant ones.
the overtranscription of a few pyrethroid resistance genes in mos-
quitoes (Wondji et al. 2009; Itokawa et al. 2010, 2011; Bariami
et al. 2012), no comprehensive screening of gene amplifications Sequencing metrics and coverage
linked to insecticide resistance has ever been conducted. Consider- The use of target enrichment technology followed by multiplexed
ing the central role of copy number variation (CNV) in adaptations Illumina sequencing allowed obtaining an average of 4.35 million
associated with a “gene dosage” effect (for review, see Kondrashov 75-bp reads per sample (Supplemental Table 1). Filtering them ac-
2012) and the richness of the Ae. aegypti genome in transposable cording to read pairing, sequence quality, and mapping quality

1348 Genome Research

www.genome.org
 Variants associated with insecticide resistance

Table 1. Deltamethrin resistance levels

R+/R− segregation

Population Group Resistance level KDTa50 KDT5095% CI LDb50 LT25 (min)c LT75 (min)c

Livp S Genome Susceptible 11.5 10.8–12.1 0.001 — —

Bora S Laboratory Susceptible 12.1 11.6–12.7 0.001 — —
Delta R Laboratory Slightly resistant 14.2 13.5–15.0 0.005 30 80
Patg S Thailand Susceptible 17.0 15.1–19.2 0.001 — —
Phet R Thailand Resistant 98.2 91.4–106.7 0.25 40 120
Nakh R Thailand Resistant 92.9 87.7–99.2 0.25 40 120
NwOr S S. America Susceptible 11.9 11.2–12.6 0.001 — —
Cayn R S. America Highly resistant >500 — 0.75 30 180
StGe R S. America Highly resistant >500 — 0.75 30 180

a
KDT50: time (min) necessary to knock down 50% of individuals. KDT50 was determined using papers impregnated with 0.05 g/100 mL deltamethrin
following WHO insecticide testing recommendations.
b
Dose of insecticide (g/100 mL) used for paper impregnation leading to 50% mortality after 1 h exposure.
c
R−/R+ segragation of resistant populations consisted in separating individuals dead at LT25 from those surviving at LT75 using impregnated paper cor-
responding to the LD50.

allowed successful mapping of 68% of reads on the Ae. aegypti ge- conserved within each region but showed marked differences
nome. The majority of mapped reads consisted in perfect matches between continents. Thai resistant populations (Nakh R and
(46.6%), whereas reads matching with one and two or more mis- Phet R) showed a strong amplification of four CCEs located in
matches represented 12.6% and 9.2% of sequenced reads, respec- supercontig 1.142 and another one in supercontig 1.1678. Two
tively. The mean coverage was homogenous across samples with CYP6s located in supercontig 1.702 and two sulfotransferases in
a total mean coverage of 83× (Supplemental Fig. 1). supercontig 1.975 were also amplified in Thai resistant popula-
tions. In contrast, resistant populations from South America
(Cayn R and StGe R) showed a marked amplification of four
Gene amplifications linked to deltamethrin resistance
CYP6s located in supercontig 1.1327. Resistant populations from
Of the 763 targeted candidate genes, 727 showed a mean coverage both regions also shared the amplifications of several CYP9Js
greater than 90 reads per kilobase per million reads (RPKM) and in supercontigs 1.1188 and 1.221. As expected, the less resistant
were considered for further analysis. Among them, 41 genes laboratory population (Delta R) exhibited fewer gene amplifica-
showed an increased coverage >1.5-fold in any resistant popula- tions associated with resistance. The amplification profile of the
tion compared to the mean coverage of all susceptible ones, to- Delta R population shared particular gene amplifications with field
gether with an increased coverage >1.0-fold in LT75 survivors (R+ resistant populations. This included a chloride channel also detect-
phenotype) compared to individuals dead at LT25 (R− phenotype). ed in Thailand together with two CYP4Js and a glycosyltransferase
These genes were considered amplified in association with delta- also detected in South America.
methrin resistance (Supplemental Table 2). These genes mostly
encoded detoxification enzymes with a significant enrichment
in P450s (Fig. 1). Most gene amplifications occurred in gene clus- Validation of gene amplifications on individual mosquitoes
ters located within a few genomic supercontigs spread across all Nine gene amplifications identified by deep sequencing belonging
chromosomes (Fig. 2). Gene amplification profiles were highly to two P450 clusters and one CCE cluster heavily impacted by gene
amplifications were cross-validated by qPCR
on individual mosquitoes. The presence of
gene amplifications was confirmed for all
targeted genes. A good correlation was ob-
served between CNVs obtained by deep
sequencing and qPCR (Supplemental Fig.
2). Individual qPCR data reveal high copy
number polymorphism within resistant
populations (Supplemental Fig. 3). Estimat-
ing the frequency of gene amplifications
within each population revealed that ampli-
fications are frequent in resistant popula-
tions and sometimes also present at low
frequencies in susceptible populations (Sup-
plemental Fig. 4). Comparing the amplifi-
cation profile of each gene across all
individuals revealed CYP6 and CCE clusters
Figure 1. Gene families affected by gene amplifications associated with deltamethrin resistance. are likely affected by single amplification
(Left) Frequency of each gene family among the 763 captured genes. (Right) Frequency of each
events (Supplemental Fig. 5). In contrast,
gene family among the 41 genes affected by genic amplifications associated with deltamethrin resis-
tance. Proportions of each gene family between all captured genes and genes affected by genic am- different copy numbers were frequently ob-
plifications were compared using a one-sided Fisher’s exact test: (∗∗∗ ) P < 0.001. served for genes belonging to the CYP9J

Genome Research 1349

www.genome.org
Faucon et al.

Figure 2. Gene amplifications associated with deltamethrin resistance. (Left) CNV profiles of genes affected by gene amplifications associated with del-
tamethrin resistance. Color scale shows (R+)/(meanS) CNV for each resistant population, and overimposed “+” marks show (R+)/(R−) CNV. (Right) Location
of gene amplifications on genomic supercontigs. Amplified genes are shown in red. Nonamplified genes are shown in maroon. Genes not included in the
capture design are shown in gray. Chromosomal locations are shown as described in Timoshevskiy et al. (2014) and Juneja et al. (2014).

cluster on supercontig 1.1188, suggesting that distinct duplication Polymorphisms identified by deep targeted sequencing
events affect this large P450 cluster.
A total of 41,469 polymorphisms supported by an important cov-
erage were called against the reference genome. Among them,
Impact of gene amplifications on transcription levels 30,400 were polymorphic among populations. A principal compo-
Comparing gene amplification levels obtained by DNA-sequenc- nent analysis (PCA) based on allele frequency variation across all
ing with transcription levels obtained by RT-qPCR for the nine can- populations confirmed the overall population structure and the
didate genes across all resistant populations confirmed that gene different genetic background of Thai, South-American, and labora-
amplifications are often associated with increased transcription tory (French Polynesia) populations (Supplemental Fig. 7). The re-
(Supplemental Fig. 6). High and significant positive correlations liability of polymorphism data was confirmed by comparing the
were observed for most genes except for two CYP9Js in supercontig frequencies of V1016I, V1016G, and F1534C kdr mutations esti-
1.1118, where positive correlations were not significant. mated by deep targeted sequencing and those obtained by qPCR
on individual mosquitoes (Supplemental Fig. 8). The presence of
distinct kdr mutations at position 1016 in Thailand and South
Association of gene amplifications with target-site mutations America was confirmed, and the F1534C mutation was also detect-
Among the nine gene amplifications genotyped by qPCR, only ed in most populations.
one was found associated with kdr mutations (Supplemental
Table 3). This association occurred in the Nakh population from
Thailand, where a positive correlation between the presence of Polymorphisms associated with deltamethrin resistance
CYP9J28 amplification and the presence of the V1016G mutation Searching for selection footprints using a Bayesian approach tak-
was found (r = 0.61; P = 0.025). As a consequence of the total segre- ing into account population structure (Foll et al. 2014) identified
gation of V1016G and F1534C mutations in Thailand, the F1534C 156 outlier loci significantly associated with deltamethrin resis-
mutation was negatively correlated to CYP9J28 amplification in tance. These outlier loci included 31 nonsynonymous variants
this population (r = −058; P = 0.036). and affected multiple supercontigs spread over all chromosomes.

1350 Genome Research

www.genome.org
 Variants associated with insecticide resistance

deltamethrin resistance (Fig. 4; Supplemental Table 4). These

variants affected 39 distinct genes mainly belonging to P450s,
UGTs, CCEs, and cuticle proteins. As for gene amplifications, allele
frequency patterns were conserved within regions but often dif-
fered among continents. Most nonsynonymous variants affecting
P450s were identified from South American or laboratory popula-
tions. These P450s mostly belong to CYP6, CYP9, and CYP12 fam-
ilies. Conversely, most variants affecting CCEs were identified
from Thai populations. Two were both affected by gene amplifica-
tions and nonsynonymous variants with CCEAE4A bearing height
distinct nonsynonymous variants associated with resistance.
Several UGTs were also affected by nonsynonymous variants,
with three showing an increased frequency across multiple re-
gions. ABC transporters and cuticle proteins were most frequently
affected in South America. Several other enzymes potentially
involved in insecticide detoxification were also affected by non-
synonymous variants, including two aldehyde oxidases, one alco-
hol dehydrogenase, and one sulfotransferase. Finally, the nicotinic
acetylcholine receptor (nAchR) AAEL004935 targeted by neonico-
tinoid insecticides and the voltage-gated sodium channel (VSGC)
AAEL006019 targeted by pyrethroid insecticides were also affect-
Figure 3. Supercontigs showing selection footprint. For each region, ed. Variants detected in the VGSC corresponded to the S989P
loci displaying a significant BayeScan3 Fsc Q-value <0.05 and both f(R+)-f and F1534C kdr mutations previously described in this species.
(S) and f(R+)-f(R−) allele frequency variation occurring in the same direc- None of the V1016I/G kdr mutations previously associated with
tion were considered as outliers. For each supercontig, outlier loci densities pyrethroid resistance was retained as best candidates. Neither
were obtained by dividing the number of outlier loci by the total length of
of them was retained by the Bayesian approach, and the V1016G
all captured genes (outlier loci per kb of captured region). Outlier loci den-
sities are shown as a blue color scale. Gray stands for an absence of outlier mutation only passed the allele frequency filtering in Thailand
loci. For each supercontig, horizontal bars show the number of captured (Supplemental Table 4).
genes from each gene family (upper bar) compared to those affected by
outlier loci (lower bar). Supercontigs are clustered according to their puta-
tive chromosomal location as described in Timoshevskiy et al. (2014) and
Juneja et al. (2014). Discussion
Deep targeted sequencing for studying
Selection footprints varied between geographical regions with out- insecticide resistance
lier loci densities ranging from 0 to 4.1 per kb of captured sequence Screening for genomic changes associated with insecticide resis-
(Fig. 3). In Thailand, supercontigs 1.142 and 1.4629 showed high tance remained challenging until the sequencing of mosquito ge-
outlier loci densities with three CCE and one ABC transporter be- nomes (Holt et al. 2002; Nene et al. 2007; Arensburger et al. 2010).
ing affected. Six other supercontigs also displayed selection foot- Afterward, the development of mosquito DNA microarrays al-
prints impacting two P450s, two other detox enzymes, one PPO, lowed screening for resistance genes based on their differential
and one cuticle protein. High outlier loci densities were identified transcription, leading to the identification of detoxification genes
in laboratory populations in supercontigs 1.371, 1.47, and 1.51 conferring resistance (David et al. 2005, 2013; Strode et al. 2008;
affecting 17 P450s and two cuticle proteins. Selection footprints Vontas et al. 2010; Edi et al. 2014). However, these screenings fo-
were less pronounced in South America but showed overlap with cused on differential transcription while genomic changes such
laboratory and Thai populations. as CNVs and polymorphisms were neglected. Then, the develop-
Of the 30,400 polymorphic loci, 3054 loci showed an allele ment of high-throughput DNA sequencing allowed screening
frequency variation between R+ phenotypes and their respective for genomic variants across whole genomes. However, such an
susceptible population ≥40% together with an allele frequency approach required a huge amount of sequences to reach the min-
variation between R+ and R− phenotypes ≥5%. These loci includ- imum genome coverage necessary for accurately quantifying CNV
ed 465 nonsynonymous variants (Supplemental Table 4). Compar- and allele frequencies, making it unaffordable when several sam-
ing the Bayesian and the frequency filtering approaches revealed ples have to be compared. Although RNA-seq is more affordable
that the Bayesian approach was overall more stringent (Supple- and can generate gene expression and variant data concomitantly,
mental Fig. 9). However, the stringency of the Bayesian approach this approach is not suitable for identifying CNVs and does not
was high in South America, medium in Thailand, and low in lab- provide reliable polymorphism data from poorly expressed genes.
oratory populations following the decreasing diversity of these In addition, allele frequencies inferred from RNA-seq are to be
groups. The frequency filtering approach showed an opposite taken with caution because they can be altered by allele-specific
stringency with more variants identified in South America than expression events (Gaur et al. 2013).
Thailand and even less in laboratory populations. Frequency filter- Recently, the development of target enrichment techniques
ing identified 16 nonsynonymous variants shared by distinct geo- allowed conducting DNA-seq focused on genomic regions of in-
graphic regions, whereas the Bayesian approach only identified terest, hence, drastically increasing the sequencing depth while
variants being specific to each region. maintaining costs at a reasonable level (Altmüller et al. 2014). By
Combining the best candidates identified by each approach targeting more than 760 genes from all protein families potentially
identified 55 nonsynonymous variants strongly associated with involved in insecticide resistance in the dengue mosquito, our

Genome Research 1351

www.genome.org
Faucon et al.

Figure 4. Best nonsynonymous polymorphisms associated with deltamethrin resistance. Only the 55 best differential nonsynonymous variants identified
from frequency-based filtering and the Bayesian approach are shown (see Methods). For each region, allele frequency variation between each resistant
population (R+ phenotypes) and their susceptible counterpart (S) are shown as a blue-yellow color scale. Blue indicates an enrichment in the reference
allele, whereas yellow indicates an enrichment in the variant allele. Variants identified by the Bayesian approach are indicated by “B” marks. For variants
passing frequency-based filtering or BayeScan3 filtering, allele frequency variation between R+ and R− phenotypes are shown as overimposed “+” marks.
Variants are grouped by gene families and are described by the following annotations: chromosomal location (according to Juneja et al. 2014; Timoshevskiy
et al. 2014), supercontig position, nucleotide change, amino acid position, amino acid change, gene accession number, and gene description. (∗ ) Genes
also found affected by CNVs linked to deltamethrin resistance. (∗∗ ) The sodium channel S729P and F1249C variants correspond to the S989P and F1534C
kdr mutations described in the literature due to changes in AAEL006019-RD transcript annotation.

study screened for both CNVs and polymorphisms across several tions originating from different geographical areas, providing the
samples. Our experimental design allowed reaching a very deep first comprehensive screening of CNVs and polymorphisms asso-
coverage (>80×) across multiple resistant and susceptible popula- ciated with insecticide resistance. Targeted DNA-seq data were

1352 Genome Research

www.genome.org
 Variants associated with insecticide resistance

consistent with qPCR data obtained from individual mosquitoes et al. 2008; Duangkaew et al. 2011; Stevenson et al. 2011; Riveron
validating the robustness of the approach. Altogether, this study et al. 2013). Altogether, these results confirm the key role of P450
demonstrates that target enrichment coupled with high-through- in the resistance of mosquitoes to pyrethroids and their propensity
put sequencing is a powerful tool for pinpointing genomic chang- to increase their expression level through gene amplification when
es associated with adaptive traits from which candidate gene undergoing a strong selection pressure from insecticides.
families are known. Three CCEs (CCEAE3A, CCEAE4A, and CCEAE6A) belonging
to a single genomic cluster were highly amplified in Thai resistant
Gene amplifications associated with pyrethroid resistance populations leading to their strong overtranscription. Highest
amplification levels were found in the Nakh R population which
Only a few CNVs were previously associated with insecticide resis- is also resistant to organophosphates (Poupardin et al. 2014).
tance, and no comprehensive screening has ever been conducted In this previous study, the overtranscription of CCEAE3A and
in mosquitoes (for review, see Bass and Field 2011). The present CCEAE6A through gene amplification was associated with teme-
study identified 41 genes affected by gene amplifications linked phos resistance at the larval stage, although resistance to pyre-
to deltamethrin resistance. Most were previously found signif- throids was also observed. Other studies performed on South
icantly overtranscribed in pyrethroid resistant Ae. aegypti pop- American and Caribbean populations confirmed the association
ulations according to data publicly available in the VectorBase of some of these CCEs with temephos resistance (Marcombe et al.
expression browser (https://vectorbase.org, data sets from Strode 2009, 2012; Saavedra-Rodriguez et al. 2014). Our data revealed
et al. 2008; Marcombe et al. 2009; Poupardin et al. 2012; David higher CCE copy number in mosquitoes surviving deltamethrin
et al. 2014). The majority of CNVs observed between resistant exposure (R+ phenotype), supporting their role in pyrethroid resis-
and susceptible populations were <2.5-fold, suggesting a predom- tance. Although the association between pyrethroid resistance and
inance of duplications. However several genes displayed CNV these CCE genes in Thailand might be due to multiresistant indi-
>fivefold in particular resistant populations, suggesting that copy viduals carrying both pyrethroid- and organophosphate-resistance
number may vary within continents. This was confirmed by genes, the role of CCEs in pyrethroid metabolism has been demon-
qPCR on individual mosquitoes showing an important copy num- strated in mammals (Hodgson 2003; Nakamura et al. 2007) and
ber polymorphism within and across resistant populations. Such mosquitoes (Somwang et al. 2011; Chandor-Proust et al. 2013).
polymorphism suggests that most amplifications are not fixed in However, no individual mosquito CCE has yet been validated as
resistant populations. The strong relationship between gene am- a pyrethroid metabolizer, and insecticide sequestration may also
plification and overtranscription was confirmed by RT-qPCR. confer resistance. Further functional studies are required for vali-
Although, overexpression can also be triggered by cis/trans regula- dating the precise role of these amplified CCEs in pyrethroid
tory elements or post-translational events, the present study sug- resistance.
gests that gene amplification is a common adaptive mechanism Two sulfotransferases were amplified in Thai resistant popula-
allowing mosquitoes to overexpress detoxification genes confer- tions. The overtranscription of these enzymes has been frequently
ring resistance to insecticides. The importance of such mechanism described in insecticide resistant mosquitoes and deserve further
may even be greater in species highly infected by transposable el- attention (Poupardin et al. 2012; Marcombe et al. 2013; Nkya
ements like Ae. aegypti because their presence is known to favor et al. 2013). Indeed, their involvement in insecticide degradation
duplication events (Nene et al. 2007). pathways has been validated in mammals (Lee et al. 2007).
The present study revealed distinct gene amplification pat- Among GSTs, only GSTE2 was found amplified in Thai resis-
terns among resistant populations according to their geographical tant populations. This GST is known for its ability to metabolize
origin. This supports that the selection of gene amplifications con- DDT in various mosquito species (Ortelli et al. 2003; Lumjuan
ferring resistance depends on the genetic background of popula- et al. 2005), and recent studies confirmed its role in pyrethroid re-
tions, their population dynamics, and on the selection pressures sistance although pyrethroid metabolism has not been observed
they undergo. Most amplified genes encoded detoxification en- (Lumjuan et al. 2011; Riveron et al. 2014b).
zymes, suggesting that gene amplification is rather linked to met- A recent study suggested that the duplication of the gene en-
abolic resistance than resistance conferred by an altered transport/ coding the Ae. aegypti voltage-gated sodium channel (VGSC) tar-
penetration of the insecticide. geted by pyrethroids contributed to resistance by maintaining
The majority of detoxification enzymes affected by gene am- together wild-type allele and kdr mutations and reducing mutant
plifications encoded P450s belonging to CYP9J and CYP6 families allele deleterious effects (Martins et al. 2013). A similar role was
that were frequently associated with insecticide resistance in mos- suggested for the duplication of the acetylcholinesterase gene
quitoes (Vontas et al. 2012; David et al. 2013). The overrepresen- (carrying the Ace1 mutation) in the resistance of An. gambiae to
tation of P450s was expected because their overexpression is carbamates (Edi et al. 2014). Although such a mechanism is likely
known to play a key role in pyrethroid resistance, and they are fre- contributing to resistance in mosquitoes, no amplification of
quently affected by copy number variants (Feyereisen 2006, 2011). genes encoding proteins targeted by insecticides were detected in
Several amplified CYP9Js were previously found overtranscribed our study. In addition, a poor association was observed between
in resistant populations and some of them (CYP9J24, CYP9J28, the presence of kdr mutations and gene amplifications affecting
CYP9J32) have been validated as able to metabolize pyrethroids detoxification enzymes in individual mosquitoes. Indeed, recom-
(Stevenson et al. 2012), supporting the importance of their am- bination events are probably countering such association, espe-
plification in resistance. Among the CYP6s amplified, most were cially as the VGSC gene and amplified detoxification genes are
previously found overtranscribed in resistant populations, and located on distant genomic regions. In addition, fitness costs asso-
CYP6BB2 was recently involved in pyrethroid metabolism (Kasai ciated with both kdr mutations and gene amplifications have been
et al. 2014). In addition, some of their close orthologs in Anopheles identified (Kondrashov 2012; Brito et al. 2013; Katju and
(CYP6P3, CYP6P7, CYP6P9, CYP6M2, and CYP6AA5) are also capa- Bergthorsson 2013; Rinkevich et al. 2013).Therefore, although
ble of degrading pyrethroids (Boonsuepsakul et al. 2008; Müller kdr mutations and gene amplifications of detoxification enzymes

Genome Research 1353

www.genome.org
Faucon et al.

co-occur in resistant populations, additive fitness costs may con- substantial contribution of other mechanisms in the resistance
tribute to segregate them at the individual level. phenotype.

Polymorphisms associated with pyrethroid resistance New tools for monitoring resistance genes
Apart from mutations affecting the proteins targeted by insecti- in mosquito populations
cides, very few nonsynonymous variants have been associated Overall, the present study confirms that gene amplifications affect-
with insecticide resistance. In this context, the present study rep- ing detoxification enzymes are good markers of metabolic resis-
resents the first DNA-seq screening of variants associated with py- tance in mosquitoes. Novel nonsynonymous variants potentially
rethroid resistance in mosquitoes. Searching for selection enhancing their ability to detoxify insecticides were also identi-
footprints identified multiple genomic regions under selection, fied. Despite significant research efforts, developing molecular
suggesting that pyrethroid resistance is polygenic. Several nonsyn- tests to track the overexpression of these enzymes in field popula-
onymous variants were associated with deltamethrin resistance, tions has proved challenging in terms of affordability, throughput,
but few were found in multiple geographic regions, confirming and reliability; and current vector population monitoring tools
the strong influence of genetic background and population history do not include metabolic resistance markers (Bass et al. 2010).
on the selection of resistant alleles. Detecting their increased copy number or the presence of specific
Although several nonsynonymous variants affecting P450s variants conferring resistance through simple PCR-based DNA as-
were identified from each region, those mostly associated with re- says opens up new perspectives for managing insecticide resis-
sistance were mainly detected in South American and laboratory tance. Indeed, such novel resistance markers will be accessible
populations. Some were previously identified in a permethrin-re- to developing countries in terms of technology and affordability
sistant strain by RNA-seq (David et al. 2014). In Drosophila mela- and will allow tracking down both target-site and metabolic-re-
nogaster, functional studies demonstrated that nonsynonymous sistance alleles in individual mosquitoes and monitoring their
mutations in the P450 CYP6A2 identified in the RDDTR resistant frequencies through time and space among and within natural
strain have a prominent role in resistance by enhancing DDT populations.
metabolism (Amichot et al. 2004). In the mosquito, Anopheles
funestus, directional selection footprints were found in the coding
sequence of two CYP6Ps capable of metabolizing pyrethroids Methods
(Riveron et al. 2014a). Most nonsynonymous variants impacting
CCEs were identified in Thailand with CCEAE4A and CCEAE3A Mosquitoes
impacted by multiple variants strongly associated with delta- Nine Ae. aegypti populations of distinct genetic backgrounds were
methrin resistance. Elevated copy numbers of these genes were used. These included four populations susceptible to insecticides
also identified in Thailand, suggesting that their amplification and five populations showing elevated resistance to deltamethrin.
may have been followed by their neofunctionalization under Two distant geographic areas where pyrethroid resistance is threat-
insecticide selection pressure (Katju and Bergthorsson 2013). ening mosquito control were studied: South East Asia (Thailand)
Interestingly, the best orthologous gene of CCEAE4A in the sheep and South America (French Guiana). Two resistant and one suscep-
blowfly, Lucilia cuprina (LcαE7), encodes an esterase associated tible population having a close genetic background were studied in
with organophosphate resistance through a single point mutation each area (see Table 2 for populations’ information). In addition,
enhancing its activity toward these chemicals (Newcomb et al. one laboratory population selected with deltamethrin for five
1997). Later, it was shown that another mutation of this gene al- generations was used together in comparison to its susceptible “pa-
lowed this enzyme to metabolize synthetic pyrethroids (Heidari rental” population. The laboratory Liverpool population used for
et al. 2005; Devonshire et al. 2007). More recently, multiple ran- public genome sequencing (Liverpool population) (Nene et al.
domly generated nonsynonymous mutations of this gene were 2007) was used as an additional susceptible population. Resistant
shown to enhance its activity toward various pyrethroids includ- populations from Thailand and French Guiana were colonized
ing deltamethrin (Coppin et al. 2012). Although additional work for two generations in the laboratory without insecticide selection
is necessary for validating the role of these nonsynonymous pressure before assessing resistance levels and sample collection.
variants, our data support the selection of detoxification enzyme
variants as being a key process in the adaptation of mosquitoes
to chemical insecticides. Deltamethrin resistance levels
Finally, multiple nonsynonymous variants were detected in For each population, the exposure time necessary to knock down
the gene encoding the voltage-gated sodium channel targeted 50% of individuals (KDT50) was determined on 2- to 4-d-old fe-
by pyrethroids (kdr mutations). Their frequencies were con- males using test tubes equipped with Whatman filter paper im-
sistent with those previously reported in Thailand and South pregnated with 0.05% deltamethrin in silicone oil according
America (Stenhouse et al. 2013; Linss et al. 2014). However, nei- to standard WHO procedure (WHO 2006). Then, a collection of
ther of the two mutations extensively used as resistance markers Whatman filter papers impregnated with different concentrations
(V1016I/G) were retained as best candidates. Although the of deltamethrin (from 0.005% to 0.75%) was used for evaluating
V1016I mutation showed high frequencies in South American re- the lethal dose for 50% of individuals (LD50) of each population.
sistant populations, it showed a poor association with deltameth- Bioassays were conducted in duplicates of 40 2- to 4-d-old females,
rin survival phenotype. The V1016G mutation was associated with and mortality was recorded after 1 h exposure to deltamethrin and
resistance in one Thai population but was not retained among the 24 h recovery without insecticide. For each resistant population,
best candidates according to our criteria. Instead, two other less- the impregnated paper killing 40%–60% mosquitoes after 1 h
known mutations (S989P and F1534C) showed a stronger asso- exposure was then used to segregate the most susceptible individ-
ciation with deltamethrin resistance. Such a result supports the uals (dead at LT25: R− phenotype) from the most resistant ones

1354 Genome Research

www.genome.org
 Variants associated with insecticide resistance

Table 2. Sample descriptions

Population Group Full name Origin History (year colonized) Deltamethrin resistance

Livp S Genome Liverpool (genome) West Africa Laboratory colony (1930s) Susceptible
Bora S Laboratory Bora-Bora French Polynesia Laboratory colony (1990s) Susceptible
Delta R Laboratory Deltamethrin R French Polynesia Bora S selected Slightly resistant
Patg S Thailand Patthalung South Thailand Laboratory colony (2013) Susceptible
Phet R Thailand Phetchaburi SW of Bangkok Field collected Resistant
Nakh R Thailand Nakhon Sawan Center Thailand Field collected Resistant
NwOr S S. America New Orleans Louisiana Laboratory colony (1980s) Susceptible
Cayn R S. America Cayenne French Guiana Field collected Highly resistant
StGe R S. America St-Georges de l’Oyapock French Guiana Field collected Highly resistant

(survivors at LT75: R+ phenotype) by adjusting the exposure time nome Analyzer IIx (Illumina) producing 76 bp reads. An average
from 10 min to 3 h. sequencing coverage of greater than 60× was expected for each
sample.

Sample preparation
Reads filtering and mapping
All sampled mosquitoes consisted of 2- to 4-d-old non-blood-fed
Ae. aegypti females grown in standardized laboratory conditions. Sequenced reads were assigned to each sample (unplexing), and
For each population/phenotype, 200 individuals were collected adaptors were removed. Overall read quality was checked for
and stored individually in silica gel. For each population, three each sample using FastQC (http://www.bioinformatics.babraham
pools of 30 females not exposed to insecticide were also collected .ac.uk/projects/fastqc). Reads were then filtered based on their
and stored in RNAlater (Life Technologies) for gene expression length, pairing, and quality using Trimmomatic (Bolger et al.
analyses. Gene amplification screening by exon capture and 2014). Parameters were set as follows: read length ≥76 bp; mean
mass sequencing was performed on gDNA extracted from two Phred quality score greater than 30; and Phred quality score greater
pools of 65 individuals per population/phenotype. Genomic than 30 across all reads (10-bp sliding window). Only paired
DNA was extracted using the PureGene Core Kit A (Qiagen) fol- reads were kept for mapping. Reads were mapped to the Ae.
lowing the manufacturer’s instructions. For each population/ Aegypti genome (AaegL2 assembly) using BWA (Li and Durbin
phenotype, gDNA from each pool were then combined in equal 2009) with default parameters implemented into a Galaxy pipe-
quantity in order to be representative of a total of 130 individuals. line (http://galaxyproject.org). Mapped reads were loaded into
Quantitative PCR validation of gene amplifications and genotyp- Genespring NGS version 12.5 (Agilent) and further filtered accord-
ing of kdr mutations was performed on gDNA extracted from ing to their mapping quality (alignment score of 95 or above). In
12 single adult females per population/phenotype using the meth- case of multiple mapped reads, only primary hits were conserved.
od described in Collins et al. (1987) and resuspended in 20 µL
nuclease-free water. Gene expression analyses were performed on Copy number variation analysis
RNA extracted from three pools of 30 adult females per popula- Coverage of target genes was quantified for each sample and nor-
tion (R and S populations) using the RNAqueous4PCR kit (Life malized according to the total number of filtered reads per library.
Technologies) according to the manufacturer’s instructions and re- Only genes showing more than 90 reads per kilobase exon model
suspended in 50 µL nuclease-free water. per million sequenced reads (RPKM) in all samples were considered
for further analysis. Detection of copy number variation (CNV)
Capture of target regions and sequencing was based on differential coverage between samples obtained
from resistant and susceptible populations. Fold changes (FC) be-
Gene capture was performed by Hybrigenics-Helixio (Clermont-
tween resistant samples (R+ and R− phenotypes) and all suscepti-
Ferrand, France) using the SureSelect target enrichment system
ble populations (meanS) were computed for each gene. Genes
(Agilent). Capture library consisted in 51,073 overlapping RNA
satisfying the following criteria in any resistant population were
probes of 120 bp (baits) targeting 3458 exons belonging to 789
considered affected by gene amplifications in association with del-
genes. The mean coverage of target regions was 4×. These genes
tamethrin resistance: [(R+)/(meanS) FC ≥ 1.5] and [(R+)/(R−) FC] >
were chosen according to their putative role in insecticide resis-
1 (i.e., 1.5-fold more copies in resistant samples and more copies in
tance. These genes include all detoxification enzymes sensu lato,
LD75 survivors compared to LD25 dead).
all cuticle proteins, all ABC transporters together with several ion
channels, redox enzymes, and synaptic proteins (list of captured
genes in Supplemental Table 5). Capture of target genes was per- Validation of gene amplifications on individual mosquitoes
formed according to “SureSelect Target Enrichment System for Gene amplifications detected by previous CNV analysis were
Illumina Paired-end Sequencing Library version 1.5” protocol. further studied by qPCR on 156 individual mosquitoes (12 from
Briefly, 3 µg gDNA were fragmented using a Bioruptor (Diage- each population/phenotype). Nine candidate genes located in
node), ligated to adaptors, and amplified by PCR using Herculase three different genomic clusters affected by gene amplifica-
II DNA polymerase (Agilent). After QC analysis, libraries were tions were selected: Supercontig 1.1188: AAEL014614 (CYP9J?),
hybridized to biotinylated baits and purified using Dynabeads AAEL014615 (CYP9J23), and AAEL014617 (CYP9J28); Super-
MyOne Streptavidin T1 beads (Life Technologies). Captured contig 1.1327: AAEL014890 (CYP6CC1), AAEL014891 (CYP6P?),
DNA fragments were amplified, purified, and multiplexed be- and AAEL014893 (CYP6BB2); Supercontig 1.142: AAEL05112
fore sequencing. Paired-end sequencing was performed on a Ge- (CCEAE3A), AAEL005101(CCEAE4A), and AAEL005122

Genome Research 1355

www.genome.org
Faucon et al.

(CCEAE6A). PCR primers targeting exonic regions were de- Departure from neutrality is supposed when the locus-specific
signed using NCBI Primer BLAST and checked for specificity component is necessary to explain the observed pattern of diver-
against the whole Ae. aegypti genome (Supplemental Table 6). sity using a reversible jump Markov chain Monte Carlo (MCMC).
Two target genes showing a constant copy number across all sam- Because the high genetic proximity between R+ and R− pheno-
ples from sequencing data (see above) were used for normali- types can bias the estimation of the neutral model, only S popu-
zation (AAEL005950 and AAEL007808). Real-time quantitative lations and R+ phenotypes from resistant populations were
PCR was performed on an iQ5 cycler (Bio-Rad). PCR reactions con- used. For the same reasons, loci showing allele frequency <5% or
sisted of 3 µL gDNA template (see above), 3.6 µL nuclease free >95% in all samples were not considered. BayeScan3 was run
water, 0.45 µL of each primer (10 mM), and 7.5 µL of SYBR with default parameters and the following hierarchical structure:
Green Supermix 2× (Bio-Rad). A dilution scale made from a pool Thailand (Patg S, Nakh R+, and Phet R+), South America (NwOr
of all gDNA samples was used for assessing PCR efficiency and S, Cayn R+, StGe R+), and Laboratory (Bora S, Delta R+). Loci
quantification. All samples were amplified in triplicates (14 popu- showing a Fsc Q-value ≤ 0.05 in any geographical group were
lations × 12 individuals × 3 replicates). After normalization, CNVs considered under selection (i.e., outlier locus). These loci were
were expressed as mean relative gDNA quantity compared to one further filtered based on their association with the resistance
individual of the Livp S population. phenotype by retaining only those showing both [f(R+)-f(S)] and
[f(R+)-f(R−)] metrics being positive (enrichment of variant allele)
Impact of gene amplifications on transcription levels or negative (enrichment of reference allele). Because the Ae. aegypti
genome is not assembled, selection footprints were looked for at
The link between gene amplifications and increased transcription the supercontigs scale. For each supercontig, the density of outlier
levels was investigated on nine genes (see above) by RT-qPCR. loci was computed by dividing the number of outlier loci by the
Total RNA samples from each population were treated with total length of captured regions within the supercontig. In order
DNase I (Invitrogen) to remove genomic DNA following the man- to avoid inferring selection footprint from limited genomic cover-
ufacturer’s instructions. Reverse transcription and qPCR reactions age, only supercontigs represented by not more than four genes or
were performed as described in Nkya et al. (2014). Data analysis from which the captured region represented >10% of supercontig
was performed according to the ΔΔCt method taking into account length were considered.
PCR efficiency (Pfaffl 2001) and using the housekeeping genes
encoding the ribosomal proteins L8 (AGAP005802) and S7
(AGAP010592) for normalization. Three technical replicates were Identifying polymorphisms mostly associated
performed per population, and results were expressed as mean with deltamethrin resistance
transcription ratio in each resistant population ± SD relative to Allele frequency-based filtering was combined with the Bayesian
the mean transcription ratio of all susceptible populations. approach for identifying polymorphisms most strongly associated
with deltamethrin resistance. Frequency-based filtering consisted
Polymorphism calling and population structure in retaining alleles satisfying the following conditions in any resis-
Polymorphisms (SNPs, multiple nucleotide polymorphisms tant/susceptible population pair from each geographical group:
[MNPs], and indels) were called against the reference genome [f(R+)-f(S) ≥ 40%] and [f(R+)-f(R−) ≥ 5%] and both metrics being
using filtered reads by using Genespring NGS version 12.5 positive or negative. Alleles passing both the frequency-based fil-
(Agilent). Calling parameters were as follows for each sample: call- tering and the Bayesian approach were retained. In addition, the
ing score of 50 or above (P-value < 10−5); ignore homopolymer 10 best candidates obtained by each approach were also retained.
or higher and surrounding positions; locus coverage of 30 or For the Bayesian approach, this included loci showing Fsc Q-values
more. Variants identified from each sample were further filtered within the best 10th percentile from each region. For frequency-
based on their coverage and strand bias across all samples (cover- based filtering, this includes alleles passing frequency-based filter-
age 30 or more; strand bias 140 or less). For each variant, allele ing and showing the five highest Abs{[f(R+)-f(S)]+[f(R+)-f(R−)]}
frequencies were estimated for each population based on the num- metric from each region. Variants most associated with deltameth-
ber of reads supporting each allele, and genic effects were comput- rin resistance were then filtered according to their genic effects and
ed against the reference genome. A principal component analysis those being nonsynonymous (i.e., affecting protein sequence)
(PCA) based on allele frequencies of all polymorphic variants were retained.
across all samples was used for inferring the genetic structure of
the studied populations and confirming the relevance of the geo- Validation of polymorphism data and association between
graphical groups. kdr mutations and gene amplifications
Allele frequencies obtained by high-throughput sequencing were
Searching for selection footprint using a hierarchical validated on kdr mutations (V1016I, V1016G, and F1534C) on
Bayesian approach 156 individuals (12 individual mosquitoes from each popula-
BayeScan3 was used for identifying loci under natural selection tion/phenotype). Genomic DNA was extracted as described above,
through a hierarchical Bayesian model, taking into account popu- and kdr mutations were detected by qPCR following the method
lation structure and usable on pools of individuals (Foll et al. described in Saavedra-Rodriguez et al. (2007). Associations be-
2014). This approach estimates the probability that each locus tween the presence of kdr mutations and gene amplifications
is subject to selection by discriminating population-specific and were tested across the five resistance populations for the three
locus-specific components of the fixation index using a logistic kdr mutations and the nine genotyped gene amplifications by test-
regression. The posterior probability of a given locus being un- ing linear correlations between the kdr gentoype (0: homozygote
der selection is assessed by defining two alternative models, one wildtype, 0.5: heterozygote, 1: homozygote resistant) and gene
including the locus-specific effect and the other excluding it. amplification levels.

1356 Genome Research

www.genome.org
 Variants associated with insecticide resistance

Data access DNA-based diagnostics for the monitoring of mosquito vector popula-
tions. Malar Res Treat 2010: 190434.
The sequence data from this study have been submitted to the Bhatt S, Gething PW, Brady OJ, Messina JP, Farlow AW, Moyes CL, Drake JM,
Brownstein JS, Hoen AG, Sankoh O, et al. 2013. The global distribution
European Nucleotide Archive (ENA; http://www.ebi.ac.uk/ena)
and burden of dengue. Nature 496: 504–507.
under accession number PRJEB7976. Bolger AM, Lohse M, Usadel B. 2014. Trimmomatic: a flexible trimmer for
Illumina sequence data. Bioinformatics 30: 2114–2120.
Boonsuepsakul S, Luepromchai E, Rongnoparut P. 2008. Characterization
Acknowledgments of Anopheles minimus CYP6AA3 expressed in a recombinant baculovirus
system. Arch Insect Biochem Physiol 69: 13–21.
This project was primarily funded by the Centre National de Brengues C, Hawkes NJ, Chandre F, McCarroll L, Duchon S, Guillet P,
la Recherche Scientifique (CNRS, INEE department) (Grant pro- Manguin S, Morgan JC, Hemingway J. 2003. Pyrethroid and
DDT cross-resistance in Aedes aegypti is correlated with novel muta-
gramme APEGE 2013, project Mosqui-Target). Additional funding tions in the voltage-gated sodium channel gene. Med Vet Entomol 17:
was received from the French Institut de Microbiologie et Maladies 87–94.
Infectieuses (grant IMMI 109764). We also thank the Laboratoire Brito LP, Linss JG, Lima-Camara TN, Belinato TA, Peixoto AA, Lima JB, Valle
d’Ecologie Alpine of Grenoble for additional funding. F.F. was D, Martins AJ. 2013. Assessing the effects of Aedes aegypti kdr mutations
on pyrethroid resistance and its fitness cost. PLoS One 8: e60878.
supported by a PhD fellowship from the Grenoble-Alpes
Chandor-Proust A, Bibby J, Regent-Kloeckner M, Roux J, Guittard-Crilat E,
University. We acknowledge support from the federative structure Poupardin R, Riaz MA, Paine M, Dauphin-Villemant C, Reynaud S,
Environmental and Systems Biology (BEeSy) of Grenoble-Alpes et al. 2013. The central role of mosquito cytochrome P450 CYP6Zs in in-
University. This work was also supported by the French-Thai coop- secticide detoxification revealed by functional expression and structural
eration programme-PHC Siam project (RESA 2013–2014) funded modelling. Biochem J 455: 75–85.
Collins FS, Drumm ML, Cole JL, Lockwood WK, Vandewoude GF, Iannuzzi
by the French embassy and the Office of the Higher Education MC. 1987. Construction of a general human-chromosome jumping li-
Commission of Thailand. We also thank the Thailand Internation- brary, with application to cystic-fibrosis. Science 235: 1046–1049.
al Development Cooperation Agency (TICA) through the STOP- Coppin CW, Jackson CJ, Sutherland T, Hart PJ, Devonshire AL, Russell RJ,
VEC programme. This work was also supported by the Center for Oakeshott JG. 2012. Testing the evolvability of an insect carboxylester-
ase for the detoxification of synthetic pyrethroid insecticides. Insect
Advanced Studies for Agriculture and Food, Institute for Advanced
Biochem Mol Biol 42: 343–352.
Studies, Kasetsart University under the Higher Education Research Corbel V, Nosten F, Thanispong K, Luxemburger C, Kongmee M,
Promotion and National Research University Project of Thailand, Chareonviriyaphap T. 2013. Challenges and prospects for dengue and
Office of the Higher Education Commission, Ministry of Educa- malaria control in Thailand, Southeast Asia. Trends Parasitol 29: 623–
tion, Thailand. W.J. was supported by the Thailand Research 633.
David JP, Strode C, Vontas J, Nikou D, Vaughan A, Pignatelli PM, Louis C,
Fund (senior research scholarship RTA 5558002 and grant for Hemingway J, Ranson H. 2005. The Anopheles gambiae detoxification
new researchers MRG 5380102) and Kasetsart University Research chip: a highly specific microarray to study metabolic-based insecticide
and Development Institute (KURDI). We also acknowledge the resistance in malaria vectors. Proc Natl Acad Sci 102: 4080–4084.
UMR 5558 and the Pôle Rhône-Alpin de Bioinformatique (PRABI) David JP, Ismail HM, Chandor-Proust A, Paine MJI. 2013. Role of cyto-
for providing access to their computational cluster. Finally, we chrome P450s in insecticide resistance: impact on the control of mos-
quito-borne diseases and use of insecticides on Earth. Philos Trans R
thank Dr. A. Bonin, Dr. M. Weill, and the anonymous reviewers Soc Lond B Biol Sci 368: 20120429.
for their constructive comments on the manuscript. David JP, Faucon F, Chandor-Proust A, Poupardin R, Riaz MA, Bonin A,
Author contributions: F.F. was involved in all aspects of the Navratil V, Reynaud S. 2014. Comparative analysis of response to selec-
study and helped draft the manuscript. I.D. contributed to field tion with three insecticides in the dengue mosquito Aedes aegypti using
mRNA sequencing. BMC Genomics 15: 174.
sampling and molecular analyses. T.G. provided technical support
Devonshire AL, Heidari R, Huang HZ, Hammock BD, Russell RJ, Oakeshott
for mosquito rearing and contributed to sample preparation. V.N. JG. 2007. Hydrolysis of individual isomers of fluorogenic pyrethroid an-
and F.B. contributed to data analyses and helped draft the manu- alogs by mutant carboxylesterases from Lucilia cuprina. Insect Biochem
script. F.C., W.J., T.C., R.G., I.D., and V.C. contributed to study Mol Biol 37: 891–902.
design and helped draft the manuscript. P.S., K.T., and R.P. con- Duangkaew P, Pethuan S, Kaewpa D, Boonsuepsakul S, Sarapusit S,
Rongnoparut P. 2011. Characterization of mosquito CYP6P7 and
tributed to field sampling. S.R. contributed to sample preparation CYP6AA3: differences in substrate preference and kinetic properties.
and helped analyze data and draft the manuscript. J.P.D. con- Arch Insect Biochem Physiol 76: 236–248.
ceived the study, contributed to field sampling and sample prepa- Edi CV, Djogbénou L, Jenkins AM, Regna K, Muskavitch MA, Poupardin R,
ration, analyzed data, and wrote the manuscript. Jones CM, Essandoh J, Kétoh GK, Paine MJ, et al. 2014. CYP6 P450 en-
zymes and ACE-1 duplication produce extreme and multiple insecticide
resistance in the malaria mosquito Anopheles gambiae. PLoS Genet 10:
e1004236.
References Feyereisen R. 2006. Evolution of insect P450. Biochem Soc Trans 34(Pt 6):
1252–1255.
Altmüller J, Budde BS, Nürnberg P. 2014. Enrichment of target sequences
Feyereisen R. 2011. Arthropod CYPomes illustrate the tempo and mode in
for next-generation sequencing applications in research and diagnos-
P450 evolution. Biochim Biophys Acta 1814: 19–28.
tics. Biol Chem 395: 231–237.
Amichot M, Tarés S, Brun-Barale A, Arthaud L, Bride JM, Bergè JB. 2004. Foll M, Gaggiotti OE, Daub JT, Vatsiou A, Excoffier L. 2014. Widespread
Point mutations associated with insecticide resistance in the signals of convergent adaptation to high altitude in Asia and America.
Drosophila cytochrome P450 Cyp6a2 enable DDT metabolism. Eur J Am J Hum Genet 95: 394–407.
Biochem 1: 1250–1257. Gaur U, Li K, Mei S, Liu G. 2013. Research progress in allele-specific expres-
Arensburger P, Megy K, Waterhouse RM, Abrudan J, Amedeo P, Antelo B, sion and its regulatory mechanisms. J Appl Genet 54: 271–283.
Bartholomay L, Bidwell S, Caler E, Camara F, et al. 2010. Sequencing Harris AF, McKemey AR, Nimmo D, Curtis Z, Black I, Morgan SA, Oviedo
of Culex quinquefasciatus establishes a platform for mosquito compara- MN, Lacroix R, Naish N, Morrison NI, et al. 2012. Successful suppression
tive genomics. Science 330: 86–88. of a field mosquito population by sustained release of engineered male
Bariami V, Jones CM, Poupardin R, Vontas J, Ranson H. 2012. Gene ampli- mosquitoes. Nat Biotechnol 30: 828–830.
fication, ABC transporters and cytochrome P450s: unraveling the Heidari R, Devonshire AL, Campbell BE, Dorrian SJ, Oakeshott JG, Russell
molecular basis of pyrethroid resistance in the dengue vector, Aedes RJ. 2005. Hydrolysis of pyrethroids by carboxylesterases from Lucilia
aegypti. PLoS Negl Trop Dis 6: e1692. cuprina and Drosophila melanogaster with active sites modified by in vitro
Bass C, Field LM. 2011. Gene amplification and insecticide resistance. Pest mutagenesis. Insect Biochem Mol Biol 35: 597–609.
Manag Sci 67: 886–890. Hemingway J, Hawkes NJ, McCarroll L, Ranson H. 2004. The molecular
Bass C, Nikou D, Vontas J, Donnelly MJ, Williamson MS, Field LM. 2010. basis of insecticide resistance in mosquitoes. Insect Biochem Mol Biol
The Vector Population Monitoring Tool (VPMT): high-throughput 34: 653–665.

Genome Research 1357

www.genome.org
Faucon et al.

Hodgson E. 2003. In vitro human phase I metabolism of xenobiotics I: pes- Nakamura Y, Sugihara K, Sone T, Isobe M, Ohta S, Kitamura S. 2007. The in
ticides and related compounds used in agriculture and public health, vitro metabolism of a pyrethroid insecticide, permethrin, and its hydro-
May 2003. J Biochem Mol Toxicol 17: 201–206. lysis products in rats. Toxicology 235: 176–184.
Hoffmann AA, Montgomery BL, Popovici J, Iturbe-Ormaetxe I, Johnson PH, Nene V, Wortman JR, Lawson D, Haas B, Kodira C, Tu ZJ, Loftus B, Xi ZY,
Muzzi F, Greenfield M, Durkan M, Leong YS, Dong Y, et al. 2011. Megy K, Grabherr M, et al. 2007. Genome sequence of Aedes aegypti, a
Successful establishment of Wolbachia in Aedes populations to suppress major arbovirus vector. Science 316: 1718–1723.
dengue transmission. Nature 476: 454–457. Newcomb RD, Campbell PM, Ollis DL, Cheah E, Russell RJ, Oakeshott JG.
Holt RA, Subramanian GM, Halpern A, Sutton GG, Charlab R, Nusskern DR, 1997. A single amino acid substitution converts a carboxylesterase to
Wincker P, Clark AG, Ribeiro JM, Wides R, et al. 2002. The genome an organophosphorus hydrolase and confers insecticide resistance on
sequence of the malaria mosquito Anopheles gambiae. Science 298: a blowfly. Proc Natl Acad Sci 94: 7464–7468.
129–149. Nkya TE, Akhouayri I, Kisinza W, David JP. 2013. Impact of environment on
Itokawa K, Komagata O, Kasai S, Okamura Y, Masada M, Tomita T. mosquito response to pyrethroid insecticides: facts, evidences and pros-
2010. Genomic structures of Cyp9m10 in pyrethroid resistant and sus- pects. Insect Biochem Mol Biol 43: 407–416.
ceptible strains of Culex quinquefasciatus. Insect Biochem Mol Biol 40: Nkya T, Akhouayri I, Poupardin R, Batengana B, Mosha F, Magesa S, Kisinza
631–640. W, David JP. 2014. Insecticide resistance mechanisms associated with
Itokawa K, Komagata O, Kasai S, Masada M, Tomita T. 2011. Cis-acting mu- different environments in the malaria vector Anopheles gambiae: a case
tation and duplication: history of molecular evolution in a P450 haplo- study in Tanzania. Malar J 13: 1–15.
type responsible for insecticide resistance in Culex quinquefasciatus. Ortelli F, Rossiter LC, Vontas J, Ranson H, Hemingway J. 2003. Heterologous
Insect Biochem Mol Biol 41: 503–512. expression of four glutathione transferase genes genetically linked to a
Juneja P, Osei-Poku J, Ho YS, Ariani CV, Palmer WJ, Pain A, Jiggins FM. 2014. major insecticide-resistance locus from the malaria vector Anopheles
Assembly of the genome of the disease vector Aedes aegypti onto a genet- gambiae. Biochem J 373(Pt 3): 957–963.
ic linkage map allows mapping of genes affecting disease transmission. Pfaffl MW. 2001. A new mathematical model for relative quantification in
PLoS Negl Trop Dis 8: e2652. real-time RT-PCR. Nucleic Acids Res 29: 6.
Kasai S, Komagata O, Itokawa K, Shono T, Ng LC, Kobayashi M, Tomita T. Poupardin R, Riaz MA, Jones CM, Chandor-Proust A, Reynaud S, David
2014. Mechanisms of pyrethroid resistance in the dengue mosquito vec- JP. 2012. Do pollutants affect insecticide-driven gene selection in mos-
tor, Aedes aegypti: target site insensitivity, penetration, and metabolism. quitoes? Experimental evidence from transcriptomics. Aquat Toxicol
PLoS Negl Trop Dis 8: e2948. 114–115: 49–57.
Katju V, Bergthorsson U. 2013. Copy-number changes in evolution: rates, Poupardin R, Srisukontarat W, Yunta C, Ranson H. 2014. Identification of
fitness effects and adaptive significance. Front Genet 4: 273. carboxylesterase genes implicated in temephos resistance in the dengue
Kondrashov FA. 2012. Gene duplication as a mechanism of genomic adap- vector Aedes aegypti. PLoS Negl Trop Dis 8: e2743.
tation to a changing environment. Proc Biol Sci 279: 5048–5057. Ranson H, Burhani J, Lumjuan N, Black WC. 2010. Insecticide resistance in
Lacey LA. 2007. Bacillus thuringiensis serovariety israelensis and Bacillus dengue vectors. TropIKA.net 1: 1–12.
sphaericus for mosquito control. J Am Mosq Control Assoc 23(2 Suppl): Rinkevich FD, Du Y, Dong K. 2013. Diversity and convergence of sodium
133–163. channel mutations involved in resistance to pyrethroids. Pestic Biochem
Lee CH, Kamijima M, Li C, Taneda S, Suzuki AK, Nakajima T. 2007. Physiol 106: 93–100.
3-Methyl-4-nitrophenol metabolism by uridine diphosphate glucuro-
Riveron JM, Irving H, Ndula M, Barnes KG, Ibrahim SS, Paine MJ, Wondji
nosyltransferase and sulfotransferase in liver microsomes of mice, rats,
CS. 2013. Directionally selected cytochrome P450 alleles are driving
and Japanese quail (Coturnix japonica). Environ Toxicol Chem 26: 1873–
the spread of pyrethroid resistance in the major malaria vector
1878.
Anopheles funestus. Proc Natl Acad Sci 110: 252–257.
Li H, Durbin R. 2009. Fast and accurate short read alignment with Burrows-
Riveron JM, Ibrahim SS, Chanda E, Mzilahowa T, Cuamba N, Irving H,
Wheeler transform. Bioinformatics 25: 1754–1760.
Barnes KG, Ndula M, Wondji CS. 2014a. The highly polymorphic
Li X, Schuler MA, Berenbaum MR. 2007. Molecular mechanisms of metabol-
CYP6M7 cytochrome P450 gene partners with the directionally selected
ic resistance to synthetic and natural xenobiotics. Annu Rev Entomol
CYP6P9a and CYP6P9b genes to expand the pyrethroid resistance
52: 231–253.
front in the malaria vector Anopheles funestus in Africa. BMC Genomics
Linss JG, Brito LP, Garcia GA, Araki AS, Bruno RV, Lima JB, Valle D, Martins
15: 817.
AJ. 2014. Distribution and dissemination of the Val1016Ile and
Phe1534Cys Kdr mutations in Aedes aegypti Brazilian natural popula- Riveron JM, Yunta C, Ibrahim SS, Djouaka R, Irving H, Menze BD, Ismail
tions. Parasit Vectors 7: 25. HM, Hemingway J, Ranson H, Albert A, et al. 2014b. A single mutation
Lounibos LP. 2002. Invasions by insect vectors of human disease. Annu Rev in the GSTe2 gene allows tracking of metabolically based insecticide re-
Entomol 47: 233–266. sistance in a major malaria vector. Genome Biol 15: R27.
Lumjuan N, McCarroll L, Prapanthadara LA, Hemingway J, Ranson H. 2005. Saavedra-Rodriguez K, Urdaneta-Marquez L, Rajatileka S, Moulton M, Flores
Elevated activity of an Epsilon class glutathione transferase confers DDT AE, Fernandez-Salas I, Bisset J, Rodriguez M, McCall PJ, Donnelly MJ,
resistance in the dengue vector, Aedes aegypti. Insect Biochem Mol Biol et al. 2007. A mutation in the voltage-gated sodium channel gene asso-
35: 861–871. ciated with pyrethroid resistance in Latin American Aedes aegypti. Insect
Lumjuan N, Rajatileka S, Changsom D, Wicheer J, Leelapat P, Mol Biol 16: 785–798.
Prapanthadara LA, Somboon P, Lycett G, Ranson H. 2011. The role of Saavedra-Rodriguez K, Strode C, Flores AE, Garcia-Luna S, Reyes-Solis G,
the Aedes aegypti Epsilon glutathione transferases in conferring resis- Ranson H, Hemingway J, Black WC IV. 2014. Differential transcription
tance to DDT and pyrethroid insecticides. Insect Biochem Mol Biol 41: profiles in Aedes aegypti detoxification genes after temephos selection.
203–209. Insect Mol Biol 23: 199–215.
Marcombe S, Poupardin R, Darriet F, Reynaud S, Bonnet J, Strode C, Scholte EJ, Knols BG, Samson RA, Takken W. 2004. Entomopathogenic fun-
Brengues C, Yébakima A, Ranson H, Corbel V, et al. 2009. Exploring gi for mosquito control: a review. J Insect Sci 4: 19.
the molecular basis of insecticide resistance in the dengue vector Somwang P, Yanola J, Suwan W, Walton C, Lumjuan N, Prapanthadara LA,
Aedes aegypti: a case study in Martinique Island (French West Indies). Somboon P. 2011. Enzymes-based resistant mechanism in pyrethroid
BMC Genomics 10: 494. resistant and susceptible Aedes aegypti strains from northern Thailand.
Marcombe S, Mathieu RB, Pocquet N, Riaz MA, Poupardin R, Sélior S, Parasitol Res 109: 531–537.
Darriet F, Reynaud S, Yébakima A, Corbel V, et al. 2012. Insecticide resis- Stenhouse SA, Plernsub S, Yanola J, Lumjuan N, Dantrakool A, Choochote
tance in the dengue vector Aedes aegypti from Martinique: distribution, W, Somboon P. 2013. Detection of the V1016G mutation in the voltage-
mechanisms and relations with environmental factors. PLoS One 7: gated sodium channel gene of Aedes aegypti (Diptera: Culicidae) by al-
e30989. lele-specific PCR assay, and its distribution and effect on deltamethrin
Marcombe S, Paris M, Paupy C, Bringuier C, Yebakima A, Chandre F, David resistance in Thailand. Parasit Vectors 6: 253.
JP, Corbel V, Despres L. 2013. Insecticide-driven patterns of genetic var- Stevenson BJ, Bibby J, Pignatelli P, Muangnoicharoen S, O’Neill PM, Lian
iation in the dengue vector Aedes aegypti in Martinique Island. PLoS One LY, Muller P, Nikou D, Steven A, Hemingway J, et al. 2011.
8: e77857. Cytochrome P450 6M2 from the malaria vector Anopheles gambiae me-
Martins AJ, Brito LP, Linss JG, Rivas GB, Machado R, Bruno RV, Lima JB, tabolizes pyrethroids: sequential metabolism of deltamethrin revealed.
Valle D, Peixoto AA. 2013. Evidence for gene duplication in the volt- Insect Biochem Mol Biol 41: 492–502.
age-gated sodium channel gene of Aedes aegypti. Evol Med Public Health Stevenson BJ, Pignatelli P, Nikou D, Paine MJ. 2012. Pinpointing P450s
2013: 148–160. associated with pyrethroid metabolism in the dengue vector, Aedes
Müller P, Warr E, Stevenson BJ, Pignatelli PM, Morgan JC, Steven A, Yawson aegypti: developing new tools to combat insecticide resistance. PLoS
AE, Mitchell SN, Ranson H, Hemingway J, et al. 2008. Field-caught per- Negl Trop Dis 6: e1595.
methrin-resistant Anopheles gambiae overexpress CYP6P3, a P450 that Strode C, Wondji CS, David JP, Hawkes NJ, Lumjuan N, Nelson DR, Drane
metabolises pyrethroids. PLoS Genet 4: e1000286. DR, Karunaratne SH, Hemingway J, Black WC IV, et al. 2008.

1358 Genome Research

www.genome.org
 Variants associated with insecticide resistance

Genomic analysis of detoxification genes in the mosquito Aedes aegypti. WHO. 2009. Dengue and dengue hemorrhagic fever. World Health Organiza-
Insect Biochem Mol Biol 38: 113–123. tion Factsheet no117 revised March 2009. World Health Organization,
Timoshevskiy VA, Kinney NA, deBruyn BS, Mao C, Tu Z, Severson DW, Geneva, Switzerland.
Sharakhov IV, Sharakhova MV. 2014. Genomic composition and evolu- WHO. 2014. Chikungunya. World Health Organization Fact Sheet N° 327.
tion of Aedes aegypti chromosomes revealed by the analysis of physically World Health Organization, Geneva, Switzerland.
mapped supercontigs. BMC Biol 12: 27. Wondji CS, Irving H, Morgan J, Lobo NF, Collins FH, Hunt RH, Coetzee M,
Vontas J, Ranson H, Alphey L. 2010. Transcriptomics and disease vector Hemingway J, Ranson H. 2009. Two duplicated P450 genes are associat-
control. BMC Biol 8: 52.
ed with pyrethroid resistance in Anopheles funestus, a major malaria vec-
Vontas J, Kioulos E, Pavlidi N, Morou E, Torre AD, Ranson H. 2012.
tor. Genome Res 19: 452–459.
Insecticide resistance in the major dengue vectors Aedes albopictus and
Aedes aegypti. Pestic Biochem Physiol 104: 126–131. Yanola J, Somboon P, Walton C, Nachaiwieng W, Somwang P,
Walker T, Johnson PH, Moreira LA, Iturbe-Ormaetxe I, Frentiu FD, Prapanthadara LA. 2011. High-throughput assays for detection of the
McMeniman CJ, Leong YS, Dong Y, Axford J, Kriesner P, et al. 2011. F1534C mutation in the voltage-gated sodium channel gene in per-
The wMel Wolbachia strain blocks dengue and invades caged Aedes methrin-resistant Aedes aegypti and the distribution of this mutation
aegypti populations. Nature 476: 450–453. throughout Thailand. Trop Med Int Health 16: 501–509.
WHO. 2006. Guidelines for testing mosquito adulticides for indoor residual spray-
ing and treatment of mosquito nets. Document WHO/CDS/NTD/
WHOPES/GCDPP/3. World Health Organization, Geneva, Switzerland. Received January 5, 2015; accepted in revised form June 30, 2015.

Genome Research 1359

www.genome.org