Discovery of super–insecticide-resistant dengue mosquitoes in Asia: Threats of concomitant knockdown resistance mutations

Abstract

Aedes aegypti (Linnaeus, 1762) is the main mosquito vector for dengue and other arboviral infectious diseases. Control of this important vector highly relies on the use of insecticides, especially pyrethroids. The high frequency (>78%) of the L982W substitution was detected at the target site of the pyrethroid insecticide, the voltage-gated sodium channel (Vgsc) of A. aegypti collected from Vietnam and Cambodia. Alleles having concomitant mutations L982W + F1534C and V1016G + F1534C were also confirmed in both countries, and their frequency was high (>90%) in Phnom Penh, Cambodia. Strains having these alleles exhibited substantially higher levels of pyrethroid resistance than any other field population ever reported. The L982W substitution has never been detected in any country of the Indochina Peninsula except Vietnam and Cambodia, but it may be spreading to other areas of Asia, which can cause an unprecedentedly serious threat to the control of dengue fever as well as other Aedes-borne infectious diseases.

Highly insecticide-resistant mosquito vectors are an emerging threat to controlling arbovirus infectious diseases.

INTRODUCTION

Aedes aegypti is a vector for a variety of arboviral infectious diseases including dengue, chikungunya, zika, and yellow fever. Dengue is the most prevalent among them; the number of dengue cases has increased 30-fold in the past 50 years (1). One of the factors for this is contributed by the geographical expansion of A. aegypti due to climate change (2). A modeling study estimated that 390 million people are infected annually (3). Thus, the World Health Organization (WHO) has designated dengue as one of the top 10 threats to global health in 2019, together with other important issues such as antimicrobial resistance and global influenza pandemic (https://who.int/news-room/spotlight/ten-threats-to-global-health-in-2019). Various new technologies, including the use of sterile insect technique and transgenic and Wolbachia-infected mosquitoes, have been developed to break the transmission of dengue virus (4–7). However, they are not ready for large-scale implementation, and there are also some arguments on the release of genetically engineered organisms to the environment (8). With no cost-effective vaccine or medication available, the use of vector control insecticides remains the only approach to manage dengue epidemics. Conversely, many A. aegypti populations have developed resistance to pyrethroids, making vector control more difficult (9).

One of the major mechanisms of pyrethroid resistance is the reduced sensitivity of the target site, i.e., voltage-gated sodium channel (Vgsc). Vgsc gene mutations that cause amino acid substitutions and confer pyrethroid resistance are collectively called knockdown resistance or kdr (10, 11). Several kdr mutations, including V253F, V410L, I1011M, V1016G, and F1534C, have been reported in A. aegypti (12–16). In Asia, V1016G and F1534C are recognized as major kdr mutations. Co-occurrences of substitutions S989P, V1016G, and F1534C are also reported in A. aegypti collected from Thailand, Myanmar, Indonesia, China, Sri Lanka, Saudi Arabia, Laos, and Malaysia (17–19). Although electrophysiological studies revealed that triple mutations of S989P, V1016G, and F1534C strongly reduced the susceptibility of mosquitoes to permethrin and deltamethrin (20), its effect on the phenotype of insecticide resistance has not been elucidated at all.

In Cambodia, there were two major dengue endemics in 2007 and 2012 with 39,618 and 42,362 reported cases, respectively. Nearly 600 deaths were also recorded within these years (21, 22). Dichlorodiphenyltrichloroethane (DDT) had been used to control dengue and malaria in this country from 1981 to 1987, followed by permethrin (from the late 1980s) and deltamethrin (from 2000) treatments by thermal fogging and ultralow volume spraying (23). It is also reported that A. aegypti collected in and around Phnom Penh city exhibited an extremely high level of resistance to several pyrethroids. The mortality rates of A. aegypti population to permethrin and deltamethrin in Phnom Penh were 0 and less than 10%, respectively, by the WHO diagnostic method (23). In 2022, Boyer et al. (24) also reported similar results using seven pyrethroid insecticides. Nevertheless, the molecular mechanism of pyrethroid resistance of A. aegypti collected from Cambodia remains unknown. It is crucial to monitor the insecticide resistance in A. aegypti collected from a dengue epidemic area and to elucidate molecular mechanisms of insecticide resistance of this important mosquito species.

In this study, we examined the relationship between Vgsc and pyrethroid resistance in A. aegypti collected in Asia. Ten previously unidentified substrains of A. aegypti having an identified kdr allele were established and the susceptibility of each strain to pyrethroids was examined. Full-length Vgsc coding DNA sequences were determined for all strains, which resulted in the identification of several previously unknown amino acid substitutions. We focused on the co-occurrence of kdr mutations that confer an extremely high level of pyrethroid resistance. Furthermore, molecular evolution processes of Vgsc to accumulate kdr substitutions were discussed.

RESULTS

Permethrin susceptibility

Permethrin susceptibility of 23 A. aegypti populations collected from Vietnam, Indonesia, Ghana, and Taiwan as well as an insecticide-susceptible strain (SMK) was evaluated (Fig. 1A). This study was conducted as part of a project to assess pyrethroid susceptibility of world Aedes mosquitoes globally (25). Diagnostic doses were established on the basis of the susceptibility of the reference Aedes albopictus (Skuse, 1895) strain HKM (25). The mortality of the SMK strain at 5.9 ng of permethrin per female, which is equivalent to 99% lethal dose (LD₉₉) of HKM, was 100% (Fig. 1A), suggesting that this dose of permethrin could detect permethrin resistance of above a certain level in A. aegypti. By contrast, the field populations showed lower mortalities across the board. When exposed to 5.9 ng per female, all populations except for three Taiwanese populations and a Ghana population (Aburi) had less than 20% mortality. Raising the dose by 10 times (59 ng) led to higher mortality for all strains. Nonetheless, only two populations from Ghana had 100% mortality. Even the Taiwanese populations were found to have surviving individuals. In particular, the populations of Hanoi 2 (Vietnam) and Simo Sidomulyo (Indonesia) showed mortality below 30% even when exposed to 59 ng of permethrin.

Fig. 1. Permethrin susceptibility of field-collected adult Aedes aegypti and its association with kdr alleles.

(A) Mortalities of adult A. aegypti after exposures to permethrin. Mosquitoes were topically exposed to two doses of permethrin, i.e., 5.9 or 59 ng, which were respectively equivalent to the LD₉₉ and LD₉₉ × 10 doses for the susceptible A. albopictus strain HKM (25). Twenty females were treated in each assay, whereby the subsequent proportion of dead mosquitoes was calculated 24 hours after the treatment. Data are the average of four replicates. Error bars indicate SEs for the mortalities of each population. (B) Comparison of permethrin susceptibility between A. aegypti and A. albopictus, which were collected from the same areas. Mortality data for A. albopictus were from the previous report (25). Except for the Taiwanese Pingtung population, all A. aegypti populations showed lower mortalities than A. albopictus. (C and D) Genotyping of Vgsc genes from dead and surviving mosquitoes after the bioassay with 59 ng of permethrin. The total number of mosquitoes tested from Hoa Kien (C) and Hanoi 2 (D) were 78 and 80, respectively.

Differences in the levels of permethrin resistance between A. aegypti and A. albopictus

There have been many reports of pyrethroid resistance in A. aegypti but relatively few in A. albopictus (9). In this study, we compared the permethrin susceptibilities between these two species collected from the water containers at the same areas and found that the level of resistance was higher in A. aegypti (Fig. 1B). This is probably due to the differences in the ecology and habitats of the two species. Because of higher preference to humans as a blood-sucking source and high habitation affinity to humans, a higher percentage of A. aegypti are generally collected indoors than A. albopictus (26). In countries where dengue fever is endemic, vectors are mainly controlled by indoor pyrethroid fogging, so it is likely that adult A. aegypti have more chances of exposure to insecticides resulting in the development of resistance. On the other hand, kdr genes, such as F1534C/S and V1016G, have also been recently detected from European and Asian populations of A. albopictus; thus, controlling this mosquito species by pyrethroids is also becoming difficult at some dengue epidemic regions (25, 27).

Associations between Vgsc substitutions and permethrin susceptibility

To evaluate the association between Vgsc substitutions and pyrethroid susceptibility, genotyping analysis was conducted on individual mosquitoes of the two Vietnamese A. aegypti populations (Hoa Kien and Hanoi 2) that were subjected to 59 ng of permethrin treatment. We first attempted to genotype three codons containing well-known substitutions, namely, S989P, V1016G, and F1534C, which have been shown to be associated with pyrethroid resistance (9, 20, 28), especially among Asian A. aegypti. In this process, we found that an additional known substitution, L982W (14), was included in the same polymerase chain reaction (PCR) product for S989P and V1016G (Fig. 1, C and D). V1016G substitution, one of the major kdr variants in Asia, was not detected from these two Vietnamese populations. Contrarily and unexpectedly, the Hoa Kien population had L982W at a high frequency of 73.1%. There was a strong genotype-phenotype correlation, where 97.6% (40 of 41) of the surviving individuals were L982W homozygotes, whereas 89.2% (33 of 37) of the dead individuals were L982W and F1534C heterozygotes (both sites) (Fig. 1C). Although the frequency of F1534C in the Hoa Kien population was 26.3%, no individuals homozygous for L982W had an F1534C, suggesting that these substitutions were in the repulsion phase (L982 + F1534C and L982W + F1534) (Fig. 1C). Assuming that there is no concomitant L982W + F1534C allele in the Hoa Kien population, most of the dead individuals are heterozygous (L982W + F1534/L982 + F1534C) and most of the surviving individuals are homozygous L982W + F1534. There was a strong correlation between homozygotic L982W genotype and phenotype (dead/survived) (P < 2.2 × 10⁻¹⁶ in Fisher’s exact test) where 97.6% (40 of 41) of the survived individuals were L982W + F1534 homozygotes, whereas no dead individual had the same genotype (Fig. 1C). In the Hanoi 2 population, the frequency of L982W was 97.5%, resulting in the fact that most individuals were homozygous for this substitution. Unlike the Hoa Kien population, there were L982W homozygotes with heterozygous or homozygous F1534C (i.e., WW/FC), indicating that the Hanoi 2 population contained concomitant L982W + F1534C alleles. The frequencies of mosquitoes having the WW/FC genotype in the surviving and dead groups were 40.9% (27 of 66) and 7.1% (1 of 14), respectively (Fig. 1D). The survival rate was much higher in the mosquito group that had at least one L982W + F1534C allele than the mosquito group that did not have concomitant L982W and F1534C substitutions in the Hanoi 2 population (Fig. 1D). Statistical analysis was conducted using Bayesian logistic analysis to evaluate the contribution of L982W to the resistance. The contribution value of L982W was 0.903 [95% confidence interval (CI), 0.644 to 0.997], which was much higher than the values for V1016G (0.122 with 95% CI, 0.003 to 0.450) and F1534C (0.146 with 95% CI, 0.003 to 0.519), suggesting strong correlation of L982W mutation with permethrin resistance.

Establishment of various resistant strains and determination of the whole Vgsc CDSs

Since genotyping studies of Vgsc showed that Hoa Kien and Hanoi 2 populations contained various Vgsc alleles, we next attempted to establish several substrains with homozygous genotypes on all of the Vgsc polymorphic loci (Fig. 2, A and B, and fig. S1). Since the frequency of the V1016G substitution was very low in the two Vietnamese populations, we also tried isolating three additional strains from a population collected from Singapore in 2016 (table S1). After the first selections, conducted on the basis of the four mutations, namely, L982W, S989P, V1016G, and F1534C, we sequenced full-length Vgsc coding regions individually using a combination of the target capture method and the next-generation sequencing (NGS) (fig. S1). When any heterozygous polymorphism was found, we repeated the isolation process using these polymorphisms. In HK-C, isolated from the Hoa Kien population that had the F1534C substitution, two different alleles were found to be included. Two HK-C–derived strains FC66 and FC213 had S66F or L213F substitutions, respectively (Fig. 2B). Similarly, in the HK-W isolated from the Hoa Kien population and having the L982W substitution, two additional alleles were found. Two strains isolated from HK-W were designated as FTW (L199F + A434T + L982W) and FWI (L199F + L982W + T1385I) (Fig. 2B). Three strains were established from the Hanoi 2 population: GY (V1016G + D1763Y), PGC (S989P + V1016G + F1534C), and FTWC (L199F + A434T + L982W + F1534C). Three strains were isolated from the Singapore population: 1534C (F1534C), PG (S989P + V1016G), and PGG (S989P + V1016G + V1703G) (Fig. 2B). Eventually, a total of 1594 mosquitoes were individually genotyped in 12 isolation processes, resulting in the establishment of 10 resistant strains having an identified kdr allele (Fig. 2B).

Fig. 2. Establishment of 10 resistant strains of A. aegypti and their resistance levels to pyrethroids.

(A) Amino acid substitutions found from Vgsc of A. aegypti. Seven amino acid substitutions discovered in this study were indicated with blue letters. Three amino acid substitutions I1011M, V1016G, and F1534C have proven their contribution to pyrethroid resistance by multiple electrophysiological studies (19). In this study, we numbered the amino acid position according to the sequence of the most abundant splice variant of the house fly Vgsc(GenBank accession nos. AAB47604 and AAB47605). (B) Establishment of 10 resistant strains having a single Vgsc allele. Substituted amino acids are indicated in blue bold. The numbers of mosquitoes genotyped were expressed in parentheses below population names. The homogeneity of isolated populations was verified via sequencing whole Vgsc CDSs of at least eight mosquitoes individually via NGS (see Materials and Methods). (C and D) RRs of 12 resistant strains with different mutant Vgsc alleles for two pyrethroid insecticides, permethrin (C) and deltamethrin (D). RRs were calculated by dividing the LD₅₀ of each strain by the LD₅₀ of the susceptible SMK strain. Error bars indicate the minimum and maximum 95% confidence interval (CI) for the RRs. RRs for SMK, MCNaeg-LIC, and MCNaeg-C are from our previous report (13).

Susceptibilities of 10 previously unestablished resistant strains to pyrethroids

Ten resistant strains were tested for the susceptibility to two representative pyrethroids permethrin and deltamethrin (Fig. 2, C and D, and fig. S2). Two strains having L199F + L982W (FTW and FWI) showed approximately the same resistance level to permethrin (52- and 75-fold) and deltamethrin (73- and 106-fold), compared with the strains having V1016G substitution; resistance ratios (RRs) of GY, PG, and PGG to permethrin and deltamethrin were 36- to 102-fold and 26- to 89-fold, respectively. The FTWC (L199F + A434T + L982W + F1534C) strain showed the incomparably highest resistance level, with RRs of 1039- and 527-fold to permethrin and deltamethrin, respectively (Fig. 2, C and D). The 50% lethal dose (LD₅₀) of FTWC to permethrin was 130 to 208 times higher than that of three strains having F1534C (1534C, FC66, and FC213). The PGC (S989P + V1016G + F1534C) strain showed the second-highest resistance level, with RRs to permethrin and deltamethrin both being approximately 160-fold. Mosquitoes were treated with piperonyl butoxide (PBO) before applying pyrethroids to inhibit the detoxification of insecticide by cytochrome P450 monooxygenases. The synergistic ratios (SRs) of PBO on permethrin toxicity varied between strains; the SRs of FC66, FC213, FTW, and FWI, all isolated from the Hoa Kien population, were 2.4 to 4.2 (table S2). Among the three strains isolated from the Hanoi 2 population, the SRs of PGC and GY were 2.5 and 3.5, respectively, and were not substantially different from those of the susceptible SMK (2.7). By contrast, the SR of FTWC to permethrin was 9.2, which is the highest value among the 11 strains tested (Fig. 2C, fig. S2, and table S2). The RRs to permethrin under PBO-treated conditions were 3.6- to 6.1-fold in the three strains containing F1534C (1534C, FC66, and FC213) and 18- to 48-fold in the five strains having V1016G (GY, PG, and PGG). These values were not significantly different from the results of previous studies conducted using congenic strains, which were closely related to a susceptible strain Rockefeller (ROCK) but were homozygous for the S989P + V1016G or F1534C Vgsc alleles (29, 30). The RRs of these two congenic strains 989P + 1016G:ROCK and 1534C:ROCK to permethrin were 40- and 7.0-fold, respectively (29, 30). The RRs of 1531C and GY to permethrin + PBO were similar to those of A. albopictus having F1534C or V1016G substitutions reported in previous studies (25). The RR of FTW (L199F + A434T + F982W) to permethrin with PBO was 33-fold, and these values increased to 300-fold with another amino acid substitution F1534C in the FTWC strain. The RR of PG to permethrin was 48-fold, whereas that of PGC having an F1534C substitution was 177-fold under pretreatment of PBO. The RRs to deltamethrin + PBO were similar to those of permethrin, and the RRs of FTWC and PGC were 217- and 173-fold, respectively, which were 3.0- and 5.2-fold higher than those of FTW and PG, respectively (Fig. 2D and table S2). With PBO treatment, 1534C, FC66, and FC213 strains, having the F1534C substitution, showed resistance to deltamethrin as well, and the RRs of 11- to 12-fold were greater than those to permethrin (3.6- to 6.1-fold). The 95% CIs of RRs for deltamethrin in PGC and FTWC overlapped, suggesting that resistance levels of these two strains to deltamethrin were not significantly different without detoxifications by P450 monooxygenases. The resistance levels of PGC and FTWC to permethrin and deltamethrin were much higher than those of MCN-LIC (V410L + S723T + V1016I + F1534C) and MCN-C (V253F + M374I + G923S + F1534C), which were previously established from a Brazilian population of A. aegypti (Fig. 2, C and D) (13).

DNA polymorphisms in Vgsc CDSs of 10 established strains

The purities of the isolated strains were confirmed by sequencing the entire Vgsc coding sequence (CDS) for at least eight individuals of each strain. Within 10 previously unestablished strains and the susceptible SMK strain, 24 synonymous and 11 nonsynonymous polymorphisms were found, compared with the reference LVP strain, which was used for the Aedes genome project (Fig. 3A). Six of the 11 nonsynonymous polymorphisms were found in this study (S66F, L199F, L213F, A434T, T1385I, and V1703G). On the basis of the polymorphisms each strain has, we predicted how each allele was created historically. It is estimated that Vgscs of both FTWC and PGC were generated not due to additive polymorphisms but due to crossing over events (Fig. 3B). A phylogenic tree created by the full-length CDSs of Vgsc resulted in both alleles having V1016G and L982W fit into the single clades, while F1534C distributed across multiple clades (Fig. 3C).

Fig. 3. Polymorphisms and amino acid substitutions in the coding region of Vgsc gene of A. aegypti strains.

(A) All synonymous (blue) and nonsynonymous (red) nucleotide polymorphisms found in entire CDSs of Vgsc genes of 10 resistant strains. Data of two insecticide-susceptible strains, LVP and SMK, and five resistant mosquitoes from our previous report are also included (13). Accession for LVP is VectorBase accession number. (B) Expected evolutionary processes of Vgsc genes obtained in this study. (C) Phylogenetic tree constructed by Genetyx software using the full-length CDSs of Vgsc gene. Numbers on branches are the 1000 bootstrap values calculated using MEGA7.

Vgsc allele frequencies in A. aegypti collected from Vietnam and Cambodia

To understand the allele frequency of Vgsc in A. aegypti collected from Vietnam and Cambodia, 13 amino acids of Vgsc were targeted and genotyped (Table 1). Another amino acid substitution, T1539A, which was found in the process of genotyping F1534C, was also included (Fig. 2A). In this study, 368 field-collected nonbreeding insects (i.e., generation zero) were examined to understand the gene frequency more accurately. Five possible amino acid substitutions, L213F, V410L, V1007G, I1011M, and T1520I, were not detected in any of the field populations of A. aegypti collected from Vietnam (Hanoi, Dak Lak, and Ho Chi Minh) and Cambodia (Phnom Penh). The frequency of the L982W substitution was high in Vietnam, accounting for more than 79% (Table 1 and Fig. 4). Furthermore, the frequency of the concomitant mutations L982W + F1534C was estimated to be 28.4 to 40.5%, 11.8 to 16.5%, and 6.2 to 7.3% in the populations from Hanoi, Dak Lak, and Ho Chi Minh City, respectively. The frequency of V1016G, one of the major kdr substitutions in Southeast Asia, was low in Vietnam, with 10.8% in Hanoi and <1% in Dak Lak and Ho Chi Minh (Fig. 4). The V1016G substitution was not always accompanied by S989P; in Hanoi, eight mosquitoes had V1016G heterozygously and three of them did not have S989P concomitantly. In total, 51 individuals of A. aegypti collected from two water pools (larvae) and a vehicle (adults) in Phnom Penh, Cambodia, were genotyped. Since the genotype composition was quite simple and all genotypes were composed of any of five alleles, the frequencies of all alleles in the Phnom Penh populations could be estimated (fig. S3). The frequencies of L199F + A434T + L982W + F1534C and L199F + L982W + F1534C in the Phnom Penh population were 63.0 and 8.8%, respectively. Furthermore, the frequency of Vgsc having S989P + V1016G + F1534C was 18.6%, resulting in >90% of the alleles in the Phnom Penh population being occupied by super-resistant–type L982W + F1534C or V1016G + F1534C (Fig. 4 and fig. S3). The frequency of the L199F substitution was also quite high in Vietnam (58.1 to 98.9%) and Cambodia (78.4%). Of 368 mosquitoes tested, 266 were homozygous at the L199F substitution and all were also having L982W substitution homozygously, suggesting that L199F is strongly linked to L982W. Conversely, of 39 mosquitoes that had both L982W and F1534C homozygously, nine had L199 homozygously, suggesting that L982W is not always accompanied by L199F when it is with F1534C (Table 1). The frequencies of A434T and T1385I in Vietnam ranged from 45.9 to 50.0% and from 2.7 to 28.0%, respectively. In this study, no amino acid substitution was detected at the I1011 location, which was inconsistent with a previous study detecting I1011V all homozygously (30 of 30) in the populations collected at Nha Trang, Vietnam (31).

Table 1. Genotype of Vgsc in the field-collected (G0) A. aegypti from Vietnam and Cambodia.

Genotype* (L199F/A434T/L982W/S989P/V1016G/T1385I/F1534C/T1539A)	Vietnam†				Cambodia (Phnom Penh)†
	Hanoi	Dak Lak	Ho Chi Minh	Total	AMPOV	CamboFarm	PPCAR	Total
FF/TT/WW/++/++/++/CC/++	2	0	0	2	0	19	2	21
FF/+T/WW/++/++/++/CC/++	0	0	0	0	0	5	2	7
++/++/WW/++/++/++/CC/+	8	1	0	9	0	0	0	0
+F/+T/+W/+P/+G/++/CC/++	0	0	0	0	0	9	1	10
+F/++/+W/+P/+G/++/CC/++	0	0	0	0	0	1	0	1
++/++/++/PP/GG/++/CC/++	0	0	0	0	0	3	0	3
FF/TT/WW/++/++/++/+C/++	4	0	6	10	5	0	0	5
FF/+T/WW/++/++/+I/+C/++	0	0	2	2	0	0	0	0
FF/+T/WW/++/++/++/+C/++	0	0	3	3	0	0	0	0
+F/+T/WW/++/++/+I/+C/++	0	1	0	1	0	0	0	0
+F/+T/WW/++/++/++/+C/++	0	23	0	23	0	0	0	0
+F/++/WW/++/++/++/+C/+A	0	1	0	1	0	0	0	0
FF/TT/WW/++/++/++/++/++	8	48	18	74	0	0	0	0
+F/++/WW/++/++/+I/+C/++	0	14	0	14	0	0	0	0
+F/++/WW/++/++/++/+C/++	0	4	0	4	0	0	0	0
+F/+T/+W/++/++/++/CC/++	0	0	0	0	1	0	0	1
FF/++/WW/++/++/II/++/++	0	14	1	15	0	0	0	0
FF/++/WW/++/++/++/++/++	0	0	18	18	0	0	0	0
FF/+T/WW/++/++/+I/++/++	1	49	12	62	0	0	0	0
FF/+T/WW/++/++/++/++/++	0	5	24	29	0	0	0	0
FF/++/WW/++/++/+I/++/++	1	4	3	8	0	0	0	0
FF/++/WW/++/++/+I/++/+A	0	5	0	5	0	0	0	0
FF/+T/WW/++/++/++/++/+A	0	5	0	5	0	0	0	0
+F/+T/+W/+P/+G/++/+C/++	1	0	0	1	0	2	0	2
+F/+T/+W/++/++/++/+C/++	1	8	0	9	0	0	0	0
+F/++/+W/++/++/++/+C/+A	0	1	0	1	0	0	0	0
+F/++/+W/++/++/++/+C/++	3	1	1	5	0	0	0	0
+F/++/+W/+P/+G/++/++/++	3	0	0	3	0	0	0	0
+F/+T/+W/+P/+G/++/++/++	0	1	0	1	0	0	0	0
+F/+T/+W/++/+G/++/+C/++	1	0	0	1	0	0	0	0
+F/+T/+W/++/+G/++/++/++	2	0	0	2	0	0	0	0
+F/++/+W/++/++/+I/+C/++	0	6	1	7	0	0	0	0
++/++/++/++/++/++/CC/++	1	0	0	1	1	0	0	1
++/++/++/+P/+G/++/+C/++	1	0	0	1	0	0	0	0
Total	37	191	89	317	7	39	5	51

*L213F, V410L, V1007G, I1011M, and T1520I were also genotyped, but none of these amino acid substitutions was detected.

†Information of the locations where each population was collected is described in table S1.

Fig. 4. Super–insecticide-resistant dengue vectors localizing in Indochina Peninsula.

Frequencies of eight substitutions and alleles with combinations of two substitutions (L982W + F1534C and V1016G + F1534C) of Vgsc in the field-collected A. aegypti collected from Vietnam and Cambodia are shown. Since the possible number of alleles is large, the exact frequency of two alleles cannot be calculated and minimum and maximum possible frequencies of these alleles are expressed for the populations from Vietnam. In the population of Hanoi, only three mosquitoes possibly had V1016G + F1534C as an allele and two of them had S989P concomitantly. In the population from Phnom Penh, the exact frequency of each allele was estimated because of the restricted number of alleles. All mosquitoes having V1016G + F1534C had S989P concomitantly in the Phnom Penh population. Please see fig. S3 for the actual frequencies of all five alleles of Vgsc in the populations from Phnom Penh. No L982W substitution was detected from the cities marked with blue plots in China, Laos, Thailand, and Malaysia although the L982 was targeted (15, 39, 40). Note that L982W has never been detected in the Indochina Peninsula except Vietnam and Cambodia.

Molecular modeling ofA. aegypti Vgsc

To understand the effects of amino acid substitutions on the affinity of pyrethroids to the sodium channel, a homology model of A. aegypti Vgsc was generated on the basis of electron microscopy crystal structure of the American cockroach Periplaneta americana (Linnaeus, 1758) NavPaS. The Vgsc model with and without amino acid substitution(s) was subjected to docking simulation with 1R-trans-permethrin. Segments 5 and 6 of each domain conform to the pore domain, which is directly involved in ion permeation. Several Vgsc substitutions such as V1016G, L1014F, and F1534C, all of which have been proven for their association with pyrethroid resistance by electrophysiological studies, are located in these segments (9, 32). In this model, 1R-trans-permethrin was located at this binding pocket, the so-called pyrethroid receptor site PyR1, as previously reported (Fig. 5A) (28). The molecular model exhibited that L982 locates at the region near V1016 and F1534 where permethrin interacts (Fig. 5, A and B). The indolyl group of the aromatic ring of L982W, however, is a steric obstacle for permethrin for approaching this position (Fig. 5C). Furthermore, in L982W + F1534C, the degree of inhibition looks much stronger (Fig. 5D). Effects of L982W with F1534C on the relationship between permethrin and Vgsc were quite similar to those of V1016G with F1534C (Fig. 5, E to G). These results strongly support the fact that the FTWC strain having Vgsc with L982W + F1534C exhibited a quite high level of pyrethroid resistance in the bioassay. We also conducted a similar docking simulation at the deduced second pyrethroid receptor PyR2 (28); however, this region has some distance with the L982, and no significant effect on the interaction with permethrin was observed.

Fig. 5. Docking simulation of A. aegypti Vgsc (with and without kdr mutations) and 1R-trans-permethrin (yellow surface in the center of the channel).

(A) Locations of the key amino acid residues L982, V1016, and F1534 and nine other amino acid residues found in this study. Note that the residues L982, V1016, and F1534 are surrounding permethrin configurating at the pyrethroid binding pocket. (B to G) Magnified models around the pyrethroid binding pocket. (B) Wild-type, nonmutated Vgsc . (C) Vgsc with L982W. (D) Vgsc with concomitant L982W + F1534C. (E) Vgsc with F1534C. (F) Vgsc with V1016G. (G) Vgsc with concomitant V1016G + F1534C. Note that concomitant substitutions L982W + F1534C severely interfere with permethrin approaching the normal binding position of Vgsc (D) compared to the Vgsc of wild type (B) and Vgsc with a single amino acid substitution L982W (C) or F1534C (E).

DISCUSSION

In the present study, the L982W substitution located at the linker helix connecting S5 and S6 in domain II of Vgsc was identified at a fairly high frequency (>78%) in A. aegypti collected from Vietnam and Cambodia. L982W had been originally identified in an A. aegypti population collected before 1998 at Long Hoa, Vietnam (Fig. 4) (14). The contribution of this substitution on pyrethroid resistance, however, has never been confirmed, and therefore, this substitution has never been highlighted as a resistance factor. Chen et al. (33) previously reported that the alanine-to-valine substitution in linker helix of the S5 and S6 in domain III of Drosophila Vgsc conferred pyrethroid resistance. They proved this by electrophysiological studies using mutant cockroach Vgsc expressed in Xenopus oocytes. Remarkably, this amino acid position is topologically at the same position as L982W in domain II, implying the important role of L982W in pyrethroid resistance (33). In this study, we established 10 A. aegypti strains that have various knockdown resistance substitutions in Vgsc and evaluated their resistance levels to pyrethroid insecticides. Strains having L982W (FTW and FWI) showed similar or even higher levels of resistance to pyrethroids than those having V1016G (GY, PG, and PGG), a typical knockdown resistance substitution in Asia. Concomitant mutations L199F + A434T + L982W + F1534C were also confirmed, and the strain that has these multiple Vgsc mutations exhibited >1000-fold resistance, which is >10 times higher than any resistance level that other field populations ever reported. By doing the bioassay with pretreatment of PBO, the association of cytochrome P450–mediated detoxification was confirmed in this resistance; however, 300-fold resistance remained, suggesting a remarkable contribution of Vgsc mutations in this resistance.

L982W is located in the linker helix, connecting the fifth and sixth segments in the second domain of Vgsc (Fig. 2A). E985 and E988 located in this region are called the inner and outer rings, respectively, and play very important roles for sodium ion selectivity and permeation rate (34). We confirmed that the 31 amino acid sequences containing these two rings and L982 (PRW…FRVL₉₈₂CGE₉₈₅WIE₉₈₈SMWDCM) are completely conserved among at least 80 insect species (fig. S4). Regarding the core sequence (L₉₈₂CGE₉₈₅WIE₉₈₈), all seven amino acids of 80 insects are completely identical with Vgsc of mammals such as rats, mice, and humans (fig. S4). It is presumed that this amino acid composition is highly important and essential to maintaining the function of Vgsc beyond the borders of vertebrates and invertebrates. Both leucine and tryptophan are hydrophobic, nonpolar amino acids, but tryptophan has an aromatic ring, which may alter the intermolecular forces with ligands (fig. S5). L1014F, the first reported and the most common kdr allele in insects, is also the substitution from leucine (32). Phenylalanine is also a hydrophobic, nonpolar amino acid and has an aromatic ring–like tryptophan (fig. S5). Although the substitution from leucine to tryptophan may have had any costs for Vgsc to maintain its function as a channel, the benefits may have outweighed the disadvantages in Vietnam and Cambodia, where strong selection pressure by pyrethroid insecticides is expected (35, 36).

Previously unknown amino acid substitutions have been identified by sequencing CDSs of Vgsc: S66F, L199F, L213F, A434T, T1385I, T1539A, and V1703G. The combination of NGS and a concentration method of CDSs using target capture probes effectively found SNPs as in previous studies (13, 37). Since all of these substitutions locate at some distance from the permethrin binding region (Fig. 5A), and since there have been no reports of amino acid substitution at these locations from other resistant pest insects, these mutations are unlikely to affect the susceptibility of Vgsc to pyrethroids (32). As for D1763Y, electrophysiological studies have ruled out its involvement in pyrethroid sensitivity (28). The FTWC strain, which showed the highest level of resistance to pyrethroids among 10 resistant strains tested, had L199F and A434T besides L982W and F1534C. L199F is located in the pore helix, which connects the second and the third segments of the first domain (Fig. 2A). This substitution is strongly linked to L982W since all mosquitoes having L199F homozygously had L982W homozygously (Table 1). Conversely, L982W is not always with L199F when it is accompanied by F1534C (Table 1). Although the role of L199F could not be derived clearly by molecular modeling analysis, it may sterically compensate for the disadvantage of Vgsc as a sodium channel caused by the L982W substitution. To evaluate the effects of each substitution on pyrethroid resistance, the first generation (G1) of the three field populations of A. aegypti, Hanoi 2, Dak Lak 6, and HCM 4, was genotyped after selections with 59 ng of permethrin. Thirteen amino acid residues that have been detected in this study or in previous reports were targeted (Fig. 2A and table S3). In the HCM 4 population, the estimated frequencies of L199F + A434T + L982W (all homozygotes) in dead and surviving mosquitoes were 2.9 and 3.1%, respectively (table S3), whereas the estimated frequency of L199F + L982W (all homozygotes) in dead and surviving mosquitoes were 5.8 and 12.4%, respectively. These results seem to imply that A434T is not involved in the pyrethroid resistance. Unfortunately, all three populations examined included 25 genotypes resulting in the restricted number of each genotype, and we could not statistically evaluate the effect of each mutant allele (table S3). Conversely, 31 of 32 (96.9%) mosquitoes from the Hanoi 2 population having at least one L982W + F1534C allele survived, which is significantly higher than the survivorship (31 of 44, 70.5%) in the group having L982W homozygously but without F1534C, suggesting a strong effect of concomitant L982W + F1534C on permethrin resistance (table S3). Nevertheless, further studies are needed to elucidate the association of L199F and A434T on pyrethroid resistance using electrophysiological studies and/or gene editing technologies.

Kawada et al. (38) conducted a nationwide survey of the pyrethroid susceptibility of A. aegypti between 2006 and 2008 in Vietnam. Larvae of A. aegypti collected from central and southern Vietnam were particularly resistant to d-T₈₀-allethrin. The frequency of V1016G was not so high throughout Vietnam, and F1534C was identified at a certain frequency although no clear correlation with pyrethroid susceptibility was observed (12). It is possible that resistance of A. aegypti to d-T₈₀-allethrin confirmed between 2006 and 2008 was mostly due to the L982W substitution, which was not focused at that genotyping study. In the present study, frequencies of L982W were higher at central (Dak Lak, 95.5%) and southern (Ho Chi Minh, 98.9%) populations (Fig. 4). We rechecked the raw data from the previous genotyping studies by Kawada et al. and calculated the frequency of the L982W in A. aegypti collected between 2006 and 2008 in Vietnam (12). Of 448 mosquitoes tested, 240 were homozygous and 50 were heterozygous for L982W, and the overall frequency of this substitution was 59.2%. It is noteworthy that the frequency of L982W was lower than that of A. aegypti collected from Vietnam in 2016 (>79%, Fig. 4) and we could not confirm any individual having L982W + F1534C homozygously implying that this allele has been selected by pyrethroid treatments and increased its frequency after the year 2008.

The L982W substitution is located near S989P and V1016G on the Vgsc, and all can be genotyped by sequencing the same PCR product of domain II. It is a sort of mystery that the L982W substitution has never been identified in the recent studies of A. aegypti from at least 16 regions of the neighboring countries: Thailand, Laos, and China, although this substitution was focused (Fig. 4) (15, 39, 40). L982W was also not detected from 442 A. aegypti collected from three regions of Nepal in 2017 and 2018 (41). Thus far, there is no information regarding the fitness cost of the L982W on the viability of mosquitoes, which we must investigate; at least in the laboratory, we do not experience difficulties in breeding the FTWC strain compared with other A. aegypti strains. It is thus imaginable that populations carrying these mutant alleles will expand and spread to the whole Indochina Peninsula and from Southeast Asia to other tropical and subtropical regions of the world in the future. A case in point is the finding of this important allele with L982W + F1534C in an A. aegypti population that invaded and bred at an international airport of Japan (42). Fortunately, the population was not sustained in the temperate climate. The expansion of these concomitant mutant alleles may face less resistance in tropical and subtropical countries and would hamper the control of A. aegypti and the diseases it vectors. Particularly in Phnom Penh city, two alleles (V1016G + F1534C and L199F + L982W + F1534C) that confer hyper-resistance to pyrethroids are expected to account for most of the population, making it nearly impossible to control A. aegypti with pyrethroids in this region. This is also consistent with the fact that A. aegypti collected in Phnom Penh showed an extremely high level of resistance to many pyrethroid insecticides (24). The only previous report on the genotyping of Vgsc in the A. aegypti population from Cambodia showed that all 10 individuals collected at Battambang city, 280 km away from Phnom Penh (Fig. 4), had F1534C with wild-type V1016 all homozygously (35). Countrywide survey of the knockdown resistance gene in A. aegypti is required in Cambodia.

We have previously synthesized a Vgsc full-length cDNA with the triple mutations S989P + V1016G + F1534C, expressed in the Xenopus oocytes, and measured its sensitivity to pyrethroids by electrophysiological studies (20). We observed that Vgsc with these mutations was much less sensitive to both permethrin and deltamethrin than those with each single mutation. Soon after our report, A. aegypti having such triple-mutant Vgsc allele was reported from Myanmar, followed by various countries including Indonesia, Saudi Arabia, Sri Lanka, China, Thailand, Laos, and Malaysia (17, 19). In this study, we have established a strain (PGC) with triple amino acid substitutions and evaluated its susceptibility to pyrethroids in vivo for the first time. The PGC strain certainly exhibited a high level of resistance to permethrin than GY, PG, and PGG strains, all having V1016G but without F1534C, strongly supporting our electrophysiological studies. On the other hand, our previous studies exhibited that the triple-mutant Vgsc allele showed 90-fold insensitivity to deltamethrin while 1100-fold insensitivity to permethrin. This contrasted with the results that the PGC strain showed similar RRs to permethrin and deltamethrin in bioassays. Alternatively, multiple electrophysiological studies suggested that F1534C confers resistance to type I but not to type II pyrethroids (20, 28, 43), but in this study, the 1534C strain as well as FC66 and FC213 strains showed more than 10-fold resistance to type II deltamethrin under the pretreatment of PBO. The remaining RRs may be due to any unknown resistance mechanism(s), but this (F1534C may confer resistance to type II pyrethroids) is consistent with the results of other reported studies (25, 29). Furthermore, our previous electrophysiological studies claimed that S989P alone does not confer permethrin susceptibility of Vgsc but decreases deltamethrin sensitivity when it is combined with the V1016G (20). The RRs of GY (V1016G + D1703Y) and PG (S989P + V1016G) strains to deltamethrin (+PBO) were 22-fold (95% CI, 18.4 to 26.4) and 33-fold (95% CI, 26.9 to 41.3), respectively. Although the 95% CIs of these RRs do not overlap, no obvious differences were observed between the RRs of these two strains (Fig. 2D). The discrepancy between electrophysiological and in vivo studies was also mentioned in some other recent studies (44–46), and the reason for this has not been implicated though. Further study is needed to bridge the gaps between the results of electrophysiological studies and the actual levels of insecticide resistance.

PBO showed a strong synergistic effect on permethrin toxicity in the FTWC strain, suggesting that detoxification by cytochrome P450 monooxygenases is also an important resistance mechanism in addition to kdr. Moreover, it was found that mosquitoes can develop extremely high levels (>1000-fold) of resistance when they have multiple resistance factors. We have previously reported that SP strain originally collected from Singapore in 2009 and artificially selected by permethrin for 10 generations in the laboratory developed 1650-fold resistance (47). In this case, increased detoxification activity by cytochrome P450 monooxygenases was involved in the resistance besides the S989P + V1016G–type kdr (47). The RR of SP to permethrin under PBO treatment was 34-fold, which was similar to that of the PG (48-fold) in this study. Given that the synergistic effect of PBO on permethrin toxicity in the SP strain was 82, FTWC can potentially and theoretically develop >24,000-fold resistance (300 × 82), if this strain has the same detoxification mechanisms as SP. Studies on Anopheles gambiae Giles, 1902 and Anopheles funestus Giles, 1900, the major malaria vectors, have also reported that increased cuticle thickness reduces the penetration of insecticides and causes resistance (48, 49). Since similar resistance mechanisms are possibly involved in the pyrethroid resistance of A. aegypti, it will be necessary to focus on such other mechanisms as well in the future.

F1534C does not confer a very high level of resistance by itself, but it causes a much stronger level of resistance when it occurs together with other amino acid substitutions such as L982W and V1016G. The analysis of the full-length CDSs of Vgsc of two highly resistant strains FTWC and PGC established in this study can decipher that the multiple resistance substitutions were accumulated by recombination events rather than by sequential mutations (Fig. 3, A and B). In FTWC and PGC, the 3′ region of Vgsc CDSs has a polymorphism that causes F1534C, but both genes also lose some polymorphisms(s) that are characteristic to FTW and PG/PGG/GY, supporting the idea that these genes were produced by crossing over events of multiple genes (Fig. 3, A and B). It is strongly suggested that genetic recombination is an efficient evolutionary means for insects to develop a high level of insecticide resistance, and thus, it was the driving force behind the rapid development of resistance in A. aegypti. There is no denying the possibility that the super-resistant alleles of Vgsc found in this study will be further recombined with each other to produce even stronger resistance genes.

In conclusion, our studies have shown that concomitant L199F + A434T + L982W + F1534C and S989P + V1016G + F1534C of Vgsc confer an eminently higher level of pyrethroid resistance than the previously reported single kdr substitutions F1534C and V1016G. Further experiments to characterize each mutation is necessary. On the other hand, the high frequency of A. aegypti, having these concomitant mutations in a region of Cambodia, shows the stark fact that threats for controlling dengue and other arboviral infectious diseases are certainly looming. We emphasize the importance of strengthening the monitoring of these mutant alleles, especially in Southeast Asia, to take appropriate countermeasures before they spread globally.

MATERIALS AND METHODS

Mosquitoes

The larvae of A. aegypti were collected from fields of four countries listed in table S1. The larvae were reared in the laboratory as previously described (47). Species of mosquitoes were identified in the adult stage on the basis of the keys and then bred in the laboratory.

Bioassays

Simplified bioassays were conducted to evaluate the susceptibility of adult female A. aegypti to a representative pyrethroid insecticide, permethrin (91.2%, trans:cis = 6:4, Sumitomo Chemical Co. Ltd., Osaka, Japan). Pools of 20 respective adult mosquitoes (4 to 6 days old) were treated in four replicates with each permethrin dose (5.9 and 59 ng) by topical applications, and mortalities were assessed as described previously (25); 5.9 ng of permethrin is equivalent to LD₉₉ of a pyrethroid-susceptible strain of A. albopictus (25). After the establishment of 10 resistant strains based on Vgsc alleles, additional bioassays were conducted using permethrin and deltamethrin (99.4%, GL Sciences Inc., Tokyo, Japan) to compare the pyrethroid susceptibility among strains as previously described (25). LD₅₀s were calculated using Finney’s log-probit mortality regression analysis (50), implemented in R version 3.3.3 (www.r-project.org). RRs were calculated as the LD₅₀ of each strain divided by the LD₅₀ of the susceptible SMK strain. The 95% CIs of RR were determined by calculating the RRs for the minimum and maximum 95% CIs of LD₅₀ values (29). Two RR values were considered significantly different if the minimum and maximum RR values did not overlap (29). The synergism of PBO (98.0%, Wako Chemical Pure Industries Ltd., Osaka, Japan) was examined to evaluate the contribution of detoxification by cytochrome P450 monooxygenases (47).

Genotyping

Genomic DNA was extracted from the legs of mosquitoes as previously described (25) and the partial genomic DNA of Vgsc was sequenced individually. To examine the effect of known kdr substitutions, S989P, V1016G, and F1534C were genotyped for dead and survived mosquitoes from two highly resistant Hoa Kien (n = 78) and Hanoi 2 (n = 80) populations used for the bioassay. After sequencing the whole CDSs of Vgsc from established strains as described below, 13 Vgsc substitutions were targeted for the genotyping studies: L199F, L213F, V410L, A434T, L982W, S989P, A1007G, I1011M/V, V1016G/I, T1385I, T1520I, F1534C, and T1539A (Fig. 2A and figs. S6 to S12). A1007G was reported in the previous report (51). The condition of the PCR and the sequencing procedure were described previously (47). The PCR elongation time was differentiated according to the length of the PCR products (60 s/kbp). Primers used for PCR and sequencing are listed in the table S4.

Mosquito selections by Vgsc genotypes and establishment of resistant strains

Since genotyping studies of Vgsc showed that Hoa Kien and Hanoi 2 populations contained various Vgsc alleles, we attempted to establish several substrains with homozygous genotypes on all of the Vgsc polymorphic loci. Since the frequency of the V1016G substitution was very low in the two Vietnamese populations, we also tried isolating three additional strains from a population collected from Singapore in 2016 (Fig. 2B and fig. S1). In this study, we numbered the amino acid position according to the sequence of the most abundant splice variant of the house fly Vgsc (GenBank accession numbers AAB47604 and AAB47605). Four (FC66, FC213, FTW, and FWI) and three strains (1534C, PG, and PGG) were established from Hoa Kien and Hanoi 2 populations, respectively. Three strains (1534C, PG, and PGG) were established from the Singapore SP strain. According to preliminary genotyping studies, a very low frequency of L982W + F1534C was found in the Hanoi 2 population; thus, 144 males were individually genotyped for L982W, S989P, V1016G, and F1534C using genomic DNA from a single hind leg. The small numbers of males having the same Vgsc genotypes (L982W/F1534C: WW/FC, S989P/V1016G/F1534C: SP/VG/CC, and S989P/V1016G: SS/VG) were each mated with 100 virgin females of the Hanoi 2 population, and offspring were used for additional isolating steps (Fig. 2B). Eventually, three resistant strains (FTWC, PGC, and GY) were established (Fig. 2B). Bioassays for permethrin and deltamethrin were conducted as described above and previously (47) for 10 resistant strains to obtain LD₅₀s.

Sequencing the whole coding region of the Vgsc gene

Genomic DNA was extracted from eight individual female mosquitoes using the MagExtractor Genome kit (Toyobo Co. Ltd., Osaka, Japan) (52). Index library construction and targeted capture of pooled libraries were conducted as previously described (52). Illumina library construction and hybridization capture was conducted with the biotinylated oligo probe designed from the A. albopictus Vgsc gene, whose exons show >92.5% homologies to the exons in A. aegypti (52). The quantified library was sequenced using the Illumina MiniSeq with the Mid Output Kit (Illumina Inc., San Diego, CA) and 151 cycles for both ends. Read pairs of 114 to 349 kbp were sequenced for each sample. Raw fastq read data were deposited to the National Center for Biotechnology Information (NCBI) Sequence Read Archive (BioProject ID: PRJNA795523). NGS-based reads were mapped to the reference Vgsc genome sequence (AALF000723-RA) and annotated for the synonymous and nonsynonymous nucleotide polymorphisms by using the automated MoNas pipeline (https://github.com/ItokawaK/MoNaS) as previously described (52). Other detailed information about NGS analysis with target capture probes is described in our previous paper (52).

Evolutionary tree

The evolutionary history was inferred using the UPGMA method (53). The percentage of replicate trees in which the associated taxa clustered together in the bootstrap test (1000 replicates) is shown (54). The evolutionary distances were computed using the Kimura two-parameter method (55) and are in the units of the number of base substitutions per site. The rate variation among sites was modeled with a gamma distribution (shape parameter = 1). This analysis involved 17 nucleotide sequences (table S5). All ambiguous positions were removed for each sequence pair (pairwise deletion option). There were 6709 positions in the final dataset. Evolutionary analyses were conducted in MEGA X (56). A phylogenic tree was constructed by GENETYX software (Ver 13, GENETYX Corp., Tokyo, Japan).

Generation of homology model and ligand docking

A homology model of A. aegypti Vgsc (Vectorbase accession: AAEL023266) was generated on the basis of the electron microscopy crystal structure of the voltage-gated sodium channel NavPaS from the American cockroach P. americana [Protein Data Bank (PDB) accession number: 6A90, DOI:10.2210/pdb6A90/pdb] (57). The model was produced using the Modeller Ver10.1 program (58). Automated docking of 1R-trans-permethrin (PDB accession number: 40159) with the model of the sodium channel was performed using the AutoDock 4.2.6 software package (59). Docking simulations were performed for 100 runs for each trial to generate histograms of the binding energies of each run. The root mean square deviation tolerance was set at 2 Å in the simulations.

Statistical analysis

The association of kdr mutations and pyrethroid resistance was analyzed via Bayesian logistic analysis using Cmdstan Ver. 2.28.2 in R Ver. 4.1.2 (https://mc-stan.org). The model formula was yi ~ Be[logit(Pi)] and Pi = a + b1x1i + b2x2i + b3x3i + b4x4i, where Be represents Bernoulli distribution, i represents an individual number, yi represents alive(1) or dead(0), x1 to x4 represent input[x1 = Hoa Kien(0) or Hanoi(1), x2 = L982(0) or W982(1), x2 = V1016(0) or G1016(1), x3 = F1534(0) or C1534(1)], a represents intercept, and b1, b2, b3, and b4 represent parameters (the ranges were defined as 0 to 1).

Supplementary Materials

This PDF file includes:

Figs. S1 to S12

Tables S1 to S5

View/request a protocol for this paper from Bio-protocol.