Dengue is the most important and widespread mosquito-borne viral infection of humans in the world [1]. An estimated 390 million cases of dengue virus (DENV) infection occur per year throughout the tropical and subtropical world. It is estimated that up to 55 % of the world’s population is at risk of infection in 128 countries, where 824 million people live in urban environments [1]. In the last 20 years, dengue epidemics have increased in both number and magnitude, due to a range expansion of the Aedes aegypti mosquito, the primary vector of dengue viruses, as well as increased trends in urbanization and global travel and weakened public health infrastructure [2, 3].

In the absence of effective therapeutic medications and vaccines, Ae. aegypti vector control is presently the only approach for preventing and controlling dengue virus transmission [4, 5]. Vector control strategies rely heavily on the application of chemical insecticides which target immature mosquitoes in their development sites and adult mosquitoes which are often targeted through the use of ultra-low-volume (ULV) and indoor space spraying [5, 6]. Due to the daytime biting behavior of female Ae.aegypti, the use of bednets as a protective barrier is not generally recommended, although the presence of insecticide treated bednets in houses has been associated with lowered dengue vector infestation indices [7]. Other insecticide-treated materials such as window and door curtains and screens and water storage container covers have also been shown to be effective at reducing household-level Ae. aegypti infestations [8–15], but the impact of such entomological reductions on the risk of dengue virus transmission remains to be determined [16].

As the importance of dengue as a public health problem has increased globally, insecticide-based vector control interventions have been widely employed. The heavy reliance on chemical insecticides to control dengue vectors has led to the development of resistance to many of the insecticides most commonly applied. In the Americas alone, resistance to multiple classes of insecticides has been widely reported in Ae. aegypti [17–23]. This is cause for alarm to those involved in dengue prevention and control, as the growing prevalence of resistance could threaten vector control efficacy [21, 23–25].

Resistance to insecticides can be caused by a variety of physiological changes within a mosquito, including structural alterations at the target site of the insecticide, increased activity of enzymes associated with insecticide detoxification, and changes to the mosquito cuticle. Non-synonymous mutations in the voltage-gated sodium channel transmembrane protein (para) can result in ‘knockdown resistance’ (kdr), in which the binding of DDT and pyrethroid insecticides are reduced at this target site [26–29]. There are several point mutations known to confer kdr-type insecticide resistance in Ae. aegypti. In the Americas, substitutions at codons 1016 (V1016I) and 1534 (F1534C) on domains II and III of the voltage gated sodium channel are strongly associated with DDT and pyrethroid resistance in Ae. aegypti [27, 30, 31].

Pyrethroids have become the most frequently used public health insecticides globally due to their low cost and low toxicity to mammals, in addition to their high residual power [32]. It is of considerable concern when kdr is found in wild populations of vector mosquitoes, given there are few suitable alternatives to pyrethroid insecticides approved for public health use. Previous research has demonstrated that kdr mutations arose and spread rapidly in Mexican Ae. aegypti populations. Between 2007 and 2009, Siller et al. [33] reported an increase in the frequency of the Ile1016 allele in several localities within the state of Veracruz, Mexico. Similar findings by Ponce Garcia et al. [34] showed over the course of 14 years, several Mexican states, including Veracruz and Yucatan, experienced a significant increase in the frequency of the Ile1016 allele. In the city of Merida, in Yucatan State, Ile1016 had not been detected in 1999, yet by 2007, it was occurring at frequencies as high as 54 %. A more recent study in the state of Guerrero, in southern Mexico, found Ile1016 at a frequency of 80 %,with all mosquitoes also containing the Cys1534 allele [35]. The F1534C substitution was initially reported in Ae. aegypti mosquitoes in Asia and first detected in North America in the Cayman Islands [30]. The detection of the Cys1534 allele in Guerrero was the first report of that particular kdr mutation in Mexico [35].

Dengue outbreaks have been reported in Mexico since 1979, and all four DENV serotypes have been reported in Mexico [36]. It is well documented that as viral serotypes are displaced by the introduction of new serotypes, the incidence of severe dengue increases [37, 38]. Yucatan State experienced a similar trend with the incidence of severe dengue increasing each time a new serotype was introduced [12]. The city of Merida is the capital and main urban center of Yucatan State, and dengue transmission in Merida is hyper-endemic. In recent decades, co-circulation of multiple serotypes and a high abundance of Ae.aegypti has increased the risk of severe dengue for the population living in and around the city center [12].

From 1998 to 2009, pyrethroids were the primary insecticides used for adult mosquito control in Mexico, and deltamethrin was the primary insecticide used for indoor space spraying when a dengue case was reported. In 2010, reports of high levels of pyrethroid resistance and a high frequency of the Ile1016 allele in Mexico [34, 35] prompted vector control authorities to modify their strategy. From 2010 up until when collections for this study were conducted in 2013, carbamates (bendiocarb, and later propoxur) had been predominantly used for indoor space spraying, although the pyrethroids bifenthrin and sumithrin continued to be used at a reduced level.

While the intensity of insecticide use for dengue vector control in the city of Merida is relatively consistent, the surrounding communities have experienced more sporadic insecticide applications. Little entomological data, including resistance data, are available from these surrounding towns. Previous research [34] has suggested a high degree of heterogeneity in insecticide resistance frequency across the Yucatan Peninsula, yet little is known about these patterns of resistance at a finer spatial scale. Spatial heterogeneity of insecticide resistance could have important implications for vector control efficacy, particularly when vector control strategies are designed to be applied across a large geographical area. The development of scientifically sound vector control and insecticide resistance management strategies for Ae. aegypti depends on understanding the patterns and drivers of spatial heterogeneity in insecticide resistance. In the present study, we characterized patterns of insecticide resistance in Ae. aegypti populations within five towns surrounding the city of Merida, Mexico, to gain a greater understanding of how resistance may be distributed at a fine geographical scale.

Our results suggest variability in both resistant phenotypes and genotypes between towns at a small geographical scale, illustrating the complex and focal nature of insecticide resistance. Our results also confirmed the association between two kdr mutations and phenotypic resistance to deltamethrin [30], and detected significant differences in kdr allele frequency at multiple spatial scales.

Low mortality was observed in all populations tested for susceptibility to chlorpyrifos-ethyl, demonstrating high levels of resistance to this organophosphate insecticide across the study area. Although all populations were resistant to chlorpyrifos, a high degree of variability was evident, with Acanceh and San Lorenzo showing the greatest resistance (<20 % mortality) and the remaining towns showing between 40 and 56 % mortality. Interestingly, susceptibility to the carbamate insecticide bendiocarb was high in all communities except Progreso, where mortality was 95 % (suggesting incipient resistance). That such a high level of resistance to an organophosphate didn’t hold true with respect to a carbamate suggests that the mechanisms underlying the resistance are unlikely to be due to ace-1 target site insensitivity, as this would normally result in cross resistance between both organophosphate and carbamate insecticides [48], nor have ace-1 mutations been widely reported in Ae. aegypti.

Mortality data from the deltamethrin bioassays also showed variation across the study sites. While all populations were categorized as resistant, the most resistant populations, San Lorenzo and Progreso, had mortalities of 63 and 73 %, respectively, while the remaining towns all had mortalities between 87 and 88 %. Significant associations were detected between deltamethrin resistance phenotype and the presence of both kdr alleles (Ile1016 and Cys1534). Phenotypic resistance was highly predictive of the presence of both alleles, however, not all mosquitoes containing a mutant allele were phenotypically resistant. This is not entirely unexpected, as both Ile1016 and Cys1534 are largely recessive alleles [27, 30]; indeed, our data supported this as over 90 % of heterozygotes for either allele were phenotypically susceptible to deltamethrin.

The Ile1016 allele is known to be associated with resistance to type I and II pyrethroids (including deltamethrin), as well as DDT [27], while the Cys1534 allele is thought to be associated most strongly with resistance to type I pyrethroids such as permethrin, as well as DDT [30]. However, recent evidence suggests that deltamethrin exposure selects for Cys1534 more rapidly than for Ile1016 [49]. Given that not all mosquitoes that were resistant to deltamethrin in the bioassays were homozygous for the kdr alleles suggests that multiple resistance mechanisms are contributing to the resistant phenotypes, most likely including metabolic mechanisms arising from the overproduction of detoxifying enzymes [50]. The role of the S989P kdr mutation is also of potential interest, as it has recently been associated with deltamethrin resistant phenotypes in Asian populations of Ae. aegypti [51, 52]. Investigation into these additional mechanisms and the role detoxifying enzymes may have on resistance to other insecticides in these populations will help to further explain the resistance patterns observed.

The population-level frequencies of both alleles were estimated from the indoor resting adult Ae. aegypti females collected from houses. Genotypic differentiation was significant between the five communities and was further detected between blocks within the same town in San Lorenzo for both kdr alleles and in Acanceh and Conkal for the Ile1016 allele. These results indicate that even at a fine geographical scale, kdr frequencies can differ significantly. This is particularly interesting in light of population genetics studies that have shown that Ae. aegypti populations in the Yucatan experience free gene flow within 180 km [53]. This suggests that the observed differences in kdr genotype are being driven by fine-scale pressures, such as focal insecticide exposure, as was hypothesized in a recent population genetics study by Saavedra et al. [47]. This could suggest an important role in the household use of insecticides in maintaining pyrethroid pressure, even when vector control programs have switched chemical groups. Indeed, the results from this study suggest that 3 years after the vector control program ceased deltamethrin use, deltamethrin resistance and a high frequency of kdr alleles persisted in the Ae. aegypti population.

Some evidence suggests that pyrethroid resistance has an associated fitness cost resulting in reduced larval development and adult longevity [54], although conflicting data suggest that the Cys1534 allele does not confer fitness costs in Ae. aegypti [55]. Evidence of duplication of the sodium channel gene has recently been reported in Ae. aegypti [56], and such an event could maintain both wild-type and mutant alleles in the population, simultaneously conferring resistance and reducing associated fitness costs. Although we currently do not know if this duplication has occurred in Ae. aegypti in Mexico, this warrants further investigation, particularly when discerning why such a high frequency of pyrethroid resistance has been maintained despite a near absence of pyrethroid pressure from vector control authorities during the period leading up to our entomological collections.

While this study presents novel findings regarding the variability of insecticide resistance in space, the interpretation of these results is limited by the cross sectional nature of the study and the small number of insecticides and resistance mechanisms investigated. It will be important to further investigate the patterns and drivers of variability in these communities to assess how insecticide use and other factors can drive resistance patterns and explain fluctuations in both space and time. The observed differences in the frequency of kdr alleles in these populations may have occurred due to pressure from widespread applications of insecticides. A recent study conducted in Martinique found that genetic patterns in Ae. aegypti were driven by insecticide pressure [57]. An alternative explanation is that the alleles were introduced through the immigration of mosquitoes from elsewhere, via containers harboring eggs or adult mosquitoes through other forms of transportation [58–60]. Understanding how these factors contribute to the establishment and maintenance of resistance is key to developing effective resistance management and prevention strategies.

The variations in both resistance phenotypes and genotypes detected in this study have practical implications for vector control operations. The observation of variability between blocks within the same community suggests that blanket vector control strategies at the provincial or even municipal levels may yield mixed results at best, and further highlights the operational challenge of managing resistance in vector populations with varying resistance profiles at a fine spatial scale. Fluctuations in resistance over time will likely further complicate the landscape, highlighting the need for systematic and routine insecticide resistance surveillance.

The limited number of suitable insecticides for vector control and the logistical constraints inherent in vector control programs make the small-scale tailoring of vector control interventions particularly challenging, but this is a challenge that must be considered given that insecticide resistance can vary over a small scale. The use of multiple active ingredients through rotations or mixtures has been suggested as an effective way to mitigate or stall the emergence of resistance [61–63], at least in the short term. The findings presented here suggest that large-scale blanket vector control strategies may have limited efficacy given that vector populations can vary widely in their susceptibility to insecticides over a small geographical scale. The development of evidence-based resistance management strategies with respect to dengue vectors should be a priority as dengue continues to increase as a global public health problem.