In the new world, the Culex pipiens complex of mosquitoes consists of Cx. pipiens L., Cx. quinquefasciatus Say, their hybrids, and an autogenous form of Cx. pipiens, Cx. pipiens form molestus Forskål. All taxa are considered efficient vectors of arboviruses such as West Nile virus [1] and St. Louis encephalitis virus [2] in North America. In terms of factors affecting their distribution in the U.S., the most important difference between Cx. pipiens and Cx. quinquefasciatus is the inability of Cx. quinquefasciatus to diapause, which limits its northward spread. The hybrid zone in the U.S. is nevertheless extensive, indicating that a high degree of gene flow has occurred since these species were introduced [3], [4]. For a discussion of theories regarding the introduction of Cx. pipiens complex mosquitoes to the New World, see [5].

In this study, we use a combination of microsatellite markers that we recently developed [6] and previously published [7]–[9] microsatellite markers to explore the genetic relationships of a set of populations across a relatively small area, the state of Colorado, USA. Colorado (CO) is bisected by the Rocky Mountains, which reach heights of over 4,000 m and therefore should influence patterns of gene flow among populations of flying insects. The dry eastern plains of the state represent expanses of low-quality habitat that make rivers important conduits for dispersal and colonization by Cx. pipiens complex mosquitoes. Also influencing their distribution is a preference for larval sites with organically-enriched stagnant water, which makes these organisms well-adapted to coexist with humans in urban and suburban habitats [10]. Within Colorado, low winter temperatures and high elevations favor Cx. pipiens, or specimens that diapause in winter, and pure Cx. quinquefasciatus specimens have not been collected in the state. However, ephemeral summertime introductions of Cx. quinquefasciatus along transportation corridors likely occur and may result in limited hybridization.

Population genetic studies of disease vectors are useful for their ability to detect and quantify genetic structure in a system. The presence of genetic structure indicates that some barriers to panmixia exist, which can lead to adaptation to local habitats. Such adaptation may play a role in the observed variation in vector competence among populations of Culex mosquitoes. For example, Reisen et al. [11] speculated that variation in Cx. tarsalis vector competence could be attributable to genetic factors, because other sources of variation had been accounted for. In addition, Kilpatrick et al. [12] found differences in the genetic background of individual Cx. pipiens complex mosquitoes that became infected with WNV versus those that did not. Richards et al. [13] asserted that environmental and biological (including genetic) causes of within- and among-population variability in Cx. quinquefasciatus vector competence have not been adequately explored and thus could account for such variability. With regard to other mosquito species, Lambrechts et al. [14] determined that vector competence for dengue viruses in Aedes aegypti was likely a function of interactions on a genetic level between vector and virus. Bataille et al. [15] argued that habitat differences drive local adaptation in Ae. taeniorhynchus, resulting in genetic divergence and site-mediated disease transmission dynamics.

Isolation by distance (IBD) is the tendency for populations to diverge genetically with linear distance. Previous work by our lab on Cx. pipiens complex mosquitoes in the central U.S. found a significant IBD effect among the 14 populations sampled, which were separated by an average distance of 112 km [4]. That study, as well as several others in different parts of the U.S. [16]–[18], [5], was on a scale encompassing several states, although work on a similar geographic scale to this study has been done in the state of California [19]. The finding of significant genetic structure at a broad scale suggests that finer scale structure (for example, on the scale of a single state) could elucidate genetic differences among groups thought to be more or less genetically homogenous across populations.

The main objective of this study was to quantify the genetic diversity and differentiation among Cx. pipiens populations in Colorado. In addition, this study sought to determine whether the new set of markers, used in combination with previously-published markers, would allow finer scale resolution of genetic differences among populations of closely-related individuals.

Our results indicate that genetic diversity is fairly consistent across populations, with no sites or groups of sites showing strikingly larger or smaller values. This is perhaps due in part to the scale at which populations were sampled, which is smaller than similar studies of Cx. pipiens complex populations in the U.S. The highest H_E values came from the two easternmost populations in the study (Lamar and Sterling), which are the two lowest-elevation sites and located on the eastern plains. This observation suggests there may be environmental factors that reduce genetic diversity of Cx. pipiens complex populations. For example, lower elevation sites may experience comparatively fewer climatic extremes, which could allow for a greater variety of successful genotypes in those areas. Conversely, it could be that weather conditions at the higher elevation sites reduce a greater percentage of diapausing populations compared to lower elevation sites, resulting in populations with moderately lower genetic diversity.

It is interesting to note that all of the differences between observed and expected heterozygosity within populations in this study were such that H_O < H_E, indicating an excess of homozygotes. In several populations, including Lamar, Pueblo, Denver, Evans, Fort Collins and Platteville, several of the differences were statistically significant and resulted in departures from HWE. One possible explanation for the observed homozygote excess is the Wahlund Effect [37], whereby differences in allele frequencies due to drift among structured subpopulations (here referred to as populations) result in an overall deficit of heterozygotes. When subpopulations are considered as a population, divergent allele frequencies cause fewer heterozygotes overall than would be expected.

The detection of Cx. quinquefasciatus alleles at the Cxpq78 locus in some of the study populations suggests that Colorado is a site of limited gene flow between the two species in the Cx. pipiens complex. The presence of Cx. quinquefasciatus alleles at low frequencies in populations close to the I-25 corridor offers a possible mechanism for their movement, which may occur by anthropogenic means. A similar scenario was proposed for the movement of Aedes aegyptii in the state of Arizona, USA, because local conditions made observed dispersal patterns unlikely by natural means [38]. The population with the highest frequency of Cx. quinquefasciatus alleles, Pueblo, has some of the warmest temperatures in the state [39]. Notably, there were no Cx. quinquefasciatus alleles detected in any of the populations west of the Rocky Mountains. This may be due to high elevation areas and dry conditions of western Colorado, which would impede dispersal through those areas. In addition, the road system on the west side of the mountains is such that there is no major north-south highway through the area, reducing chances of human transport.

Our study has shown that Cx. pipiens populations in Colorado are not panmictic. For example, most of the pairwise F_ST comparisons were significant (Table 4), except for populations located in close proximity to each other. Also, there was a significant IBD effect, indicating that genetic differentiation increases with distance. The pattern of data points shown in Figure 3 is likely related to our sampling methods, because the two groups of points in the lower part of the plot are from pairs of populations located comparatively close to each other. Data points from pairs of populations that are further apart generally share significant pairwise F_ST values and are located in the upper right portion of the plot. The AMOVA analyses showed that moderate but statistically significant amounts of genetic variation can be attributed to grouping populations either by their position relative to the mountains or by river basin. These findings are consistent with the results of the analysis of genetic distances that produced the neighbor joining tree (Figure 2). The tree shows particularly good support for grouping populations on either side of the mountains, as well as the groups of populations in north-central and southeast Colorado, and is generally in agreement with the geographic locations of populations in this study. Sterling was an exception and was located between the western sites and other eastern sites, perhaps due to it experiencing more gene flow from migration than other eastern populations, as suggested by the Structure results.

Although many of the larger differences among sites are between populations on either side of the mountains, comparisons with the Lamar population produced the largest pairwise F_ST values. Taken with its higher genetic diversity, low elevation and because it occupies its own cluster in Structure, Lamar appears different from the other CO populations. Whereas Denver, which also occupied its own cluster in Structure and had relatively high genetic diversity, receives a high volume of vehicle traffic and represents a large area of suitable habitat, Lamar is smaller and comparatively remote. These features could account for Lamar’s genetic characteristics, despite its location near a potential dispersal corridor in the Arkansas River. In order to determine whether the genetic differentiation observed in Lamar disproportionately influenced the results of the AMOVAs, analyses were performed excluding this population. The fixation indices were still statistically significant, but their values were smaller, indicating that while Lamar is differentiated from the rest of the populations, it does not account for all of the structure in this system. The neighbor joining tree is consistent with this finding, in that Lamar is located on one end of the tree, but does not have a particularly long branch length.

Despite evidence of statistically significant amounts of genetic differentiation, our results indicate some amount of gene flow among all collections. The mechanism for facilitating the movement of Cx. pipiens mosquitoes around the significant physiographic barriers present in the state is unknown at this time, although other studies have suggested anthropogenic or weather-induced movement, as well as dispersal routes around the mountains [40], [38], [41]–[43]. The frequency of significant pairwise F_ST comparisons in the current study is in contrast to findings of a study of the congener Cx. tarsalis, which did not find significant differentiation among populations sampled from a smaller, five county area in north-central Colorado [41]. The difference could be due to life history characteristics of the two species, as Cx. tarsalis is thought to be routinely capable of dispersing long distances, and thrives in habitats such as those provided by agricultural irrigation.

The Structure results indicated that K = 5, and were obtained with the parameters set such that the program used only genetic information, not each individual’s population of origin. This finding is in contrast with other studies of the Cx. pipiens complex across larger areas [16]–[18], [4], [44] that discerned taxon-level, but not population-level genetic structure. We propose that our findings are the result of the additional microsatellite markers we used, which have allowed finer-scale resolution of genetic relationships among this group of closely related organisms.

Recent studies on Culex mosquitoes in Colorado deployed CDC light traps and focused on the congener Cx. tarsalis, which is also an efficient vector of West Nile virus in the western U.S. [45], [46], [41], [47]. Among findings of the above studies was a positive association between river corridors and the transmission of West Nile virus, and a greater abundance of Culex mosquitoes at lower elevations (1200–1600 m). Epidemiological work conducted on WNV in Colorado has included mapping the locations of disease cases [48]. For example, Figure S1 shows the occurrence of WNV neuroinvasive disease cases by county in Colorado from 2003–2010. When this map is examined with Figure 1, it is apparent that more cases of WNV occur in the lower elevations than in the mountains [41], [48]. For example, the area around Grand Junction is lower in elevation than surrounding counties. In contrast, areas of high elevation, corresponding to the gray shading in Figure 1, have comparatively fewer cases. Although many of these cases are assumed to be infected by Cx. tarsalis, most of the shaded counties in Figure S1 (e.g. Larimer, Boulder, Mesa, Pueblo) have urban areas that possess suitable habitat for Cx. pipiens complex mosquitoes [49], [50], and we collected about equal numbers of both species at our collection sites. Future disease modeling efforts that include data on both primary vectors, Cx. tarsalis and Cx. pipiens, will prove more realistic.

Since dispersal and demographic changes are difficult to track in small organisms such as mosquito vectors, population genetics analyses can be employed to inform researchers and disease modelers about these epidemiologically important characteristics. In our study, we performed such analyses for Cx. pipiens populations in Colorado, and suggest our work as an example of how genetic data can complement epidemiological data in characterizing a disease vector. Future studies that incorporate population genetics, and epidemiological data on human disease will lead to comprehensive models of disease transmission.