Range expansion of Culex quinquefasciatus and Culex pipiens hybrids across mid-latitudes of North America

Abstract

We investigated recent reports of a northward expansion of Culex quinquefasciatus and hybrids of the Culex pipiens species complex, important vectors of West Nile virus in mid-latitudes of North America. Because Cx. quinquefasciatus more readily feeds on both birds and mammals, its movement into higher latitudes may increase the risk of WNV spillover from avian reservoirs to humans. Using an Ace2 PCR assay, we identified species and detected hybridization in mosquito specimens from 26 sites across the continental U.S. Our results reveal a strong latitudinal gradient in hybrid index values, consistent with climatic filtering of overwintering traits such as diapause. We detected both northward expansion of Cx. quinquefasciatus alleles and southward introgression of Cx. pipiens, with admixture occurring beyond previously defined hybrid zone boundaries. Hybrid zone structure varied regionally: the East Coast exhibited sharp latitudinal structuring of hybridization patterns; the Central U.S. showed broader corridors of admixture; and the Mountain/Southwest and West Coast zones of secondary contact displayed patchy Cx. quinquefasciatus distributions consistent with a mosaic hybrid zone. These patterns suggest incomplete reproductive isolation, with limits to interbreeding likely shaped by ecological barriers, such as winter survival constraints, and region-specific colonization histories. As climate change relaxes overwintering barriers and urbanization alters host and habitat availability, this hybrid zone may become increasingly dynamic and spatially complex. By updating the distribution of Cx. quinquefasciatus and hybrids, this study provides critical data for tracking range shifts, improving vector surveillance, and refining our understanding of WNV risk. More broadly, it advances integrated approaches to public health by linking mosquito ecology and evolution to emerging disease risk in both human and wildlife populations.

1. Introduction

The Culex pipiens species complex includes several globally important mosquito vectors of zoonotic diseases, including Cx. pipiens, Cx. quinquefasciatus, Cx. australicus, Cx. globocoxitus, Cx. torrentium, and the urban-adapted form Cx. pipiens molestus. In North America, Cx. pipiens, Cx. quinquefasciatus, and Cx. pipiens molestus are especially important for West Nile virus (WNV) transmission. Cx. pipiens and Cx. quinquefasciatus were likely introduced independently to North America via human-aided dispersal during the colonial period, with Cx. pipiens arriving from Europe and Cx. quinquefasciatus from Africa or the American tropics [[1], [2], [3], [4], [5], [6], [7]]. Cx. pipiens is primarily found at higher latitudes, while Cx. quinquefasciatus occupies southern regions, with a hybrid zone between approximately 33° and 36°N latitude [1,[8], [9], [10], [11], [12]] that represents secondary contact between evolutionarily distinct lineages [[1], [2], [3]].

Where these species meet, they form a broad hybrid zone that provides a natural system for investigating the genetic and ecological drivers of range expansion and the consequences of hybridization [[2], [3], [4]]. Although broad patterns of distribution and ecological differentiation appear generally stable [8,10,[13], [14], [15], [16]], there are indications of zone instability, and the impacts on WNV transmission are not straightforward. While introgression may facilitate the transfer of adaptive traits—such as insecticide resistance, altered feeding behavior, or diapause—the effects of trait recombination are complex and likely context dependent. For instance, Cx. quinquefasciatus acts as a bridge vector by feeding on both avian and mammalian hosts [2,11,17], and if mammal-feeding traits from Cx. quinquefasciatus introgress into diapausing genotypes, this could increase the risk of virus spillover in temperate regions. On the other hand, introgression of diapause traits could support vertical transmission of WNV at higher latitudes [18], potentially intensifying early-season transmission pressure on susceptible bird species [19]. Indeed, WNV has been linked to persistent population declines in several North American birds [20,21], underscoring the broader significance of hybrid zone dynamics under a One Health framework. Nonetheless, the consequences of hybridization remain highly speculative, and resolving hybrid boundaries and tracing trait movement across the hybrid zone remains a significant challenge.

There is growing evidence that the Cx. pipiens–Cx. quinquefasciatus hybrid zone is undergoing range shifts resulting in a northward expansion of Cx. quinquefasciatus alleles and increasingly diffuse zones of secondary contact [10]. Specific drivers, limits, and consequences of these expansions remain poorly understood. One reason for this gap is that identifying members of the Cx. pipiens complex is challenging due to morphological similarity, particularly in regions of sympatry. A range of methods has been used to differentiate these taxa, from morphometrics and allozyme electrophoresis to microsatellites and sequence-based assays targeting cytochrome c oxidase subunit I (CO1) and acetylcholinesterase 2 (Ace2). More recent tools such as genome-wide markers [[22], [23], [24], [25], [26]] and AI-based computer vision tools (e.g., the IDX platform developed by Vectech Inc. Baltimore, MD), offer high resolution but require large budgets and are still in the development phase. In this study, we selected the Ace2 PCR assay as a cost-effective, validated method for distinguishing species and hybrids at scale, with the bonus utility of simultaneously acting as training data for our sister projects aimed at detecting context-dependent fine-scale patterns of introgression using whole genome sequencing, and refining AI-based identification tools for morphologically similar taxa with IDX systems (Vectech Inc.) [[27], [28], [29]].

This study aims to update the geographic distribution of Cx. quinquefasciatus and Cx. pipiens hybrids in North America to improve understanding of hybrid zone dynamics and recent range shifts. By documenting current hybrid patterns, this study lays the groundwork for more precise assessments of WNV enzootic transmission dynamics and provides foundational data for developing new diagnostic tools that will enhance vector surveillance and would be capable of tracking ecologically and epidemiologically important traits under changing environmental conditions.

2. Material and methods

2.1. Mosquito collection and identification

Adult mosquitoes of Cx. pipiens, Cx. quinquefasciatus and their hybrids were collected through collaborations with mosquito abatement agencies from 26 localities from a range of latitudes and habitats across North America in 2023–2024, with one additional sample collected for a different project in 2018 from Salt Lake City, Utah. Adult mosquitoes were collected in CO2 traps following local protocols and stored frozen until sorting under the microscope, at which point they were identified as Cx. pipiens s.l. species complex using morphological keys [30], then placed in falcon tubes with silica gel beads and shipped to VecTech Inc. for imaging with the IDX system (Vectech Inc. Baltimore, MD) as part of a broader initiative to develop the use of sensitive convolutional neural networks (CNNs) for distinguishing morphologically similar taxa [[27], [28], [29],31]. An average of 20 mosquitoes were chosen per locality for imaging, and then shipped singly on ice to The Connecticut Agricultural Experiment Station (Gloria-Soria lab) and Utah State University (Saarman lab) for DNA extraction and PCR assay, respectively. We performed Ace2 PCR assays on additional Utah samples not available for imaging from the Ute Tribal Lands, Moab, and Salt Lake City (2018 only).

2.2. Species-specific PCR assay

Mosquitoes were classified using the Ace2 PCR assay [9], which distinguishes Cx. pipiens, Cx. quinquefasciatus, and heterozygous individuals. DNA was extracted using DNeasy Blood and Tissue kits (Qiagen, Valencia, CA) and stored at −20 °C until PCR. In order to score polymorphisms in the second intron of the Ace2 gene that are diagnostic of members of the Cx. pipiens complex as well as hybrids [9], we completed PCR reactions using species-specific forward primers ACEpip and ACEquin with the universal reverse primer B1246s detailed in Table 2 of Smith and Fonseca, 2004 [9].

Each reaction contained 1 uL of extracted DNA, 0.4 uL of primer ACEpip, 0.8 uL each of primers ACEquin and B1246s, and 10 uL 2× MyTaq™ Red Mix (Meridian Life Science, Memphis, TN) in a total volume of 20 uL. Samples were heated to 95 °C for 5 m, followed by 35 cycles of 95 °C for 30 s, 55 °C for 30 s, and 72 °C for 60 s, with a final extension of 72 °C for 5 m. DNA fragments were visualized on 1.5 % agarose gels pre-stained with SYBR™ Safe DNA Gel Stain (Thermo Fisher Scientific, USA) in 1× TAE buffer at 130 V for 30 m.

2.3. Hybrid index estimates and tests for Hardy-Weinberg equilibrium (HWE)

For each sampling site, we calculated a hybrid index (H-index) as (2p + h) / (2p + 2 h + 2q), where p, h, and q represent the number of individuals identified as Cx. pipiens, hybrids, and Cx. quinquefasciatus, respectively. This index ranges from 0 (all Cx. quinquefasciatus) to 1 (all Cx. pipiens), with intermediate values reflecting degree of admixture (the mixing of previously separated populations). Hybrid indices are widely used to summarize ancestry proportions in mosquitoes and related systems [3,9,32]. In this study the H-index is derived from the diagnostic Ace2 locus and therefore reflects ancestry at a single marker rather than genome-wide admixture. We interpret H-index variation as a site-level summary of relative ancestry suitable for detecting broad spatial patterns and ecological correlates, while recognizing that it does not capture the full genomic complexity of hybridization. To allow comparability with earlier work predating diagnostic PCR assays, we also compared Ace2-based H-indices to hybrid indices derived from diagnostic visible marker (DVM) ratios reported in prior studies [8,12,18,32].

Deviations from Hardy-Weinberg equilibrium (HWE) were assessed at the Ace2 locus, with genotype counts for each sampling site. Sites with only a single genotype class were excluded from analysis. For each remaining site, allele frequencies were calculated and used to estimate expected genotype frequencies under HWE. Observed and expected counts were compared using Fisher's exact tests, implemented in R (v4.4.0) with simulated p-values (100,000 replicates) to account for small sample sizes and low expected counts.

In addition to site-level tests, pooled HWE tests were conducted within geographic zones (East Coast, Central, Mountain/Southwest, and West Coast) to evaluate broader-scale patterns by summing genotype counts across sites. Because HWE assumes random mating within the unit of analysis, no selection, and no substructure, pooled tests are sensitive to Wahlund effects arising from spatial or temporal structure and heterogeneous collection methods [[32], [33], [34]]. We therefore interpret zone-level deviations primarily as evidence of population structure and non-random mixing at regional scales, rather than as site-specific departures from equilibrium.

2.4. Data compilation and hybrid classification across studies

To enable historical comparisons, we incorporated previously published datasets that varied in their methods for classifying individuals within the Cx. pipiens species complex. These approaches included the morphometric ratio of male genitalia known as DV/D ratios alone [8,32], DV/D ratios combined with allozyme data [12], Ace2 PCR combined with microsatellite genotyping [10,11], Ace2 PCR combined with DV/D ratios [3], and Ace2 PCR alone [18,[35], [36], [37]]. A summary of classification methods is provided in Table S1. The complete dataset, including species and hybrid counts, sampling locations, hybrid index calculations, and explanatory footnotes, is available in Table S2. Where relevant, footnotes indicate exclusion criteria, laboratory-only confirmations, and potential cases of misidentification. For consistency across studies, we retained the original species and hybrid/intermediates classifications and calculated standardized hybrid index (H-index) values based on reported counts with the same equation as used for the Ace2 genotypes.

2.5. Geographic mapping and predicted hybrid index based on habitat suitability

Occurrence data for Cx. pipiens and Cx. quinquefasciatus were compiled from new records from this study, previously published studies [3,8,[10], [11], [12],18,32,[35], [36], [37]], and public repositories such as VectorBase [38] and VectorMap (http://vectormap.si.edu; collected 8 June 2024), following Gorris et al. [39]. Hybrid and parental species occurrence were based on the classification system described above, with details available in Table S1, Table S2 in the supplementary material.

Species distribution models (SDMs) generated via MaxEnt [40,41] from Gorris et al. [39] were used to predict habitat suitability and identify potential hybrid zones and expected levels of hybridization per sampling site. SDMs were thresholded at a suitability value of 0.5, and overlap areas were defined and merged using a buffering approach to account for dispersal and spatial uncertainty. For each sampling site, we quantified local habitat context by buffering the site by 30 km and categorizing the area into Cx. pipiens-only (A_pip), Cx. quinquefasciatus-only (A_qui), and predicted overlap (A_overlap) zones. We then calculated a predicted hybrid index (H-index_pred) for each site as: H-index_pred = (A_pip + 0.5 × A_overlap)/(A_pip + A_qui + A_overlap). This formula assumes Cx. pipiens-only areas contribute fully toward Cx. pipiens ancestry (H-index_pred = 1.0), Cx. quinquefasciatus-only areas contribute fully toward Cx. quinquefasciatus ancestry (H-index_pred = 0.0), and overlap areas contribute intermediate ancestry (H-index_pred = 0.5). Sites falling outside predicted ranges (SDM > 0.5) of both species were assigned H-index_pred = NA.

2.6. Statistical analysis of spatial and temporal hybridization patterns

To evaluate spatial and temporal contributions to hybridization patterns, we fit linear models relating empirical H-index to either latitude or predicted H-index_pred as response variables. Spatial and temporal effects were assessed by including geographic zone and sampling period as interacting factors to allow comparison of slope variation by interacting term and overall model fit based on adjusted R² and AIC values. Spatial and temporal effects were assessed in separate models due to sample size constraints.

3. Results

3.1. Genotype composition, hybrid index, and HWE

Observed genotype composition and hybrid index generally aligned with species distribution model (SDM) predictions, with hybridization most frequent in predicted overlap zones (Fig. 1, Table 1). However, admixture was also detected north of the expected range of Cx. quinquefasciatus and south of the expected range of Cx. pipiens, and hybridization levels varied markedly among nearby sites in several regions.

Fig. 1.

Predicted distributions and observed genotype composition of Culex pipiens and Cx. quinquefasciatus in North America. (A) Binary SDM predictions from MaxEnt (thresholded at suitability >0.5; Gorris et al. 2021), showing predicted presence of Cx. pipiens (purple), Cx. quinquefasciatus (green), and overlap zones (orange) representing potential hybrid areas. (B) Observed genotype composition at sampled sites. Pie charts indicate proportions of Cx. pipiens (purple), Cx. quinquefasciatus (green), and hybrid genotypes (orange); fixed genotype sites are shown as solid circles. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Table 1.

Summary of observed genotype composition at each sampling site and within connecting geographic zones (West Coast, East Coast, Central, and Mountain/Southwest). For each site and zone, the table reports the number of Cx. pipiens (pp), hybrid (pq), and Cx. quinquefasciatus (qq) genotypes, total sample size (N), latitude and longitude, observed hybrid index (H-index), predicted hybrid index (H-index_pred) based on local habitat suitability and overlap predictions from Gorris et al. (2021), and results from Fisher's exact tests for deviations for Hardy-Weinberg equilibrium (HWE p). H-index ranges from 0 (all Cx. quinquefasciatus) to 1 (all Cx. pipiens), with intermediate values reflecting admixture proportions.

Zone (Site, Year)	pp	pq	qq	N	Latitude	Longitude	H-index	H-indexpred	HWE p
West Coast	84	12	1	97	–	–	–	–	1.00
(ByroWA, 2023)	37	2	0	39	46.19	−119.90	0.97	0.52	1.00
(SuttCA, 2023)	9	8	1	18	39.17	−121.60	0.72	0.49	1.00
(LincCA, 2023)	38	2	0	40	38.90	−121.31	0.98	0.46	1.00
Mtn/Southwest	144	13	81	238	–	–	–	–	1.00E-05
(CachUT, 2023)	20	0	0	20	41.80	−111.82	1.00	1.00	NA
(BoxEUT, 2024)	20	0	0	20	41.70	−112.16	1.00	1.00	NA
(OgdeUT, 2024)	7	0	0	7	41.20	−112.05	1.00	0.71	NA
(SaltUT, 2023)	2	11	7	20	40.75	−111.97	0.38	0.70	0.83
(SaltUT, 2018)	13	2	0	15	40.75	−111.90	0.93	0.70	1.00
(FortCO, 2023)	19	0	0	19	40.38	−104.79	1.00	1.00	NA
(UteTUT, 2024)	18	0	0	18	40.33	−109.89	1.00	1.00	NA
(VernUT, 2024)	28	0	0	28	40.26	−109.35	1.00	1.00	NA
(ProvUT, 2024)	12	0	0	12	40.20	−111.63	1.00	0.84	NA
(MoabUT, 2024)	5	0	0	5	38.60	−109.57	1.00	NA	NA
(StGeUT, 2023)	0	0	38	38	37.18	−113.32	0.00	0.42	NA
(PhoeAZ, 2024)	0	0	20	20	33.45	−112.07	0.00	0.00	NA
(MariAZ, 2023)	0	0	16	16	33.07	−111.97	0.00	0.00	NA
Central	18	6	51	75	–	–	–	–	2.00E-05
(CookIL, 2023)	17	0	0	17	42.03	−87.93	1.00	0.50	NA
(DalFTX, 2024)	1	2	16	19	32.42	−96.94	0.11	0.00	0.66
(CollTX, 2023)	0	3	17	20	30.60	−96.27	0.08	0.00	1.00
(SlidLA, 2024)	0	1	18	19	30.33	−89.75	0.03	0.00	1.00
East Coast	57	8	39	104	–	–	–	–	1.00E-05
(BarnMA, 2023)	12	0	0	12	41.79	−69.99	1.00	0.50	NA
(HuntNJ, 2023)	16	2	1	19	40.54	−74.83	0.89	0.79	0.66
(SomeNJ, 2023)	12	3	0	15	40.53	−74.59	0.90	0.65	1.00
(RockMD, 2023)	17	3	0	20	39.06	−77.13	0.93	0.56	1.00
(AnasFL, 2023)	0	0	19	19	29.90	−81.40	0.00	0.00	NA
(MiDaFL, 2023)	0	0	19	19	25.80	−80.20	0.00	0.00	NA

In the East Coast and Central zones, hybridization levels were highly variable, with several sites showing greater hybridization than anticipated. On the East Coast, sites in New Jersey showed high H-index values (0.89–0.93), reflecting strong Cx. pipiens ancestry and the presence of hybrids. In contrast, Florida sites were fixed for Cx. quinquefasciatus (H-index = 0.00), consistent with SDM predictions for the far southeastern portion of the range. In the Central zone, hybrid genotypes were common in Texas and Louisiana, where sites exhibited low H-index values (0.03–0.11), reflecting substantial Cx. quinquefasciatus ancestry. Further north in Illinois, values were moderate (H-index = 0.24).

In the Mountain/Southwest zone, Cx. quinquefasciatus alleles were also common, but hybridization varied among sites. Northern Utah locations (Cache, Box Elder, Ogden, and Provo) were fixed for Cx. pipiens genotypes (H = 1.00), whereas Salt Lake City exhibited a mixed composition. Notably, temporal comparison of Salt Lake City samples indicated instability: the 2018 sample was dominated by Cx. pipiens (H = 0.93), while the 2023 sample showed higher hybridization levels (H = 0.38). In contrast, Moab (southeastern Utah) was fixed for Cx. pipiens (H = 1.00), despite its proximity to mixed sites. Further south, sites in Arizona and St. George (southern Utah) were fixed for Cx. quinquefasciatus (H = 0.00), marking the southern limit of Cx. pipiens in this region.

Along the West Coast, Cx. quinquefasciatus alleles were widespread but unevenly distributed. In Washington (Byron), Cx. quinquefasciatus alleles were detected (H = 0.97), marking one of the northernmost occurrences of admixture observed in this study. In central California, nearby sites showed contrasting levels of hybridization, highlighting patchiness at fine spatial scales. Despite the predicted presence of both species in this region, hybridization patterns were inconsistent across short distances.

Significant deviations from HWE were detected in pooled zone-level analyses, except on the West Coast. West Coast samples were consistent with HWE (p = 1.0), while the Mountain/Southwest (p = 1 × 10⁻⁵), Central (p = 2 × 10⁻⁵), and East Coast (p = 1 × 10⁻⁵) zones showed significant departures (p < 0.001), consistent with non-random mating or recent admixture. In contrast, individual sites with adequate sample sizes exhibited genotype frequencies consistent with HWE, suggesting local variation in hybridization dynamics (Table 1).

3.2. Spatial and temporal hybridization patterns

Hybridization patterns derived from both newly generated and historical published data revealed strong spatial structure and evidence for range expansion over time. Occurrence records compiled from this study, the published literature, and public databases indicated that hybrid clusters were historically restricted but have expanded both northward and southward in recent decades (Fig. 2). Compared to hybrid zone boundaries inferred from records prior to 1960, contemporary records (2020–2024) show new hybrid occurrences at both higher and lower latitudes, indicating a marked broadening of the hybrid zone.

Fig. 2.

Geographic scope and temporal trends in Culex species distributions and hybrid zones. Occurrence points for (A)Culex pipiens and (B)Cx. quinquefasciatus include records from this study, the published literature, and public databases (VectorBase and VectorMap, as compiled by Gorris et al., 2021). (C) Hybrid occurrence records are from this study and the literature; full criteria, references, and dataset details are available in Table S1, Table S2. Hybrid clusters represent locations where combinations of hybrids (hyb), Cx. pipiens (pip), and Cx. quinquefasciatus (qui) co-occur. Symbols indicate cluster composition: triangles (hyb + pip), inverted triangles (hyb + qui), and circles (hyb + pip + qui). Shading indicates inferred hybrid zone extent over three temporal periods (1940–1960, 1990–2019, and 2020–2024), showing a pattern of hybrid zone expansion both northward and southward through time.

Geographic scope and temporal trends in Culex species distributions and hybrid zones. Occurrence points for (A)Culex pipiens and (B)Cx. quinquefasciatus include records from this study, the published literature, and public databases (VectorBase and VectorMap, as compiled by Gorris et al., 2021). (C) Hybrid occurrence records are from this study and the literature; full criteria, references, and dataset details are available in Tables S1 and S2. Hybrid clusters represent locations where combinations of hybrids (hyb), Cx. pipiens (pip), and Cx. quinquefasciatus (qui) co-occur. Symbols indicate cluster composition: triangles (hyb + pip), inverted triangles (hyb + qui), and circles (hyb + pip + qui). Shading indicates inferred hybrid zone extent over three temporal periods (1940–1960, 1990–2019, and 2020–2024), showing a pattern of hybrid zone expansion both northward and southward through time.

This expansion has occurred alongside substantial spatial heterogeneity in admixture levels within and among zones. In the Central and East Coast zones, hybridization levels varied markedly, with notable admixture observed at southern sites (e.g., Texas and Louisiana) and in New Jersey, while sites in Florida remained fixed for Cx. quinquefasciatus (Table 1, Fig. 1). Sites in the Mountain/Southwest and West Coast regions, for example, exhibited local variability in hybridization, with hybridization evident in Salt Lake City but absent in nearby Moab (Table 1, Fig. 1). Along the West Coast, Cx. quinquefasciatus alleles were detected within areas of predicted range overlap (Fig. 1) [37], but were considerably farther north than previously reported in other hybrid classification efforts [12,18,32].

Spatial and temporal linear models relating empirical H-index to either predicted H-index_pred or latitude indicated that east-west spatial structure is a stronger and more consistent predictor than sampling period (Fig. 3). For the H-index_pred model, geographic zone explained more variance (adjusted R² = 0.66, AIC = −73.1) than sampling period (adjusted R² = 0.49, AIC = −22.3). Similarly, for the latitude model, geographic zone again explained more variance (adjusted R² = 0.69, AIC = 578.4) than sampling period (adjusted R² = 0.59, AIC = 611.9). Spatial models indicated a stronger relationship between latitude and H-index in the East Coast (R² = 0.82) and Central (R² = 0.66) zones, indicating strong latitudinal structuring of hybridization patterns. In contrast, the relationship was weakest in the West Coast zone (R² = 0.37). The Mountain/Southwest zone showed an intermediate pattern (R² = 0.62), despite high sampling density, with apparent patchiness in the transition from Cx. quinquefasciatus- to Cx. pipiens-dominated populations (Fig. 1). Temporal effects were more subtle and mainly detectable in the most recent decades, as shown in Supplementary Fig. S1.

Fig. 3.

Geographic structure of hybridization patterns in North America. Relationships between empirical hybrid index (H-index) and either (A) predicted hybrid index (H-index_pred), derived from species distribution model overlap, or (B) latitude are shown across geographic zones (West Coast, Mountain/Southwest, Central, East Coast). Zone explained a larger proportion of variance and had lower AIC values than temporal models (adjusted R² = 0.66–0.69 vs. 0.49–0.59; AIC = −73.1 to 578.4 vs. -22.3 to 611.9), indicating strong east-west structuring of hybridization patterns. Occurrence records are from this study and the literature; full criteria, references, and dataset details are available in Table S1, Table S2. These results suggest that ecological and climatic gradients associated with both latitude and east-to-west geography play a dominant role in shaping hybridization across North America.

Geographic structure of hybridization patterns in North America. Relationships between empirical hybrid index (H-index) and either (A) predicted hybrid index (H-index_pred), derived from species distribution model overlap, or (B) latitude are shown across geographic zones (West Coast, Mountain/Southwest, Central, East Coast). Zone explained a larger proportion of variance and had lower AIC values than temporal models (adjusted R² = 0.66–0.69 vs. 0.49–0.59; AIC = −73.1 to 578.4 vs. -22.3 to 611.9), indicating strong east-west structuring of hybridization patterns. Occurrence records are from this study and the literature; full criteria, references, and dataset details are available in Tables S1 and S2. These results suggest that ecological and climatic gradients associated with both latitude and east-to-west geography play a dominant role in shaping hybridization across North America.

4. Discussion

4.1. Broad ecological gradients and hybrid zone structure at large spatial scales

Our results indicate strong latitudinal patterns in ancestry, consistent with prevailing theory positing that opposing selective pressures imposed by winter conditions maintain the Cx. pipiens–Cx. quinquefasciatus hybrid zone boundaries across large spatial scales. Latitude alone explained hybrid index variation as well or better than predicted habitat overlap (Fig. 3), underscoring the importance of climatic factors in shaping gene flow between Cx. pipiens and Cx. quinquefasciatus.

A key ecological barrier to gene flow is thought to be their overwintering biology: Cx. pipiens enters diapause to survive cold temperatures in temperate regions, while Cx. quinquefasciatus lacks diapause and is limited to subtropical and tropical climates [8,13,16]. As a result, diapause has long been proposed as a key selective barrier [10]. Laboratory crosses have shown that non-diapause phenotypes are dominant, and diapause expression only increases when backcrossing (mating of a hybrid with the Cx. pipiens parental population) restores >75 % Cx. pipiens genomic background [13], suggesting that maintaining the diapause phenotype in colder environments requires strong selection against non-diapause genotypes [10]. At northern latitudes, selection could limit the persistence of Cx. quinquefasciatus alleles and hybrids unable to overwinter. In contrast, in southern regions, physiological trade-offs associated with diapause could reduce the fitness of Cx. pipiens alleles and hybrids expressing diapause-linked traits. Together, these opposing selective pressures likely filter pure parental forms and advanced backcrosses (hybrids mated back to one of their parental populations) from the hybrid zone margins, constraining the movement of key traits like diapause and other genes involved in the physiological trade-offs of diapause [10,42,43]. However, according to hybrid zone theory, the boundary would remain porous and variable across the genome [[44], [45], [46]], with the average rate of introgression (the incorporation of genetic material from one population into another through repeated hybridization) beyond these boundaries shaped by regional climate, countervailing forces such as local adaptation or positive selection for traits like insecticide resistance, and the genetic architecture of traits (e.g., recombination rates) [47].

4.2. East–West variation across contact zones

In addition to broad-scale ecological gradients, our results revealed distinct regional patterns in hybrid zone structure across North America. Zone identity explained a large proportion of variance in hybrid index in the full model (adjusted R² = 0.66–0.69; Fig. 3), reflecting strong east–west differences. These patterns were further clarified by considering sampling density and predicted distributions (Fig. 1, Supplementary Fig. S2), suggesting that regional variation reflects not only ecological factors—such as winter severity, elevation, and habitat structure—but also differences in the timing and sequence of introduction and secondary contact.

Along the East Coast, hybridization followed a steep latitudinal gradient (R² = 0.82), with pure Cx. quinquefasciatus populations persisting in Florida and hybrid index increasing sharply northward. This sharp transition may reflect stronger latitudinal climatic gradients in the eastern U.S. than in the west, and possibly earlier introduction of both Cx. pipiens along the northern East Coast and Cx. quinquefasciatus along the southern East Coast, potentially dating to the transatlantic slave trade [4,6].

In the Central U.S., the hybrid transition also has a strong latitudinal signal (R² = 0.66) but is broader and more diffuse, with hybrids detected farther north and south. This pattern may reflect weaker environmental gradients and ongoing expansion of the hybrid zone along the Mississippi River corridor.

In the Mountain/Southwest, the hybrid transition still displayed a latitudinal signal (R² = 0.62), but in contrast to the East Coast and Central regions, was narrow and abrupt, despite dense sampling relative to habitat availability (Fig. 1). Patchy hybrid index values in this region, especially in the Salt Lake City area (Table 1, Fig. 1), suggest strong local ecological filtering, likely driven by topographic and climatic heterogeneity across short spatial scales. This abrupt and patchy transition is also consistent with strong selection against non-diapause genotypes at higher altitudes, particularly in areas farther from the Great Salt Lake that may experience more extreme winter temperatures. Historical contingencies may also have contributed to the observed patchiness. For example, the establishment of Cx. quinquefasciatus in northern Utah may reflect urban colonization and priority effects, whereby the timing and route of introduction shaped which genotypes became established and persisted. This aligns with models of introgression at invasion fronts [48] and the coupling of ecological traits with barrier loci in heterogeneous environments [43]. In this context, microhabitat variation, historical routes of colonization, and the timing of secondary contact likely interact in complex ways to shape the observed spatial mosaic.

On the West Coast, hybridization was broad and spatially diffuse, with the weakest latitudinal signal (R² = 0.37). Extensive sampling, especially in California, rules out sampling bias as an explanation. Instead, the shallow gradient likely reflects coastal moderation of winter temperatures, which allows Cx. quinquefasciatus to persist farther north through the Central Valley and into central Washington. This broad, patchy pattern may also reflect weak selection for diapause in coastal and lowland areas, where winter temperatures are less severe. The diffuse nature of the transition on the West Coast may thus reflect both environmental permissiveness and the endpoint of a broader westward expansion of both species from initial introduction points on the East Coast, with increasingly complex hybrid zone dynamics arising in regions of recent or repeated contact.

Consistent with this interpretation, tests for HWE revealed significant deviations at the zone level (p < 0.001) in all but the West Coast (p = 1.0), despite general conformity at individual sites. This pattern has been discussed previously by Cornel et al. [32], who reported that California populations showed no HWE deviations, in contrast to pronounced Wahlund effects in South Africa. The moderate climate in the West may explain it as an outlier in the tests for region-wide HWE, as weaker winter selection likely reduces the strength of non-random mating. Together, results reinforce that gene flow remains spatially limited and heterogeneous even in the absence of strong reproductive isolation.

4.3. Variation through time: The 1940s to present

Although spatial structure was the strongest predictor of hybridization (Fig. 2, Fig. 3), temporal trends were also evident. Compared to pre-1960 records, contemporary data (2020–2024) show both northward and southward expansion of hybridization. Detection of Cx. quinquefasciatus alleles in Washington and Utah, and Cx. pipiens alleles in Texas and Louisiana, suggests introgression into regions once dominated by pure parental types—echoing findings by Kothera et al. [10] of southward expansion relative to Barr's [8] early delineation.

Time period also explained variation in hybrid index across latitude and predicted hybrid index (adjusted R² = 0.49–0.59), with the most recent timeframe showing a steeper relationship (Supplemental Fig. S1). Although these patterns suggest expansion of the hybrid zone over time, they should be interpreted with caution, as uneven sampling across time periods and variation in hybrid classification methods may have led to underreporting of hybrids in earlier datasets. Nonetheless, the data in hand point to the action of dynamic processes, including range expansion and localized introgression, layered atop broader ecological structure.

Together, these spatial and temporal contrasts demonstrate that hybrid zone structure is shaped not only by stable ecological gradients but also by historically contingent and dynamic processes operating at regional and local scales.

4.3.1. Limitations and confidence

Our analysis integrates both new and historical data, resulting in several limitations that should be noted. First, sample sizes were small in some regions, particularly the Midwest, which reduces statistical power and may underrepresent local variation. Second, historical records varied in collection methods and diagnostic criteria, making comparisons across time imperfect despite our efforts to standardize. Finally, confidence in broad-scale patterns should be viewed heuristically. The overall latitudinal cline is well supported, and East Coast structure is consistent across independent datasets, but regional contrasts elsewhere remain more tentative and will require denser, more consistent sampling to confirm. These caveats highlight the need for cautious interpretation of local patterns, but they do not detract from the broader insights our results provide into how ecological gradients and historical contingencies shape hybrid zone structure across North America.

4.3.2. Synthesis and alignment with hybrid zone theory

The spatial and temporal complexity of hybridization patterns observed across North America aligns with multiple models of hybrid zone structure. Classic tension zone theory posits that hybrid zones may be maintained by a balance of dispersal and selection against hybrids [42], while the coupling hypothesis emphasizes the role of linkage among barrier loci and ecological traits in reinforcing reproductive boundaries [43]. These frameworks help explain broad-scale latitudinal gradients and zone-wide deviations from HWE observed in our data.

At finer scales, our results are also consistent with mosaic hybrid zone theory, which emphasizes ecological heterogeneity and habitat partitioning as drivers of spatially variable introgression [44]. The sharp transition observed in the Mountain/Southwest and the broader, spatially diffuse pattern on the West Coast both suggest that local adaptation and microhabitat variation act to restrict gene flow beyond what would be expected from climate alone. This interpretation is further supported by the patchiness in hybrid index values in areas like Salt Lake City and central California, as well as the zone-level HWE deviations observed outside the West Coast.

Similar patterns have been reported in other Culex hybrid zones. On the East Coast, researchers have repeatedly documented fine-scale spatial and temporal variation in population genetic structure and vector competence within Cx. pipiens complex populations, associated with urban, suburban, and rural habitat types [37,49,50]. Likewise, in Chicago in the Central U.S., Kothera et al. [11] observed spatially structured variation in host use, admixture, and ecological associations along urbanization gradients. These cases underscore the broader pattern that habitat variation, microclimate, and landscape structure impose additional constraints on gene flow, even in the absence of strong intrinsic reproductive isolation.

5. Implications and future directions for one health

This study provides a foundation for advancing both fundamental research and applied strategies within a One Health framework. On the applied side, future work will build on our standardized image-labeled dataset to train AI-based identification models using IDX system images. These models will improve resolution of fine-scale hybrid zone structure, particularly in complex mosaic regions where molecular sampling is logistically limited. By linking AI-generated species identifications to ecological and epidemiological data, future surveillance programs could more accurately estimate West Nile virus (WNV) infection rates by vector population. Looking forward, AI-based identification will only reach its full potential if paired with systematic, large-scale datasets that capture variation across both geography and time. Models such as convolutional neural networks require not only high-quality, expertly validated images for training but also dense spatiotemporal sampling to ensure that seasonal, latitudinal, and ecological gradients are well represented. For cryptic species complexes like Culex pipiens, this entails constructing image-labeled libraries that span diverse hybrid backgrounds, environmental contexts including variable microhabitats and population densities, and seasonal cohorts. Such structured data collection would improve the system's ability to detect range shifts in cryptic species and hybrids under field conditions and enable surveillance programs to monitor with finer resolution. Similar applications are being explored in cryptic species such as the Anopheles gambiae complex in Africa, where early studies suggest that AI methodologies hold promise, though results remain under active investigation. When properly trained, these approaches not only enhance species resolution but can also uncover subtle morphological traits previously overlooked, which may inform future taxonomic keys and advance both applied surveillance and fundamental systematics. Furthermore, by detecting nuanced morphological population structure, AI tools can provide new insights into biogeography, population shifts, and historical migrations or expansions of species. While not a substitute for genomic analyses, the integration of high-resolution morphological AI with molecular data offers a particularly powerful approach to linking genotype, phenotype, geography, and evolutionary history.

With additional genome-wide analysis becoming increasingly accessible [7,[24], [25], [26]], this system also offers a powerful model for investigating the permeability of reproductive barriers under contemporary selection. In particular, diapause remains a central, yet unresolved, trait that appears to govern north-south gene flow in the Cx. pipiens complex. Identifying the genetic basis of diapause—and understanding how it interacts with traits like host use and insecticide resistance—is critical to predicting hybrid zone stability, adaptive trait spread, and long-term vector dynamics. Because diapause-linked alleles are favored in colder regions while introgression proceeds more freely elsewhere in the genome, these regions should exhibit elevated levels of differentiation. This system offers a rare opportunity to identify such genomic islands and investigate the mechanisms leading to seasonal adaptation, physiological tradeoffs, and constraints on range expansion under changing environmental conditions. On the applied side, targeting these loci could inform innovative control strategies, including gene drive or symbiont-based methods designed to suppress overwintering capacity in key transmission zones.

More broadly, this and similar studies position the Cx. pipiens hybrid zone as a globally relevant natural experiment linking ecological and evolutionary theory to emerging One Health challenges. As climate change and urbanization reshape species distributions, systems like this offer rare opportunities to study how ecological gradients, local adaptation, and hybridization interact to influence ecological processes such as pathogen transmission risk. These insights have implications for both human and wildlife health, improving our ability to predict where and when mosquitoes are most likely to amplify WNV in avian hosts or serve as bridge vectors to humans.

6. Conclusions

Our findings strongly suggest that regional hybrid zone structure in the Cx. pipiens complex emerges from the interaction of two key processes: broad-scale environmental gradients and fine-scale ecological heterogeneity and historical contingency. We suggest that broad-scale environmental gradients and diapause-associated selection generate predictable hybrid zone boundaries, while ecological heterogeneity and historically contingent factors produce mosaic patterns and geographic variation. By integrating new and historical data, we also document temporal dynamics indicative of both hybrid zone expansion, with admixture now extending northward into the Pacific Northwest and northern Utah, and southward into warmer regions such as Texas and Louisiana. These spatial and temporal patterns position this hybrid zone as a powerful model for exploring eco-evolutionary responses to colonization, range shifts, and climate adaptation in real time. Emerging tools such as AI-based identification may further enhance resolution in detecting hybrid dynamics and complement genetic approaches in this system, especially when paired with systematic, large-scale datasets. As mosquito distributions will continue to shift with climate and land use change, this system offers a critical lens into how local adaptation, gene flow, and overwintering dynamics shape vector populations—and, in turn, influence avian reservoir systems and the risk of zoonotic disease transmission to humans. With similar hybrid zones now documented in South America, South Africa, Australia, and Europe, the Cx. pipiens species complex provides a globally relevant model for linking evolutionary biology to One Health challenges.

The following are the supplementary data related to this article.

Supplementary Fig. S1

Temporal structure of hybridization patterns in North America.

Supplementary Fig. S2

Occurrence of Cx. pipiens, Cx. quinquefasciatus and hybrids in North America from 1940 to 2024.

Table S1

Summary of classification methods and thresholds to identify Culex pipiens, Culex quinquefasciatus, and their hybrids across North America from the 1940s to the present. Further details can be found in the source publications.

Table S2

Summary of published observations of Culex pipiens, Culex quinquefasciatus, and their hybrids across North America from the 1940s to the present.

CRediT authorship contribution statement

Norah Saarman: Writing – original draft, Visualization, Validation, Supervision, Software, Resources, Project administration, Methodology, Investigation, Funding acquisition, Formal analysis, Data curation, Conceptualization. Katelyn Graybeal: Writing – review & editing, Methodology, Investigation, Data curation. Tyler Seeley: Writing – review & editing, Methodology, Investigation, Data curation. Emily Calhoun: Writing – review & editing, Methodology, Investigation, Data curation. Eric Jenkins: Writing – review & editing, Methodology, Investigation, Data curation. Andre De Lima Moraes: Writing – review & editing, Visualization, Software, Resources, Methodology, Investigation, Formal analysis, Data curation. Roy Faiman: Writing – review & editing, Software, Resources, Project administration, Funding acquisition, Data curation, Conceptualization. Hannah Markle: Writing – review & editing, Investigation, Data curation. Rachael Pellegrini: Writing – review & editing, Investigation, Data curation. Skylar Arent: Writing – review & editing, Investigation, Data curation. Andrea Gloria-Soria: Writing – review & editing, Supervision, Resources, Project administration, Methodology, Investigation, Funding acquisition, Data curation, Conceptualization.

Declaration of generative AI and AI-assisted technologies in the writing process

During the preparation of this work, the authors used GPT-4o to assist with grammatical corrections and word choice to improve language and readability. Following its use, the authors reviewed and edited the content as needed and take full responsibility for the final version of the publication.

Declaration of competing interest

Hannah Markle and Roy Faiman are employed by Vectech, Inc., a Public Benefit Corporation that develops AI-based vector surveillance tools. Vectech is a for-profit entity with a public benefit mission to support pest control organizations and reduce vector-borne disease. This affiliation did not influence the study design, data collection and analysis, decision to publish, or preparation of the manuscript. No competing financial interests exist.

Appendices

Appendix A. Full R analysis pipeline applied to the Ace2 genotype dataset generated in this study (new data only). Source code and data are available on GitHub.

Appendix B. Full R analysis pipeline applied to the Ace2 dataset from this study combined with previously published datasets (integrated dataset). Source code and data are available on GitHub.

Data availability

Genotype data, hybrid index values, and R scripts are in the Supplementary Data and at https://github.com/saarman/ace2. Additional materials, including raw lab gel images, are available on request.