A continental-scale survey of Wolbachia infections in blue butterflies reveals evidence of interspecific transfer and invasion dynamics

Abstract

Infections by maternally inherited bacterial endosymbionts, especially Wolbachia, are common in insects and other invertebrates but infection dynamics across species ranges are largely under studied. Specifically, we lack a broad understanding of the origin of Wolbachia infections in novel hosts, and the historical and geographical dynamics of infections that are critical for identifying the factors governing their spread. We used Genotype-by-Sequencing data from previous population genomics studies for range-wide surveys of Wolbachia presence and genetic diversity in North American butterflies of the genus Lycaeides. As few as one sequence read identified by assembly to a Wolbachia reference genome provided high accuracy in detecting infections in host butterflies as determined by confirmatory PCR tests, and maximum accuracy was achieved with a threshold of only 5 sequence reads per host individual. Using this threshold, we detected Wolbachia in all but 2 of the 107 sampling localities spanning the continent, with infection frequencies within populations ranging from 0% to 100% of individuals, but with most localities having high infection frequencies (mean = 91% infection rate). Three major lineages of Wolbachia were identified as separate strains that appear to represent 3 separate invasions of Lycaeides butterflies by Wolbachia. Overall, we found extensive evidence for acquisition of Wolbachia through interspecific transfer between host lineages. Strain wLycC was confined to a single butterfly taxon, hybrid lineages derived from it, and closely adjacent populations in other taxa. While the other 2 strains were detected throughout the rest of the continent, strain wLycB almost always co-occurred with wLycA. Our demographic modeling suggests wLycB is a recent invasion. Within strain wLycA, the 2 most frequent haplotypes are confined almost exclusively to separate butterfly taxa with haplotype A1 observed largely in Lycaeides melissa and haplotype A2 observed most often in Lycaeides idas localities, consistent with either cladogenic mode of infection acquisition from a common ancestor or by hybridization and accompanying mutation. More than 1 major Wolbachia strain was observed in 15 localities. These results demonstrate the utility of using resequencing data from hosts to quantify Wolbachia genetic variation and infection frequency and provide evidence of multiple colonizations of novel hosts through hybridization between butterfly lineages and complex dynamics between Wolbachia strains.

Keywords: Lycaeides; Wolbachia; GBS data; geography of infection; host parasite interactions; infection acquisition.

Fig. 1.

Range maps of the 5 nominal species of Lycaeides in the United States with the 107 sampled locations plotted as site numbers corresponding to Table 1. The dense sampling in the southwestern United States is expanded in the lower left. The inset square indicates the Verdi, Nevada sampling area, including sites 66–76, and is also expanded (bottom, middle). Sample locations in Alaska are illustrated in the map on the lower right.

Fig. 2.

Accuracy and error rates of comparing bioinformatics results to previous PCR-based studies for detecting putative Wolbachia infections in the genome for 129 individuals (shown here for a threshold of varying number of reads of length greater than 80 bp). We used a threshold read depth of 5 for classifying an individual as infected, as it had the highest accuracy of 96.9% correspondence with the PCR-based results, while still maintaining a low FNR (classifying an individual as not being infected when the individual is inferred to be infected from PCR-based analysis). FPRs (compared to PCR-based results) were generally low. Note that the X-axis is on a log-scale.

Fig. 3.

Bubble plots indicating the proportion of infected individuals in a population across the 107 sampled locations. Most populations in the western United States are mostly or wholly infected ($>95\%$), while the L. samuelis populations in the east show low to no infection ($<5\%$). Inset plots zoomed into regions of interest for visibility. The white square indicates the Verdi, NV sampling area, and is expanded (bottom, middle). Inset plot is a histogram of infection frequencies across 107 sampling localities using a threshold of a minimum of 5 sequence reads of at least 80 bp.

Fig. 4.

Demographic histories (left) and haplotype networks (right) for each major strain (wLycA, wLycB, wLycC). Population sizes were estimated using BEAST 2 (Bouckaert et al. 2014). The median mutation-scaled effective population size (dashed line) and 95% credible interval (central posterior density, shaded region) for each strain is presented over time (measured in substitution rate). For simplicity, we assume equal substitution rates across strains to aid interpretation. Ninety-five percent parsimony networks show observed haplotypes in red and inferred haplotypes in blue with numbers of individuals observed possessing each haplotype in parentheses. Haplotypes are 115 bp in length.

Fig. 5.

Plot of PCoA of Wolbachia haplotypes (76 in total) based on uncorrected pairwise distances among haplotypes. Colored dots represent 115 bp haplotypes in the 3 major strains (blue: wLycA, orange: wLycB, green: wLycC). Strain wLycA was found mostly in the L. melissa, L. idas, and L. samuelis populations continent-wide. Strain wLycB was mostly found in the L. melissa populations in the western Great Basin. Strain wLycC was found exclusively in the L. anna populations and in the hybrids between L. melissa and L. anna. Black dots represent haplotypes found as singletons and not considered part of the 3 main strains (see Table 1 and Supplementary Fig. 3).

Fig. 6.

Pie charts showing the distribution of haplotypes from all 3 strains (row-wise: wLycA, wLycB, wLycC). Haplotypes A1 and A2 are present in 90% of individuals infected with strain wLycA. The label “OtherA” corresponds to rare haplotypes in wLycA (A3–A19). Haplotypes B5, B9 and B10 make up 78% of all infections in the wLycB strain. Haplotype C1 makes up for 98% of all infections in the wLycC strain, and all other wLycC haplotypes are found in localities that also include haplotype C1. Pies are only shown if a given haplotype is present in the population.