Environmental adaptation in an asexual invasive insect

Abstract Parthenogenetic reproduction is generally associated with low genetic variance and therefore reduced ability for environmental adaptation, and this could limit the potential invasiveness of introduced species that reproduce asexually. However, the hemlock woolly adelgid is an asexual invasive insect that has spread across a large geographic temperature gradient in its introduced range. Consequently, this insect has shown significant variation in cold hardiness among populations. We hypothesized that the increased cold hardiness of northern populations represents an adaptation to the colder temperatures. To test this, we collected individual adelgid from populations spanning their invaded range and inoculated them into a common thermal environment. We then experimentally sampled the supercooling point of the progeny of these adelgids and compared these results with tests of the supercooling point of adelgid sampled directly from their source populations. The results showed that the same significant differences in supercooling that was found among geographically distinct populations existed even when the adelgid was reared in a common environment, indicating a genetic basis for the variation in cold hardiness. These findings support the hypothesis that the adelgid has adapted to the colder environment as it has expanded its distribution in its invaded range.


| INTRODUCTION
Invasive species are a major ecological concern due to their detrimental effects on biodiversity, ecosystem functioning, disturbance regimes, etc. (Liebhold, MacDonald, Bergdahl, & Mastro, 1995;Pimentel et al., 2001). However, few species introductions result in established populations due to the numerous factors that inhibit the successful colonization of new habitat (Liebhold et al., 1995;Sakai et al., 2001).
A major hurdle is that most introduced species are not well adapted to their new environment. This can result in reduced fitness and eventual extinction of the introduced population. For this reason, the genetic makeup of the founding population-including the amount of additive genetic variation-and consequently their ability to rapidly adapt to the new environment are important factors in both the successful colonization of new habitat and expansion of their distribution following establishment (Carroll, Dingle, Famula, & Fox, 2001;Lambrinos, 2004;Lee, 2002).
The process of environmental adaptation is driven by directional selection (Cano, Escarre, Fleck, Blanco-Moreno, & Sans, 2008;Lambrinos, 2004;Maynard-Smith, 1980), and a species response to selection is a function of their additive genetic variation (Fisher, 1930). Not all species respond to selection in the same manner.
For example in asexually reproducing organisms, environmental adaptation presents a challenge because unlike sexual reproduction Data Archive Location: Dryad or GitHub (Upon acceptance for publication). the lack of recombination reduces genetic variation under directional selection (Burger, 1999;Charlesworth, 1993;Crow, 1992).
For this reason, it is reasonable to suggest that parthenogenetically reproducing organisms might be a lower risk for becoming invasive, particularly in regions where environmental variation is high. While previous studies have demonstrated adaptation in asexual organisms, the majority of these involved relatively stable conditions (Grapputo, Kumpulainen, Mappes, & Parri, 2005;Liu & Trumble, 2007;Llewellyn et al., 2003;Sandstrom, 1996;Sunnucks, Chisholm, Turak, & Hales, 1998). Few studies have experimentally tested the ability of an asexual invasive species to adapt to a stochastic environment. There is evidence of such adaptation in certain phenological traits of the asexual alpine grasshopper (Chorthippus cazurroi) when reared under variable conditions in the lab; however, these results were generally not evident in wild populations (Laiolo & Obeso, 2015). A more probable example is that of the invasive parthenogenetic insect, the hemlock woolly adelgid (Adelgis tsugae Annand). This insect has shown significant differences in cold hardiness among populations spanning a large geographic gradient of cold temperature extremes (Skinner, Parker, Gouli, & Ashikaga, 2003). However, because cold tolerance in insects is generally induced by exposure to progressively colder temperatures (Elkinton et al., 2016;Zachariassen, 1985), it is unclear whether the variation represents evolutionary adaptation (genetically based) or environmental acclimation (phenotypic). Subsequent work on the hemlock woolly adelgid by Butin, Porter, and Elkinton (2005) provided evidence of a genetic basis for the regional variation in cold tolerance in this insect. Their study was limited in spatial scale though and examined adelgid from just two locations (one from a northern population, and one from the Mid-Atlantic region). Yet, combined these prior studies on this insect provide evidence of environmental adaptation despite its obligately asexual life history.
The purpose of this study was to experimentally test for adaptation in an asexual species, using the hemlock woolly adelgid as a model system. We focused on a specific functional trait, cold tolerance, which is an important adaptation to northern environments. Differentiating between adaptation and acclimation requires experimentally testing adelgid from across their geographic range with methods to control for the sources of variation (adaptation vs. acclimation). Therefore, we employed a robust method involving the artificial inoculation of adelgid from populations spanning their introduced range into a common thermal environment.
We used a quantifiable measure of cold hardiness, freeze avoidance, as our physiological trait. Freeze avoidance constitutes the F I G U R E 1 Four hypotheses for determining the influence of adaptation and environmental acclimation on a functional reaction norm, modified after Conover and Schultz (1995). N i = individuals from a natural population, T i = individuals from a transplant or common garden experiment.  (Zachariassen, 1985). Production of cryoprotectants is a physiological response. While some degree of cold tolerance in insects is present constituently, a large portion is induced by exposure to increasingly cold temperatures. Generally, the quantity of cryoprotectants determines the level of protection (Lee, Chen, & Denlinger, 1987).
We initiated a series of a priori hypotheses following from a modified version of those outlined by Conover and Schultz (1995, box 2) for a common garden style experiment. Our hypotheses are demonstrated in Figure 1 (boxes a-d, respectively) and cover four potential outcomes: (1) a null model where neither acclimation nor adaptation influences the physiological reaction norm, (2) variation in the reaction norm due to acclimation, with no evidence of adaptation, (3) adaptation as the sole source of variation in the reaction norm, and (4) both acclimation and adaptation influencing variation.

| Model species
The hemlock woolly adelgid ( Figure 2) is an invasive insect introduced to the eastern United States near Richmond VA, in 1951 from Japan (Havill, Montgomery, Yu, Shiyake, & Caccone, 2006). By the 1980s, it had spread to the New England states, and its range currently spans from southern Maine to northern Georgia. The insect is native to Asia as well as the Pacific Northwest of the United States (Havill et al., 2006). In both of these native habitats, the insect is largely benign. In its invaded range, however, the adelgid has caused widespread mortality of hemlock trees.

| Regional variation in supercooling
To determine the regional variation in cold hardiness of hemlock woolly adelgid, we tested the supercooling point of individuals collected from 14 different sites spanning their introduced range. A list of the locations can be found in Table 1. Sites were sampled in  T A B L E 1 Locations of adelgid sampling sites, their regional designation, sampling dates, experimental use, and the mean minimum yearly temperature for each site packaging tape. The thermocouples were placed into a container submerged in a supercooling bath (Neslab RTE-140), and the temperature of the bath was slowly reduced from a starting point of 5°C down to −35°C over the course of 3 hours (≈1°C change every 5 minutes).
The temperature of each adelgid was recorded in one-second intervals using a multi-channel thermocouple recorder (Physitemp Inc., NJ, USA). When an adelgid freezes, the heat of fusion produces an obvious spike in the temperature. We used the temperature in the second prior to the thermal spike as the supercooling point. For each of the 14 sampling sites, we tested approximately 25 adelgids. All samples were tested within 48 hours of collection and were maintained at a constant 12°C for much of that time.
To analyze the data, each of the regional source populations was assigned to one of three geographic categories, specifically North, In addition, a one-way analysis of variance was used to test for differences in the means of each region. This was followed with a post hoc multiple comparisons test using a Tukey-Kramer HSD to determine pairwise differences. All analyses were completed using the R statistical software version 3.2.1 (R Core Development Team, 2015).

| Common garden experiment
To separate the influence of environmental acclimation from the role of genetic adaptation in the cold hardiness of hemlock woolly adelgid, we transplanted gravid, sistens generation adelgids from 10 sites (latitudinal range: 34.7858-42.5719) across their introduced range into a common garden experiment (Table 1). It is important to note that in this experiment, the term common garden is referring to a common Afterward the bags and the inoculum were removed.
To determine the cold hardiness of these adelgids, we collected samples from the common garden experiment in December of 2015.
These samples represented the F 2 sistens generation that arose from the progrediens that were established in the common garden the previous spring. A benefit of this method is that it can help to avoid potential complications due to maternal effects on cold tolerance, because the maternal generation never experienced extreme cold. Once Supercooling temperature (°C)

Density
samples were collected, we tested their supercooling point using the exact same procedures as outlined for the regional samples. As with the regional samples, those from the common garden were tested within 48 hr of collection and were maintained at a constant 12°C for most of that time. We had scheduled a second sampling date for February 2016; however, we were unable to complete the February sample due to an extreme cold temperature event (approx. −25°C, 13 February 2016) which resulted in >90% adelgid mortality in all samples at the common garden site.
To examine differences in the cold hardiness of the adelgid, we grouped the data by region of the source population (North, Atlantic, South) and examined the form and spread of the distributions (see Table 1). We also used a one-way analysis of variance with region as a factor to determine differences in the means among regions. This was followed by a post hoc Tukey-Kramer HSD test to examine pairwise differences. Again, all analyses were completed using the R statistical software version 3.2.1 (R Core Team, 2015).

| Common garden
Data from the common garden was compiled into the same North, Atlantic, and South classifications as was used in the regional analyses,

| Comparison of the common garden with the source populations
The significant variation that was seen in the December 2015 sample of adelgid from throughout their introduced range when exposed to ambient conditions was also maintained when individuals from those same populations were reared in a common environment. This is evident by the approximately horizontal lines connecting the mean supercooling temperatures of adelgid from their regional source populations with those from the same populations from the common garden ( Figure 5). We used a multi-way ANOVA to test for differences in the main effects of treatment (source population, common garden), and region (North, Atlantic, South), as well as the treatment × region interaction. We found significant differences in both main effects, and no significant difference in the interaction term, indicating that all sites responded to the treatment in the same manner (Table 2). A follow-up post hoc analysis showed no significant differences in the F I G U R E 4 The distribution of supercooling points of hemlock woolly adelgid collected from source populations in three different regions of their introduced range and reared in a common environment showed significant differences between those from the Southern versus Northern populations, as well as the North versus Atlantic. Adelgids were inoculated into a common garden in spring of 2015 and sampled for supercooling point the following December F I G U R E 5 Significant differences in the supercooling point of adelgid from different source populations are maintained when they're reared in a common environment, indicating the populations have adapted to their regional environments. No significant differences in the pairwise comparisons among the three regions (connecting lines have slopes ≈ 0) indicate that environment does not explain these differences. See Figure 1 for an overview  (Lundmark & Saura, 2006;Niklasson & Parker, 1994). However, its close association with humans is likely a large contributing factor in its wide geographic distribution (Niklasson & Parker, 1994). A couple well-known examples of adaptive evolution in parthenogenetic species include the root-knot nematode (Meloidogyne sp.), (Castagnone-Sereno, 2006), and the simultaneous evolution of multiple adaptive traits in clonal lab populations of Daphnia Magna (Boersma, Meester, & Spaak, 1999).
While winter temperature is the selective factor, it is not clear whether extreme cold tolerance was present in the founding population, or whether it was acquired via mutation in the invasive population. Butin et al. (2005) laid out a mathematical generalization for mutation to act as a source of adaptation. However, the acquisition of adaptive mutations in a strictly parthenogentic population can be problematic (Garrish & Lenski, 1998;Hill & Robertson, 1966;Wagner & Gabriel, 1990), and Butin et al. (2005) acknowledged this issue. A more parsimonious explanation is that genes conferring extreme cold tolerance were present in the founding population, and as the adelgid expanded its range northward the colder environment selected for the most cold-tolerant of those individuals. In its native range, this insect exists across a gradient of thermal extremes similar to that of the eastern US. Therefore, it is reasonable to expect that such extreme cold tolerance could have been present in the founding population.
One of the concerns when testing for evidence of environmental adaptation is the potential for maternal effects to influence the results. Our experiment was designed to avoid maternal effects by using the progrediens generation to inoculate the common garden, and sampling from the subsequent sistens. This ensured that the maternal generation was not exposed to extreme cold temperatures, because eggs of the progrediens are laid in the spring and hatch in the early summer. The progrediens give rise to the sistens generation in early fall, and it was those sistens adults that we sampled throughout the winter. Of course this design does not guarantee that maternal effects were not present. Indeed, in some instances maternal effects have been shown to occur across generations (i.e., the F2 generation). This is sometimes referred to as grand maternal effects. For example, exposure to cold temperature in some Drosophila spp. can occasionally influence the cold hardiness of both the F1 and F2 generations, but the effect on the F2 generation is dependent upon specific environmental context (Watson & Hoffmann, 1995).
Generally, grand maternal effects seem to be uncommon in most insects (Mousseau & Dingle, 1991;Mousseau and Fox 1998). We were unable to find any research that specifically tested for any type of maternal effects in the hemlock woolly adelgid; however, studies have shown an absence of grand maternal effects in the closely related Aphidoidae (Andrade & Roitberg, 1995;McLean, Ferrari, & Godfray, 2009;Via, 1991;Zehnder & Hunter, 2007 Although the means of the two groups were significantly different, there were no within-group differences (i.e., North from the source populations were not different from the North reared in the common garden, etc.) (Tukey HSD, α = 0.05). T A B L E 3 Tukey-Kramer HSD post hoc analysis of pairwise differences in the supercooling point of adelgid sampled from their source population versus those from the same populations when reared in a common environment There are a few scenarios by which the adelgid could continue its northern range expansion. First, it is possible that the adelgid is yet to reach their physiological limit of extreme cold tolerance. According to our results and those of others (Butin et al., 2005;Parker, Skinner, Gouli, Ashikaga, & Teillon, 1998;Skinner et al., 2003), the maximum cold tolerance for northern adelgid is between −30 and −35°C. Such temperatures are not uncommon minimum winter temperatures in the northern range of hemlock trees. Even at such extreme temperatures, however, some adelgid may survive (approx. 1% survival according to Skinner et al., 2003). As new populations of hemlock woolly adelgid can arise from a single individual (Tobin, Turcotte, & Snider, 2013), it is plausible that the adelgid could continue to expand its range. This would be a slow northward progression, compared to the rapid range expansion that occurred in warmer regions. Alternatively, a more coldtolerant ecotype could be introduced from their native range in Asia.
Extreme cold temperatures in their native range are similar to those in the northern extent of hemlock trees in eastern North America.