Environmentally triggered variability in the genetic variance–covariance of herbivory resistance of an exotic plant Solidago altissima

Abstract The variability in the genetic variance–covariance (G‐matrix) in plant resistance and its role in the evolution of invasive plants have been long overlooked. We conducted an additional analysis of the data of a reciprocal transplant experiment with tall goldenrod, Solidago altissima, in multiple garden sites within its native range (USA) and introduced range (Japan). We explored the differences in G‐matrix of resistance to two types of foliar herbivores: (a) a lace bug that is native to the USA and recently introduced to Japan, (b) and other herbivorous insects in response to plant origins and environments. A negative genetic covariance was found between plant resistances to lace bugs and other herbivorous insects, in all combinations of garden locations and plant origins except for US plants planted in US gardens. The G‐matrix of the resistance indices did not differ between US and Japanese plants either in US or Japanese gardens, while it differed between US and Japanese gardens in both US and Japanese plants. Our results suggested that the G‐matrix of the plant resistance may have changed in response to novel environmental differences including herbivore communities and/or other biotic and abiotic factors in the introduced range. This may have revealed a hidden trade‐off between resistances, masked by the environmental factors in the origin range. These results suggest that the stability of the genetic covariance during invasion, and the environmentally triggered variability in the G‐matrices of plant resistance may help to protect the plant against multiple herbivore species without changing its genetic architecture and that this may lead to a rapid adaptation of resistance in exotic plants. Local environments of the plant also have a critical effect on plant resistance and should be considered in order to understand trait evolution in exotic plants.


| INTRODUC TI ON
Plants are usually attacked by multiple herbivores, and the herbivore communities differ geographically (e.g., Anstett, Naujokaitis-Lewis, & Johnson, 2014;Craig, 2016;Strauss & Irwin, 2004). Therefore, different plant populations are subjected to different local selective pressures by herbivores. Plants either evolve resistance to specific herbivores through a pairwise (co)evolutionary arms race or respond simultaneously to multiple herbivores through diffuse (co)evolution (Agrawal, 2005;Berenbaum & Zangerl, 2006;Strauss, Sahli, & Conner, 2005). Diffuse (co)evolution occurs if the presence of a third species indirectly alters the magnitude and/ or direction of natural selection or the response to selection in a pair of interacting species (Iwao & Rausher, 1997;Stinchcombe & Rausher, 2001;Strauss et al., 2005). In other words, regardless of how many species are involved, if (co)evolutionary interactions between the pair are not influenced by the presence of a third species, the interactions are considered pairwise rather than diffuse (Wise & Rausher 2013). One of the major factors that contributes to the trajectory of diffuse (co)evolution in plant resistances is the genetic correlations among plant resistances toward different herbivores. For example, a negative genetic correlation between two plant resistance traits can constrain evolutionary responses of the traits against two herbivores while each herbivore species can also impose directional selection toward higher resistance.
These variances and covariances in plant resistance can be summarized in a genetic variance-covariance matrix (G-matrix). This is a fundamental parameter in evolutionary quantitative genetics because G-matrix constrains evolutionary responses of plants to natural selection (Lande, 1979). However, the extent of G-matrix constraints on adaptation in plant resistance is still largely unknown.
The breeder's equation in quantitative traits is defined as follows: ∆z = Gβ, where ∆z is the changes in the mean value of traits across one generation, G is the G-matrix, and β is the vector of selection gradients (Lande, 1979;Lande & Arnold, 1983). Thus, phenotypic responses would vary owing to the G-matrix even under equal selection gradients. Exploring population divergence patterns in G-matrices can improve our understanding of the mechanisms underlying phenotypic divergence and how the response to selection might be constrained by genetic architecture during adaptive evolution (Eroukhmanoff, 2009;Schluter, 1996;Steppan, Phillips, & Houle, 2002). Several studies have found that the structure of the G-matrix is stable (i.e., no changes in its direction and strength of variance and covariance) over ecological timescales of a few generations, while it is unstable in evolutionary timescales of hundreds of generations (Björklund, 1996;Cano, Laurila, Palo, & Merilä, 2004).
On the other hand, recent studies have indicated that G-matrix can change very rapidly over a few generations (Eroukhmanoff & Svensson, 2011;Phillips, Whitlock, & Fowler, 2001;Sgro & Blows, 2004;Uesugi, Connallon, Kessler, & Monro, 2017). In addition to the genetic changes in the G-matrix, environmental conditions such as temperature and interacting species may also influence the structure of the G-matrix because the environment affects gene expression (Bégin & Roff, 2001;Czesak & Fox, 2003;Wood & Brodie, 2015).
Thus, environments are important in determining the correlations among resistances (Wood & Brodie, 2015). Understanding whether the G-matrix in plant resistance is stable in different local populations and their environments is critical for making inferences about the evolution of plant resistance to herbivores. Reciprocal transplant experiments in which plant individuals from more than two populations are grown in their own environment and in the environments of the other populations, allow a clear separation of genetic and environmental effects on traits (Kueffer, Pyšek, & Richardson, 2013;Nuismer & Gandon, 2008). These experiments provide the ability to test whether G-matrix varies among different populations or environments.
Plant invasions are excellent systems for studying how G-matrix varies genetically or environmentally across local populations under different abiotic and biotic selection regimes (Eroukhmanoff & Svensson, 2011). Invasive plants may evolve rapidly in response to changes in biological interactions (Mitchell et al., 2006). The underlying mechanisms of this rapid evolution in resistance during invasion may be based on both the response to altered selection of a single resistance trait and changes in the G-matrix of multiple resistance traits. Franks et al. (2012) compared secondary compounds in Melaleuca quinquenervia between its native range in Australia and invasive range in the USA and found that the genetic variances and covariances were reduced in the invasive range. They also found differences in the G-matrices of plants between the invasive and native populations. The loss of genetic variation may be evidence of recent adaptation as well as founder effects (Franks et al., 2012). In addition, the reduction of negative genetic correlation may be a result of release from an evolutionary constraint affecting multiple resistance traits. This may potentially lead to the rapid evolution of plant resistance during invasion. Therefore, it is critical to explore the genetic and environmental variation in the G-matrices between native and introduced plant populations in order to understand the rapid evolution of introduced genotypes in the invasive ranges.
Tall goldenrod, Solidago altissima, is an herbaceous perennial native to old-field habitats in North America. Several studies have found large genetic variability in goldenrod's resistance to insect herbivores (Craig, Itami, & Craig, 2007;Maddox & Root, 1987;Uesugi, Poelman, & Kessler, 2013;Utsumi, Ando, Craig, & Ohgushi, 2011). In Japan, S. altissima was introduced approximately 100 years ago and it has extensively invaded abandoned fields across the country (Shimizu, 2003). The lace bug, Corythucha marmorata (Hemiptera: Tingidae) (Figure 1), is one of the major herbivorous insects feeding on leaves of S. altissima in its native range of North America (Cappuccino & Root, 1992). It was introduced to Japan in 2000 and is now a dominant herbivorous insect in Japan.
Although the specific resistant trait in S. altissima against lace bugs is unknown, secondary chemical compounds rather than physical traits are likely responsible for the resistance. This is because we did not find any relationships between physical traits (e.g., leaf trichome or leaf toughness) and resistance to lace bugs (Sakata, Yamasaki, & Ohgushi, 2016). In addition, Uesugi and Kessler (2016) found that Japanese S. altissima with low resistance to lace bugs showed lower production of leaf secondary metabolites such as diterpene acids, which may also affect resistance to other herbivorous insects. Diverse taxa of herbivorous insects were observed feeding on the plant in the USA, but very few taxa were observed in Japan (Sakata, Craig, Itami, Yamasaki, & Ohgushi, 2017). However, lace bug density was higher in Japan compared to the USA. Our previous study examining the relationship between plant resistances to lace bugs and other foliage feeding insects in multiple gardens in the USA and Japan showed an antagonistic relationship between them, which differed in strength among gardens (Sakata et al., 2018). These results suggest that the response of S. altissima to selection by lace bugs may differ between environments including differences in the herbivorous insect communities. However, it is not clear whether a negative genetic covariance actually exists and/or that degree differs among plant origins and environments.
Thus, we hypothesized that a negative genetic covariance exists between resistance to lace bugs and other herbivorous insects and that its degree differs across native and introduced ranges and aimed to test this in this study. We conducted an additional analysis of the data of the resistances of S. altissima to herbivorous insects in a reciprocal transplant experiment with multiple replicates within the native and introduced ranges (Sakata et al., 2018), and specifically asked whether G-matrices of the plant resistance differed between (a) origin populations of S. altissima in its native and introduced ranges, and (b) environments, which are reflected as the locations of the gardens, in its native and introduced ranges.

| Multiple reciprocal transplant experiment
From June to August 2013, we collected rhizome segments of S. altissima belonging to 10 genotypes from clumps at least 5 m apart from two populations in the USA (Minnesota, Kansas), and three populations in Japan (Saga, Shiga, Yamagata). Lace bugs were abundant on S. altissima in populations of Kansas, Saga, and Shiga, while they were absent or at low densities in populations of Minnesota and Yamagata (Sakata et al., 2017(Sakata et al., , 2016. The rhizome segments were planted in a greenhouse at the Center for Ecological Research, Kyoto University, Japan and at the Research and Field Studies Center, University of Minnesota Duluth (Table 1), followed by cultivation for two growing seasons to remove historical effects. In April 2015, the rhizomes were cut into 6 cm long segments with an average diameter of 5 mm; 25 ramets of each genotype were planted in pots and grown in the green house, and in June 2016, five ramets of approximately the same size from each genotype (250 plants in total) were planted in larger sized pots with potting soil and placed randomly in each of the five gardens (Table 1) (see Sakata et al., 2018 for detailed methods for cultivating the plants in the gardens).
At the end of July, the number of leaves damaged by galls, mines, and chewing damage, excluding lace bug damage (which we term "other herbivore foliage damage") were recorded for each ramet.
Lace bug herbivory can be distinguished from other insect herbivory by their yellow feeding scars. The level of lace bug damage was assessed by assigning the damaged leaves to four levels: (a) no damage, (b) <33% damage, (c) 33%-66% damage, and (d) >66% damage of total leaf area. Subsequently, we counted the number of leaves in each damage level, added the values of all four levels (which we term "lace bug damage"), and divided that with the total number of leaves to calculate the average damage value per leaf for each plant. For other herbivore foliage damage, we simply divided other herbivore foliage damage by the total number of leaves. We also counted the total number of leaves on all plants in each garden. For further analyses, we used these two values as resistance indices for herbivorous insects.

| Computation and comparisons of G-matrix in plant resistance
Genetic variances and covariances of the two resistance indices (i.e., lace bug and other herbivore herbivores) for ( we reran MCMC analyses for 65,000 steps with 15,000 discarded as burn-in, and chains were thinned by selection every 5,000 steps, yielding a total of 10 data points to estimate the distribution of G-matrices on the observed scale. Note that similar results were obtained with the posterior distribution of 1,000 matrices. As a result of estimating the G-matrices based on clonal replicates, they are broad-sense genetic parameters that include additive and nonadditive genetic effects plus shared environmental effects (Lynch & Walsh, 1998). Although using clonal replicates overestimates the genetic variances, studies have found that the breeding design does not affect the magnitude of genetic correlations between resistances (Leimu & Koricheva, 2006).
To test whether the structure of G-matrix of S. altissima has been genetically and/or environmentally altered following introduction to Japan and to explore potential constraints on trait evolution, we compared the G-matrices as follows: (a as recommended by Krzanowski (2000). To test if the matrices share the same subspace, we followed Aguirre et al. (2014) and constructed a randomized set of matrices (H). We assumed that all matrices were sampled from the same group and combined the 10 G-matrices created by the MCMC analyses for each group and randomly assigned individuals to one of the groups and constructed G-matrices from the vectors of breeding values (genotype). If the randomized and observed eigenvalues of H were the same, we concluded that the matrices share the same subspace. All statistical analyses were conducted in R 3.3.2 (R Development Core Team, 2016).

| G-matrices among plant resistances to foliar herbivores
We obtained four G-matrices in terms of all combinations of plant origins and gardens (Table 2). All covariances between resistance indices were significantly different from zero. There were no large differences between the MCMC analyses with 1,000 samples and 10 samples (  Table 2). The genetic variances of resistance to lace bugs were larger in Japanese gardens than in US gardens, whereas the genetic variances of resistance to other herbivorous insects were larger in US gardens than in Japanese gardens (Table 2).
According to the Krzanowski subspace method, the 95% HPD

| D ISCUSS I ON
The results showed that a negative genetic covariance was detected between lace bugs and other herbivorous insects, and the G-matrix of the resistance indices did not differ between US and Japanese plants either in US or Japanese gardens, while it differed between US and Japanese gardens in both US plants and Japanese plants. This suggests that the G-matrix may be stable during an invasion and variable among environments, which could be important mechanisms allowing for the rapid adaptation in invasive plants. in the genetic covariance detected in our study may be due to this categorization. However, we note that even using this categorization of herbivores the negative genetic covariance was detected.

| Negative genetic covariance between plant resistances to different herbivorous insects
Alternatively, the negative genetic covariance between herbivores may be due to behavioral avoidance of herbivory caused by other species (Ando, Utsumi, & Ohgushi, 2017;Halitschke, Hamilton, & Kessler, 2011;Poelman & Kessler, 2016). Note that even when lace bug damage was most prominent in Japanese gardens, undamaged leaves were unlikely to be scarce for other herbivores.  significance. On the other hand, in the Japanese gardens with high abundance of lace bugs, the difference in the resistance as damage rate in the US plants may have been magnified, which allowed us to detect the negative covariance. Contrarily, in the case of Japanese plants, they have recently received a directional selection only toward the increase in lace bug resistance following the recent lace bug invasion. As a result, a negative genetic covariance might be realized along the antagonistic relationship of resistances in gardens of both countries.
An alternative explanation for the difference in covariance of resistance on US plants in US and Japanese gardens is that difference in the lace bug phenology between locations may cause difference in plant quality caused by other insect herbivory. Helmberger, Craig, and Itami (2016) reported that lace bugs perform better on droughtstressed S. altissima due to mobilization of structural nitrogen increasing the nutritional quality of stressed tissues. In addition to drought stress, stress from early herbivory may positively affect lace bug performance, as early herbivory has been reported to increase nitrogen and increase herbivory by later herbivores (Danell & Huss-Danell, 1985). Because lace bugs emerge later in the season in the USA compared to Japan (Y.S. personal observation), there may be a positive effect of stress due to previous damage by other herbivores on the lace bugs in the USA.

| Stability of the G-matrix during invasion
Interestingly Alternatively, the G-matrix of the amount of each chemical compound such as the secondary metabolites may be more labile than  Another factor that might have led to the similar G-matrices for US and Japanese populations is continuous migration between populations in the two countries that may homogenize the genetic variation between ranges (Guillaume & Whitlock, 2007). However, the divergence of US and Japanese S. altissima populations was revealed by a neutral molecular genetic analysis, and the Japanese populations used in this study were genetically similar and likely shared a common origin from a single or a small number of US populations (Sakata, Itami, Isagi, & Ohgushi, 2015). It is, therefore, unlikely that migration between populations caused the similar G-matrices in the USA and Japan.

| Environmentally triggered variability in the G-matrix between native and introduced ranges
Although the structure of G-matrix did not differ between US and Japanese plants, it differed between US and Japanese gardens in both US plants and Japanese plants (i.e., as the subspace and a stable set of eigenvectors of the G-matrices were not shared between gardens). This indicates that environmental differences influence the magnitude and the sign of G-matrix. Together with our former study (Sakata et al., 2018), the genetic variances of both lace bugs and other herbivore resistances were larger in gardens with higher levels of insect damage. In fact, a higher genetic variance of resistance to lace bugs was found in Japanese gardens, and a higher genetic variance of resistance to other herbivorous insects was found in US gardens. This could be the primary mechanism producing the difference in the G-matrix between US and Japanese gardens. Although the genetic covariance between lace bugs and other herbivorous insects was relatively larger in the Japanese gardens compared to the US gardens, the low damage rate by other herbivorous insects in the introduced range compared to the native range (Sakata et al., 2018)  Our results are consistent with the notion that environmental variation is important for structuring trait correlations (Wood & Brodie, 2015).

| CON CLUS IONS
The main reason why the G-matrices differed between environments but it did not differ between plant origins may be that the genetic variance of lace bugs became large in Japanese gardens due to extremely high lace bug density, which can cause increase in gene expression and/or induction for lace bug resistance. Strauss et al. (2005) argued that changes in community composition can alter the G-matrix of a trait under selection by one interactor in the context of diffuse evolution. Unexpectedly, our results suggest that the genetic covariance can be stable during invasion and may not impose a large constraint on the evolution of defense in invasive plants. This is likely because this stability in the trade-off between plant defenses is favored in highly heterogeneous herbivory environments. Tradeoffs may be an underlying adaptive mechanism that evolved under spatiotemporal variation in the community structure of herbivores.
In addition, the environmentally triggered variability in G-matrices of plant resistance may promote plastic adaptation that increases resistance to specific herbivores in invasive plants. Although rapid evolution in invasive plants has been reported in many species (Mitchell et al., 2006), the mechanisms producing it is still poorly understood. These two characteristics of the G-matrices that we found may be key mechanisms in invasive plants that allow them to quickly adapt to the new range. In this study, we compared the G-matrices between native and introduced ranges, based on data from multiple gardens within each range that cover a wide range of different environments. This has enabled us to determine that novel environmental factors including the herbivore communities and/or other biotic and abiotic factors in introduced ranges can alter the G-matrices of plant resistances. Moreover, this allowed us to detect a hidden trade-off between the two resistances, which may have been masked by the environmental factors such as the composition of insect community in the origin range. In future, the G-matrix of each population should be measured to more accurately understand the stability and variability of G-matrices among different environments. The local environment, including the herbivore community, has a critical effect on plant resistance, and it should be considered in order to understand the trait evolution of invasive plants.

ACK N OWLED G M ENTS
We greatly appreciate T.

CO N FLI C T O F I NTE R E S T
None declared.

AUTH O R CO NTR I B UTI O N S
TPC and YS designed and planned the experiment. YS, MI, TPC, and JKI conducted the field experiment. YS and SU analyzed the data. All authors contributed to writing and editing the manuscript.

DATA AVA I L A B I L I T Y S TAT E M E N T
All data used in this manuscript are deposited to Dryad https://doi.