Evolution of invasion syndrome in invasive goldenrod is not constrained by genetic trade-offs
Data files
Jun 07, 2024 version files 473.04 KB
-
EvolApp_2023_Gmatrix_data.csv
-
EvolApp_2023_goldenrod_ped_indiv.csv
-
README.md
Abstract
A suite of plant traits is thought to make weed populations highly invasive, including vigorous growth and reproduction, superior competitive ability, and high dispersal ability. Using a breeding design and a common garden experiment, we tested whether such an “invasion syndrome” has evolved in an invasive range of Solidago altissima, and whether the evolution is likely to be genetically constrained. We found an overall shift in invasive phenotypes between native North American and invasive Japanese populations. The invasive populations were taller and produced more leaves, suggesting a superior ability to exploit limited resources. The populations also produced more allelopathic compounds that can suppress competitor growth. Finally, invasive populations produced more seeds, which are smaller and are released from a greater height, indicating a potential for superior dispersal ability than the native populations. Quantitative genetics analyses found a large amount of additive genetic variation in most focal traits across native and invasive populations, with no systematic differences in its magnitude between the ranges. Genetic covariances among three traits representing invasion strategies (leaf mass, polyacetylene concentration and seed size) were small. The R metric, which measures the effect of genetic covariances on the rate of adaptation, indicated that the covariance neither constrains nor accelerates concerted evolution of these traits. The results suggest that the invasion syndrome in S. altissima has evolved in the novel range due to ample additive genetic variation, and relatively free from genetic trade-offs.
README: Evolution of “invasion syndrome” in invasive goldenrod is not constrained by genetic trade-offs
https://doi.org/10.5061/dryad.n8pk0p34b
This dataset contains multivariate phenotypic traits of Solidago altissima from its native and invasive ranges measured in a common garden experiment. Plants were sourced from three populations per range, crossed with a partial diallel design, and phenotyped in a common garden. We found evidence of evolutionary increase in multiple traits associated with invasiveness, whose evolution did not seem to be constrained by genetic tradeoffs among traits.
Description of the data and file structure
EvolApp_2023_Gmatrix_data.csv contains trait data collected for each of the 1324 individual samples of S. altissima plants. Rows are separated into Type = "Plant" or "seed". "Plant" rows show measurements of all traits for each individual plant, except for "seed.area". Because seeds were bulk germinated for each maternal family, we do not know the size of seeds for particular individual plant that was grown in the greenhouse. "NA" in the seed.area column indicates the lack of seed data in "Plant" row. "Seed" rows present only seed.area, measured for 3 individuals per maternal family. All other trait values are shown as "NA".
EvolApp_2023_goldenrod_ped_indiv.csv contains pedigree information for individual plants used in the study.
Column abbreviation for EvolApp_2023_Gmatrix_data.csv:
population | source populations where seeds of S. altissima was originally collected. |
---|---|
origin | Native or Introduced |
animal | individual plant ID |
cross_block | block in which F1's were crossed. There were 5 individuals within a block, all fully crossed without self-fertilisation. |
sire | identity of pollen donor |
dam | identity of maternal plant |
rep | replicates (1 or 2) per maternal family. For each cross, we planted 2 individuals. |
fam | identity of cross family, where offspring of reciprocal crosses belong to the same family |
recipro | identity of maternal family. |
tray | Tray number where F2's were grown in the common garden. 17 pots were placed in one tray. |
tray.pos | Position of plants within each tray, either E = edge, or C = centre. |
bay | Bay (or block) number within a greenhouse. |
growth.rate | Relative growth rate was determined as ([height at week 10 – week 8]/14 days). "NA" indicates data not available due to early death of the individual plant or missing data. |
height | final height at harvest in cm. "NA" indicates data not available due to early death of the individual plant or missing data. |
sla | Specific leaf area: [fresh leaf area/dry mass] of 3 leaves per plant. "NA" indicates data not available due to early death of the individual plant or missing data. |
leaf.mass | Leaf dry biomass at harvest, in g. "NA" indicates data not available due to early death of the individual plant or missing data. |
rhizome.mass | Rhizome dry biomass at harvest, in g. "NA" indicates data not available due to early death of the individual plant or missing data. |
flower.mass | inflorescence dry biomass at harvest, in g. "NA" indicates data not available due to early death of the individual plant or missing data. |
flower.day | Number of days from transplant to first flowering. "NA" indicates data not available due to early death of the individual plant or missing data. |
flower.duration | Number of days between the date of first flower opening and the last flower senescing. "NA" indicates data not available due to early death of the individual plant or missing data. |
polyacetylene | Concentration (in UV absorbance) of polyacetylenes in root tissue, measured using HPLC, per mg of fresh weight. "NA" indicates data not available due to early death of the individual plant or missing data. |
seed.area | Seed size, measured as area of the seed in photographs. "NA" indicates data not available for "Plant" type data. |
Type | Type of data. Plant = traits measured at individual plant level, and Seed = seed area measured for three seeds per cross family. |
Methods
Study system
The tall goldenrod, Solidago altissima L. (Asteraceae), is a perennial forb native to eastern North America (Werner et al., 1980) and is a dominant species of old fields and disturbed habitats. Solidago altissima was first introduced into Japan in the late 1890s as an ornamental plant but only became widespread across the country since the 1980s (Fukuda, 1982). Current molecular data suggests that invasive Japanese S. altissima populations are likely to be introduced primarily from south-eastern North America (Sakata et al., 2015), which may overlap with the distribution of a putative S. altissima variety, pluricephala that has also invaded other parts of Asia (Semple et al., 2015). The Japanese populations seem to be founded by multiple introduction events, resulting in similar levels of genetic diversity within a population at neutral markers as in the native populations (Sakata et al., 2015; Uesugi et al., 2020). Moreover, broad-sense genetic variation for herbivore resistance traits did not differ between the US and Japanese populations (Sakata et al., 2020), suggesting that evolution in introduced Japanese populations may not be genetically constrained. However, additive genetic variation and covariation among traits associated with competitive and dispersal abilities has not been examined previously.
In its native range, a diverse group of specialist and generalist herbivores keeps the S. altissima populations in check (Root, 1996), whereas in the invaded Japanese range, the plant generally escapes herbivory, except from Uroleucon nigrotuberculatum (a Solidago specialist aphid introduced in 1990’s, Cappuccino, 1987; Sugimoto & Matsumoto, 2000) and Corythucha marmorata (an Asteraceae specialist lacebug introduced to Japan in 2000, Kato & Ohbayashi, 2009; Sakata et al., 2014). Upon establishment in a field patch, S. altissima can rapidly grow tall and shade out neighbouring plants (Root, 1996), while suppressing germination and growth of competitor plants through allelopathy (Kobayashi et al., 2008; Johnson et al., 2010; Uesugi et al., 2019). It produces several compounds of polyacetylenes in roots, which are released into the soil at a concentration known to inhibit the growth of several other plant species (Kobayashi et al., 2008; Johnson et al., 2010; Uesugi & Kessler, 2013; Uesugi et al., 2019). Within a patch, S. altissima primarily spreads through rhizome production (Hartnett & Bazzaz, 1983), but colonization of new patch relies on wind-born seeds that are produced in abundance (Uesugi et al., 2020).
Experimental design
Seed sources and quantitative genetic breeding design:
Seeds were collected in 2016 from three native populations in the south-eastern United States within the known range of S. altissima var. pluricephala where invasive Japanese populations are likely to have originated (Durham, NC, Spartanburg, SC, and Murrells Inlet, SC; Sakata et al., 2015, Suppl. Table 1). Three invasive populations in Japan were sampled across a similar latitudinal range (Utsunomiya, Shizuoka, and Otsu). A principal component analysis (PCA) of variation in 19 WorldClim climatic variables among sampled populations indicated that invasive Japanese populations generally experience wetter summers and drier winters than native US populations (PC1, explaining 52.9 % of variance). Within each range, our study populations varied across a temperature gradient (PC2, explaining 35.5% of variance), with both ranges spanning similar PC2 coordinates (Suppl. Fig. 1, Suppl. Table 1 & 2).
Within each population, we collected seeds from ~50 maternal plants, which grew at least 10 m apart in the field to minimize the probability of sampling the same genetic individual multiple times. We germinated ~10 seeds per maternal plant in a common greenhouse environment (temperature at 18-30 °C and relative humidity at 60%) at Monash University, Victoria, Australia, and grew a single individual per maternal family in 12-cm pots with potting mix with Osmocote. Each population resulted in 18-35 parental (P1) individuals (Suppl. Table 1).
P1 individuals from each population were split into groups of five individuals (4-7 replicated groups per population) and crossed within a group using a partial diallel design with reciprocals but no selfing (Lynch & Walsh, 1998). Prior to the anthesis, we bagged branches of inflorescence (~ 5cm long) with nylon bags to avoid accidental pollen transfer. When > 50 % of the flowers on a branch had opened, we removed the branch of the sire plant, and rubbed it against an intact branch on a dam plant. Each plant sired four other individuals within the group and received pollen from the four. We also made a self-pollination cross as a negative control (because S. altissima is a self-incompatible species, self-crosses did not produce any viable seeds). This resulted in total of 760 crosses (20 crosses per group x 38 groups) across the populations.
Common garden experiment:
In 2017, three viable achenes from each cross were selected, and photographed under a dissecting scope for later examination of achene size (see “Seed dispersal ability” below). Seeds of the F1 generation were germinated as above, and two seedlings per cross were transplanted to individual 10-cm diameter pots. A lack of germination from some crosses resulted in a total of 1324 experimental individuals. Potted plants were placed in evenly spaced trays of 17 pots, across six benches (“blocks”) in the greenhouse as above. Trays were rotated weekly to mitigate effects of microclimate within the greenhouse until flowering.
Trait measurements
We measured 10 traits that are thought to mediate plant invasiveness, including growth rate, maximum stem height, specific leaf area (SLA), leaf mass, rhizome mass, inflorescence mass, days to first flower, flowering duration, root polyacetylene concentration, and seed size.
Growth, morphology, and phenology:
We estimated the relative growth rate of individual plants by measuring height on week 8 and week 10 after transplanting, which corresponded with the period of rapid vertical growth. Relative growth rate was determined as ([height at week 10 – week 8]/14 days). In week 15, we collected three fully expanded leaves from each plant at the height of 60 cm above the base. Each set of three leaves was scanned together, dried for 48 hours at 50 °C and weighed. Leaf area was measured using ImageJ (version 1.51), and specific leaf area (SLA) was calculated as [fresh leaf area/dry mass].
To estimate flower onset and flowering duration, we checked plants daily during the flowering period (week 14 through 22) and marked the dates we observed the first flower and the last senesced flower. Flowering duration for each plant was calculated as days from flowering onset to final day of flowering. We harvested plant biomass between weeks 22 and 25 as individual plants finished flowering. Aboveground biomass was separately harvested for leaves, stems, inflorescence, and ramets. We used inflorescence mass as a proxy for seed production (Root, 1996). We were unable to directly estimate seed production because S. altissima is an insect-pollinated, self-incompatible species, and does not naturally set seeds in the greenhouse. Belowground rhizomes were harvested by removing roots and washed in water. All harvested samples were dried for 48 hours at 50 °C before weighing.
Polyacetylene analysis:
Root samples were collected for polyacetylene analysis in week 15 by removing a subsample of root tissues from each plant. Root samples were flash frozen in liquid nitrogen and stored in -80 °C for later analysis. Following Uesugi et al. (2019), approximately 200 mg fresh weight of root tissue per sample was crushed with mortar and pestle in liquid nitrogen, sonicated in extraction buffer (1ml of 90% methanol) for 6 min, and left in the dark at room temperature for 24h. Samples were centrifuged, and 0.5 ml aliquot was filtered with 0.45 mm syringe filter. The samples were analysed with high-performance liquid chromatography (HPLC) at Monash University using Agilent Infinity 1260 equipped with C18 reserve-phase column (ED-C18, 2.7μm, 150×3.0mm). The elution method was: 0–5min, 0–20% of acetonitrile; 5–25min, 20–95% of acetonitrile and 25–30min, 95% of acetonitrile, with a flow rate of 0.5ml min–1 and injection volume of 2µl. Peaks of polyacetylene compounds were identified using UV spectra and quantified at 254 nm (Uesugi et al., 2019).
Seed dispersal ability
Dispersal ability of wind-dispersed seeds increases with the time they spend airborne (i.e., longer time taken to drop to the ground; Cody & Overton, 1996; Tabassum & Bonser, 2016). To test if seed morphology correlates with drop time, we conducted a separate experiment, where we measured the drop time, achene size and pappus length for 60 randomly selected seed samples. We measured the time taken for a seed to fall a 50 cm distance down a glass cylinder (10 cm diameter) by videotaping and analysing the footage with BORIS software (Friard & Gamba, 2016). Each seed sample was dropped twice, and the average was used for subsequent analysis. The seed was then photographed under dissecting scope, and the achene area (hereafter referred as ‘seed size’ for simplicity) and the maximum pappus length were estimated using ImageJ.
Seed size and pappus length were positively correlated (linear regression: coefficient = 0.56, t = 4.7, P < 0.0001). A multiple regression analysis with model selection using partial F-test showed that seed size alone was the best predictor of seed drop time (coefficient = -0.75, t =2.8, P = 0.006, R2 = 0.14). Thus, we used seed size as a proxy for seed dispersal ability (i.e., smaller seed size increases seed dispersal ability) in the subsequent analyses.
Seed size of plants used in the common garden experiment was estimated using the photographs taken prior to seed germination (see above) using ImageJ. For each cross family, the size of the three largest seeds out of five were measured. Because seeds were bulk germinated for each maternal family, we could not directly link the seed measurement with a specific individual plant sample used in the experiment.
Statistical analysis
All statistical analyses were undertaken using R in the RStudio environment (version 3.4.3) (R Core Team, 2017).
Divergence in multivariate invasive phenotypes:
We tested for overall trait differences between native and invasive populations using a multivariate linear mixed model, fitted with the lmer function of the lme4 package (Bates et al., 2014). To include seed size in the multivariate model, we used the family means for all trait values in this analysis. The response variables included standardized values (scaled to mean = 0, standard deviation = 1) of 10 focal traits. We used trait, range (native or introduced), and their interactions as fixed effects, and dam and sire as random effects. In this model, we also included source populations’ climatic variables representing variation in temperature within each range (PC2 in Supp. Table 1 & 2) as a fixed covariate to control for clinal divergence within each range (Colautti et al., 2009, Woods & Sultan, 2022). In a separate model, we tested variability among populations within each range by modelling as above, but populations as nested within a range. PC2 of source populations was removed from this model, as the values were identical for all individuals within a population. Significance of fixed effects was evaluated using Wald χ2 tests.
All interaction terms that included trait effect were significant (see Results), indicating that traits were differentially affected by plant origin. Thus, we conducted univariate analyses with linear mixed models as above with range and PC2 as fixed effects, and dam and sire as random effect. We then contrasted populations in univariate models, where population was modelled as nested within a range as a fixed effect, and conducted post hoc tests using the emmeans function (emmeans package).
Estimation of univariate genetic variances:
To explore the amount of genetic variance in each of the 10 focal traits within each population, we first conducted univariate analyses by fitting a linear mixed model with ASReml-R (Butler et al., 2009). Data were scaled to mean = 0, and standard deviation = 1 prior to analysis (Bolstad et al., 2014). Additive genetic variance in each trait was estimated from pedigree information (i.e., parental identities in the crosses). We also included family identities (where offspring of reciprocal crosses belong to the same family) to estimate non-additive genetic variance, which includes variance due to dominance and epistasis (Lynch & Walsh, 1998). Block (i.e., greenhouse benches) was included as a fixed effect to control for variability in greenhouse environment.
Correlation between genetic variance and the magnitude of trait divergence:
We tested if the traits that harbour more additive genetic variance (VA) diverged more between native and invasive ranges (Opedal et al., 2023). The magnitude of trait divergence was estimated as the standardized range effect for each trait from the above analysis (“Divergence in multivariate invasive phenotypes”). The amount of VA for each trait was determined as the median of the mean VA for the six populations calculated above (“Estimation of univariate genetic variances”). We explored the relationship between divergence and VA using a linear regression analysis.
Estimation of genetic (co)variances among invasive traits:
Genetic covariances among traits can either constrain or facilitate trait divergence between the ranges. Our limited sample sizes for each population precluded us from examining the multivariate genetic (co)variance structure for all 10 traits, thus we used a subset of traits to examine genetic variances in a multivariate sense. We selected three traits representing invasive strategies—leaf mass representing exploitative competition, polyacetylene concentration for interference competition, and seed size for dispersal ability—to test whether genetic covariances among these traits would constrain evolution of invasion syndrome. These traits showed significant or marginal levels of additive genetic variation in all six populations with univariate analyses (Fig. 2), allowing us to estimate genetic covariances among them.
We generated genetic (co)variance matrices (G-matrices) for each of the six populations in a Bayesian framework using an animal model with the MCMCglmm package (Hadfield, 2010). As in the univariate analysis with ASReml-R, random effects included pedigree information and family identities, and block effect was modelled as a fixed effect. The model in matrix form was:
where y was the vector of observations on multivariate traits, was the design matrix for the fixed effect () of blocks within the greenhouse, and were design matrices for the random effects estimating additive genetic variance (; from the pedigree information) and non-additive genetic variance (; modelled as ‘family identity’ as above), respectively. The residual () was an unstructured covariance matrix containing the variances of (and covariances between) leaf mass and polyacetylenes, but because we were unable to assign seed size to individual plants (seeds were germinated in bulk for each maternal family), we fixed the covariance between seed size and all other traits to zero.
We used half-Cauchy distributions with weakly informative priors based on phenotypic variances of the three traits. Posterior distributions were sampled from 3,150,000 Markov chain Monte Carlo (MCMC) iterations sampled every 3000 iterations after an initial burn-in of 150,000 iterations. We checked convergence from plots of traces and posterior distributions, and calculated autocorrelations between samples (all were below the recommended level of 0.1, yielding an effective sample size close to 1,000 for each parameter). We used 1,000 MCMC samples to estimate the mean and 90% HPD intervals for additive genetic covariances of the three traits for each population.
Does genetic covariances constrain evolution in an invasive range?
The theory of resource allocation trade-offs predicts that evolution of invasion syndrome may be constrained by negative genetic correlations among invasive traits in the introduced range. In contrast, their positive genetic correlations would facilitate concurrent evolution of invasive phenotypes. We tested how the genetic covariance structure might affect population’s “evolvability”—an evolutionary trajectory of a population relative to the direction favoured by selection (Hansen & Houle, 2008) in the invasive range of S. altissima (hereafter, “novel selection”). A relative size of evolvability estimated with an observed full G matrix, over evolvability estimated with a partial G matrix without genetic covariances (R-score), would indicate the contribution of covariances on multivariate trait evolution (Agrawal & Stinchcombe 2009).
Following Hansen and Houle (2008), evolvability in the direction of selection gradients can be calculated as:
where is the angle between the vector representing a selection gradient, , and the evolutionary response, i.e., Based on the observed pattern of phenotypic divergence in the focal traits between native and invasive S. altissima populations, we assumed that novel selection would favour increased values of leaf mass and polyacetylene concentration, and decreased values of seed size that enhance dispersal ability when smaller. Because the novel selection gradients, , are unknown in this system, we simulated them by randomly generating a selection gradient for each trait, while setting their signs corresponding to the above expectations (Hangartner et al., 2020). To estimate the uncertainty in evolvability estimates, we used 1,000 MCMC samples of G in each population and generated for each of the G samples. The metric R (Agrawal & Stinchcombe, 2009) was calculated as: [(evolvability calculated with full G matrix) / (evolvability with all off-diagonal elements in G set to zero)]. R < 1 implies that covariances constrain the evolution of traits in the direction of selection, whereas R > 1 implies that covariances facilitate evolution. We calculated 1,000 samples of R values and tested against 1 using 90% HPD intervals.
Estimation of genetic covariances among reproductive traits:
Examining the relationship between VA and trait divergence revealed that rhizome mass and seed size each had ample VA within each population but showed relatively low magnitude of trait divergence between the ranges (see Results, Fig. 3). As a post hoc test, we explored the hypothesis that evolution of rhizome mass and seed size may be constrained by genetic trade-offs between other reproductive traits, such as days to first flower. We selected flowering phenology as a trait representing sexual reproduction, as later flowering plants tend to achieve larger aboveground biomass and produce more inflorescence (and are therefore assumed to produce more seeds) than early flowering plants (see Results). The flowering trait also had high VA within each population (except for Utsunomiya), allowing us to estimate co-variances with rhizome mass and seed size. The same analyses of genetic covariances and R-scores for this set of traits were conducted as above.