Study species
Senecio vulgaris (Asteraceae) is a self-compatible annual herb that likely originated in southern Europe [38], and has been introduced to Asia, America, South Africa, New Zealand and Australia [39]. It was first reported in northeast China in the nineteenth century. Currently, S. vulgaris is distributed in northeast and southwest China and mainly inhabits ruderal and agricultural environments [40]. Genetic analyses based on microsatellite markers suggest multiple introductions of S. vulgaris to China and significant genetic differentiation among populations in China [37].
Field investigation
Field investigations were conducted in five populations sampled in Europe and 11 populations in China (Table 1). The European sampling sites are distributed in England and Switzerland. The Chinese sampling sites are distributed in Heilongjiang, Inner Mongolia and Jilin in northeast China and Yunnan, Sichuan and Guizhou in southwest China. At each sampling site, we randomly sampled more than 30 individuals and measured the three following phenotypic traits: plant height, number of branches and number of capitula.
Common garden experimental design
Mature seeds were collected from 30 groundsel plants within each of six native populations in Europe, three introduced populations in northeast China and three populations in southwest China in 2005 (Table 1). For each population, all mature seeds were mixed together and stored at 4°C in envelopes. Fifty high-quality seeds from each population were used for the common garden experiments.
Beginning on 26 April 2006, we conducted a common garden experiment at the University of Neuchatel, Switzerland, to evaluate the differences in growth and reproduction of offspring sampled from native and introduced ranges. All seeds were placed in 0.5% formaldehyde for 15 min, rinsed with water and then immersed in 1 mg/L gibberellin for 24 h. After the seeds were cleaned with water, all 50 high-quality groundsel seeds from each population were sown in 50 pots (one seed per pot). Thus, the experiment consisted of 600 pots separated from one another by 0.5 m.
To evaluate plant growth and reproduction in different stages, we measured the height and number of capitula of each plant 45, 51, and 57 days after the start of the experiment. Once a capitulum matured, the whole individual was harvested. At harvest, we measured the plant height, number of capitula. Then, the plant was oven dried at 80°C for 48 h to a constant weight, and we measured the dry mass.
Statistical analysis
To test the effects of range (native or introduced) and latitude on the phenotypic traits, we fitted linear mixed models by using plant height and dry mass as response variables and fitted generalized linear mixed models by using number of branches and number of capitula as response variables. In both model fittings, range and latitude were set as fixed effects without interaction, and population was set as a random effect. These models were defined as ‘range and latitude’ models.
To specifically test whether latitude affected estimates of the difference between native and introduced populations, we compared two models: 1) the previously described ‘range and latitude’ model including range and latitude without interaction and 2) the ‘range-only’ model excluding latitude. The -2 residual log-likelihood of the two models were compared and tested with X2 test (chi-square test) in linear mixed model fitting. The deviance of the two models were compared and tested with X2 test in generalized linear mixed model fitting. We used the parameter of range estimated in each model as the effect size for range and compared this value between the two models.
The sampling sites in China aggregate into two regions (northeast and southwest China). Therefore, we replaced the two fixed effects (range and latitude) with one fixed effect (region: Europe, northeast China or southwest China) in the ‘range and latitude’ models to compare the difference between native and introduced populations. These models were defined as ‘region’ models. All analyses were performed in R [41]. The linear mixed models were fitted using the function ‘lme’ and the generalized linear mixed models were fitted using the function ‘glmer’ in the package lme4.