Spartina alterniflora Loisel., also referred to as smooth cordgrass, is a long-lived grass that can grow between 1 and 3 meters tall. It can flower annually with up to one inflorescence per stem (Chen et al., 2021). Highly viable seeds enable S. alterniflora to spread over long distances with ocean currents (i.e., sexual reproduction), and vigorous rhizomes contribute to local expansion (i.e., clonal growth) (Daehler and Strong, 1994). Spartina alterniflora originates from the Atlantic and Gulf coasts along the eastern coasts of the Americas but has evolved into a highly problematic invader worldwide (Strong and Ayres, 2013). In 1979, S. alterniflora was brought into China for the first time and has spread rapidly (Xu and Zhuo, 1985). Now, S. alterniflora is found along the entire coast of China, making it the largest area invaded by a species in the world (Zhang et al., 2017).

In 2018, we collected seeds from seven S. alterniflora populations from September to October (end of growing season) at different geographic sites from 21°N to 38°N, covering the entire distribution of coastal China ( Figure 1 ; Supplementary Table S1 ). Because S. alterniflora populations in China resulted from a single genetic admixture occurrence which resulted in multiple novel recombinant genotypes (Qiao et al., 2019), each population was treated as a single geographic genotype. At each site, we selected two subsites, separated by 2-3 km, and gathered seeds from five 0.5 m × 0.5 m quadrats at each subsite. The quadrats were spaced at least 30 meters apart and derived from distinct clones, thus treating the seeds gathered from each quadrat as separate seed families. In total, we collected mature, whole inflorescences from 70 quadrats. Although seeds from the same plant may only be half-sibs, such collections are a well-established method for examining the impact of genetics on traits. We individually placed the filled seeds from each quadrat into sealed plastic bags containing 10 PSU seawater and stored them at 4°C.

To assess plasticity and integration of functional traits of S. alterniflora, two common gardens were built at the S. alterniflora distribution margins in China. They were located at Guangdong (GD, low latitude, 21.12°N, 110.31°E) and Shandong (SD, high latitude, 37.69°N, 118.84°E), respectively, in which we cultivated in parallel plants of the same seed families collected in the seven populations of S. alterniflora ( Figure 1 ; Supplementary Table S1 ). Precipitation may not be the most important abiotic factor influencing plant growth and distribution in regularly inundated coastal wetland (Pennings and Bertness, 2001). Additionally, the temperature could be the most critical factor to S. alterniflora growth (Liu et al., 2020a), development (Chen et al., 2021), and reproduction (Liu et al., 2020b), because temperature has direct effects on physiological function (Pau et al., 2011). Therefore, we particularly wanted to focus on the effects of differences in temperature between gardens, and minimized differences in other environmental conditions, such as precipitation, herbivory, and edaphic condition. To shield the plants from rain, we covered the top of the common gardens with transparent plastic film to shield the rain. To keep out insect pests, we also covered the lower parts of the sides with a 1-m high plastic film and the remaining parts with insect-proof gauze. The latter allowed for sufficient air circulation to keep the ambient temperature consistent with the temperature of the surrounding environment. We initiated the germination of seeds from every family in March 2019, the seeds were nurtured in a greenhouse at Xiamen in Fujian (24°N) until they reached a height of 7-8 cm. Subsequently, we transferred the seedlings to their respective garden locations. Each common garden was equipped with 10 individual rectangular plastic pools (dimensions: 1.1 m in length, 0.8 m in width, and 0.3 m in depth). Inside each pool, there were 7 plastic pots (dimensions: 18 cm in diameter, 24 cm in height) filled with a mixture consisting of 50% Jiffy’s peat substrate (Jiffy Products International BV, Moerdijk, Netherlands) and 50% vermiculite by volume. Two seedlings from each of the ten seed families of each of the seven populations was randomly assigned to two plastic pots in each pool (2 seedlings × 10 seed families × 7 populations × 2 gardens = 280 seedlings). The pools were filled with water that we adjusted to 10 practical salinity units (PSU) by dissolving sea salt into it, and the water level in the pots was maintained at the same height as the soil level. Fresh water was added to the pools every second day, and the salt water was entirely replaced monthly to keep the salinity stable. Above conditions were within the range experienced for salt marshes invaded by S. alterniflora, although it does not completely mimic the conditions in nature, the continuously waterlogged soil corresponds to what the plants experience in nature due to regular flooding. In October 2019 (end of growing season), we measured the plants (see the section ‘Trait measurements’).

We measured four traits that relate to performance and fitness of S. alterniflora: two vegetative growth traits (plant height and shoot density), one key phenological trait (days until flowering), and one reproductive trait (inflorescence biomass). To record the date on which the first shoot in each pot began flowering, we checked the pots every four days during the entire growing season (April to October), and we calculated the first flowering day represented by the days elapsed since January 1st (Chen et al., 2023). A shoot was considered flowering once the inflorescence had protruded from the topmost leaf and visible pollen was present (Chen et al., 2021). In October, in each pot, we tallied the shoots that were taller than 25 cm and calculated shoot density by dividing the shoot count by the area of the pot. There are often many very short shoots in S. alterniflora stands, yet their contribution to overall standing biomass is negligible (Morris and Haskin, 1990; Liu et al., 2020b). These shoots were almost never able to complete their life history in this study. In each pot, the height of the three tallest shoots was measured. The inflorescences of these shoots were then collected, dried at 60°C until their mass remained constant, and weighed (Liu et al., 2020b). Therefore, the inflorescence biomass was on a per capita basis, i.e., averaged across three inflorescences from the tallest stems. The traits data of two pots from the same seed family were averaged before statistical analyses.

Local environmental conditions at the original sites could have driven the variation in phenotypic plasticity. Therefore, we obtained long-term climatic data (including MAT, AGDD, MCDT, MWDT and TAR as listed in Supplementary Table S1 ) for every sampled site covering the years from 1981 to 2020, which were sourced from the China Meteorological Data Service Center (CMDC, http://data.cma.cn) (Chen et al., 2023). This duration covers nearly all of the years that S. alterniflora spread vigorously and occupied almost the entire Chinese coastline following its introduction in 1979. We calculated Pearson coefficients for every pairwise combination of the climatic variables to determine collinearity among them, and to reduce the collinearity among these variables, we performed the principal component analysis (PCA) on the climatic variables. Finally, within each common garden, we placed a HOBO temperature logger (MX2202) at a height of one meter above the ground at the garden’s center to record air temperature every 10 minutes throughout the experiment. The recorded temperature data were used to calculate the five variables above ( Supplementary Table S1 ) during the experiment for both common gardens.

We assessed normality via Shapiro-Wilk’s test and homogeneity of variance via Levene’s tests. Since results of these tests were not well satisfied, log[x]-transformed data was applied in following analyses when necessary. To test whether inflorescence biomass had higher plasticity than the other traits (hypothesis 1), firstly, we investigated the phenotypic differences of each trait between common gardens and among populations, via using two-way ANOVAs incorporating common gardens and populations and their interaction as fixed effects. Secondly, phenotypic plasticity of each trait was calculated as phenotypic plasticity index (PIv) for each seed family, PIv is calculated as the difference between the maximum and minimum values between gardens, divided by the maximum value; the PIv varies from 0 to 1, where 0 denotes an absence of plasticity and 1 denotes the greatest plasticity (Valladares et al., 2007; Ren et al., 2020). This approach is a robust, simple and widely used index when two environments are considered (Valladares et al., 2006). We then applied one-way ANOVAs followed by Tukey-Kramer HSD post-hoc tests to compare the PIv values among the four traits including data from all seven populations. Finally, in order to accounts for population variations in PIv values, we created a mixed effect model for PIv values via R function “lmer” from the “lme4” package (Bates et al., 2015). This model employed trait, population, and interaction of trait and population as fixed factors, and treated seed family nested within each sampling subsite as a random effect.

To test whether phenotypic plasticity was stronger for populations from higher latitudes than other ones (hypothesis 2), firstly, we detected whether there was variation in plasticity among populations, via creating mixed effect models for each measured trait with garden site, latitude of origin, and interaction of garden site and latitude of origin as fixed factors, and subsite and seed family as random effects (due to multiple levels being sampled in a single site). Secondly, we regressed the PIv of each trait on latitude of origin. Further to compare the slopes of these regressions, we performed a linear regression model with PIv as the response variable and interaction between latitude of origin and trait as the predictor variable. Finally, to test the effects of local environmental conditions on traits plasticity (hypothesis 2), we regressed the PIv of each functional trait with climate PC1 at sites of origin.

To test whether climatic conditions contribute more to phenotypic plasticity than phenotypic integration (hypothesis 3), we performed regression analyses with phenotypic plasticity as dependent variable and phenotypic integration as independent variable in both gardens. The degree of phenotypic integration for each trait was assessed by counting the significant correlations (P < 0.1; Pearson’s correlation) it had with every other trait. In this study, for each of the seven populations in both gardens, we computed phenotypic integration per trait (n = 4 traits × 7 populations × 2 gardens = 56). Mean values of phenotypic plasticity of each trait for each population were log-transformed before analysis (log[x]) to achieve better normality in error distribution and consistency in variance.

We performed all analyses with R statistical software (R Development Core Team, 2018).

Our results showed that all measured traits exhibited considerable phenotypic variation between common gardens. Plasticity differed by trait, and we found the highest plasticity in inflorescence biomass, which serves as an indicator of fitness (Johnson et al., 2010). Higher plasticity in traits that contribute to fitness may increase environmental tolerance and thereby widen the niche breadth (Poorter et al., 2012; Castillo et al., 2014). However, we found relatively low plasticity in the first flowering day. Flowering phenology is quite plastic and usually speeds up at higher temperatures (Menzel et al., 2006; Richardson et al., 2006; Augspurger and Zaya, 2020), surprisingly, its plasticity was low for S. alterniflora. Having said this, many studies on variation in phenology under climate warming have shown that it is particularly the spring-flowering species that speed up their phenology, not the autumn-flowering species (like S. alterniflora) (Miller-Rushing and Primack, 2008; Alecrim et al., 2023). On the other hand, physiological adaptations to extreme climatic conditions, such as high temperature in the low-latitude garden in the present study may constrain S. alterniflora growth (Liu et al., 2020a), thus reduce the need for plasticity in phenological traits (Cooper et al., 2019). Besides, we found plant height and shoot density had similar plasticity as time to flowering.

We found phenotypic variance increased with latitude, there were positive relationships between plasticity of plant height, shoot density, and inflorescence biomass with population latitude. These results were consistent with the CVH, which stated that organisms from higher latitudes have greater phenotypic plasticity (Molina-Montenegro and Naya, 2012). Differentiation in plasticity among populations may indicate genetic control of plasticity in invasive S. alterniflora to a certain extent. We found significantly stronger evidence for genetic based variation in phenotypic plasticity of the inflorescence biomass than plant height and shoot density. Previous common garden experiments found that reproductive traits of S. alterniflora exhibited contrasting latitudinal clines between common gardens, but the plant height or shoot density showed similar clines (Liu et al., 2017; Qiao et al., 2019). These results also indicated stronger latitudinal variation in plasticity of reproductive traits than the vegetative ones. In this study, the significant clinal variation in plasticity of inflorescence biomass indicated a potentially adaptive phenotypic plasticity in this trait. However, this clinal variation in plasticity was not ubiquitous among traits; we found no latitudinal cline in the first flowering day plasticity. Results of mixed models further supported this finding, indicating significant interactions between garden site and population latitude affecting plant height, shoot density, and inflorescence biomass, but not first flowering day. Studies of phenological plasticity have yielded mixed results. For example, Ren et al. (2020) also found no latitudinal cline in the PIv of flowering time, however, Cooper et al. (2019) found that phenological plasticity tended to decrease with increasing latitude. These contrasting results may be due to differences in plant life forms and habitats. In our study, we found no difference in flowering time plasticity across latitudinal populations. Although many studies have suggested the high genetic diversity of S. alterniflora populations in China (Wang et al., 2012; Qiao et al., 2019), the genetic differentiation for flowering time plasticity across populations may be not significant, as reported in another Spartina species, Spartina densiflora (Castillo et al., 2018).

Temperature at the site of origin was the primary factor affecting phenotypic variation. Spartina alterniflora, in the United States (native range), occurs at higher latitudes than in China (invasive range) (Liu et al., 2020b). In China, however, S. alterniflora experienced greater environmental fluctuations at high latitudes than the native range (higher TAR) (Liu et al., 2020a; Chen et al., 2023). We found that greater fluctuation in temperature (higher TAR) at higher latitudes was indeed associated with greater plasticity in plant height, shoot density, and inflorescence biomass of northern populations in invasive China. The northern populations with higher phenotypic plasticity exhibited higher overall performance compared to those with lower plasticity. Phenotypic plasticity is often regarded as highly beneficial for plants (Baker, 1974), as it is believed to enhance environmental adaptability, or fitness homeostasis (Valladares et al., 2014). This suggests that phenotypic plasticity of S. alterniflora is a potentially adaptive response to novel environmental conditions in the invasive range. Additionally, lower temperature (i.e., mean annual temperature, growing degree days, coldest daily temperature, and warmest daily temperature) may also relate to higher plasticity of northern populations, because populations that experience predictable winter freezing may show greater plasticity (Cooper et al., 2019). Although S. alterniflora occupies lower latitudes in the invasive range than in the native range, the lower latitude portions of both ranges have similar ambient temperatures (Liu et al., 2020a). Climatic conditions in these low latitudes exceed the optimum temperature for vegetative growth and may limit the aboveground activity of S. alterniflora (Liu et al., 2020a). Therefore, reduced plasticity in southern S. alterniflora populations may be due to physiological constraints related to poor heat tolerance, as stress can limit plant phenotypic plasticity because of higher cost (Stotz et al., 2021). We found no relationships between plasticity of the first flowering day and climatic variables in S. alterniflora, indicating no local adaptation, as a previous study has found parallel latitudinal clines in first flowering day of this species in different common gardens (Chen et al., 2023). Such patterns suggest that different genotypes of S. alterniflora maintain similar degree of phenotypic response to environmental variation. This variation in plasticity should be subject to selection.

Understanding local adaptation to past long-term climatic condition allows us to contextualize contemporary ecological variation and predict distribution dynamics under changing environments (Woods et al., 2012). Latitude represents a proxy for multiple variables, and plastic responses of plants have been documented with respect to many abiotic factors, including temperature, precipitation, and soil (Dudley, 2004). Temperature is a key factor in affecting plastic adaptation along clines (Atkin et al., 2006) and has received particular interest in the context of global warming. Since our results suggested that both low and fluctuating temperatures favor high plasticity, therefore, when under a continuous and steady warming in the future, the advantages of high plasticity in northern S. alterniflora populations may be reduced (Norberg et al., 2012).

Phenotypic integration plays a crucial role by enabling organisms to coordinate their traits or by affecting a single trait under given environmental conditions, thereby enhancing their acclimation ability. The four traits that measured in this study represented the key growth, development, and reproduction in plant life history, can provide a comprehensive representation of phenotypic integration in plants. Although many studies used more than five traits to calculate phenotypic integration (Pigliucci and Marlow, 2001; Zimmermann et al., 2016; Matesanz et al., 2021), they also categorized the traits into growth, development, reproduction, and so on. There was a weak effect of phenotypic integration on phenotypic plasticity of S. alterniflora in this study. Changes in environmental conditions can affect both the trait values and trait-trait correlations (Wood and Brodie, 2015). We found no substantial association between phenotypic integration and plasticity in both common gardens. Only a marginally positive relationship between integration and plasticity in the low-latitude garden. High temperature in the common garden at low-latitude could be stressful for S. alterniflora, such positive effect could be essential for plants to deal with stressful environment. Greater phenotypic traits integration contributed to higher level of trait plasticity, thereby enhancing the flexibility and acclimation of organisms in response to environmental conditions (Ghalambor et al., 2007; Zimmermann et al., 2016; Matesanz et al., 2021). A previous study has suggested that the effects of internal constraints to plasticity depend on environmental conditions, with stronger effect within stressful environments. Such phenomenon may be due to the physiological limits on plants growth (Steinger et al., 2003). Besides, phenotypic variation in traits within a given environment may also diminish the effect of phenotypic integration on plasticity (Matesanz et al., 2021). Our results suggest that phenotypic integration does not universally influence trait plasticity. Nevertheless, the determinants and mechanisms underlying phenotypic plasticity are complex and far from resolved (DeWitt et al., 1998). Further empirical or theoretical evidences are needed to clarify the specific role of integration as a potential constraint on plasticity (Valladares et al., 2007).