Angiosperms are believed to have emerged initially in the tropics and expanded their distribution range polewards through diverse mechanisms, for example polyploidization-driven cold-tolerance evolution. Reversed expansion from temperate to pantropic climate through a polyploidization-driven shift in heat-tolerance remains largely unknown. Here, we found autopolyploidy in relation to the global expansion of Solidago canadensis from its temperate-climate native range in North American to hot-summer climate in an introduced range. Our cytogeographical study of 2062 accessions from 471 locations worldwide demonstrates that ploidy levels correlates negatively with latitude and positively with average temperature. An isotherm-dependent shift of the climate niches at the threshold of 20-24ºC between geo-cytotypes can be attributed mainly to autopolyploidy-driven differentiation of heat tolerance; only polyploids and not diploids are able to complete sexual reproduction, germinate and grow in the hot-summer climate of low latitudes. Ploidy-dependent fertility appears to play a key role in the hot-summer introduced range in the northern hemisphere through both pre-adaptation and rapid post-introduction adaptive evolution of delayed flowering and improved heat-tolerance during embryo development. MaxEnt model predicts continuous expansion of this weed under global change. These results provide new insights into the mechanisms governing autopolyploidy-driven backward range-expansion of plant species from temperate origins.

Field sampling

We collected seeds and leaf samples from up to 21 (median = 5) individuals from 78 locations in the native (NA) ranges (64 from the US, 14 from Canada) and 72 locations in the introduced (IN) ranges (4 from Japan, 1 from Russia, 3 from Europe, 1 from India, 63 from China), as well as two fields in China (Lianyungang, Jiangsu and Kunming, Yunnan), where the species was cultivated for sale as cut flowers (Supporting Information Table S1-3). At each site, sampling plots were separated by 10 m to minimize the chance of sampling multiple ramets from individual genets. For populations from India and 40 Chinese populations, leaf samples were collected from plants which were generated from field-collected rhizomes grown in a common garden. Leaf samples were silica-dried to allow DNA extraction for molecular identification of the species (ITS and psbA-trnH sequencing), and seed samples were air dried at room temperature prior to experiments.

Molecular identification

To confirm the correct identification of each of the 152 collected populations, we sequenced the ribosomal ITS (608bp) and psbA-trnH (213bp) intergenic spacer of up to five individuals from each population (Table S1, S2). To this end, we extracted total DNA from the dried leaves using a Plant Genomic DNA kit (DP305, TIANGEN BIOTECH CO., LTD, Beijing, China). The detailed information on primers and PCR reaction system and cycle conditions is presented in the Supplementary Information. The ribosomal ITS and chloroplast psbA-trnH were sequenced and edited using the DNAstar EdiSeq software. As the reference for alignment, we extracted the data for S. canadensis (ITS, HQ142590.1; psbA-trnH, KX214929.1) from GenBank and aligned them with the ITS and psbA-trnH sequences of our samples using the Clustal X 1.83 software. The ITS1, 5.8S and ITS2 gene boundaries and psbA-trnH were determined from ITS and psbA-trnH sequences of the Solidago genus in GenBank. To further verify the species identity of our samples, we conducted a BLAST search in the database of the National Center for Biotechnology Information (NCBI) to determine the species with the most similar identities (Table S1, S2). In addition, we built phylogenetic trees (Fig. S1, S2) for our samples with the ITS and psbA-trnH sequences, using the MEGA 4.1 Beta 3 software based on the principle of maximum parsimony method. We assessed branch support with 2,000 bootstrap replicates.

Determination of ploidy level

First, as a (diploid) reference for the ploidy-level determination, we assessed the exact chromosome numbers of the progeny from each of the 13 accessions/seed families of the Canadian population CA09 (Table S3, Fig. S3a), using mitotic root-tip squashes under a microscope. We grew up plants from each seed family/accession in the greenhouse, except for the Chinese and Indian populations for which field samples of maternal plants were used for ploidy examination. Fresh leaf material was held in self-sealing plastic bags containing silica and transported to the laboratory for ploidy level determination. We applied a modified flow cytometry method by Galbraith et al. (1983) for the determination of ploidy level. Briefly, 0.5 g fresh leaf tissue was chopped up with a razor blade in 2 ml of ice-cold Galbraith buffer (PH7.0, 45mM MgCl2, 30mM sodium citrate, 20mM 4-morpholinepropane sulfonate, 0.5% w/v polyvinylpyrrolidone and Triton X-100 1 mg/ml) in a vitreous Petri dish held on a chilled brick. The resultant fine slurry was filtered through a 50μm microfilter into 2 ml polystyrene tubes and incubated at 4 oC for 5min. The final filtrate was treated with 10 μL/ml of RNase A for 10 min at 4°C. Then, 200 μL/ml of a propidium iodide (PI) staining solution was added and incubated at 4°C for 30 min in darkness. The fluorescence strength of PI-DNA conjugates was measured by flow cytometric (BD FACS, USA) using a 488 nm laser and 590/40 emission filter (Arumuganathan & Earle 1991). The diploid CA09 population (Table S3) was used as an external reference. Flow-cytometry data acquisition and analysis was done with the software FloMax version 2.0 (Quantum Analysis, Münster, Germany). Only histograms with CVs smaller than 10% and with a minimum peak height of 50 particles were accepted. Other histograms were discarded and corresponding samples were designated no signal (Suda and Travnicek 2006). To ensure the measurement quality, the external reference material (CA09, the fluorescence intensity scale at position 50, Fig. S3a) were measured and analyzed at regular intervals in the analysis process to verify the corresponding position of the peak of new samples that indicates their ploidy levels. Accordingly, each sample was identified as diploid (2x), tetraploid (4x) or hexaploid (6x). Based on the combination of cytotype (2x/4x/6x) and geographical range (native [NA] / introduced [IN]), we assigned each accession to one of the following six geo-cytotypes: NA2x, NA4x, NA6x, IN2x, IN4x or IN6x. As some of the 152 populations included more than one cytotype, the total number of population × cytotype combinations was 233.

We also extracted data on chromosome numbers of 321 S. canadensis populations with known locations from ISI-indexed journals, books and databases of cytological indices (Index of Plant Chromosome Numbers [IPCN], International Organization of Plant Biosystematics [IOPB] Chromosome Number Reports). Detailed information about the published ploidy records are listed in Table S4. In total, our study includes information on the 1172 newly collected accessions/seed families for which we determined ploidy (Table S3) and 890 previously published ploidy records in 471 locations worldwide (Table S4). We used these data points to create a cytogeographical map in ArcGIS 10.0 (Environmental Systems Research Institute, Redlands, California, USA). For each of the 471 locations, we extracted the mean July temperature of each sample site using ArcGIS 10.0. Using the temperature of the hottest month (July) in the northern hemisphere, we try to guide the hypothesis that the thermotolerance evolution of S. canadensis may occur in populations with different ploidy levels. A Tukey test was used to detect differences among the different geo-cytotypes separated by the latitude distribution and mean July temperature of sample site. To examine whether the latitude distribution and mean July temperature were affected by geocytotypes, we used two-way ANOVA, with explanatory variables being origin (native, introduced), ploidy level (diploid, tetraploid, hexaploid) and their interaction.

Seed germination at different temperatures

Since not all 150 populations had enough seed, which were collected in field, only 30 native and 56 introduced cytotype populations representing all geo-cytotypes were tested in this experiment (Table S3 E1). We bulked the seeds from up to five accessions (if available) from each population, and tested their germination under two temperature regimes with a temperature of 25°C (12h/12h, day/night) and an extreme temperature of 39°C (12h/12h, day/night), respectively (the two temperatures were based on Li, 2011). A total of 150 seeds from each population were divided into three sets as three replicates. The 50 seeds were placed on wet filter paper in petri-dish (diameter = 9 cm) and kept in the growth chamber (2 m long × 0.8 m wide × 0.8 m high; Ningbo Jiangnan Biological Instrument factory, Zhejiang, China). The photon flux density was 250 μmol m−2 s−1, with an 85% relative humidity (RH) in the chamber. The germination process was observed every day lasted for 15 days, during which the filters were kept wet. Petri-dishes were randomly distributed in chambers and their position was changed every day. Seeds infected by fungi or bacteria were removed and not considered for calculations, but the same number of new seeds were added. A seed was considered as germinated if the radicle grew longer than seed length. The seed germination inhibition rate was calculated using the following equation: Inhibition rate = [(a-b)/a] × 100%, with “a” and “b” representing mean germination rate of population at 25°C and 39°C. One-way ANOVA was used for data analysis with Duncan’s parametric tests being applied to compare the differences of the germination inhibition rate among geo-cytotypes.

Common garden comparative experiments

To examine ecological adaptation and reproduction of the different geo-cytotypes, two comparative common garden experiments were conducted in the introduced range (IRCG) at Nanjing Agricultural University (32º02′N, 118º50′E; Nanjing, Jiangsu, China) and in the native range (NRCG) at the University of Georgia (33º56′N, 83º22′W; Athens, GA, USA), respectively. The IRCG experiments started on December 3, 2011 and December 5, 2012, while NRCG ones did not begin until December 10, 2012. At IRCG, the mean annual rainfall was 1090.7 mm, annual mean temperature was 15.6°C, mean max and min temperatures in January were 7.2°C and -0.7°C, and mean max and min temperatures in July were 32.2°C and 24.9°C. A total of 507 accessions propagated from the 30 native and 56 introduced populations were used in IRCG (population details in Table S3 E1). At NRCG, mean annual rainfall was 1250.0 mm, annual mean temperature was 15.9°C, mean max and min temperatures in January were 13.9°C and 1.4°C, and mean max and min temperature in July were 32.8°C and 20.8°C. A subset of plants from 26 populations, including all geo-cytotypes, was grown in NRCG (population details in Table S3 E2). At both common gardens, seeds from up to 5 maternal plants of each population were germinated in seedling trays filled with a 1:2 mixture of soil and sterilized compost in a greenhouse (18-25oC) in December. To minimize maternal effects associated with the quality of sampled seeds, at least 30-day-old seedlings with similar heights, basal diameters and leaf numbers were selected from each maternal plant per population and individually transplanted into 11-cm-wide, 8-cm-deep, 1 L plastic pots containing the same growing medium. Potted plants were randomly placed in a glasshouse supplemented with artificial light for 16 h during the day and watered as needed, usually every 2-4 days. Each accession was grown individually outdoors in a 5-L plastic pot filled with the same growing medium in March (three month after planting, the daily average temperature was above 10 oC). Trays and pots were reassigned randomly to new positions weekly. Pot management in two common gardens was the same.

To determine heat and frost tolerance with different geo-cytotypes (Table S3 E1), seedlings of S. canadensis accessions were transplanted to 1-L plastic pots filled with a 1:2 mixture of soil and sterilized compost 30 days (January 2, 2013) after sown in the greenhouse at IRCG. Two months later (March 3, 2013), the plants were divided into three sets for different experiments.

The first set of 516 plants (six offspring individuals of 1 to 6 accessions from each population) was used to determine the heat tolerance (beginning on March 3, 2013). Plants were kept with saturated soil moisture and subjected to 45±1°C for 10h in a growth chamber (E-36HO, PERCIVAL, U.S.), and assessed leaf heat damage. Seedlings were assessed according to a heat-damage index: 0) for healthy and normal seedlings; 1) when the tip of leaves was slightly burned; 2) 1-2 leaves were wilted; 3) 3-4 leaves were wilted; 4) > 4 leaves were wilted; and 5) all leaves wilted. Heat-tolerance index (HI) was calculated as 1-Σ[(number of seedling corresponding grade) / ( maximum grade total number of seedling)].

The second set of 774 plants (nine offspring individuals of 1 to 3 accessions from each population) was used to examine physiological responses to high temperature (beginning on March 5, 2013). The 3rd to 5th leaves from the shoot apex of plant were sampled and immediately frozen in liquid nitrogen and stored at -80°C until use after exposing to 45±1°C for 0, 2 and 4h, respectively (room temperature was about 28°C). The reactive oxygen species (ROS) levels (i.e. H2O2 content), activities of antioxidant enzymes including Cu/Zn superoxide dismutase activity (Cu/Zn-SOD), peroxidase activity (POD), catalase activity (CAT), glutathione reductase activity (GR), contents of non-enzymatic antioxidants (ASA), and osmoregulatory substances (i.e. soluble proteins) were assayed using commercial reagent kits purchased from Nanjing Jiancheng Bioengineering Institute (Nanjing, China).

The third set of 430 plants (five offspring individuals of 1 to 5 accessions from each population) was used to estimate freezing tolerance (beginning on March 9, 2013). We examined the relative PIABS (performance index on absorption basis) using chlorophyll-fluorescence measurements (Pocket-PEA, Hansatech Instruments LTD., King’s Lynn, Norfolk, PE32 1JL. UK) (Xie et al. 2015). The initial PIABS was measured after 30 min dark adaptation at room temperature. The upper, middle and lower parts of both sides of veins in each leaf (the fifth mature leaf) were measured. The freezing PIABS was measured at the same part of each leaf after the exposure of detached leaves to -8°C for 30 minutes. The relative PIABS = freezing PIABS/initial PIABS (Xie et al. 2015). The relative PIABS represents the freezing tolerance of S. canandensis.

One-way ANOVA was used to analyze the differences of heat-damage index, freezing tolerance (relative PIABS), inhibition rate of germination, ROS level, SOD, POD, CAT, GR, ASA, and soluble proteins among different geo-cytotypes. Duncan’s test was applied when assumptions of normality and homogeneity of variance were fulfilled; otherwise non-parametric ANOVA (Dunnett’s C test) would be used. In addition, to confirm the effect of environmental temperature factors on plant heat tolerance, we used canonical correspondence analysis (CCA) to assess the relationship between “species factors” and “environmental factors” as implemented in CANOCO for Windows 4.5 (Ter Braak & Smilauer 2002). For species factors, we included the heat-damage index, ROS level (H2O2 content) and antioxidative parameters. For environmental factors, we included latitude and six bioclimatic variables, including annual mean temperature (bio 1), isothermality (bio 3), minimum temperature of coldest month (bio 6), mean temperature of driest quarter (bio 9), mean temperature of warmest quarter (bio 10) and mean temperature of coldest quarter (bio 11). Furthermore, because “species factors” also were significantly affected by “ploidy level”, we also included “ploidy level” as an environmental factor.

To examine reproductive biology, we kept S. canadensis plants in both common gardens, with 860 pots of plants involved in IRCG (March 30, 2012 and 2013; Table S3 E1) and 260 pots in NRCG (April 4, 2013; Table S3 E2). In both common gardens, we assessed flowering phenology (onset, peak and end of flowering) and the number of viable seeds produced per plant. The latter was based on the number of seeds per plant and the germination rate at 25°C for each population (the details are described as above). One-way ANOVA was used to analyze the data for first flowering dates, flowering duration and number of viable seeds among the six geo-cytotypes. Duncan’s test or Dunnett’s C test was applied to separate means following the principle described above.

An additional set of measurements was taken only in IRCG in 2012. Because plants from some of native-range North American populations failed to produce viable seeds, we tested for pollen viability (Table S3 E5), stigma receptivity (Table S3 E6), germination rates of pollen grains on stigmas (Table S3 E7) and pollen-tube growth (Table S3 E7) for these native populations. As S. canadensis is self-incompatible and the different cytotypes are not cross-compatible (Yao 2014), we always used pollen and stigmas from different plants of the same cytotype in a population. We also recorded the mean and maximum daily temperatures during embryo development of plants from native populations to assess the relationship between embryo development and high temperatures (Table S3 E3).

To assess the impact of high temperatures on seed sterility more accurately, additional measurements were taken also in IRCG in 2014 by transferring offspring of 48 maternal plants from 16 populations (Table S3 E4) that did not produce viable seeds in IRCG to a climatic chamber (Room C3006 in Collage of Life Science, Nanjing Agariculture University) with a temperature regime of 24/18°C (12h/12h, day/night, the photon flux density was 250 μmol m−2 s−1) before flowering in July 2014. These populations included four NA2x, four NA4x, four NA6x and four IN2x populations. In addition, offspring of 24 maternal plants from 4 populations each of IN4x and IN6x that could produce viable seeds in IRCG were also transferred to the climatic chamber with a temperature regime of 36/30°C (12h/12h, day/night) before flowering in Oct. 2014 (Table S3 E4). An identical set of plants that included the above 24 populations was also prepared and placed outdoors in the common garden to serve as controls. The daily management of all experimental plants was the same to ensure their normal and comparable growth. Plants belonging to the same cytotype and grown under the natural and controlled conditions were artificially pollinated with the pollen from each other. Seeds per plant were harvested when mature and the number and germination rate of well-developed seeds were determined and used to estimate the number of viable seeds.

Potential distributions of the different cytotypes

To predict the potential distribution of S. canadensis under the current and potential future (2050) climate, based on a climate-change scenario, we used MaxEnt climatic niche models (Young and Evangelista 2011) and ArcGIS software. For these models, we used 5354 occurrence data of S. canadensis, including new data from our own research (n=150) (Li 2011, Lu et al. 2020), data from the literature (n=321) (Table S4), records downloaded from the Global Biodiversity Information Facility (GBIF; n=4784) and additional records from Chinese literature (n=101). We used MaxEnt Jackknife tests (test gain > 0.7) to select 9 key bioclimatic variables (Elith et al. 2006) out of the 19 variables available in WorldClim (http://www.worldclim.org) (Table S5, Fig. S8). The nine bioclimatic variables are annual mean temperature (bio 1), isothermality (bio 3), minimum temperature of coldest month (bio 6), mean temperature of driest quarter (bio 9), mean temperature of warmest quarter (bio 10), mean temperature of coldest quarter (bio 11), precipitation of driest month (bio 14), precipitation of driest quarter (bio 17) and precipitation of coldest quarter (bio 19). The WorldClim Current Conditions dataset (bioclimatic data averaged across the period 1950-2000) at a spatial resolution of 10 minutes was used to model the current climatic niche of S. canadensis. To forecast the potential future suitable ranges for infestation, we used the climate dataset of WorldClim for the year 2050 based on the Representative Concentration Pathway (RCP) scenario 4.5 (http://tntcat.iiasa.ac.at:8787/RcpDb). It is a moderate scenario where the total radiative forcing stabilizes before 2100 (Dyderski et al. 2018).

For the predictions of the climatic niche modelling, occurrence data for S. canadensis was randomly divided according to the algorithms of the MaxEnt software into training data (75% of occurrence points) for model prediction and test data (25% of occurrence points) for model validation. All additional parameters were set to their default values. A receiver operating characteristics (ROC) curve plot was built, and the scores of the area under the curve (AUC) indicated a good (AUC values at 0.8–0.9) to excellent (AUC values of 0.9–1.0) model fit (Fig. S9). The forecasted climatic suitability of different areas around the globe was classified into five categories using natural segmentation method based on clustering analysis in ArcGIS software: unsuitable (0–0.2327), low suitable (0.2327–0.5), margin suitable (0.5–0.6), suitable (0.6–0.7) and high risky suitable (0.7–1.0) (Kigen et al. 2019).

Supporting Information Materials and Methods

Molecular identification

A pair of primers (upstream: 5'-CGTAACAAGGTTTCCGTAG-3', downstream: 5'-TTATTGATATGCTT AAACTCAGCGGG-3') were used to amplify ITS rDNA gene and another pair of primers (upstream: 5’-CCGCCCCTCTACTATTATCTA-3’, downstream: 5’-TCTAGACTTAGCAGCTATTG-3’) were used to amplify psbA-trnH intergenic spacer. Polymerase chain reaction (PCR) was carried out in 50μL containing 2μL of 10μmol/L primers, 10×buffer 5μL, 25mM Mg2+ 4μL, 2.5mmol/L dNTPs 4μL, 50ng/μL DNA templates1μL and 5U Taq DNA polymerase 0.5μL. The PCR cycle was 95°C for 4 min followed by 30 cycles of 95°C for 30s, 52°C for 60s, 72°C for 60s, and a final extension step at 72°C for 7 min. PCR products were separated on 1% agarose gels and stained with ethidium bromide. The PCR products were purified and sequenced.

Common garden sites and materials

Two pot common garden greenhouse pot experiments were established at Nanjing Agricultural University [Nanjing, Jiangsu, China (32º2’N, 118º50’E), December 2011] and University of Georgia [Athens, GA, USA (33º56’N, 83º22’W), December 2012]. Seeds from all maternal plants of 86 populations were germinated in seedling trays filled with 1:2 mixture of soil and sterilized compost in the introduced common garden. For all of the populations at least five individual plants were originally collected each sampling plot but a few populations only included less than five individuals each cytotype through detection of ploidy. The populations with less individuals, all characterized according to cytotype by detection of their ploidy, were considered in the experiment and their lesser number did not significantly affect the obtained results according to statistical tests (Supporting Data 1).

Measurement of physiological and biochemical characteristics associated with seedling thermotolerance.

The activity of Cu/Zn superoxide dismutase (Cu/Zn SOD, EC.1.15.1.1) was measured by test Kit (Nanjing Jiancheng Bioengineering Institute, China; A001-4). The activity of peroxidase (POD, EC 1.11.1.7) was .measured by Guaiacol method (Li et al. 2013). Catalase activity (CAT, EC.1.11.1.6) was .measured by test Kit (Nanjing Jiancheng Bioengineering Institute, China; A007-1-1) (Yong et al. 2008). Glutathione reductase (GR, EC.1.6.4.2) was measured by test Kit (Nanjing Jiancheng Bioengineering Institute, China; A062-1-1) (Li et al. 2017). Protein content was detected by the method of coomassie brilliant blue (Deng 2013). The H2O2 concentration was detected by monitoring the absorbance of the titanium- peroxide complex at 415 nm, following the method of Patterson (Brennan & Frenkel 1977).

Reproductive traits

Growth of all potted plants was regularly monitored in both the native and introduced common gardens. Flowering phenology was assessed after onset of flowering (first flower head open), peak flowering (>50% flower head open) and end of flowering (>50% flower heads closed); in each plant flowering dynamics was recorded for every shoot (capitulescence) at 3-d intervals. The duration of flowering per plant (days between onset and end of flowering) was determined. Infructescences of all flowering individuals were bagged before seeds matured and dispersed. After seed matured, infructescences per plant were harvested, dried naturally and weighed. The number of seeds of each plant was estimated from the average 500-seed weight of five samples per population. The number of viable seeds was estimated by the number of seeds and the germination rate at 25°C per population. Five individuals per population were recorded and measured (Table S3 E1, E4).

Cross-compatibility experiments

A group of seedlings representative of the six geo-cytotypes was sown every 40 days (beginning Jan. 15, 2012) to ensure synchrony of their flowering periods. Two plants matching in flowering period from the six geo-cytotypes were selected at the early flowering stage for reciprocal crossings.

The positive and reverse crosses were made in all possible combinations between paired individuals belonging to the six different geo-cytotypes (three each from native and introduced ranges). Three populations were selected from each cytotype except for single introduced diploid (Table S3 E7). For each cross, three individual plants from each population were used; each one was considered a replication. Additionally, two kinds of artificial self-pollinations were defined as controls: those of different individuals within each cytotype and those of different florets of same individual.The same procedure was used to select the individuals for artificial self-pollination.  For same individual, artificial self-pollinations were made between different clonal shoots from a maternal plant, different capitula on the same shoot and different tubular florets within a single capitulum.

The bisexual tubular florets in the capitulum were carefully removed and the emasculation capitula were isolated with paper bags. After 5-7 days, plants were artificially pollinated when the stigma of ligulate flowers grew out of the petals and showed a “Y” shape. The pollen of selected male parents was collected with a brush and applied to the column head of the emasculation female parent and the paper bags were replaced quickly after pollination. Capitula were collected at 1h, 2h, 4h, 8h, 24h, 48h, and 72h after pollination (Fig. S5d), and thereafter gathered every 3 days, fixed with Carlo fixation liquid for further observation of pollen tube growth and embryo development. The mature seeds from each cross-combination were carefully harvested and their germination was tested at 25°C.

At the full-bloom stage of the S. candensis, six capitula from each of five populations per geo-cytotype were selected to determine their pollen vitality with the benzidine method (King 1960) (Table S3 E5). Briefly, new blooming capitula, collected between 9:00 and 11:00hrs in a sunny day, were placed into concave slides, followed by the addition of reagent I (0.5% biphenyl 10ml, 0.5% alpha naphthol 10ml, 0.25% sodium carbonate 10ml) and 0.3% hydrogen peroxide. Capitula were mashed and the mixture was transferred into a 1.5ml centrifuge tube and incubated for 50 min at 30°C. Stained pollen grains were observed and photographed with a microscope (Olympus BX53). Vigorous pollen grains stained black while the undynamic pollen grains stained yellow. Ten visual fields of each slide were randomly selected for assessment.

To determine stigma receptivity, capitula were also collected as described before (Table S3 E6). The florets were placed into concave slides, followed by the addition of the staining benzidine-hydrogen peroxide solution (1% benzidine: 3% hydrogen peroxide: H2O = 4:11:55 volume ratio), and observed under the same microscope. Vigorous stigma stained blue and produced a large number of bubbles. Nine florets per geo-cytotype were observed and counted. Additional capitula were soaked in 70% ethanol for 30 min and for additional 30 min in 90% ethanol after fixation with Carnoy's fluid for at least 24 hour. Then, the capitula were preserved in 70% ethanol. Before observation, capitula were rehydrated sequentially in 50%, 30%, 10% ethanol and distilled water for ten min each, softened in 6 mol/L NaOH for eight hours, washed three times with distilled water and stained with 0.1% aniline blue for 12 hours. The number of pollen grains germinating on the stigma and pollen tubes growing into the ovary were counted under the fluorescence microscope (Carl Zeiss, Axio Observer Z1), after making microscopic slides of tissue samples. Thirty single flowers of each cross combination were observed and counted.

After artificial pollination, capitula were sampled and fixed in formalin–acetic acid–alcohol (Berlyn & Miksche 1976) (FAA) at different embryonic developmental stages. Before observation, capitula were immersed orderly in 50% alcohol, 30% alcohol, 10% alcohol and distilled water every ten min. Individual flowers were carefully separated from capitula with forceps and immersed and softened in 2.5 mol·L-1 NaOH solution for six hours, and washed three times with distilled water. The single flowers were carefully placed on a slide and stained by aniline blue - lactic acid - phenol dye solution (Hu 1994) for five min. Before dropping 40% glycerol on individual flowers, redundant dye solution was absorbed by filter paper. Single flowers were covered with slide cover slips. The number of normal embryos at the different development stages was counted under the light microscope (Olympus BX53) (Table S3 E3). The normal development and abortion process of embryo is illustrated in Fig. S6: After pollination, the developing proembryo could be observed within 24 hours, globular embryos, the spherical embryo formed in 3~5 days after pollination (DAP), the heart-shaped embryo formed in 6~8 DAP, the torpedo-shaped embryo formed in 20~25 DAP and gradually matured. However, in the case of the following, we considered to be abnormal: 1, the development of embryos is seriously lagging. 2, the periphery of the embryo is transparent and the embryo atrophy or cluster. 3, the embryo is severely atrophied and only some small irregular clusters can be observed inside the peel. 4, only pericarp, no embryos and clusters can be observed inside the peel.

Prediction of potential geographical distribution of S. canadensis

Occurrence data

We searched published information on chromosome numbers of S. canadensis in relation to location in ISI-indexed journals, books and in cytological indices [Index of Plant Chromosome Numbers (IPCN); International Organization  of Plant Biosystematics (IOPB) Chromosome Number Reports] (Table S4) and sampled wild populations in the native and introduced range (Table S3).

Environmental variables, isotherms and processing of environmental data

The “Current Conditions” (Current Conditions 1950~2000) bioclimatic dataset was used to extract the current ecological niche of S. canadensis. For simulation of the future suitable region, we used the future climate dataset in the RCP4.5 scenarios (http://tntcat.iiasa.ac.at:8787/RcpDb) provided by the Coupled Model Intercomparison Project (CMIP). The RCP 4.5 is developed by the MiniCAM modeling team at the Pacific Northwest National Laboratory's Joint Global Change Research Institute (JGCRI). It is a stabilization scenario where total radiative forcing is stabilized before 2100 by employment of a range of technologies and strategies for reducing greenhouse gas emissions and has a higher priority.

We simulated the distribution of suitable region in the future climate change scenarios based on the current ecological niche of S. canadensis. The future climate dataset of RCP4.5 was used to drive the ecological niche modeling, before that it should be treated with the following: (1) the value of each environmental variable corresponding to all records in the current climate data layer was extracted by the spatial data processing tools of ArcGIS, (2) the value was assigned a value of background pixels at corresponding position in the future climate data layer.

Data processing

Occurrence data of S. canadensis was randomly divided according to the algorithms of the software (MaxEnt) itself into training data (75% of occurrence point data used for model prediction) and test data (25% of occurrence point data used for model validation) with all additional parameters set as default.

A receiver operating characteristics curve (ROC) plot was built by plotting the sensitivity values and the false positive fraction for all available probability thresholds (Manel et al. 2001) (Fig. S9). The area under the curve (AUC) was a measure of the area under the ROC, ranging from 0.5 to 1.0. The higher the AUC score, the better the model is at predicting presence/absence, indicating the environmental variables have high correlation with the predicted distribution of species, and prediction of the model is of high validity. The predicted model can be invalid (AUC value is lower than 0.6), poor (AUC value is 0.6–0.7), fair (AUC value is 0.7-0.8), good (AUC value is 0.8–0.9) and excellent (AUC value is 0.9–1.0) (Swets 1988). Using a heuristic estimation during training of the model and by means of a jackknife test, the importance of each environmental variable in the model was evaluated (Fig. S8).

The model was visually exhibited in ArcGis (version 10.0) to produce a single final distribution map. Different map colors corresponded to different fitting indices in the potential distribution map. The analysis base map was obtained from the National Fundamental Geographic Information System (http://nfgis.nsdi.gov.cn/).

There are some missing values in the dataset:

1.) Six missing values will be finded in antioxidative parameters of cytotypes among native and invasive S. canadensis after heat exposure (45 ºC for 0, 2 and 4 h respectively) （File: Heat_tolerance.xlsx，Sheet: antioxidative parameters）

2.)  In the parts of cross experiment, 29 replications were collected in some combinations，which should be 30 replications in plan (Fig. S5h and S 6a, b, c). Only 19 replications of the "IN6x X NA4x" ombinations were collected in the observation of germinated pollen grains on stigmas (Fig S 6a) and  11 replications of the "IN4x X NA4x" ombinations were collected in the observation of pollen tubes entering into the ovaries (Fig S 6b).

3. ) ITS sequences and psbA-trnH intergenic spacer sequences of 152 S. canadensis populations were submit the data in a format of ".txt". You could read and align them using MEGA 4.1 Beta 3 software.