1. Estimating the role of specific processes in the spread of alien species necessitates the determination of introduction pathways and source populations of invaded areas. Alien grasses in the genus Lolium that have become widely invasive in Japan provide provides a unique opportunity to estimate the expansion process through a direct comparison between source and naturalised populations because the introduction pathways, contaminants in grain commodities and commercial cultivars for fodder crops or revegetation materials are well-known. Therefore, by directly comparing between source and naturalised populations, we estimated the introduction pathways and whether adaptative evolution occurred in Lolium species on sandy coasts in Japan.

2. Lolium individuals sampled from naturalised populations in croplands, seaports and sandy coasts were compared with those from two introduction sources for morphological and genetic variations based on a genome-wide single nucleotide polymorphism analysis and a common garden experiment. Furthermore, we conducted a reciprocal sowing experiment between a cropland and sandy coast.

3. Populations naturalized in croplands were closely related to the cultivars, whereas those naturalized in seaports and sandy coasts were associated with contaminants. This pattern is common in western Japan. Therefore, these results indicate that the cropland and sandy coast populations are derived from cultivars and contaminants, respectively. The reciprocal transplant experiment clearly demonstrated the home site advantage; populations derived from croplands yielded higher floret numbers than those derived from other habitats at the cropland site; however, such an advantage was not detected with the coastal site population. Sandy coast populations had higher survival rates than those from croplands at coastal sites. Port populations exhibited a similar tendency as sandy coast populations, indicating that contaminants may be originally adapted to salty and dry environments, such as that in sandy coasts.

4. Synthesis. We demonstrated that two congeneric species with different ecological characteristics were introduced through multiple introduction pathways and spread into different habitats. A direct comparison between source and naturalised populations can greatly advance our understanding of the patterns and processes of biological invasions.

Plant materials

For contaminant seeds from imported grains, we obtained three 20-kg samples of five wheat classes imported between 2006 and 2007: three classes of U.S. wheat (Hard Red Winter wheat: HRW, Western White wheat: WW and Dark Northern Spring wheat: DNS), one class of Canadian wheat (No.1 Canada Western Red Spring wheat:1CW), and one class of Australian wheat (Australian Standard White wheat: ASW) (Table S1). Furthermore, we obtained three types of fodder crops imported from Australia between 2016 and 2017: ground barley, wheat straw and oat hay (Table S1). The seeds of Lolium species were selected from each sample, according to the method described by Shimono et al. (2010).

For cultivated varieties, we purchased 1 kg bags each of Italian ryegrass seeds and perennial ryegrass seeds from Snow Brand Seed Co., Ltd. (Hokkaido, Japan) and Takii Seed Co., Ltd. (Kyoto, Japan) in 2012. We selected the most currently available commodity, ‘Common’ (a generic term, not a cultivar name), for fodder crop or revegetation material (Japan Grassland Agriculture and Forage Seed Association, 2010). Additionally, eight Japanese cultivars of Italian ryegrass (Hanamiwase, Akatsuki, Waseyutaka, Inazuma, Tachimusya, Satsukibare, Mammos B and Jiant) were kindly provided by the Institute of Livestock and Grassland Science, National Agriculture and Food Research Organization in Japan. The following foreign cultivars were obtained from GenBank, National Agriculture and Food Research Organization (NARO): the United States cultivar of Italian ryegrass, Westerwold ryegrass (GULF); New Zealand cultivar of Italian ryegrass, Westerwold ryegrass (MOTO 191, TETRAPLOID); and Australian cultivar of annual ryegrass, Wimmera ryegrass (Table S2).

Additionally, seeds were collected from naturalised populations in three habitat types–cropland, seaport and sandy coast–in each of the three geographic regions of Japan (Kanto, Kansai and Kyushu) (Table S3). All the selected seaports (Kobe, Kashima and Hakata ports) are major international trading ports where more than 1,000 kt of imported grain is unloaded every year. Matured seeds were individually collected from 10 plants at intervals of at least 2 m from each habitat from May to June, 2017.

Common garden experiment

On October 2017, 10–30 seeds were randomly selected from each sample and germinated on two pieces of moistened filter papers placed in 9-cm plastic petri dishes in an incubator set at 25°C for a 12-h light period and at 15°C for a 12-h dark period. One week after germination, the seedlings were individually transplanted into pots (10.5 cm in diameter, 9 cm in depth) filled with potting soil (Tachikawa Heiwa Nouen Co., Tochigi, Japan) and placed in a glasshouse under natural sunlight at the experimental farm of Kyoto University, Kyoto, Japan (35.032121°N, 135.783453°E). The plants were regularly watered.

To investigate whether phenological and morphological differences exist between individuals with different provenances and habitats, the heading  (ear emergence from the flag leaf sheath) date and the following morphological characteristics at 3 weeks after heading were measured: number of tillers, length and width of the longest culm, inflorescence length and width of the longest culm and number of spikelets of the longest culm. In addition, spikelet length, glume length, number of florets per spikelet, first floret length and awn length were measured using a spikelet placed in the middle of the inflorescence of the longest culm. Awn length was graded on a scale of 1 to 3 depending on the length (1: no awn, 2: <5 mm and 3: ≥5 mm). The ratio of glume to spikelet length was subsequently calculated.

Reciprocal transplant experiment

Habitat-specific adaptations were examined in a reciprocal transplantation experiment. The experimental farms of Kyoto University (35.032121°N, 135.783453°E) and Shimazaki Sandy Coast (34.7319463°N, 136.5287767°E) were selected as representatives of the cropland and sandy coastal habitats, respectively, and were used in a reciprocal transplant experiment (Fig. S1). From May to June 2017, bulk seeds were harvested from more than 30 individuals in naturalised populations in three habitat types: cropland, seaport and sandy coast, in each of the three geographic regions of Japan (Kansai, Kanto and Kyushu). Seeds were stored in an airtight container containing silica gel at room temperature (20°C–28°C) until September 2017..

Fifty seeds from each population were sown in plastic trays (length: 26.5 cm, width: 18.3 cm and depth: 7.7 cm) filled with soil from bare land in each habitat. Eight trays were prepared for each population, three of which were set on croplands and five on the sandy coast on 29 September 2017 which is the typical germination period for Lolium species. The trays were covered with fine mesh nets to prevent further seed input and removal by wind or animal dispersal. The trays were buried at depths where the rims were at the same level as the soil surface. Seedling emergence, survival and flowering were recorded at 1-week intervals through July 2018, except for the period from December to March, during which observations were made every two weeks. Individuals were harvested three weeks after heading, and the number of culms,  the number of spikelets per culm and the number of florets per spikelet in the middle of the inflorescence were measured. The total number of florets produced by surviving individuals was calculated as follows: number of culms × number of spikelets per culm × number of florets per spikelet. Survival rate was defined as the ratio of the number of surviving individuals to the number of emerged seedlings.