Introduction

Ecological divergence among populations is frequently attributed to local adaptation to contrasting environments and driven mainly by divergent natural selection1,2,3. During the process of ecological adaptation, phenotypic divergence may result in gradual establishment of reproductive isolation as a consequence of divergent natural selection and thus lead to ecological speciation2,4,5,6. Phenotypic variation influences the survival of plants and a range of interactions between plants and environments on one hand and is the primary source of information for classifying plants and determining their evolutionary relationships on the other1,2,3,7. Therefore, investigations on phenotypic diversity, including phenological and morphological variation, have significant implications for the understanding of evolutionary processes and mechanisms of biological diversity1,8,9,10,11. To date, extensive studies have been conducted on phenotypes and the roles of natural selection on phenological and morphological variation1,3,4,6,12. However, the contributions of phenotypic changes to ecological adaptation and speciation in particular environments, the ecological and genetic factors contributing to phenotypic changes as well as the link between the evolution of divergent phenotypes and the build-up of reproductive isolation remain less clear6,8,9,13. To address these issues, it is essential to perform studies by integrating multiple approaches, including phenotypic characterization, population genomics and assessment of fitness3,9,11. Of various approaches, in situ observation and common garden experiment are classical and irreplaceable methods but have been paid less attention due to the manipulative and observational difficulties, particularly for natural populations of non-model species1,9,14,15,16.

Oryza rufipogon (Griff.) and O. nivara (Sharma and Shastry) are the most closely related species and the direct progenitors of domesticated Asian rice (O. sativa L.)17,18,19,20,21. Oryza rufipogon is perennial, largely cross-fertilized, photoperiod sensitive and inhabits persistently wet areas across southern China, South and Southeast Asia, Papua New Guinea and Northern Australia; while O. nivara is annual, predominantly self-fertilized, photoperiod insensitive and occurs in seasonally dry habitats with more restricted distribution in South and Southeast Asia19,21,22,23,24. Earlier studies have found significant differentiation between O. rufipogon and O. nivara for numerous phenotypes at both local and regional scales, involving flowering phenology, life history, mating system, and many morphological features20,25,26,27,28,29,30,31,32,33,34,35. However, several studies revealed a continuum of a few phenotypic traits, particularly for life history, both between and within species20,32,36, and geographic isolation has been proposed to be a main factor driving species divergence between the two species23,26,32,37. These studies raise a question regarding the relative roles of habitats and geographic isolation play driving phenotypic divergence between the two species23,26,32,37. Moreover, as an accurate indicator of the completion of speciation and a mechanism for maintaining species identity2,6,38, flowering time difference was considered as a major driver of reproductive isolation and thus species divergence between O. rufipogon and O. nivara27,35,37. Although the flowering time of the wild rice species has been amply documented in literatures20,25,26,27,28,37,39,40,41,42,43, it is still less clear about their variation pattern and plasticity across populations because, to our knowledge, no research on flowering phenology of any Oryza species has been undertaken based on in situ observation for its entire life cycle. Also, few comparative studies of morphological variation have been made on the two wild rice species in terms of ecological adaptation and divergence. Given information available on phenology, morphology and habitat preference of the two species is incomplete, an in-depth study on variation patterns of flowering phenology and morphological traits of O. rufipogon and O. nivara in different environments/habitats would provide critical insights into the understanding of processes and mechanisms of species divergence and speciation of these wild species in general.

Sri Lanka exhibited diversified and well-understood ecosystems that consist of three main ecological zones (wet, dry, and intermediate) that were clearly defined based on the annual rainfall patterns across the island44,45,46 (Fig. 1). Of five Oryza species recognized in Sri Lanka, O. rufipogon and O. nivara are the most common and widely distributed species23,45,47. Importantly, the two wild rice species display a parapatric distribution that reflects local environmental differences, with O. rufipogon confined to the wet zone and O. nivara to the dry zone of the country23,45,47,48. This is in marked contrast to the distribution pattern in many other areas in which the two species are largely sympatric despite different habitats22,23,27,36,49. In addition, based on microsatellite markers, our preliminary study of the O. rufipogon and O. nivara populations in Sri Lanka found relatively high level of genetic diversity for both species and a significant genetic differentiation between species48. Therefore, the well‐understood climatic zones in Sri Lanka and their close match with the distribution of two wild rice species, along with relatively characterized habitats (Table S1) and genetic background48,50, provide a unique system to investigate adaptive divergence and ecological speciation in plants.

Fig. 1
figure 1

Geographic locations of six O. rufipogon (blue and dot) and eight O. nivara (red and triangle) populations sampled in this study. Population codes with superscripts a and b represent the populations used for the in situ observation and common garden experiment, respectively. Detailed information on all populations is provided in Table S1. The wet, dry, and intermediate zones are well defined previously, with the amount of annual rainfall being more than 2,500 mm (wet zone), 1,750 to 2,500 mm (intermediate zone) and less than 1,750 mm (dry zone)44. The map was extracted from https://images.app.goo.gl/2MAHwhhJ5owH3ib99 and figure was generated by the first author using the ArcGIS 10.3 (https://developers.arcgis.com/).

Full size image

In the present study, we took advantage of the unique system of wild rice in Sri Lanka to investigate the variation patterns of flowering phenology and morphological traits and their implications for ecological adaptation and divergence of O. rufipogon and O. nivara. By incorporating in situ observation and common garden experiment into an integrative assessment, we aimed (1) to comparatively evaluate the life cycle of O. rufipogon and O. nivara. We were particularly interested in knowing the differentiation and plasticity in flowering phenology of the two species and whether flowering time is an isolation mechanism for species divergence; (2) to address the variation pattern of morphological traits between and within species and the impact of different geographic and environmental factors on the variation of the two species; (3) to determine whether the phenotypic differentiation between the two species is attributed to divergent natural selection, which is usually expected for ecological divergence and speciation2,3,51. These studies not only gain insights into adaptation and speciation process of two wild rice species but also provide an informative case contributing a better understanding of ecological adaptation and speciation in general. In addition, investigations on wild rice in Sri Lanka would facilitate to develop efficient strategies in rice breeding and conservation of wild rice resources because wild rice species is an important gene pool for rice breeding and the threats to their existence is increasingly serious due to global climate change and urbanization17,21,23,28,52.

Materials and methods

Fourteen natural populations including six O. rufipogon and eight O. nivara populations in Sri Lanka were chosen for in situ observation and common garden experiment. The O. nivara populations were confined to seasonally dry habitats in dry and intermediate zones and the O. rufipogon populations inhabited in wide ditches with variable depths of water or in marshy fallows adjacent to deep water in wet and intermediate zones (Fig. 1, Table S1). For each population, a random seed/plant sample of 10 to 16 individuals, spaced at least 5 m apart, was collected to minimize multiple sampling of the same genet.

In situ observations

The in situ observation was performed for two O. rufipogon populations (SL02-R and SL14-R) in the wet zone and two O. nivara populations (SL07-N and SL16-N) in the dry zone (Fig. 1). Two O. rufipogon populations (SL02-R, SL14-R) were representatives of the typical populations in the wet zone (Fig. 2a and b) and two O. nivara populations (SL07-N and SL16-N) were representative of the typical populations of the dry zone (Fig. 2c and d). Oryza rufipogon adapts to the habitat in wet zone where they experience a consistent supply of water throughout the year. Its flowering time and seed production can span a longer period due to stable wet conditions, which allows O. rufipogon to have multiple reproductive phases within a year or extend its flowering across seasons. In contrast, O. nivara occurs in the habitats of the wet zone and were usually found in the fields that remained dry until the first rain in October or November. For the O. nivara populations, seeds germinated at the start of the rainy season, and the plants flowered and produced seeds before the beginning of the dry season45,46,47,50.

Fig. 2
figure 2

The growth form (upper panels) and habitat (lower panels) of four O. rufipogon and O. nivara populations used for in situ observations. (a, b) Two O. rufipogon populations, with SL02-R growing around a river of 3 ~ 5 m wide and 1 ~ 3 m deep (a) and SL14-R in an abundant marshy land, along with other hydrophytic weeds and fairly dense (b). (c, d) Two O. nivara populations, with SL07-N growing nearby a road and separated by paddy fields (c) and SL16-N distributes near to lake area and mix grows with other shrubs and weeds (d).

Full size image

We established three 1 m2 quadrats (plots) for all four populations, with quadrats setting up at 3 m intervals around the area of peak plant abundance. Observations were recorded separately in three plots as replicates at two-week intervals for over an entire growth season (March 2016 - June 2017).

Common garden experiment

The common garden experiment involving five O. rufipogon and seven O. nivara populations was conducted for two consecutive years (2016–2017) at the Faculty of Agriculture, University of Ruhuna (latitude 06.08 °N and longitude 80.56 °E) in Sri Lanka (Fig. 1, Table S1). Detailed information on all populations and their habitats is provided in Table S1. Seeds were collected from the natural populations and germinated seedlings were established in the pots. Seedlings were carefully uprooted from their natural habitats and transplanted in cemented pots (40 cm length × 40 cm width × 45 cm height) containing sandy loam soils free from weeds. Fifteen individuals from each population were planted following a randomized complete block design and kept in open field conditions. Recommended doses of chemical fertilizers were applied at the time of transplanting stage and the water level in the pot was maintained at 15 ± 5 cm during the entire period of plant growth except for transplanting time.

We selected four populations from each species (Fig. 1) for phenological observations. Phenological traits were evaluated in two-week intervals for each plant, involving the performance at two growth stages (flowering and seed setting) similar to the in situ observation. To investigate the morphological variation, we measured nine traits for five O. rufipogon and seven O. nivara populations (Fig. 1) using three randomly selected tillers per plant following the descriptors for wild and cultivated rice40,53.

Phenotypic measurements and data analysis

For phenological variation, we recorded the date of first flowering and seed setting because flowering time is a crucial adaptive trait that enables plants to coordinate their reproductive processes with favorable environments and seed setting ensures plants to produce fully developed seeds9,41. For morphological variation, we measured nine traits (Table S2) that were supposed to be either taxonomically significant or adaptive25,27,28,29. Of these traits, four including anther length (ANL), panicle neck spikelet length (PNSPL), panicle exsertion (PE) and awn length (AWL) are related to reproduction, three flag leaf traits (FLL, FLW, FLA) are associated with photosynthesis processes, and the other two, culm length (CL) and culm diameter CD), are involved in habitat preference. We assessed ten individuals for in situ observation and fifteen individuals in common garden experiment. Three randomly selected tillers per plant were measured following the descriptors for wild and cultivated rice27,40,53.

All phenotypic analyses were conducted using free software R, version 4.3.1 with supplementary packages and Minitab 17. Phenotypic differentiation between species was also assessed by a standard principal component analysis (PCA)25,54. Univariate ANOVAs were used to examine the effects of species on the variables using population means and to assess the effects of populations within species using trait values of individuals within populations. To gain a deeper understanding of the impact of habitat and climate on phenotypic variation, we conducted a correlation analysis using altitude, rainfall, and temperature as environmental factors because these factors vary across Sri Lanka and affect the survival and distribution of plant species55. The altitude, rainfall and temperature data were obtained from National Meteorological Department of Sri Lanka (https://www.meteo.gov.lk/). Specially, altitude presents a natural gradient where multiple environmental factors change concurrently, including temperature, oxygen levels, solar radiation, atmospheric pressure, and even soil properties. Different altitudes often harbor distinct ecological niches. Rainfall directly influences water availability in an ecosystem and water availability can influence reproductive period, breeding strategies, and even seed germination. Temperature affects directly the developmental and growth rates and specific life stages and a species has a range of temperatures where they perform optimally. The t-test and F test were conducted to test for trait divergence between environmental variables and the differences between species, respectively25,56. Morphological similarity and physical distances of pairwise populations were calculated using the dist function and the distGeo functions in R57, respectively. The Pearson correlation coefficient between morphological similarity and physical distance was calculated using the corr.test function in R57.

We performed the QST-FST analysis, an important and widely used method for testing whether natural selection or genetic drift is the major cause of population divergence for phenotypic traits27,29,56,58. FST represents the amount of genetic variance that can be explained by population structure measured by genetic loci/markers, while QST is an analog of FST but measures genetic differentiation between populations/species for phenotypic trait59,60. We used the FST value published previously for the same set of populations based on SSR markers48 and calculated QST values for nine quantitative traits following the procedure in Guo et al.29 to determine whether the QST values were smaller or larger than FST values. QST was calculated using the technique described by Spitze61. The variance components between species (VB) and within species (VW) for each quantitative trait. QST values were obtained by 1,000 bootstrap iterations with each involving resampling across individuals within populations. Hence, QST can be computed using the formula QST = VB / (VB + 2VW), and the QST distribution produced from the 1,000 bootstrap iterations was used to assess the importance of the disparity between QST and FST. If QST is significantly larger than FST, natural selection would be the main factor driving species divergence, whereas a neutral genetic differentiation would be supported if FST is significantly greater than QST58,60.

Results

Flowering phenological variation between and within species

We investigated the flowering phenology based on two sets of investigations (in situ observation and common garden experiment). First, we performed in-situ observation of the life cycle of two O. rufipogon and two O. nivara populations for an entire year and observed substantial differences in flowering phenology between species, including first heading and flowering duration. It is apparent from Fig. 3 that O. rufipogon populations exhibited a much longer life cycle (7 to 10 months) than that of O. nivara populations (4 to 5 months). In addition, two peaks of sexual reproductions were found for the O. rufipogon populations but only one for the O. nivara populations. Unlike two O. nivara populations that showed almost the same pattern, a slight difference in life cycle was observed between two O. rufipogon populations (SL02-R and SL14-R), likely reflecting the difference of habitats in two sites because SL02-R is situated near the river where water is available year-round, while SL14-R is distributed across marshy land where floods is periodical (Table S1).

Fig. 3
figure 3

Life cycle of O. rufipogon and O. nivara populations based on in situ observation for the natural populations. The suffixes, R and N, in population codes represent O. rufipogon and O. nivara, respectively.

Full size image

Next, based on common garden experiment, we revealed a different pattern of phenological variation to that of the in situ observation. As shown in Fig. 4, all O. nivara populations exhibited continuous flowering and seed setting throughout the years with almost all tillers entering the flowering stage, a pattern contrasting to that found in the in situ observation in which much shorter life cycle was observed for the O. nivara populations. For the O. rufipogon populations, a life cycle completed within 7–8 months without two peaks of sexual reproduction found in the in situ observation, probably due to different management in the common garden experiment such as frequent watering and weeding. We also found very little variation among populations within species for either O. rufipogon or O. nivara, probably because of the same management condition. These observations indicated that the flowering phenology of wild rice can vary slightly depending on the growing conditions and that O. nivara in Sri Lanka may not be strictly an annual plant and can survive persistently when water is readily available.

Fig. 4
figure 4

Flowering phenological variation of O. rufipogon and O. nivara populations based on common garden experiment. The suffixes, R and N, in population codes represent O. rufipogon and O. nivara, respectively.

Full size image

An unexpected finding is that flowering time of two species is largely overlapped for either observed populations in the fields or manipulative populations in the common garden despite different variation patterns from two sets of investigations (Figs. 3 and 4). The overlapped flowering phenology between species and high level of plasticity of flowering time have important implications for divergence and origin of O. nivara (see Discussion).

Morphological variation between and within species

We examined variation patterns of morphological traits between and within species based on in situ observation and common garden experiment. First, we performed the analyses of nine morphological traits recorded in in situ observation. The ANOVA indicated significant differences between species for all traits except for flag leaf angle (FLA) and awn length (AWL) (Fig. 5; Table S3). Notably, O. rufipogon exhibits significantly larger values than O. nivara for seven traits that diverge between species. Significant differentiations were also observed between populations within species for O. rufipogon (six traits) and O. nivara (four traits), with three traits (FLL, FLW, and AWL) that diverged significantly in both species (Fig. 5; Table S3). The PCA analysis further supported the ANOVA results that two distinct groups corresponding to two species are apparent, with intraspecific variation larger in O. rufipogon than in O. nivara (Fig. 6a). All traits except for two (AWL and FLA) exhibited high loadings in PC1 (> 0.3) that explains majority of the total variance (52%) (Table S4), suggestive of substantial differentiation between the two species despite high level of variation within species. However, AWL exhibited high loading on PC2 (0.7) (Table S4), highlighting its potential utility in distinguishing traits at the intra-species level.

Fig. 5
figure 5

Variation patterns of nine morphological traits between and within O. rufipogon and O. nivara populations based on in situ observations. The bars indicate the mean ± standard deviation (SD) of trait variation. The white and grey columns represent the O. rufipogon and O. nivara populations, respectively.

Full size image
Fig. 6
figure 6

Principal component analysis (PCA) of O. rufipogon (blue) and O. nivara (red) populations based on in situ observation (A) and common garden experiment (B).

Full size image

Next, we analyzed the variation pattern of morphological traits between and within species based on common garden experiment. We found significant differentiation between O. rufipogon and O. nivara for all nine traits except for two flag leaf traits (FLL and FLW) and significant variation among populations within either O. rufipogon or O. nivara (Fig. 7; Table S5), largely consistent with the results from in situ observation on natural populations. PCA also shows clear separation of two species, with PC1 and PC2 accounting for 38.17% and 20.01% of the variance, respectively (Fig. 6b). However, in contrast to the in situ observation on natural populations, larger variation within species occurred in O. nivara than in O. rufipogon (Fig. 6b), with the factor loadings of many traits (Table S6) different from those from in situ observation. These results might reflect potential impact of growing/manipulative conditions on the variation patterns of morphological traits and suggest high plasticity for most morphological traits of the wild rice species.

Fig. 7
figure 7

Variation patterns of nine morphological traits among populations within O. rufipogon and O. nivara. The bars indicate the mean ± standard deviation (SD) of trait variation. The white and grey columns represent the O. rufipogon and O. nivara populations, respectively.

Full size image

Correlation between morphological traits and environmental variables

To determine the factors that might be associated with morphological differentiation between the two species, we conducted analysis for correlation between each of nine traits and three environmental variables (altitude, rainfall, and temperature) that were potentially variable across Sri Lanka55. We found that all traits except for FLL significantly correlated with at least one of three environmental factors for O. rufipogon, with two traits (CD and ANL) affected by all three factors; whereas all traits of O. nivara were significantly correlated with at least one environmental factor and four traits of them (CL, CD, AWL, and PNSPL) correlated with all three factors (Table 1), implying that almost all trait were sensitive to the environments or habitats. Specifically, variation of seven, five, and three traits correlated with altitude, rainfall, and temperature, respectively, in O. rufipogon; while that of six, seven, and seven traits correlated with altitude, rainfall and temperature, respectively, in O. nivara (Table 1). These results suggest that environmental factors play important roles in shaping the variation patterns of morphological traits and contribute to the differentiation between O. rufipogon and O. nivara.

We further tested whether positive correlation exists between morphological similarity and geographic distance for pairwise comparisons within and between species to assess the relative roles of habitat and geography. It is shown that no significant correlations were found between morphological similarity and geographic distance for population pairs within O. nivara (R2 = 0.0746, P = 0.9553) and between O. rufipogon and O. nivara (R2 = 0.0004, P = 0.9125), contrast to the pairs within O. rufipogon (R2 = 0.1987, P = 0.0429) (Fig. S1). These results indicate that the environment/habitat plays more important roles than geographic distance in the morphological divergence between species.

Tests for the role of natural selection in morphological divergence between species

To determine whether natural selection or genetic drift is the primary driver of phenotypic differentiation between the two species, we performed the QST-FST comparison based on measures of nine traits obtained from common garden experiment. By calculating QST values for each trait and the FST value of 0.129 estimated using SSR markers48, we detected significantly larger QST values than the FST value for seven out of nine morphological traits (Table 2), suggesting that natural selection plays an important role in the divergence of these seven traits between O. rufipogon and O. nivara. Consistently, the observation that habitat instead of geographic distance correlated with the divergence for population pairs of two species in most morphological traits (Fig. S1) suggest that morphological differentiation between O. rufipogon and O. nivara is adaptive or under natural selection.

Table 1 Correlation (r) between morphological traits and environmental variables based on common garden experiment.
Full size table
Table 2 The QST values for individual traits and Z test for QST-FST comparison.
Full size table

Discussion

Ecological adaptation or speciation is a consequence of phenotypic and genetic differentiation, which is driven by divergent selection under different environments1,3,6,38. Extensive studies have revealed significant differences between O. rufipogon and O. nivara in flowering time and multiple traits at regional and local scales and indicated that the differentiations were usually associated with habitat differences, as expected in ecological adaptation and speciation27,30,32,33,35,37,62. However, studies on the relative roles of adaptive plasticity and genetic variation adaptive to the environments as well as the impact of environment factors on phenotypic variation were limited in wild rice because previous studies have predominantly been conducted in artificial settings, such as greenhouses, or in areas that did not accurately reflect the natural habitats of the species20,26,27,28,32,42,49,63. Here, we integrated the in situ observation in natural habitats and manipulative experiment in the common gardens to address the variation and plasticity of phenology and morphological traits under different temporal and environmental conditions, and to examine how seasonal fluctuations and environmental variables impact the phenological and morphological variation of two wild rice species in Sri Lanka.

Flowering phenology is a key factor in the process of ecological adaptation and speciation and has been studied extensively in plant species8,9,56,64,65,66. The adaptation of plants to local environments entails the selection of flowering phenology, which is particularly important for the overall fitness of plants because reproductive success depends largely on the duration of the growing season and the timing of flowering8,66,67,68 (flowering phenology). Consequently, flowering phenology has been considered to be a magic trait that facilitates genetic differentiation and triggers the divergence of multiple traits in response to the environments16,69,70. Although substantial investigations of flowering time have been conducted on rice cultivars20,41, flowering phenology and its variation within and between wild rice species remain largely elusive despite some efforts20,32,35,42,43,50,63. In the present study, we found varying degrees of phenological variation between species, between populations within species, and between in situ observation and common garden experiment, which may be caused by both genetic elements and environmental variables21,23,27,30,32,42. First, we found significant difference between species in flowering phenology, as revealed by either in situ observation or common garden experiment. For in situ observation, O. rufipogon exhibited a prolonged life cycle with two flowering peaks while O. nivara displayed a notably shorter life cycle with a single flowering peak, as reported in Sandamal et al.50. This reflects the differences in typical environments/habitats where O. rufipogon (in wet zone) and O. nivara (in dry zone) grow. The bimodal versus unimodal reproductive phases might arise from different adaptive strategies in response to ecological pressures. However, common garden experiment showed a different pattern of flowering phenology, particularly for O. nivara that flowers continuously for almost an entire year. This suggests that the flowering phenology or life cycle of O. nivara is highly flexible or plastic depending on environmental conditions, with water availability being a primary determinant21,22,23,50,71. It also implicates that O. nivara, at least for the Sri Lankan populations, might not be an annual plant in the strict sense because it can consistently survive and succeed when ample water is available. To some extent, flowering phenology of O. rufipogon can also alters depending on environments, consistent with previous reports42,43. Together, these findings suggest that flowering phenology (or life cycle) of two wild rice species is highly plastic and vary depending on environment and management practices, demonstrating the crucial role of plasticity in adaptation of species to varying ecological conditions and could potentially influence their speciation processes14,43,72,73. Second, we unexpectedly found that flowering time of the two species in Sri Lanka overlapped to a large extent based on in situ observation and common garden experiment (Figs. 3 and 4), which is incongruent with previous reports that significant difference in flowering time was observed at local scale26,27,35,37,49. This raises an interesting question about the role that flowering phenology plays during the divergence O. rufipogon and O. nivara because divergence in flowering time is critical to maintain reproductive isolation between species and has been considered to be a main driver of speciation in the two Oryza species27,35,37 and many other plant groups8,16,56,69,74,75. Numerous studies indicated that the divergence or separation between O. rufipogon and O. nivara is apparent genetically and morphologically at the same locality due to strong barriers to gene flow (e.g., differences in flowering time and mating system)27,37,49,76. Our previous study on the Sri Lankan populations of the two species uncovered significant genetic differentiation between species and hypothesized that the Sri Lankan O. nivara might be introduced from the Indian subcontinent to the island48, implying well-established reproductive isolation (i.e., weak if any gene flow) between the two species in Sri Lanka. Consequently, the primary factor to maintain the divergence of the two species in Sri Lanka may not be flowering time but rather environments because habitat selection is able to restrain gene flow mediated by pollen dispersal and seed migration3,27,32,37. Finally, the observed difference of the Sri Lankan wild rice in flowering phenology and particularly the high levels of plasticity, reflect the strategy how the two species have adapted to their respective environments. Take O. nivara as an example, a significantly shorter life span was observed in natural habitats (roughly six months) (Fig. 3) than in the common garden (the entire year) (Fig. 4). It is understandable that, as a derivative species, O. nivara populations evolved a shorter life cycle to cope with the habitat shifts, which took place in both Sri Lanka and many other areas27,30,37,71. Such a strategy to adapt to dry and stressful habitats through phenological plasticity enables these species to coordinate their reproductive cycles with favorable environmental conditions and has been reported in many other plant species16,56,69,74,75.

Phenotypic divergence is the first step towards local adaptation and speciation and has thus been extensively explored across plant species1,3,9,11,51. In the present study, based on both in situ observation and common garden experiment, we demonstrated two distinct morphological groups, corresponding to O. rufipogon and O. nivara, and substantial phenotypic variation within species in many traits, consistent with previous studies20,25,27,28,29,42,50. Specifically, O. nivara has a shorter culm, early flowering, shorter and less exserted panicles with more compactness and has more spikelets per panicle with shorter anthers. In addition, we examined the role of spatial isolation in the population/species divergence in terms of isolation by distance3,77 because the association of morphological divergence with both habitats and geography was reported previously25,27,28,49,50,78,79. Analysis of correlation between morphological divergence and habitat and geographic distance suggests that habitat difference rather than spatial isolation is the main explanation for the phenotypic divergence between O. rufipogon and O. nivara in Sri Lanka. Particularly, our results indicate a substantial link between the variation of morphological traits and several major climatic/environmental factors including altitude, temperature and rainfall, consistent with previous studies22,25,42,71. In response to these factors, both species have evolved distinct phenotypes to adapt to their environments and expanded in range, which was supported by previous argument that significant range expansions took place for the two species in Sri Lanka since the last interglacial around 120–140 thousand years ago48. For instance, predominantly vegetative propagation, longer and more spreading culms make O. rufipogon better adapt to permanently inundated habitats; whereas being annual, shorter and less decumbent culms of O. nivara are more suitable for seasonally dry habitats25,27,50. It is noted that our previous study based on SSR markers on the same set of Sri Lankan populations detected significant genetic differentiation between species and indicated that isolation by adaptation (IBA) rather than isolation by distance (IBD) contributed to the genetic divergence48. Therefore, congruent with ecological speciation24,27,40,48, the differences in genetic constitution and morphological traits between O. rufipogon and O. nivara are mainly explained by distinct habitats they grow and have important implications for their ecological adaptation and evolutionary dynamics.

Natural selection is a fundamental mechanism of evolution that shapes the genetic makeup of populations over time, allowing for adaptations to specific environmental conditions1,11,51. Although studies of phenotypic divergence between O. rufipogon and O. nivara have been conducted both on the global and at the local scales25,27,28,49,50,78, most of them were unable to examine if the divergence was adaptive or under natural selection. Based on 60 samples that presented two species morphology, Guo et al.29 detected significant differentiation between species for 10 morphological traits and suggested that they were adaptive. By further studying six sympatric pairs of O. rufipogon and O. nivara populations using variance analyses and QST-FST comparison, Cai et al.27 revealed significant differentiation between species for 14 of 18 quantitative traits and demonstrated the role of natural selection during phenotypic divergence between species. Nevertheless, it is unknown if similar divergence patterns for morphological traits occur and whether natural selection has played a role in morphological divergence between the two species in Sri Lanka in which parapatric rather than sympatric distribution was found for the O. rufipogon and O. nivara populations (Fig. 1). Here we identified seven phenotypic traits that exhibited significant differences between the two species and demonstrated that their divergences between species were most likely to be driven by natural selection, as reported previously in studies of sympatric populations of O. rufipogon and O. nivara27. These traits are related to reproduction and habitat preference and might have evolved in response to varying biotic and abiotic factors, such as reproductive success, water availability, resource competition, or climatic factors. Collectively, the survival and persistence of O. rufipogon and O. nivara in Sri Lanka are affected by both genetics and environments and natural selection plays an important role in local adaptation and divergence. Further research should focus on elucidation of the genetic basis for the observed phenotypic variation and exploration of the effects of environmental conditions on variation of these traits.