Rapid evolutionary divergence and ecotypic diversification of germination behavior in weedy rice populations


  • Han-Bing Xia,

    1. Department of Ecology and Evolutionary Biology, Key Laboratory for Biodiversity Science and Ecological Engineering, Fudan University, Ministry of Education, Handan Road 220, Shanghai 200433, China
    Search for more papers by this author
    • These authors contributed equally to this work.

  • Hui Xia,

    1. Department of Ecology and Evolutionary Biology, Key Laboratory for Biodiversity Science and Ecological Engineering, Fudan University, Ministry of Education, Handan Road 220, Shanghai 200433, China
    Search for more papers by this author
    • These authors contributed equally to this work.

  • Norman C. Ellstrand,

    1. Department of Botany and Plant Sciences, Center for Conservation Biology, and Center for Invasive Species Research, University of California, Riverside, CA 92521-0124, USA
    Search for more papers by this author
  • Chao Yang,

    1. Department of Ecology and Evolutionary Biology, Key Laboratory for Biodiversity Science and Ecological Engineering, Fudan University, Ministry of Education, Handan Road 220, Shanghai 200433, China
    Search for more papers by this author
  • Bao-Rong Lu

    1. Department of Ecology and Evolutionary Biology, Key Laboratory for Biodiversity Science and Ecological Engineering, Fudan University, Ministry of Education, Handan Road 220, Shanghai 200433, China
    Search for more papers by this author

Author for correspondence:
Bao-Rong Lu
Tel: +86 21 65643668
Email: brlu@fudan.edu.cn


  • Feral plants have evolved from well-studied crops, providing good systems for elucidation of how weediness evolves. As yet, they have been largely neglected for this purpose. The evolution of weediness can occur by simple back mutations in domestication genes (domestication in reverse). Whether the evolutionary steps to weediness always occur in reverse remains largely unknown.
  • We examined seed germination behavior in recently evolved weedy rice (Oryza sativa f. spontanea) populations and their coexisting cultivars in eastern and north-eastern China to address whether ‘dedomestication’ is the simple reverse of domestication.
  • We found that these weedy populations did not diverge from their progenitors by reverting to the pre-domestication trait of seed dormancy. Instead, they have evolved a novel mechanism to avoid growing in inappropriate environments via changes in critical temperature cues for seed germination. Furthermore, we found evidence for subsequent ecotypic divergence of these populations such that the critical temperature for germination correlates with the local habitat temperature at latitudinal gradients.
  • The origins of problematic plant species, weeds and invasives, have already been studied in detail. These plants can thus be used as systems for studying rapid evolution. To determine whether and how that evolution is adaptive, experiments such as those described here can be performed.


Feral plants are descendants of domesticated plants that have evolved the ability to persist and reproduce without direct human care. Feral plant evolution, also known as ‘dedomestication’, can occur either directly from domesticated ancestors (‘endoferality’) or by hybridization of domesticates with wild relatives (‘exoferality’) (Gressel, 2005a). Dedomestication may produce problematic plants such as invasives and weeds. Ellstrand et al. (2010) identified 13 well-documented cases of problematic plants that evolved from domesticated ancestors. Because such feral plants evolved from well-studied taxa, they should be good model systems for elucidating how weediness or invasiveness evolves. Furthermore, like invasive species, because their histories are generally well known, they can be used as systems for studying rapid adaptive evolution. As yet, they have been largely neglected for this purpose; most studies have focused on simply documenting plants as feral, rather than how ‘ferality’ evolved (Ellstrand et al., 2010).

Often part of the story is obvious. Many feral plants have evolved shattering (seed dispersal) from nonshattering crop ancestors (Ellstrand et al., 2010). However, the trait of shattering alone may rarely suffice to permit a seed dispersing ‘crop’ to persist and reproduce without human intervention. A shattering ‘crop’ will need more features to persist in the agro-ecosystem, such as the ability to survive in soil seed-banks, to avoid human weeding, and to successfully compete with coexisting crops. These selective pressures probably serve as the driving force for the adaptive evolution of feral plants. An example is weedy rice (Oryza sativa f. spontanea), some populations of which evolved directly from cultivated rice (Oryza sativa) (Cao et al., 2006; Reagon et al., 2010; Thurber et al., 2010). Weedy rice is vegetatively very similar to cultivated rice but has some key differences: shattering seed dispersal, red pericarp pigmentation, and the ability of seeds to persist in the soil (Delouche et al., 2007). Weedy rice is a noxious weed of cultivated rice world-wide (chapters 16–21 in Gressel, 2005b; Delouche et al., 2007), both causing crop yield losses and degrading the quality of rice when co-harvested (Gressel, 2005b; Delouche et al., 2007).

Weedy rice is polyphyletic. Some populations are descended directly from cultivated rice, both from indica-type (mostly pan-tropical) cultivars (Londo & Schaal, 2007) and from japonica-type (mostly temperate) cultivars (Cao et al., 2006, 2009; Vaughan et al., 2008), as well as indica × japonica hybrids (Ishikawa et al., 2005). Furthermore, some populations have an exoferal origin as hybrid descendants of hybrids between cultivated rice and its wild ancestor Oryza rufipogon (Londo & Schaal, 2007; Suh, 2008), from which cultivated rice was domesticated from O. rufipogon c. 10 000 (8000–11 500) yr ago in the middle and lower parts of the Yangtze River valley (Normile, 1997; Zong et al., 2007). The exoferal lineages regain the strong seed-shattering and seed-dormancy traits of the wild ancestor that were lost. However, during the domestication process, the wild ancestor lost seed shattering and dormancy, resulted in cultivated rice. The former is commonly grown in pan-tropical regions and the latter in temperate regions where no wild ancestors are found (Vaughan et al., 2008).

Recent genomic studies (Gross et al., 2010; Reagon et al., 2010; Thurber et al., 2010) have focused on some simple genetic changes that have occurred to account for the evolution of shattering and grain pigmentation in US weedy rice populations. However, the weedy rice material examined evolved so long ago that evolutionary insights from these studies are limited (Londo & Schaal, 2007).

To better understand the rapid evolution of weediness in weedy rice, more recently evolved populations should be studied. Thus, we focused on some newly evolved populations whose origins are better known (Cao et al., 2006; Xia et al., 2011). After many years of successful suppression, within the last 20 yr weedy rice emerged as a problem in the rice fields of north-eastern and eastern (henceforth NEE) China following changes in cropping style and reduced weed control (Yu et al., 2005; Cao et al., 2006; Xia et al., 2011). Motivated by this pest’s resurgence, Cao et al. (2006) sought its evolutionary origin, comparing marker loci from weedy rice populations collected from China’s Liaoning province with those of wild O. rufipogon as well as selected japonica and indica accessions. The weedy rice populations had the closest affinity to a local Liaoning cultivar (a japonica type). The authors concluded that ‘weedy rice populations from Liaoning most probably originated from Liaoning rice varieties by mutation and intervarietal hybrids’.

Given that the evolutionary pathway to weediness in these populations is now known, we sought to examine how weediness evolved in them. Specifically, we sought to determine the evolution of seed germination behavior in these populations in relation to their responses to germination temperatures by comparing them with their cultivated ancestors, as well as weedy rice populations from elsewhere.

Nearly all mature rice cultivars have no seed dormancy and germinate easily under an array of conditions (Grist, 1996; also, this study). By contrast, until recently, strong seed dormancy was thought be the rule for weedy rice (Oard et al., 2000; Gu et al., 2003, 2005a; Gianinetti & Cohn, 2008), with a few exceptions (e.g. Schwanke et al., 2008). What is the nature of dormancy in the new weedy populations? For our purposes, we follow Baskin & Baskin’s (2004) definition of seed dormancy: ‘A dormant seed (or other germination unit) is one that does not have the capacity to germinate in a specified period of time under any combination of normal physical environmental factors (temperature, light/dark, etc.) that otherwise is favourable for its germination’ (similar to that of Finch-Savage & Leubner-Metzger, 2006). However, c. 60% of plant species do not have seed dormancy (Baskin & Baskin, 1998), and seed germination of these species must be regulated by other mechanisms. Previous studies indicated that optimal seed germination of many nondormant species was closely associated with their habitat temperatures. For example, seed germination of spices located at high latitudes requires a higher temperature than that of spices located at low latitudes (Fenner & Thompson, 2005).

A recent report on nondormant weedy rice (Delouche et al., 2007) prompted us to survey populations of weedy rice in China and elsewhere for nondormancy. Our preliminary field survey of weedy rice in temperate regions in NEE China suggested extremely low or no seed dormancy for nearly all populations (H. B. Xia et al., pers. obs.). We further studied these populations and asked the following questions. Does the nondormancy found in these recently evolved temperate populations in NEE China hold true for other, older temperate weedy rice populations, for example in North America and southern Europe? What fraction of their nondormant seeds can survive through winter? Given that these populations span a considerable latitudinal gradient, have they evolved a corresponding range of temperature-dependent germination cues matching their eco-geographic distribution? To address these questions, we first assessed seed dormancy for weedy rice populations collected from different regions, then tested the winter survival of nondormant weedy rice seeds from temperate regions, and finally examined germination from nondormant temperate weedy rice populations under different temperatures.

Materials and Methods

Weedy and cultivated rice sampling

We determined the nature of primary seed dormancy in 35 weedy rice Oryza sativa f. spontanea L. populations with various origins (Supporting Information Table S1). We collected seeds directly from natural populations in China (P1–P25) in 2004–2007 by sampling from at least 30 maternal plants at each site, and placed individual seed families in separate bags. The others were donated by various collaborators.

Of the Chinese populations, we randomly selected four (P1, P10, P11 and P20; Table S1), representing four provinces, to attempt to induce secondary dormancy and for burial experiments. We selected 18 NEE Chinese populations from six regions with a range of latitudes (c. 32–48°N) to examine their seed germination under a temperature gradient (Fig. 1, Table S1). Eighteen rice cultivars (C1–C18; Table S1) coexisting in the same fields as the 18 aforementioned weedy populations were collected to test their germination responses to the same temperature gradient.

Figure 1.

Map of the 18 weedy rice (Oryza sativa f. spontanea) populations (circles) and 18 coexisting rice (Oryza sativa) varieties in the same field collected in China and used for the gradient-temperature seed germination experiments, representing a latitudinal gradient. Six weather stations are represented by triangles.

Seed dormancy determination

Testing for primary dormancy  The seed collections came from different environments and from different years. To eliminate maternal effects, environmental effects, and effects of varying seed age and storage, we first grew all our weedy rice collections in a common garden in a standard paddy field on Fudan University’s campus in Shanghai, China, in 2007. For each of our collected populations, we randomly chose one seed to plant from each of 15 different maternal seed families. We randomly chose 15 seeds from each of the donated populations.

Mature seed families were harvested from 10 to 15 individuals of each population and used to measure successful germination fractions. Exactly 6 d after harvest, 90 seeds representing each sampled population were divided into three replicates and placed on moist filter paper in Petri dishes. The dishes were put into a growth chamber at a constant temperature of 28°C with a light : dark cycle of 16 : 8 h; these conditions are known to be ideal for cultivated rice seed germination. Successful germination was recorded 18 d later following Baskin & Baskin (1998).

Testing for secondary dormancy in recently evolved populations  Two months after initial collection, weedy rice seeds from P1, P10, P11 and P20 (200–800 per population) from NEE China were exposed to shock treatments to induce secondary dormancy: (1) in soil at 4°C for 30 d; (2) in soil at −20°C for 20 d; (3) in soil at −20°C for 100 d and then at 4°C for 20 d; (4) under 2-cm-deep soil for 20 d; and (5) on moist filter paper at 4°C for 30 d. For the soil treatments, we used ordinary moist rice paddy soil transferred from a rice field. Following the treatments, seeds were placed on moist filter paper in Petri dishes which were put into a growth chamber at a constant temperature of 28°C with a light : dark cycle of 16 : 8 h to examine dormancy following the procedure of Baskin & Baskin (1998).

Seed burial experiment

To determine seed survival through winter, seeds from four populations were buried in a Fudan University fallow rice field which had been allowed to dry to soil moisture contents of 20–40% and subsequently harrowed. Seeds from population 11 were tested from November 2005 to February 2006; seeds from P1, P10 and P20 were tested from December 2006 to March 2007. For each population, eight bags (replicates), each containing 50–80 intact seeds, were buried c. 2 cm deep. Two bags from each population were dug up after 20, 40, 80 and 100 d of burial and subjected to germination conditions following Baskin & Baskin (1998) as detailed above. The ground temperature was recorded daily during the experiments conducted in the campus of Fudan University, Shanghai.

Seed germination under a temperature gradient

To determine how germination varies with temperature, we germinated seeds of the 18 rice varieties collected from varying latitudes in a dark incubator with controlled temperature regimes at 11, 12, 14 and 28°C, respectively. To estimate the finer scale critical temperature (CT) that might prevent maladaptive premature germination in the soil seed bank, seeds collected directly from 18 weedy rice populations from varying latitudes (Fig. 1) were placed in a dark incubator with controlled temperature regimes at 8, 9, 10, 11, 12 and 14°C, respectively. For each treatment, 90 seeds divided into three replicates were subjected to germination conditions separately, following Baskin & Baskin (1998) with a light : dark cycle of 16 : 8 h, as described above. CT was arbitrarily determined as the temperature at which 10% of the seeds germinated. Two-way ANOVA was conducted to analyze the effects of temperature and source population latitude on seed germination. For detailed analysis of differences in germination fraction among the six groups of weedy rice populations from sampling sites over diverse latitudes, one-way ANOVA was conducted using the Duncan model in spss ver. 12.0 software (2003; SPSS Inc., Chicago, IL, USA).

Correlation of the critical temperature and habitat temperature

To test whether CT for seed germination under laboratory conditions varied with the habitat temperature (HT) of the weedy rice source locations, the correlation of CT or HT with the latitude of the collection site was analyzed using a linear regression model (Johnson & Wichern, 1998). We judged HT to be the average temperature of 18 d following rice harvest, because weedy rice seeds require that much time to ripen to achieve full germination. Thus, we estimated HT by calculating the 10-yr (1995–2004) ground temperature average for the 18 d post-harvest for the sample collection locations. The temperature data was collected from six weather stations of the China Meteorological Data Sharing Service System (http://cdc.cma.gov.cn/) that geographically correspond to the extent of weedy rice source regions (Fig. 1). Student’s t-test was used to analyze the consistency of the two correlation slopes between CT or HT and latitude (Zhang & Zhang, 2002). In addition, an ANCOVA of HT (used as a covariate) with CT (used as a dependent variance) was conducted to confirm the consistency of HT and CT. All the calculations were performed using the software spss ver. 12.0.


Dormancy of cultivated and weedy rice seeds

All rice cultivars tested germinated at 28°C after being collected directly from the field (upper graph, Fig. 2). In accordance with previous observations (e.g. Grist, 1996), we did not observe primary seed dormancy in cultivated rice.

Figure 2.

Germination ratios of 18 rice (Oryza sativa) varieties and their corresponding 18 weedy rice (Oryza sativa f. spontanea) populations from various latitudes at different temperatures. Each graph represents germination at temperatures of 11, 12, 14 and 28°C for cultivated rice (upper panels), and 8, 9, 10, 11, 12 and 14°C for weedy rice (lower panels), respectively. The vertical bars indicate the standard error of the mean. The dashed line in each chart indicates the critical temperature (CT), at which 10% of seeds germinated. Cultivars and weedy populations were arranged from left to right to correspond to north to south. For detailed locations, refer to Fig. 1.

By contrast, the germination behavior at 28°C of weedy rice seeds from 35 geographically diverse populations (Table S1) varied from 1 to 100% (Fig. 3). In general, weedy rice populations from the temperate regions had high seed germination percentages, mostly close to 100%, but with a few as low as c. 50%. The average for the temperate populations was 86.8% (Fig. 3); that is, weak or no primary seed dormancy. By contrast, the tropical weedy rice populations had very low seed germination ratios, ranging from c. 1 to 20%, with an average of 8.9% (Fig. 3), indicating relatively strong primary seed dormancy.

Figure 3.

Germination ratios at 28°C of 35 weedy rice (Oryza sativa f. spontanea) populations from temperate (circles) and tropical (triangles) regions 6 d after harvest used to evaluate primary seed dormancy. Refer to Supporting Information Table S1 for the specific origin of each population. Vertical bars indicate ± SE of the mean.

Seeds from four randomly selected temperate populations (P1, P10, P11 and P20) were subjected to treatments to test for induced secondary dormancy. We did not detect induced secondary seed dormancy. Average seed germination ratios following these treatments were about the same as those observed for primary germination for these populations (Table S2).

Survival of weedy rice seeds through winter in the soil

Seed burial experiments revealed that seeds from temperate weedy rice populations survived and germinated through Shanghai’s winter conditions at high enough ratios to sustain their populations. In the first-year experiment (P11 only), c. 30% of the weedy rice seeds survived after burial in soil for 100 d (Table 1). The average temperature was 7.1°C with a minimum of −4.0°C during this experiment (17 November 2005 to 25 February 2006). Similarly, in the second-year experiment (P1, P10 and P20), c. 32–84% of the weedy rice seeds survived to germinate after burial for a maximum of 100 d (Table 1). During this experiment (15 December 2006 to 23 March 2007), the average temperature was 8.3°C with a minimum of −2.0°C. Most remaining nongerminating seeds at the end of the experiments were rotten.

Table 1.   Seed survival and subsequent germination ratios of four randomly selected weedy rice (Oryza sativa f. spontanea) populations after varying times buried in the soil
PopulationSeed germination ratio (%) after different burial periods (d)a
  1. aNumbers in parentheses indicate ranges; most nongerminating seeds had rotted in the soil.

  2. bData not available because of insufficient seeds for this population.

P197.5 (96.9–98.1)83.1 (80.9–85.1)88.2 (83.3–93.1)83.7 (80.0–89.4)
P1058.9 (53.4–64.4)45.0 (44.3–45.7)31.3 (30.4–32.5)b
P1164.0 (46.9–81.1)37.2 (31.0–43.4)38.1 (32.0–44.2)29.8 (12.0–47.6)
P2066.6 (64.0–69.2)56.1 (55.6–56.5)55.7 (53.7–57.7)32.0 (29.4–34.5)

Seed germination under different temperatures

Not one rice cultivar in this study germinated at 11°C (upper graph, Fig. 2). Most showed some germination at 12°C; at 14°C all varieties showed some germination (but only half reached 50% germination ratios). At temperatures between no germination (11°C) and complete germination (28°C), rice cultivars had similar seed germination behavior which did not co-vary with their latitude of origin.

Similar to cultivated rice, weedy rice seed germination responses were more or less uniform at the most extreme temperatures tested. At the lowest temperature (8°C), the samples showed very little (P2 and P3) or no (the others) germination; at the highest temperature (14°C), all samples showed very high germination (> 90%) (lower graph, Fig. 2). In contrast to cultivated rice, weedy rice seed germination was significantly influenced by the experimental temperature (P < 0.001), the latitude of the sampling site (P < 0.001), and their interaction (P < 0.001), based on the two-way ANOVA. Weedy rice seed germination varied significantly and systematically under the intermediate temperature regimes (9, 10, 11 and 12°C) (Table 2) such that the populations from the higher latitudes (P1–P3 and P8–P10) germinated more readily at lower temperatures (starting at 9°C) than the populations from the mid or lower latitudes. Indeed, germination for these high-latitude populations increased with increasing temperature to a maximum ratio at 14°C. In contrast, populations from the lowest latitudes collected (P17–P25) showed substantial germination at 11°C and above; their germination ratios also increased to a maximum at 14°C. Populations from the mid range of samples (P11–P13) showed a germination response intermediate between those of the others (lower graph, Fig. 2). Overall, as a population’s latitude increased, its ability to germinate at lower temperatures also increased. The following section examines this relationship more closely.

Table 2.   Differences in the germination ratios of different groups of weedy rice (Oryza sativa f. spontanea) populations along a latitudinal gradient over different temperature regimes as indicated by the one-way ANOVA (Duncan model)
Weedy rice populationExperimental temperature (°C)
  1. Different capital letters indicate significant differences among populations at different latitudes. A significant difference among all populations is indicated by the P-values (last row). The degrees of freedom between groups and within groups and the total degrees of freedom were 5, 12 and 17, respectively.

Group 1: P1–P3NSAAAANS
Group 2: P8–P10ABABA
Group 3: P11–P13BCAA
Group 4: P17–P19BDBAB
Group 5: P20–P22BCDABB
Group 6: P23–P25BDBB
P1–P25: P value0.1230.0000.0000.0350.0160.062
(F value)(2.191)(12.818)(76.036)(3.498)(4.404)(1.953)

Correlations among CT, latitude and HT in weedy rice

Linear regression analysis revealed a significant negative correlation between CT (the temperature at which there was 10% seed germination) determined in the temperature gradient experiment (lower graph, Fig. 2) and the source latitude for the weedy rice populations (r2 = 0.672, < 0.001) (Fig. 4). The same relationship was found between HT at the weedy rice collection sites and source latitude (r2 = 0.798, = 0.017) (Fig. 4). Based on these regressions, the CT required for weedy rice seed germination was 2–3°C higher than the HT (Fig. 4), indicating that HT in every location immediately after rice harvesting was too cold to allow germination of the local weedy rice population.

Figure 4.

Changes in critical seed germination temperature (CT; red circles) and habitat temperature (HT; blue triangles) for 18 weedy rice (Oryza sativa f. spontanea) populations across a latitudinal gradient. The y-axis represents the temperature: either CT (y = −0.156x + 16.330; r2 = 0.672) or HT (y = −0.220x + 15.682; r2 = 0.798). Bars indicate ± SE.

The ANCOVA in which HT was used as a covariate and CT as a dependent variable showed that HT had no significant interaction with CT (= 0.339), indicating that HT and CT had the same tendency to change with latitude. Further analysis demonstrated that the slopes generated from CT and HT vs latitude were essentially consistent (Student t-test: = 0.252, df = 20, > 0.05) (Fig. 4).


We confirmed that there was strong seed dormancy for diverse weedy rice populations from tropical regions. By contrast, we found that temperate weedy rice populations (from China, South Korea and Italy) had little or no primary dormancy. One previous study also found that certain weedy rice populations from temperate rice planting regions have either extremely weak or no seed dormancy (Delouche et al., 2007).

Differences in seed dormancy in tropical vs temperate weedy rice may have to do with their different evolutionary origins. The temperate populations examined here (Cao et al., 2006) and others (e.g. Ishikawa et al., 2005; Londo & Schaal, 2007; Reagon et al., 2010; Thurber et al., 2010) are known to have evolved directly from domesticated rice (without hybridization with a wild ancestor). Like most cereals, cultivated rice typically has no seed dormancy (Chang & Yen, 1969; Cai & Morishima, 2000; Gu et al., 2004, 2005b). By contrast, the seed dormancy of tropical weedy rice may be attributable to a hybrid ancestry involving a wild species that donated genes for dormancy (Cai & Morishima, 2000; Gu et al., 2003; Veasey et al., 2004).

Our germination experiments on recently evolved temperate weedy rice populations from NEE China demonstrated that a variety of treatments did not induce secondary dormancy in these weedy rice seeds. Similarly, our burial experiments revealed that seeds from these populations survived to germinate through the winter at high enough ratios to sustain their populations. This finding is supported by those of other studies in northeastern China, where > 20% of weedy rice seeds (probably, but not definitely, nondormant) survived after winter burial (Meng, 2004; Yuan et al., 2006). Clearly, nondormant temperate weedy rice seeds can survive through winter in soil seed-banks until they receive appropriate germination cues.

Further experiments revealed that weedy rice seeds rely on a CT as a germination cue. Specifically, our results show that, at the time at which these weedy rice seeds disperse, the local temperature is too low for germination. The local average temperature gradually decreases after rice is harvested in autumn. The temperature remains below the CT from autumn until the next spring. Weedy rice seeds germinate when high enough temperatures return at that time.

We found evidence for ecotypic differentiation for germination temperature cues with positive correlations among a population’s CT, its local HT and its latitude. Populations from cooler, more northerly latitudes tended to germinate at lower temperatures than those from warmer, more southerly latitudes. By contrast, we did not find such a relationship for rice cultivars from corresponding latitudes.

Interestingly, all weedy populations germinated at temperatures too low for substantial germination of the cultivars. Most weedy populations had high germination ratios at 12°C, a temperature at which most cultivars showed low germination ratios. At 14°C, all weedy rice populations had > 90% germination ratios, whereas not one tested cultivar had a germination ratio as high as 90% at that temperature. These weedy rice populations have diverged evolutionarily from their cultivated ancestors with regard to seed-germination temperature cues. With a lower CT than rice culitvars, weedy rice germinates early and has a head start on directly seeded cultivated rice, resulting in a competitive advantage over rice in the field.

The nondormant weedy rice populations that we studied evolved an adaptive mechanism different from those of other weedy rice populations to regulate their seed germination to avoid growing in inappropriate environments. This evolution occurred rapidly. According to local agricultural extension services, weedy rice was first found in the sampled regions < 20 yr after rice farmers started to move away from transplanting to direct seeding and other no-till technologies. With a novel mechanism to inhibit germination under unfavorable conditions, it is not clear that continued evolution to acquiring true dormancy would present any further adaptive advantage.

Our data challenge the view that the evolution of endoferality involves back mutation at domestication loci (Gressel, 2005a); that is, domestication in reverse. During rice domestication, the evolutionary loss of seed dormancy represents the key seed germination difference between cultivated rice and its wild progenitors (Gu et al., 2004, 2005b; Veasey et al., 2004; Vaughan et al., 2008). However, the weedy rice populations that we examined evolved weediness without reverting to the seed dormancy of their wild ancestors. It appears that the immediate progenitors of weedy rice first obtained seed shattering and subsequently evolved other key traits that enhanced their ability to survive under natural conditions.

Recent genomic work in weedy rice also supports our conclusion that the evolutionary pathway to ferality need not be domestication in reverse. Most rice cultivars have white grains; nearly all weedy rice populations exhibit red pigmentation. Gross et al. (2010) examined the genetic basis of red pigmentation in US weedy rice populations that had evolved ferality before their introduction to North America. Confirming that these populations are descended from Asian cultivated rice, they found that ‘reversion of domestication alleles does not account for the pigmented grains of weed accessions,’ nor does introgression from a wild relative. Instead, they concluded that novel allelic change must account for the red pigmentation. Examining the evolution of shattering in the same populations, Thurber et al. (2010) came to a similar conclusion. They found that the nonshattering allele for cultivated rice that distinguishes it from its shattering wild progenitors was present in its weedy descendants, despite the ability of those descendents to easily disperse their seeds. Clearly, a mutation at another locus is responsible for the evolution of shattering in the weedy populations. Therefore, weedy rice pigmentation and shattering evolved via phenotypic convergence with the wild ancestor without genotypic convergence. In our case, weedy rice evolved a wholly novel seed germination behavior phenotype, different from those of both its immediate cultivated rice progenitors and its more distant wild rice ancestors. In all three cases, dedomestication proceeded without evolution at the loci involved in domestication.

Our results not only demonstrate the evolutionary divergence of weedy rice from cultivated rice, but also reveal ecological diversification of temperature cues for individual weedy rice populations to match local climatic conditions. We found that CT covaried with both HT and the latitude at which the weedy rice populations were sampled. The significant correlation between CT and HT across a wide range of recently evolved weedy rice populations indicates the rapid evolution of local adaptive differentiation of this mechanism across temperate rice-planting regions in China. Because our seed germination data from cultivars collected from geographically corresponding sites did not show any geographic or temperature-dependent pattern, introgression from local rice varieties into weedy populations can be ruled out as an evolutionary mechanism for this diversification.

Previous studies have documented high levels of variation within and among weedy rice populations (Delouche et al., 2007), but we believe ours is the first to demonstrate that adaptive differentiation has occurred rapidly. In a review of ferality, Warwick & Stewart (2005) called for studies on the speed of dedomestication. In the case of weedy rice, the close relationship between CT and local HT evolved in less than two decades.

Evidence is accumulating suggesting that rapid local adaptive evolution may be a common feature of both weeds and invasives (e.g. Maron et al., 2004; Bossdorf et al., 2005; Keller et al., 2009; but see Keller & Taylor, 2008). Common garden experiments have demonstrated that two other plants with crop ancestry evolved ecotypic differentiation in less than a century: weedy rye (Secale cereale) of the western USA, and California’s wild radish (Raphanus sativus), known as both agricultural weeds and invasives of less managed ecosystems (Burger, 2006; Burger et al., 2006; Hegde et al., 2006; Ridley & Ellstrand, 2009). Both of these evolved local adaptation in less than a century. Furthermore, common garden experiments that compare invasive plants to those collected from the native range have revealed the evolution of latitudinal differentiation for some invasives. The invasive populations of the shrub Hypericum canariense in the USA have been shown to be derived from a single source population in the Canary Islands but, despite genetic bottlenecks, have still evolved adaptive latitudinal differences in flowering phenology in less than half a century (Dluglosch & Parker, 2008). Only a handful of experimental studies report no evidence for adaptive evolution for invasive populations relative to their ancestral populations (e.g. Brodersen et al., 2008).

Because the origins of problematic plants species have already been studied in detail, these plants can be used as systems for studying rapid evolution. Feral plants are particularly good systems for this purpose if they diverged in sympatry with their progenitor, but caution must be employed when determining whether that evolutionary divergence is adaptive or not (Keller & Taylor, 2008). History, contingency, chance, and gene flow can be alternate explanations to apparent adaptation. Experimental approaches such as those described for weedy rice and the other examples given above can help distinguish among the alternatives.


This study was supported by the Chinese Ministry of Science and Technology (Grants. 2011CB100401) and the Nature Science Foundation of China (grants 30871503 and 30730066), as well as a John Simon Guggenheim Memorial Fellowship and a United States National Science Foundation OPUS Grant awarded to N.C.E. (DEB 1020799). We thank Prof. H. S. Suh of Yeungnam University, South Korea and Dr B. Basso of the University of Milan, Milan, Italy, for their donations of some weedy rice accessions.