Gene flow from cultivated rice to the wild species Oryza rufipogon under experimental field conditions


  • Zhi Ping Song,

    1. Ministry of Education Key Laboratory for Biodiversity Science and Ecological Engineering, Institute of Biodiversity Science, Fudan University, Shanghai 200433, China;
    2. School of Life Sciences, Wuhan University, Wuhan 430072, China
    Search for more papers by this author
  • Bao-Rong Lu,

    1. Ministry of Education Key Laboratory for Biodiversity Science and Ecological Engineering, Institute of Biodiversity Science, Fudan University, Shanghai 200433, China;
    Search for more papers by this author
  • Ying Guo Zhu,

    1. School of Life Sciences, Wuhan University, Wuhan 430072, China
    Search for more papers by this author
  • Jia Kuan Chen

    Corresponding author
    1. Ministry of Education Key Laboratory for Biodiversity Science and Ecological Engineering, Institute of Biodiversity Science, Fudan University, Shanghai 200433, China;
    Search for more papers by this author

Author for correspondence: Jia Kuan Chen Tel: +86 02165642468 Fax: +86 02165642468 Email:


  • •   Here, the gene flow from a cultivated rice variety (Minghui-63) to common wild rice (Oryza rufipogon) was investigated to assess the biosafety risk associated with the environmental release of transgenic varieties.
  • •   Four experimental designs differing in the spatial arrangement of the Minghui-63 and O. rufipogon plants were used in experiments conducted in an isolated rice field in Hunan Province, southern China, where O. rufipogon occurs naturally.
  • •   Natural hybridization events between the two species were detected by scoring a simple sequence repeat (SSR) molecular marker. A total of 296 hybrids were identified from 23 776 seedlings that were randomly germinated from > 80 000 seeds collected from O. rufipogon. The occurrence of the crop-to-wild gene flow was significantly associated with wind direction and frequencies of the gene flow, which decreased significantly with distance from the pollen sources. The maximum observed distance of gene flow was 43.2 m.
  • •   The results indicated that gene flow from cultivated rice to O. rufipogon occurred at a considerable rate. Therefore, isolation measures should be considered when deploying transgenic rice in the sympatric regions of the wild rice, and when establishing in situ conservation of O. rufipogon. The experimental system in this study can be used for biosafety assessment of transgene escape of other wind-pollinated crops.


Many major crops are sexually compatible with their weedy types and wild relatives, leading to so-called crop-to-weed or crop-to-wild gene flow in their sympatric regions (Arnold, 1992, 1997; Rieseberg, 1995; Arriola & Ellstrand, 1996). Gene flow from crop taxa can add new genes into wild populations, which then can re-assort into novel combinations and therefore may have a substantial impact on the evolution of wild populations (Arriola & Ellstrand, 1996; Arnold, 1997; Ellstrand et al., 1999; Jarvis & Hodgkin, 1999). This might result in more aggressive weeds (Colwell et al., 1985; Langevin et al., 1990; Arias & Rieseberg, 1994; Arriola & Ellstrand, 1996) and/or lead to extinction of rare and endangered species (Ellstrand & Elam, 1993; Rhymer & Simberloff, 1996). Transgenic crop varieties are now grown worldwide and the area planted to these varieties continues to increase rapidly. Transgenic crops can greatly benefit humanity in terms of world food security, but there are biosafety concerns related to tranegene escape from genetically modified crops to wild populations through crop-to-wild gene flow (Dale, 1992; Gray & Raybould, 1998). A better understanding of crop-to-wild gene flow is therefore essential for ecological risk assessment of transgene escape (Raybould, 1999; Auberson, 2000), in addition to its importance for strategic conservation of irreplaceable genetic resources (Bartsch et al., 1999; Storfer, 1999).

Rice is one of the most important crops in the world, providing staple food for over 50% of the world's population. In order to feed the increasing global population, there is a need for rice varieties with enhanced yield, resistance to pests or pathogens, and tolerance to environmental stress. It is very likely that such varieties will be produced by biotechnology, including genetic engineering, and that these varieties will be widely grown. The common wild rice Oryza rufipogon is the putative ancestor of the Asian cultivated rice (O. sativa), and the most important genetic resources for rice improvement in terms of its accessibility from gene transfer through sexual means (Oka, 1988). It is widely distributed in South and South-east Asian countries, including China where O. sativa is a major staple food and extensively cultivated. Natural hybridization between O. rufipogon and O. sativa has been reported in many locations (Langevin et al., 1990; Majumder et al., 1997; Suh et al., 1997; Lu, 1999; Song et al., 2002), indicating that O. rufipogon has high compatibility with cultivated rice. Therefore, transgenes from the transgenic rice might be captured by this wild rice through outcrossing. The resulting interspecific hybrids might serve as a bridge from which transgenes could further spread to other wild relatives. Little is known about the frequency and range of gene flow between the two species, although several reports mentioned the occurrence of gene flow between the cultivated and wild rice. Understanding of gene flow between the cultivated rice and O. rufipogon would allow us to minimize ecological risks caused by the transgenic escape by properly regulating and managing the transgenic rice by, for example, establishing a safety distance.

Gene flow can be estimated using direct or indirect methods. Pollen and seed, for flowering plants, are the two major vectors of gene flow. In direct methods, observations of pollen flow and/or seed dispersal are used to estimate range of gene flow (Nilsson et al., 1992). Much of the work on gene flow has been conducted using indirect methods, such as allele frequencies or paternity analysis where different isozyme systems or DNA assays were applied (Snow & Parker, 1998). Data of gene flow measured under natural conditions is significant for risk assessment of transgenes escape. However, it is somehow difficult to determine precisely gene flow under field conditions, even using specific molecular markers, because gene flow can be influenced by a number of factors such as timing, population size and background (Bossart & Prowell, 1998; Jarvis & Hodgkin, 1999). A designed experimental system simulating natural conditions can minimize the influence of such factors and allow us to measure the frequency and range of gene flow with the assistance of specific molecular markers easily (Crawley et al., 1993; Messeguer et al., 2001).

This study reports experiments estimating gene flow from cultivated rice to O. rufipogon conducted under typical cultivation conditions in which distribution patterns of wild rice in variable sizes of patches in the vicinity of cultivated rice are commonly found in Asian countries, particularly in regions with moderate scale of rice cultivation by small farmers. The experiments made use of simple sequence repeat (SSR) markers, that have been applied in many other studies of gene flow (Wu & Tanksley, 1993; Chase et al., 1996), to identify the incidence of hybridization. The objectives of the study were to investigate the frequency and range of gene flow from cultivated rice to O. rufipogon under the typical cultivation conditions described above.

Materials and Methods

Plant materials

The perennial wild rice Oryza rufipogon Griff. was collected from the Chaling population (26°50′ N, 113°40′ E) of Hunan Province, China and used as the pollen recipient. Vegetative stocks were collected to produce sufficient tillers to maintain individuals with uniform genotypes and consistent growing and reproductive performance in the experiments. Seeds of the rice variety Minghui-63 (a semidwarf cultivar with a growth period of 90–95 d) was donated by Prof. S. M. Mu from Hubei Academy of Agricultural Science (Wuhan City, China) and used as the pollen donor. Minghui-63 was selected due to its special features, such as perfect compatibility with O. rufipogon (Song et al., 2002), production of a relatively large number of pollen grains, and insensitivity to temperature and photoperiod. The flowering time of Minghui-63 was easily adjusted by different sowing time to match that of O. rufipogon, for which the flowering time was difficult to modify.

Experimental design

Four types of experimental populations were designed in this study (Fig. 1a–d), that is, central population combination (CPC), encircle population combination (EPC), alternating population combination (APC), and unidirectional population combination (UPC). The design mimicked the natural occurrence of O. rufipogon as variable sizes of patches in the rice ecosystems in Asia. The CPC was designed to determine the influences of wind and sizes of pollen source to the interspecific gene flow. The EPC was designed to compare the location effect of recipient (with the CPC) on the gene flow. The APC aimed to measure the maximum gene flow under the typical cultivation conditions. The UPC allowed us to gain information on distance of gene flow in a relatively small-scale, and combined with the CPC, UPC, to evaluate relationship between distance and gene flow.

Figure 1.

Illustration of experimental designs. (a) Central population combination (CPC); (b) Encircled population combination (EPC); (c) Alternating population combination (APC); (d) Unidirectional population combination (UPC). Stars represent Oryza rufipogon; shaded circles or bars indicate Minghui-63.

Central Population Combination (Fig. 1a) This design included five populations, CPC-1, CPC-2, CPC-3, CPC-4, and CPC-5. The CPC consisted of a central circular plot of Minghui-63 with a radius of 0.3 m (CPC-1), 0.6 m (CPC-2), 1.2 m (CPC-3), 2.4 m (CPC-4), or 4.8 m (CPC-5), surrounded by concentric circles of O. rufipogon. In each CPC, five O. rufipogon circles surrounded the Minghui-63 plot, the distance of the O. rufipogon circles from the edge of the Minghui-63 plot were from r to 5r, and O. rufipogon individuals within each circle were separated by r distances where r is the radius of the plot of Minghui-63. In CPC-4 (Minghui-63 plot radius = 2.4 m) and CPC-5 (Minghui-63 plot radius = 4.8 m), uniformly 1.2 m distant apart O. rufipogon circles were planted additionally within 6 m from Minghui-63 plot, and distance between O. rufipogon individuals within the additional circles was uniformly 1.2 m. Thus there were three and four additional O. rufipogon circles in the CPC-4 and CPC-5, receptively.

Encircled Population Combination (Fig. 1b) This design included 3 populations, EPC-1, EPC-2, and EPC-3. The EPC was composed of a 1-meter-diameter Minghui-63 circular plot encircled by a polygon-shaped plot of O. rufipogon with 19 individuals spaced uniformly 0.6 m apart. The distances between Minghui-63 circle and the margin of O. rufipogon core were 1.2 m (EPC-1), 2.4 m (EPC-2), and 3.6 m (EPC-3).

Alternating Population Combination (Fig. 1c) This design had only one population with three replicates. The APC consisted of a plot of 5 × 5 m2 with alternating rows of O. rufipogon and Minghui-63. Distance between the rows was 0.5 m and between hills in rows was 0.3 m.

Unidirectional Population Combination (UPC, one population) In the UPC, there was a 6 × 10 m plot of Minghui-63. Fifteen rows of O. rufipogon with 4.8 m between each row were planted downwind from the Minghui-63 plot. Each row consisted of five individuals spaced 2 m apart. Only one unidirectional population combination was constructed due to space limitations of the experimental area.

The study site was a rice field about 3 km from the Chaling O. rufipogon population. It was surrounded by hills running in a north–south orientation, which assured that there was no alien pollen contamination from other rice fields in the experimental system and minimized the impacts on O. rufipogon reproductive performance due to habitat change. The distance between the experimental population combinations was > 50 m to avoid interference of each other. The CPC, EPC, APC experiments were conducted in 1999 and 2000. The UPC was conducted in 2000 at the same site, based on the observations in 1999, in which a wind direction from north to south was measured during the flowering period. The O. rufipogon samples were planted in March of each year to produce tillers before construction of experimental populations. Five O. rufipogon tillers of about the same height were sampled together as one individual, and transplanted into the experimental plots in mid-July Minghui-63 seeds were germinated in mid-June and seedlings were transplanted with 10 × 18 cm spacing in mid-July. The density of the cultivar plot in all designs held constant. Fields management was conducted similar to that of normal rice cultivation.

Survey of flowering time and seed set

The dates of first and last flowering were recorded to estimate flower period of the cultivated and wild rice species. The percentage of flowering panicles was calculated over 10 randomly scored individuals every day. The peak of flowering period was defined as the duration when over 20% panicles flowered in a day. The diurnal anthesis was recorded as the time from the first to last flowering. Over 40 randomly selected panicles of O. rufipogon from natural and experimental populations were enclosed with a net-bag to calculate seed sets.

Identification of Minghui-63, O. rufipogon, and hybrids

In order to accurately identify hybrid offspring of the two species, codominant simple sequence repeat (SSR or microsatellite) markers were used. One SSR primer pair, RM44 (Forward: acgggcaatccgaacaacc/Reverse, tcgggaaaacctaccctacc) was selected from a total of 64 SSR primer pairs screened. RM44 amplified alleles from the two species that were easily distinguished by electrophoresis using 3.4% agarose gels. Minghui-63 was found to have a consistent slow-migrating allele (S) (a relatively longer DNA fragment) of about 120 bp (Lane 1 in Fig. 2). The sampled O. rufipogon showed a consistent genotype of a fast-migrating allele (F) (a relatively shorter visible DNA fragment) about 110 bp (Lanes 2–26 in Fig. 2). The artificial hybrids of the two species showed a fixed heterozygote with an FS (Lanes 27–30 in Fig. 2).

Figure 2.

Simple sequence repeat (SSR) amplification products generated with the specific-primer RM44. M, DNA ladder; 1, Minghui-63; 2–26, Oryza rufipogon; 27–30: artificial hybrids between Minghui-63 and O. rufipogon.

Determination of gene flow frequency and data analysis

The mature seeds were collected in mid-Nov from O. rufipogon panicles. Seeds randomly selected at each interval from the collected seeds of the CPC-1, CPC-2, CPC-3, EPC-1, EPC-2, EPC-3, and UPC O. rufipogon populations were analyzed using the RM44 SSR marker (Table 1). The seeds collected from the CPC-4 and CPC-5 populations were selected to detect the gene flow frequency according to the eight compass directions from East to North-east. Approx. 1000 seeds were randomly selected from each replicate of the APC O. rufipogon population. After storage at 4°C for 1 month and heat shock at 50°C for 24 h to break dormancy, the sampled seeds were germinated at alternating temperatures of 30°C during the day and 25°C during the night. Seedlings were planted in a glasshouse. Leaf samples were collected from individuals for SSR examination approx. 50 d after seed germination. An additional germination trial using the same procedure as above was conducted for the retained seeds after the first germination. DNA was extracted using the protocol of Doyle & Doyle (1987) and PCR reactions were performed following the description of Wu & Tanksley (1993). Gene flow frequencies were calculated of the number of seedlings with the FS heterozygote SSR pattern divided by the total number of seedlings examined. An ANOVA procedure was used to analyze the effect on gene flow of the size of the pollen source in the CPC and EPC experiments, with the radii of plots of pollen source as independent variants and distances from pollen source as covariants. The data of gene flow measured at four geographical locations (North, West, South, and East relative to pollen source) within 6 m apart from pollen source in the CPC-3, CPC-4, and CPC-5 were also analyzed using ANOVA to test the effect of wind direction on gene flow, with locations as independent variants and distances from pollen source as covariants. The data of gene flow obtained at the same direction of the UPC experiment from the CPC, EPC, and UPC experiments were consolidated to conduct regression analysis for the relationship between gene flow and distance. All analyses were performed using the STATISTICA for Windows software package (StatSoft Inc., 1995).

Table 1.  Number of tested seedling and identified hybrid at each distance from the cultivar, Minghui-63 in the CPC, EPC, and UPC experiments
Distance (m)Design CPC-1CPC-2CPC-3CPC-4CPC-5EPC-1EPC-2EPC-3UPC
  1. Number in parenthesis was the hybrid, ‘–’ means no design at such distance.

0.3  98 (1) 
0.6198 (1)461 (6) 
0.9305 (6)
1.2140 (3)210 (4)  715 (9)  166 (2)945 (2)1223 (18) 
1.5245 (4) 
1.8175 (2)  381 (7) 
2.4104 (2)  993 (20)  409 (2)  236 (2)654 (9) 
3.0162 (3)193 (3) 
3.61206 (13)  247 (4)  34 (1)1372 (6)
4.2  475 (3)
4.81772 (14)  445 (5)876 (6)      2 (0)482 (6)
6.01597 (22)1592 (5)
7.2  243 (3)262 (1)
9.6  460 (6)128 (1)
12.0  575 (6)
14.4384 (3)  37 (0)
19.2355 (0)165 (2)
24.0440 (1)  30 (0)
28.8  62 (0)
33.6136 (2)
38.4  79 (1)
43.2  85 (0)
48.0  65 (0)
52.8120 (0)
57.6  89 (0)
62.4    0
67.2    0
72.0    0


Flowering time and seed set

The peak flowering period of O. rufipogon was from mid-September to early November, although its flowering time can last from early September to late November in natural populations. The flowering time of O. rufipogon completely overlapped that of Minghui-63, which flowered between mid-September and mid-October. Minghui-63 flowered from 09 : 30 to 14 : 40 h, which was also within that of O. rufipogon, from 10 : 30 to 16 : 30 h.

Seed set of O. rufipogon in this experiment was nearly the same as that examined in the original Chaling population (74.27 ± 3.58 vs 72.38 ± 5.71, N= 45). Only O. rufipogon seeds produced during the flowering period that overlapped with Minghui-63 were collected, although more panicles were produced by each O. rufipogon individual. The sampled mature seeds germinated perfectly with a total rate of 80% (74.53 ± 5.13% at the first germination attempt and 4.66 ± 0.21% at the second germination attempt, N= 50) after their dormancy breaking. These optimal conditions allowed us to obtain a maximum gene flow under optimized conditions in this study.

Determination of gene flow frequency

A total of 23 776 seeds randomly selected and germinated from over 80 000 O. rufipogon seeds collected from the experimental populations were examined by the species-specific SSR marker RM44, and 296 seedlings were found to be hybrids. The number of analyzed seedlings and identified hybrids in the CPC, EPC, and UPC experiments was showed in Table 1. In addition, 39 hybrids were also found in the APC experiment from 2264 examined seedlings.

The maximum gene flow frequency in CPC was 2.75% at the interval of 1.2 m from Minghui-63 in the CPC-5 plot, and that in EPC was 2.94% at the distance of 3.6 m from the pollen source in the EPC-2 plot, which was also the highest frequency of hybridization found in this study. The gene flow occurred in the UPC experiment with a frequency below 1.5% at all distances (Table 1). The frequency was also high in the APC plot (2.19 ± 0.037%, N= 3), which was expected to have the highest hybridization frequency, due to the zero isolation between the pollen source and recipient stigmas (Fig. 1c).

It was found that hybridization occurred nonrandomly in the CPC experiments. Analysis of hybridization rates in different directions revealed that wind significantly influenced the distribution of the crop-to-wild gene flow (ANOVA, F= 4.89, P= 0.042) and more hybridization events occurred in O. rufipogon plants in the southward direction which matched the prevailing wind direction observed during the flowering period (Fig. 3). Hybridization rates from the CPC-3, CPC-4, and CPC-5 plots within 6 m of Minghui-63 were calculated to test the effects of pollen source sizes on gene flow, because densities and distances of O. rufipogon from Minghui-63 were the same in these plots. The ANOVA of these data showed that the gene flow frequency was not correlated with the area of pollen source (F = 0.668, P= 0.532). Analysis of the data obtained from the CPC-1, CPC-2, CPC-3, CPC-4, and CPC-5 plots showed the similar results (F = 0.456, P= 0.806).

Figure 3.

Distribution of gene flow within 6 m of pollen source (Minghui-63) based on the data obtained from the CPC-3, CPC-4, and CPC-5 experiments. N, W, S, and E: locations of Oryza rufipogon relative to pollen source.

Although the highest gene flow were observed at the 3.6 m interval, the gene flow frequencies at other distance intervals in the EPC experiment were not significantly higher than those at the same distance intervals in the CPC experiment. This means that location of recipient plants did not significantly influence the rate of interspecific gene flow. Data from the EPC experiment (EPC-1 to EPC-3) did not show a significant size effect of pollen source on gene flow (ANOVA, F= 5.144, P= 0.061).

In the CPC experiments, hybrid seeds were produced at all the downwind distances except for the distance of 19.2 m from Minghui-63, demonstrating that the range of pollen dispersal of Minghui-63 was at least 24 m (CPC-5). In the UPC experiment, the farthest O. rufipogon plants were 72 m from the Minghui-63 plants. In this experiment, the longest distance where hybrids were found was 43.2 m from the Minghui-63 pollen source. One hybrid in 79 mature seeds harvested from O. rufipogon was identified at this distance (Table 1).

Because no significant size effect occurred in the CPC and EPC experiment, the data of gene flow from the UPC experiment could be combined with that from the same directions of the CPC and EPC experiments to evaluate the effects of distance from pollen source on gene flow. The analysis performed with the combined data from these experiments showed that the frequency of hybridization decreased with the increasing of distance from the pollen sources (R2 = 0.634, P < 0.001) (Fig. 4).

Figure 4.

Relationship between gene flow and distance from pollen source (Minghui-63) obtained by regression analysis of the integrated data of the central population combination (CPC), encircled population combination (EPC), and unidirectional population combination (UPC) experiments.


The present study showed that gene flow from cultivated rice to its wild relative O. rufipogon occurred at a significant level, which confirms that occurrence of gene flow from cultivated to wild rice is beyond question (Ellstrand et al., 1999; Jarvis & Hodgkin, 1999). This study was carried out under experimental conditions where the factors that influenced gene exchange were optimized. It was fortunate that the weather during anthesis in this study was also suitable for fertilization and seed development, resulting in a high seed set that allowed more viable hybrid seeds to be produced. According to previous studies, O. rufipogon populations showed significant differences both in seed set and outcrossing rate due to geographical differentiation or introgression with cultivated rice, and introgression of cultivated rice increases the outcrossing rate of O. rufipogon (Morishima et al., 1961; Oka & Chang, 1961). This would increase the possibility of gene exchange between cultivated and wild rice species. The sampled Chaling O. rufipogon population was suggested to be an introgressed population (Song et al., 2001b), and under natural conditions, introgression between cultivated rice and O. rufipogon was frequently observed (Z.P. Song, personal observation). It was therefore not surprising that relatively high seed set and high hybridization rate were observed in this study. Although cultivated rice and O. rufipogon often occur sympatrically, the flowering time of cultivated rice does not always perfectly match that of O. rufipogon, and the two species do not always grow immediately adjacent to each other. These factors will reduce interspecific gene flow in many natural populations (Arriola & Ellstrand, 1997; Messeguer et al., 2001). The results of this study might represent the maximum frequency of gene flow between the two species under cultivation conditions.

The maximum frequency of gene flow from cultivated rice to O. rufipogon in this study (2.94%) is much higher than that reported by Messeguer et al. (2001) between cultivated rice varieties (0.5%) under similar conditions. This is mainly attributable to the differences of breeding systems between these species involved. Cultivated rice is a self-pollinated species and its floral structure, such as short stigmas, limits the occurrence of outcrossing (Messeguer et al., 2001). In contrast, O. rufipogon has a significantly variable outcrossing rate from 5% to nearly 60% (Oka & Morishima, 1967). Its floral structure includes long and outstretched stigmas, increasing the opportunity of capturing foreign pollen grains. The frequency of gene flow from cultivated rice to O. rufipogon is significantly lower than that reported for the weedy rice (red rice) (52.18%) by Langevin et al. (1990). Genetic distance of the cultivated rice with O. rufipogon and weed rice, resulting in differences on sexual compatibility, perhaps account for this difference. Notably, Oka & Chang (1961) once reported the hybridization rate in the wild rice (O. perennis or O. sativa f. spontanea) from Taiwan population was up to 30.7% when the wild relative was grown surrounded by a glutinous cultivated variety. The used wild rice plants by Oka & Chang might be O. sativa f. spontanea, that is, weedy rice rather than true O. rufipogon. It is also possible that this result might not be representative because of the small sample size (45/114 seeds) and lack of replication.

This study indicated that gene flow was not randomly distributed across the recipient range, but was significantly influenced by wind. This observation is similar to that reported by Messeguer et al. (2001). It is obvious that wind can significantly affect pollen flow and that downwind recipient plants may capture more foreign pollen grains, thus enhancing gene flow (Timmon et al., 1995). However, this study also showed that, besides the prevailing wind direction some hybridization occurred at other directions including the direction opposite of that of the wind. This pattern of gene flow could be attributable to the influence of microclimate, which is influenced by the microlandscape. The experimental site in this study was surrounded by hills running in a north–south direction. The wind was mainly from north to south and but occasionally was reversed, and whirlwinds also occurred occasionally. The whirlwind might result in randomly horizontal distribution of gene flow and therefore have slight effects on gene flow being found (Rognli et al., 2000; Staniland et al., 2000), but have significant effects of spatial distribution of pollen flow (Jackson & Lyford, 1999). The pollen flow of Minghui-63 directly measured in the same experimental system also showed the same pattern (Z.P. Song, unpublished data).

This study revealed that cultivated rice pollen could move downwind at a distance over 43 m, which is much wider than the isolation distance of 10 m generally recommended for hybrid seed production (Khush, 1993). On the other hand, the distance of pollen flow of cultivated rice found in this study was significantly shorter than that from investigation of rice pollen flow using male sterile cultivated variety as pollen recipients (43.2 m vs 100 m) (Y.G. Zhu, unpublished data). These disagreements were mainly due to differences in specific features between the cultivated varieties used by different researchers, and might also be attributable to different approaches applied for study. Gene flow is usually lower than pollen flow including the frequency and range due to the factors that influence the success of hybridization (Snow & Parker, 1998). Theoretically, the true range of pollen flow of Minghui-63 should be wider than 43 m because pollen may move further if the wind speed is strong enough (Jackson & Lyford, 1999). Pollen flow of Minghui-63 was observed up to 110 m when wind velocity was 10 m s−1 (Z.P. Song, unpublished data). The flowering pattern of rice and short longevity of rice pollen might be accounted for the limited distance of 43 m observed for hybridization in the present study (Messeguer et al., 2001; Song et al., 2001a). Consequently, as compared with the similar studies of other crops and their relatives (Klinger et al., 1991; Amand et al., 2000; Rognli et al., 2000; Saeglitz et al., 2000; Wilkinson et al., 2000), the gene flow between cultivated and wild rice occurred over a relatively short distance. These differences indicate that pollen flow or gene flow is associated with specific floral syndromes and pollination systems, in addition to climate conditions.

The frequency of gene flow significantly decreased with the increasing of distance away from the pollen source (Fig. 4), as has been found in other studies (Klinger et al., 1991; Rognli et al., 2000; Messeguer et al., 2001). From the relationship between gene flow and distance, the predictable distance of gene flow with a certain frequency or the frequency of gene flow at a certain distance can be easily obtained. The highest gene flow examined in this study was not at zero distance but instead at a distance of 3.6 m. One possible explanation for this observation is that O. rufipogon plants with an average height of 194.37 ± 5.65 cm are taller than Minghui-63 (64.27 ± 1.87 cm). The significant differences in height between the cultivated and wild rice species resulted in spatial isolation between flowers even though the two species were grown side by side.

Although the data from this study did not show significant influence of gene flow by the size of pollen sources, the scale-dependent effect of donor plot size on the frequency of gene flow (Amand et al., 2000) cannot be neglected. The inconsistent results might due to the difference of density between pollen donor (Minghui-63) and pollen recipient plants (O. rufipogon). Even in the CPC-1 design which had the smallest pollen source (20 Minghui-63 individuals), each individual had an average of eight flowering panicles and more than 20 spikelets per day discharged pollen grains. On the recipient plants, planted at a lower density, each individual had only about five panicles per individual and fewer spikelets flowered (4–9 flowering spikelets per day). One pollen grain is theoretically sufficient to fertilize one O. rufipogon spikelet, which has only a single ovule. Consequently, the large number of pollen grains of Minghui-63 could in principle fertilize all of the O. rufipogon spikelets. The relative inadequate data from the CPC-1–3 may influence the analytical results because the recipient plants did not have sufficient distance from the pollen sources. This suggests that further experiments with a larger scale are needed to address this question.

This study clearly showed that genes of cultivated rice could move to its sympatric wild relative O. rufipogon at a relatively high frequency within a certain distance. This suggests that some O. rufipogon populations might have introgressed genes from cultivated rice, because under natural conditions these populations co-occur with cultivated rice with a distance of less than 100 m. A sufficient isolation distance should be established for in situ conservation of O. rufipogon to minimize the contamination of the wild rice by cultivated rice pollen flow. This study also demonstrates a high probability that transgenes from genetically engineered rice varieties will escape to O. rufipogon populations in the areas where the wild and cultivated rice coexist. For containment of transgenic rice, spatial isolation between transgenic rice and their wild relatives should be seriously considered, in addition to other effective approaches. Importantly, a close monitoring of transgene escape and persist in environment should be implemented. It seems to be very difficult to completely avoid transgenes escape through gene flow, thus the safest strategy for management of transgenic rice is to avoid planting such rice varieties in regions where wild rice occurs. Moreover, a larger scale experiment to determine gene flow frequencies under natural conditions should provide more predictive results for the biosafety assessment, but this study clearly indicates that transgene escape will occur if transgenic rice and its close wild relatives are grown together. Therefore, questions of ecological consequences of such crop-to-wild gene flow need to be urgently addressed both for effective conservation of genetic resources and appropriate assessment of the risk caused by gene escape from transgenic crops. The experimental systems established in this study will be useful tools for such studies, especially for wind- and self-pollinated crop species.


The National Nature Science Foundation of China (Grant no. 39893360 and Grant no. 30125029), and Shanghai Commission of Science and Technology supported this research. We thank Bo Li for his assistance in data analysis, Guihua Liu, Xiaofan Wang, and Yong Wang for their assistance in the field experiments, Baofeng Jin and Yongchun Jing for their assistance in PCR examinations, and Dr M. Cohen for his valuable comments on this manuscript.