Gene Flow from Cultivated Rice (Oryza sativa) to its Weedy and Wild Relatives

Abstract

• Background and Aims Transgene escape through gene flow from genetically modified (GM) crops to their wild relative species may potentially cause environmental biosafety problems. The aim of this study was to assess the extent of gene flow between cultivated rice and two of its close relatives under field conditions.

• Methods Experiments were conducted at two sites in Korea and China to determine gene flow from cultivated rice (Oryza sativa L.) to weedy rice (O. sativa f. spontanea) and common wild rice (O. rufipogon Griff.), respectively, under special field conditions mimicking the natural occurrence of the wild relatives in Asia. Herbicide resistance (bar) and SSR molecular finger printing were used as markers to accurately determine gene flow frequencies from cultivated rice varieties to their wild relatives.

• Key Results Gene flow frequency from cultivated rice was detected as between approx. 0·011 and 0·046 % to weedy rice and between approx. 1·21 and 2·19 % to wild rice under the field conditions.

• Conclusions Gene flow occurs with a noticeable frequency from cultivated rice to its weedy and wild relatives, and this might cause potential ecological consequences. It is recommended that isolation zones should be established with sufficient distances between GM rice varieties and wild rice populations to avoid potential outcrosses. Also, GM rice should not be released when it has inserted genes that can significantly enhance the ecological fitness of weedy rice in regions where weedy rice is already abundant and causing great problems.

INTRODUCTION

With the rapid advance of transgenic biotechnology, more and more transgenic crop varieties are being released into the environment and entering commercial markets (Barber, 1999; Fernadez‐Cornejo and McBride, 2000; Huang et al., 2002). Undoubtedly, biotechnology and transgenic crops have provided new opportunities for global food security and new developments in life sciences. However, the release and use of transgenic products have also caused tremendous concerns about biosafety (Crawley et al., 2001; Ellstrand, 2001; Prakash, 2001; Snow, 2002). The potential ecological risks associated with transgene escape through gene flow (or cross‐pollination) are foremost among these concerns (Lefol et al., 1996; Ellstrand et al., 1999; Amand et al., 2000; Halfhill et al., 2001; Lavigne et al., 2002).

When alien transgenes escape to, and express normally, in weedy or wild relatives of transgenic crop species, transgenes may persist and disseminate within the weedy or wild populations through sexual reproduction and/or vegetative propagation. If the transgenes are responsible for resistance to biotic and abiotic stresses (such as disease and insect resistance, drought and salt tolerance, and herbicide resistance) that can significantly enhance the ecological fitness of weedy and wild populations, the escape of these transgenes will probably cause ecological problems, e.g. producing aggressive weeds. Such weeds might get out of human control, and result in unpredictable damage to local ecosystems (Ellstrand, 2001; Snow, 2002). Alternatively, when transgenes escape to and persist in populations of wild relative species, the fast dissemination of the transgenic hybrid individuals might contaminate the original wild populations. Sometimes, the aggressive spreading of hybrids with better ecological fitness could even lead to the extinction of endangered wild species populations in local ecosystems (Kiang et al., 1979; Ellstrand and Elam, 1993). Therefore, a better understanding of gene flow, including its frequencies and directions, between crops and their wild relatives will facilitate the effective management and safe use of transgenic crops.

Rice (Oryza sativa L.) is one of the most important of the world’s cereal crops, providing staple food for nearly one‐half of the global population (Lu, 1998). Particularly in Asia, rice is considered as the number one crop, both in consumption and cultural terms. Rice is also one of the earliest of the world’s crop species to which transgenic biotechnology has been effectively applied for genetic improvement (Ajisaka et al., 1993; Yahiro et al., 1993; Tyagi and Mohanty, 2000). Although no transgenic rice varieties have yet been officially approved for extensive commercial cultivation in the world, genes conferring traits such as high amounts of beta‐carotene (provitamin A), high protein content, disease and insect resistance, virus resistance, herbicide resistance, and salt tolerance have been successfully transferred into different rice varieties through transgenic techniques (Ajisaka et al., 1993; Yahiro et al., 1993; Matsuda, 1998; Datta et al., 2002; Potrykus, 2002). Some of these transgenic rice breeding lines or varieties have been released into the environment for testing (Messeguer, 2001; Huang, 2002; Jia and Peng, 2002). It is apparent that, as an important world cereal crop, transgenic rice varieties will sooner or later be released into the environment for commercial production, and probably within the near future.

It is therefore imperative to be aware of and assess the potential biosafety problems caused by transgenic rice before its mass release into the environment. As the relatives of cultivated rice, weedy and wild rice species (e.g. Oryza rufipogon Griff., O. nivara Sharma et Shasry and O. sativa f. spontanea) are commonly found and even coexist in rice farming systems in many Asian, Africa and American countries (Baki et al., 2000; Noldin, 2000; Chen et al., 2001). Many of these wild relatives contain the same AA genome and are highly compatible sexually with the cultivated rice (Lu et al., 2003). Thus, prediction of gene flow frequency between rice and its weedy and wild relatives becomes one of the important components for ecological risk assessment of transgenic rice, because gene flow is the primary step from which potential ecological consequences of transgene escape may follow.

Herbicide‐resistant crops and molecular markers, particularly the simple sequence repeats (SSR) markers are commonly used for the detection of crop‐to‐crop and crop‐to‐wild gene flow frequencies under different conditions, and have been proved to be very effective (Messeguer et al., 2001; Reboud, 2002; Tranel et al., 2002; Song et al., 2003a). The objective of this study was to determine gene flow from cultivated rice to weedy rice and perennial common wild rice under special field conditions mimicking their natural occurrence in Asia, using herbicide‐resistant rice and the SSR molecular finger printing as markers.

MATERIALS AND METHODS

Plant materials

One accession of the perennial wild rice (Oryza rufipogon), 13 accessions of weedy rice (O. sativa f. spontanea, O. spontanea for short), one cultivated rice variety (Minghui‐63) and one F₅ rice breeding line (Nam29/TR18) were included in the experiments (Table 1). Oryza rufipogon was collected as vegetative stocks from Chaling in Hunan Province in China, and its tillers were transplanted to produce sufficient individuals with uniform genotypes. Seeds of O. spontanea were obtained from the Wild Crop Germplasm Bank, College of Natural Resources, Yeungnam University in Korea, comprising 13 accessions distributed in different locations over Asian and American countries. Considering actual agricultural and environmental impacts, both ecotypes with long grain (indica) and short grain (japonica), and of various plant heights and heading dates were used for the study. Seeds of Minghui‐63, a semi‐dwarf rice cultivar with a growth period of approx. 90–95 d, were donated by Professor S. M. Mu from Hubei Academy of Agricultural Science in Wuhan, China. The herbicide‐resistant rice line, Nam29/TR18, containing the bar gene and expressing resistance to the herbicide Basta (glufosinate ammonium), was developed through a single cross between an elite rice variety Nam29 (Li et al., 1996) and the transgenic rice line TR18 provided by the National Biotechnology Institute of RDA, Korea, in 1998. The wild and weedy rice taxa were used as pollen recipients, whereas the cultivated rice was used as the pollen donor.

Table 1.

Accessions of wild, weedy and cultivated rice used in the experiment, with their origin and characteristics

Species	Strain name	Accession   no.	Country of origin	Ecotype	Heading in experimental fields* (date)	Average rate of  interspecific   hybrid (%)
O. rufipogon	Puye‐CL	CL‐01	China	–	Beginning of September	1·21–2·19%
O. spontanea	TKN7‐3	YW1404	Nepal	Japonica	27 Aug.	ni†
O. spontanea	Slashare	YW2246	Korea	Japonica	1 Aug.	0
O. spontanea	Namseonjizang 1946	YW1296	Korea	Indica	22 July	ni
O. spontanea	Danyangaengmi 40–2	YW2556	Korea	Japonica	23 July	ni
O. spontanea	Goseongaengmi 2	YW1242	Korea	Indica	30 July	ni
O. spontanea	US2	YW1392	America	Indica	23 July	ni
O. spontanea	Lu‐tao	YW1415	China	Japonica	20 Aug.	0
O. spontanea	Seongjuaengmi 8	YW2257	Korea	Japonica	12 Aug.	0·026
O. spontanea	C9576	YW1388	Japan	Indica	20 Aug.	0
O. spontanea	Galsaegshare	YW2247	Korea	Japonica	1 Aug.	0·046
O. spontanea	Ch79–1	YW1393	China	Indica	20 Aug	0
O. spontanea	Heidiaogu	YW1396	China	Indica	1 Aug.	0·011
O. spontanea	W1714	YW1390	Brazil	Indica	1 Aug.	0
O. sativa	Minghui‐63		China	Indica	Beginning ofSept.	–
O. sativa	Nam29/TR18		Korea	Japonica	5 Aug.	–

* As field observations taken independently from different experimental sites in Korea and China.

^† Hybridization was not investigated because the flowering time of these accessions was not synchronized with Nam29/TR18 in the experimental field.

Experimental design

The experiments were conducted independently at two paddy‐field sites, at Kyongsan in South Korea (35°51′N, 128°50′E) and Chaling in Hunan Province, China (26°50′N, 113°40′E). To mimic the respective crop–weedy and crop–wild growing patterns that are naturally found in the rice farming ecosystems in Asia, two types of experimental design were established by constructing two different populations under special cultivation conditions (Figs 1 and 2).

Fig. 1. Field experimental layout of the crop–weedy mixture population, in which the circles represent hills of Nam29/TR18 and the crosses represent a mixture of weedy rice and Nam29/TR18. Each replication consisted of 13 blocks (A1–A13), each with one of the 13 weedy rice accessions randomly selected.

Fig. 2. Field experimental layout of the crop–wild alternating population, in which the circles represent hills of Minghui‐63 and the crosses represent O. rufipogon. The two replications were not adjacent to each other.

Crop–weedy mixture (CWM) population.

The CWM population was constructed in a 5 × 20 m² paddy field with three replications in South Korea (Fig. 1). The experimental plot was designed as complete random blocks where Nam29/TR18 was planted with 35 hills per row, and mixed with one of the 13 weedy rice accessions in each block. In general, each block consisted of eight weedy rice plants, and one weedy rice plant was included in every fourth hill in each row as a mixture with the cultivated rice Nam29/TR18 as indicated in Fig. 1. Seedlings of each weedy rice accession were transplanted together with the cultivated rice into the plot at the same time, accounting for approx. 10 % of the total CWM population. To ensure a sufficient population of the pollen donor, approx. six or seven rows of Nam29/TR18 were planted as a protecting zone surrounding each replication. Distance between rows was 30 cm, and that between hills was 15 cm within rows.

Crop–wild alternating (CWA) population.

The CWA population was constructed in a plot 5 × 5 m² in a paddy field with two replications in China (Fig. 2). Tillers or seedlings of O. rufipogon and Minghui‐63 were planted into the plots at the same time in alternating rows. Two or three tillers were planted in each hill of Oryza rufipogon and four or five seedlings of O. sativa. Distance between the rows was 50 cm, and that between hills was 50 cm for O. rufipogon and 30 cm for Minghui‐63 within rows (Fig. 2).

Identification of interspecific hybrids

For identification of hybrids between Nam29/TR18 and weedy rice, seedlings generated from seeds of different weedy rice accessions were sprayed with the herbicide Basta at about the three‐to‐four leaf stage as a preliminarily screening. The surviving seedlings with resistance to the herbicide Basta were considered as hybrids and their genomic DNA was subject to PCR detection of the herbicide resistant bar gene to confirm their hybridity. The primer pair for PCR analysis (P1: 5′‐CGAGAACCG CAGGAGTGGA‐3′; P2: 5′‐CCAGAAACCCACGTCATG CC‐3′) was designed based on the sequence of the bar gene included in the transgenic rice TR18. The PCR reaction system included 1X buffer, 2·0 mm MgCl₂, 0·2 mm dNTP (each), 1 µm of each primer, 1·25 U Taq polymerase and 50 ng of template DNA, in a total volume of 12·5 µL. PCR amplification was performed on a GeneAmp 2400 at 95 °C for 3 min for initial denaturation, followed by 45 cycles of 95 °C for 1 min, 58 °C for 1 min, 72 °C for 2min, and terminated by a final extension at 72 °C for 5 min. PCR products were electrophoresized on 1·2 % agarose gel containing Et‐Br and visualized on an Alpha Imager^TM 1200. Seedlings with the detected DNA fragment from the bar gene were confirmed as true hybrids.

For accurate identification of hybrids between O. rufipogon and cultivated rice, the co‐dominant simple sequence repeats (SSRs) were used as molecular markers. The SSR primer pair, RM44 (forward: ACGGGCAATCCGAACAACC/reverse: TCGGGAAAA CCTACCCTACC) was selected from a large number of screened primer pairs to meet the objectives. DNA samples were extracted applying the protocol of Doyle and Doyle (1987), and the PCR reactions were performed following the description of Wu and Tanksley (1993). The RM44 amplified polymorphism alleles from the two species and the alleles were easily distinguishable with the electrophoresis in 3·4 % agarose gels, where O. rufipogon presented a consistent genotype by visualizing a fast‐migrating allele (F) and Minghui‐63 presented a slow‐migrating allele (S). The hybrids between the two species showed the stable heterozygous FS alleles.

Determination of gene flow

For crop‐to‐weedy gene flow detection, seeds were collected from different accessions of weedy rice that mixed with Nam29/TR18 in the CWM population at the mature stage. A total of approx. 2000–3500 seeds were collected from the blocks of each weedy rice accession. The seeds were treated at 50 °C for 10 d to break the dormancy, and sown in a tray with soil in a glasshouse. Seedlings were sprayed with the herbicide Basta at about the three‐to‐four leaf stage. Gene flow frequencies were estimated by calculating the number of surviving seedlings 10 d after the spray, against the total number of seedlings germinated. The final gene flow frequencies from Nam29/TR18 to weedy rice were obtained by comparing the results from herbicide spray and molecular confirmation data from PCR analysis.

For crop‐to‐wild gene flow detection, seeds were harvested at maturity from O. rufipogon in the CWA population. A total number of 1000 seeds were randomly selected from each of the replicates, given that O. rufipogon usually produces limited fertile seeds. 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 (day) and 25 °C (night). Leaf samples from germinated seeds were collected from individual seedlings for SSR examination. Gene flow frequencies were estimated by the calculation of the number of seedlings with the FS heterozygote SSR pattern divided by the total number of seedlings examined.

Results

Gene flow frequency from cultivated rice to weedy rice

Approx. 5000–10 000 seeds were collected from each weedy rice accession in the three replications of the CWM populations, but seeds from only eight accessions that had synchronous flowering time with the pollen donor Nam29/TR18 were germinated under glasshouse conditions to produce seedlings. The average frequencies of weedy rice seedlings with herbicide resistance were found to be very low, in general, and variable among different blocks, but with no significant differences among the three replications. Among the eight plots examined, interspecific hybrids with the herbicide‐resistant rice were only detected in three weedy rice accessions and not in the five others (Table 1). This can most likely be attributed to the different heading dates and flowering times of the weedy rice accessions, and also to the differences in plant height. Molecular data from PCR analysis confirmed all the herbicide‐resistant individuals contained the bar gene, as demonstrated by the presence of the specific DNA fragment (lane 1 and lanes 3–7 in Fig. 3). The experimental results indicate that the detectable rate of herbicide‐resistant gene flow from the rice breeding line to weedy rice plants varied between 0·011 and 0·046 %.

Fig. 3. PCR detection of the bar gene in the surviving weedy rice seedlings that were resistant to the herbicide Basta. Lane 1, Nan29/TR18 F₅ with the bar gene; lane 2, weedy rice (bulk DNA sample of YW2257, YW2247 and YW1396); lanes 3–7, YW2257‐HR (herbicide‐resistant) seedling‐1, YW2257‐HR seedling‐2, YW2247‐HR seedling‐1, YW2247‐HR seedling‐2 and YW1396‐HR seedling. M = DNA ladders.

Gene flow frequency from cultivated rice to O. rufipogon

A total number of more than 30 000 seeds was harvested from O. rufipogon individuals in the two replicates of the CWA population, but only approx. 3000 seeds were randomly selected for the SSR analysis due to workload constraints. The seeds were germinated under glasshouse conditions. Of the 2246 seedlings (that germinated from the 3000 seeds) examined by the species‐specific SSR marker RM44, 39 seedlings were found to be hybrids with the FS heterozygous alleles (lanes 17–20 in Fig. 4), compared with the F allele of O. rufipogon (lanes 2–16 in Fig. 4) and the S allele of Minghui‐63 (lane 1 in Fig. 4). The frequencies (1·21–2·19 %) of interspecific hybrids with Minghui‐63 varied significantly between the two replicates (Table 1). The relatively high gene flow frequency from cultivated rice to O. rufipogon was not unexpected in this experimental population, given the high compatibility between cultivated rice and O. rufipogon.

Fig. 4. SSR amplification products generated by the specific‐primer RM44. Lane 1, Minghui‐63; lane 2, O. rufipogon (parent); lanes 3–16, O. rufipogon (selfed); lanes 17–20, interspecific hybrids between Minghui‐63 and O. rufipogon. M = DNA ladders.

DISCUSSION

The perennial common wild rice O. rufipogon and weedy rice (sometimes referred to as red rice, black rice or shattering rice) are widely distributed in Asian countries. The former species is found from southern China through South and Southeast Asia down to northern Australia, occurring in lakes, ponds, along rivers, ditches, and at the edges of and/or within farmers’ rice fields (Vaughan, 1994; Lu et al., 2003). Weedy rice is found with various frequencies in different rice ecosystems occurring together with cultivated rice (Vaughan, 1994; Abdullah et al., 1996; Watanabe et al., 2000). Both species are weeds in some rice farming ecosystems, particularly in regions with relatively traditional forms of rice cultivation. Weedy rice causes the more severe damage to rice production in more extensive regions by reducing rice yield and quality, particularly in rice ecosystems where direct seeding is adopted (Suh et al., 1992; Suh and Ha, 1993; Abdullah et al., 1996; Chin, 1997). In addition, weedy rice infests several upland crops such as jute, maize, soybean and vegetables, causing general weed problems (Baki et al., 2000).

In China, O. rufipogon is mainly found in the southern provinces, such as Guangdong, Guangxi, Hainan and Yunnan, in rather low frequencies. It does not pose weed problems in China, unlike in many other Asian countries where it does cause some yield loss. O. rufipogon is, however, regarded as an endangered species (Song et al., 2003b). The occurrence of weedy rice is not yet a problem in China, but with the gradual adoption of direct seeding for rice cultivation it will almost certainly become as serious a problem in China as in many other countries (Suh et al., 1992; Suh and Ha 1993; Abdullah et al., 1996). Korea has a wide distribution of weedy rice, and the occurrences of this species in farmers’ fields range from 0·6 % to as much as 23 %. A higher rate of weedy rice infestation is found in direct‐seeded rice fields than in the transplanted fields (Suh et al., 1997a, b), causing both yield loss and reduced quality (due to mixture) of rice varieties. Previous studies have estimated that weedy rice occurring in direct‐seeded rice fields may cause rice yield losses of up to 22·1 % (Lee et al., 1987). In general, the reduction of rice yield caused directly by weedy rice is calculated to be 5–10 % of total Korean rice production every year. Traditionally, Korean farmers practiced rice transplanting, but in recent years the government has encouraged a shift to direct‐seeding to save labour and to reduce soil erosion. This has resulted in considerable increases of weedy rice in farmer’s fields (Suh et al., 1997a, b). Recently, hybrid rice with the trait of herbicide resistance has been under development in China and Korea, for the purpose of simplifying hybrid‐seed production procedures with selective herbicides. This has drawn our attention to the need to consider the long‐term impact of hybrid rice with transgenic herbicide resistance on agricultural and environmental systems, especially in regions where weedy rice infestation is serious and will increase with the spread of direct‐seeded rice cultivation.

With the current concerns of weed problems caused by wild rice, particularly in rice farming ecosystems, one of the major fears is whether the engineered alien genes in transgenic rice varieties will escape to their wild and weedy relatives through gene flow, and enhance the fitness of the wild relatives. The experimental data given here clearly indicate the possibility of gene flow from cultivated rice varieties to wild and weedy species, although with different frequencies. The gene flow frequency from cultivated Minghui‐63 to wild O. rufipogon in an alternating cultivation model was detected to be approx. 1·1–2·2 %. This result is similar to an earlier study of gene flow from cultivated rice to O. rufipogon reported by Song et al. (2003a), in which variable frequencies of gene flow were detected from the same cultivated rice variety to wild O. rufipogon under different field experiments, with the highest frequency up to 2·94 %. These frequencies are significantly high in terms of transgene escape if the cultivated transgenic rice varieties are grown in the vicinity of wild rice species. Therefore, to prevent or minimize transgene escape to wild relatives, it is recommended that isolation zones be established between transgenic rice varieties and O. rufipogon with sufficient space or with some tall plants to act as pollen traps, at least until more effective methods are available.

The detected gene flow frequencies from the cultivated rice line Nam29/TR18 to various weedy rice accessions were very low, ranging from 0·011 to 0·046 %. Some weedy accessions did not show any indication of gene flow from the cultivated rice line, most probably due to unsynchronized flowering time and different plant heights. A great variation in heading date was observed, with some weedy rice accessions being approx. 4–22 d earlier or later than Nam29/TR18, and plants of weedy rice accessions were approx. 10–60 cm taller than Nam29/TR18. Hence the low frequency of gene flow observed in the present experiment is not surprising. Messeguer et al. (2001) also detected similar amounts of gene flow from transgenic herbicide‐resistant rice to non‐transgenic counterparts, ranging from 0·01 to 0·53 % in their experimental fields in Italy and Spain. Cross‐pollination in rice takes place to a certain extent, and the frequency largely depends on the climatic and varietal differences. Lord (1935) found the frequency of natural outcrossing between different rice varieties in Sri Lanka to range between 0·34 and 0·67 %. Oka (1988) found that the degree of outcrossing is generally higher in indica cultivars and wild species than in japonica cultivars. Furthermore, it is important to point out that in natural conditions when a weedy rice strain occurs within in a cultivated rice crop, the flowering time, plant height and other features of the weedy rice will tend to shift to a point that is similar to the cultivated rice varieties growing in the same field, because of their cross‐pollination and genetic recombination. Therefore, the outcrossing rate between weedy and cultivated rice in large populations might be more significant than the data observed in this experiment. Rice cultivars cross easily with their related weedy forms (red rice) found in direct‐seeded paddy fields and produce viable and fertile hybrids, and the hybridization rates could range from 1·08 to 52·18 % (Langevin et al., 1990).

In addition, the data collected from our experiment of only one generation with a limited number of plants in the small‐scale plots might not give a sufficiently true reflection of the gene flow frequency in general. When weedy rice occurs simultaneously and consistently with a cultivated rice variety in the same field, the frequency of hybrids resulting from gene flow could accumulate and increase through generations. If transgenic rice varieties are released into environments where weedy rice occurs abundantly, the transferred alien genes could spread out and accumulate in weedy populations. Therefore, it is not recommended to release transgenic rice with genes that can significantly increase weediness and protect against weed control (such as herbicide resistance) into regions where weedy rice is already a serious problem.

This study clearly provides a good example of a simple way to estimate or predict gene flow from transgenic rice to weedy and wild rice species using herbicide‐resistant genes and molecular markers. Gene flow was detected in two experiments conducted at different localities where weedy rice or wild rice is found under natural conditions. Although very low in comparison with the gene flow of the herbicide‐resistant character from cultivated rice to O. rufipogon, it was nonetheless shown that gene flow between the cultivated and weedy rice species definitely exists. It is important to point out that transgene escape may be much more serious a problem where large‐scale transgenic rice production consistently takes place in regions where weedy rice occurs abundantly. Manual control of weedy rice is impractical because of the difficulties in distinguishing between the two species prior to the heading stage. Herbicides are not effective against weedy rice because of its close relationship with cultivated rice. Given the fact that weedy rice is already causing problems in rice fields of more than 50 countries in Asia, Africa and Latin America, and may cause even more trouble when rice cultivation practices change to direct seeding in the future, considerable attention should be paid to herbicide‐resistant genes being transferred into rice varieties. If weedy and wild rice species were to receive herbicide‐resistant genes, perhaps in addition to other transgenes from cultivated counterparts through gene flow which can significantly enhance the ecological fitness of the weeds, it may eventually produce so‐called ‘super weeds’ in rice ecosystems.

In short, as in many other crop species, transgene escape from cultivated rice varieties to their weedy and wild relatives through gene flow has become an indisputable fact. There is, therefore, an urgent need for a thorough assessment of the ecological consequences of transgene escape, including such aspects as the ecological fitness of the hybrids and progeny of cultivated and wild rice, the destiny and establishment of escaped genes in wild populations, and their impact on general biodiversity.

ACKNOWLEDGEMENTS

Dr J. Keeley of the University of Sussex, UK, kindly edited the English of this manuscript. The support of NSFC for Distinguished Young Scholars (grant no. 30125029), Shanghai Commission of Science and Technology (grant no. 02JC140022 and 03dz19309) and 211 project (Biodiversity and Regional Ecosafety) is gratefully acknowledged.

Received: 10 January 2003; Returned for revision: 17 June 2003; Accepted: 15 September 2003    Published electronically: 5 November 2003