Introduction

The unpredicted potential environmental impact caused by transgene flow from genetically engineered (GE) crops to their cross-compatible wild relatives has stimulated tremendous debates and studies over the last decades1,2,3. Wild relative populations that have acquired a strongly fitted transgene through gene flow likely changes their evolutionary potential, resulting in unwanted environmental/ecological consequences4,5,6,7. It is therefore essential to properly assess the environmental impact of transgene flow before the commercialization of any GE crops. One of the key points to assess such environmental impact is to determine fitness of a transgene introgressed into wild populations7, 8, provided that the frequency of crop-to-wild gene flow is known. Many studies have been carried out to determine the fitness effect of a transgene under the controlled field-experimental conditions, involving crop-wild hybrid lineages. These studies included hybrid descendants derived from crosses between squash-wild gourd9, 10, maize-teosinte11, cultivated-wild sunflowers12, and cultivated-wild/weedy rice13,14,15,16,17,29, 30. In contrast, herbicide-resistance transgenes are considered to be neutral or even have costs in plants under the herbicide-free environmental conditions31, 32. Therefore, wild rice populations that have acquired an herbicide-resistance transgene through gene flow may not have any benefit due to the lack of selective pressure from herbicides33. Nevertheless, a recent study reported significantly increased fecundity and ratios of tryptophan concentration and photosynthesis in crop-weed hybrid lineages containing an herbicide-resistance transgene, at the absence of herbicide16. This transgene is an engineered rice endogenous epsps (5-enolpyruvoylshikimate-3-phosphate synthase) gene that overexpresses EPSPS. The GE rice was originally developed to confer tolerance to the glyphosate herbicide34, but unexpectedly the epsps transgene provided substantial benefits to WDR for plant growth and seed production16. This phenomenon poses a question about the transgene that overproduces EPSPS. Will the epsps herbicide-resistance transgene introgressed into any of the wild rice species produce the same fitness effect to their populations?

The perennial common wild rice (O. rufipogon Griff., referred to as WR hereafter) is the direct ancestor of Asian cultivated rice and one of the wild species in the genus Oryza (Poaceae). WR is distributed in the tropics and subtropics of monsoon Asia, with its northernmost border in Jiangxi province of China35. WR can reproduce sexually through seeds, or asexually through propagule or ratooning36, 37. It is widely recognized that WR is important germplasm for the genetic improvement of cultivated rice. For example, the well-known hybrid rice breeding program was benefited from the discovery and use of a male sterility (ms) gene from WR38, 39, demonstrating the importance of germplasm in WR gene pool40. However, WR is under threats due to the rapid growth of human population and urbanization, dramatic change in agriculture land uses, and intensive human disturbances41, 42. Also, massive and continued introgression of cultivated rice genes and transgenes into WR have posed a great challenge on the existence of WR populations3, 43. Altogether, identified gene flow from cultivated rice to WR populations in the controlled experiments44, 45 and population genetic studies46 indicated the high probability of transgene introgression from GE rice to WR. In addition, the diverse genetic variability among WR populations may result in different fitness responses of WR recipients to the same transgene, as revealed in a recent study of crop-weed hybrids containing insect-resistance transgenes18.

We produced F1-F3 crop-wild isogenic hybrid lineages with or without the epsps transgene, derived from artificial crosses between an epsps GE rice line and two WR populations. The objectives of this study were to address the following questions. (1) Does the over-expressing epsps transgene change the life-cycle traits of crop-wild hybrid descendants in the glyphosate-free environment? (2) Does the epsps transgene increase the fecundity and over-winter survival of crop-wild hybrid descendants? (3) Does the genetic background of transgene recipients affect the fitness of the epsps transgene in different types of rice parents, including WR, WDR, and cultivated rice? The answer of these questions will help us to appropriately estimate the potential environmental/ecological consequences caused by introgression of the overexpressing epsps herbicide-resistance transgene into WR populations, likely also to predict the potential consequences for other crop-wild transgene introgressions.

Results

More tillers and earlier flowering with increased seed sets in transgenic hybrid lineages

F1 and F2 hybrid lineages with the epsps transgene showed a greater number of tillers and earlier flowering time than their isogenic controls without the transgene (Fig. 1a, b, c and d; Tables S1–S3). Consequently, ratios of seed sets in the transgenic hybrid lineages showed significant increase, compared to their non-transgenic counterparts (Fig. 1e and f; Table S1). An obvious negative correlation was observed between ratio of seed set and days from seed germination to flowering in F2 hybrid descendants (Fig. S1a and b). In addition, no differences were observed for plant height between transgenic and non-transgenic hybrid lineages (Table 1).

Figure 1
figure 1

Number of tillers per plant (a: WR1-F2; b: WR2-F2); flowering time (c: WR1-F2; d: WR2-F2); and seed set ratios (e: WR1-F2; f: WR2-F2) of F2 transgenic (solid squares, dark grey columns) and non-transgenic (empty triangles, white columns) hybrid lineages in pure planting. Solid triangles, empty circles, and solid diamonds (in c and d) indicate days at which 1%, 30%, and 50% plants being flowered, respectively. The comparisons were made between transgenic and non-transgenic F2 hybrid lineages in pure-planting based on independent t-test (N = 6). Bars represent standard error. * or ** indicates significances at the levels of P < 0.05 or P < 0.01, respectively.

Full size image
Table 1 Two-way ANOVAs for the effects of transgene (transgenic vs. non-transgenic), wild parent (WR1 vs. WR2), and their interactions on life-cycle fitness related traits of F1-F3 rice crop-wild hybrid descendants.
Full size table

Two-way ANOVAs showed significant effects of a transgene (T) and wild parents (WR) on the number of tillers per plant (at different growth stages), days to flowering, and ratios of seed sets in both F1 and F2 hybrid descendants. However, no significant interaction effect was detected (Table 1). In the pure planting mode, significant increases were detected for the number of tillers per plant in F1 transgenic hybrids throughout the growth stages (with 19~38% increase); but significant increases were mainly detected at the early growth stages in F2 transgenic hybrid lineages (with 16~17% increase) (Fig. 1a and b; Table S1). In the mix-planting mode with 30 × 30 cm spacing between plants, significant differences were also detected for the number of tillers per plant, but with more prominent differences between transgenic and non-transgenic hybrid lineages (with 26~49% increase in F1 and 9~24% increase in F2) (Fig. 2a and b; Tables S2 and S3). Noticeably, the increase in number of tillers was more pronounced in the 30 × 30 cm mix-planting plot than that of other plots with lower densities (Fig. 2a and b; Tables S2 and S3). In addition, transgenic hybrid plants flowered significantly earlier (3 days for F1 and 9~15 days for F2) than their non-transgenic counterparts (Fig. 1c and d; Table S1). Consequently, transgenic hybrid plants had significantly higher ratios of seed sets (29% in F1 derived from WR1 and 29–30% in F2) than their non-transgenic counterparts in pure planting (Fig. 1e and f; Table S1). The ratios of seed sets were negatively correlated with the days to flowering in F2 hybrid lineages with or without the transgene (Fig. S1a and b), suggesting the possible contribution of earlier flowering to greater seed sets.

Figure 2
figure 2

Number of tillers (a, b), panicles (c, d), and filled seeds (e, f) per plant in F1 transgenic (dark grey columns) and non-transgenic (white columns) crop-wild hybrids derived from WR1 (left panel) and WR2 (right panel), in mixed-planting plots (30 cm, 40 cm, 50 cm plant spacing) with different densities. The comparisons were made between transgenic and non-transgenic F1 hybrids in mix planting based on paired t-test (N = 6). Bars represent standard error. * or ** indicates significances at the levels of P < 0.05 or P < 0.01, respectively.

Full size image

Increased fecundity in transgenic hybrid lineages

Both F1 and F2 transgenic hybrid lineages showed significantly increased fecundity as indicated by the number of panicles and seeds per plant compared to their non-transgenic counterparts in mix-planting plots with 30 × 30 cm spacing (Figs 2c, d, e, f and 3b and d; Tables S2 and S3). The hybrid lineage with WR1 as the wild parent also showed significantly increased fecundity in pure planting (Fig. 3a and c; Table S1). In addition, no differences were detected for 1000-seed weight between transgenic and non-transgenic hybrid lineages (Table 1). Two-way ANOVAs showed a significant effect of transgene (T) on the number of panicles per plant in F1 and F2 crop-wild hybrid descendants (Table 1). Wild parent (WR) had significant effect on the number of seeds per plant in F1-F2 and 1000-seed weight in F2 hybrid descendants (Table 1). No significant interaction effect was detected (Table 1). In the pure planting mode, 17~22% (F1) and ~13% (F2) increases were detected in number of panicles per plant in transgenic hybrid lineages (Fig. 3a; Table S1); meanwhile, ~27% increase was detected for the number of filled seeds per plant in F2 transgenic hybrid lineages (Fig. 3c; Table S1). In the mix-planting mode with different densities, significant increase in number of panicles (with 10~33% increase) and filled seeds (with 21~38% increase) per plant was mainly observed in the 30 × 30 cm plots in F1 and F2 transgenic hybrid lineages (Figs 2c, d, e, f and 3b and d; Tables S2 and S3). It seems that with the increase in cultivation densities, the extent of increases in panicles and filled seeds per plant between transgenic and non-transgenic plants became more substantial, suggesting the competitive effect for fecundity (Fig. 2c, d, e, and f).

Figure 3
figure 3

Number of panicles (a, b) and filled seeds (c, d) per plant of F2 transgenic (dark grey columns) and non-transgenic (white columns) crop-wild hybrid lineages in pure-planting (a, c) and mix-planting plots (b, d). The comparisons were made between transgenic and non-transgenic F2 hybrid lineages in mix planting (30 cm) based on paired t-test (N = 6). Bars represent standard error. * or ** indicates significances at the levels of P < 0.05 or P < 0.01, respectively.

Full size image

In addition, the two transgenic events (EP3 and EP4) showed similar extent of glyphosate resistance and increased number of panicles and filled seeds per plant (Table S4), suggesting the observed differences between transgenic and non-transgenic lineages were not the result of transgene insertion effect.

Enhanced buried-seed germination and over-winter regeneration in transgenic hybrid lineages

The transgenic hybrid lineages had higher germination ratios for soil-buried seeds (F3) and tiller regeneration ratios overwinter (F1 and F2) than non-transgenic counterparts, especially for those derived from WR1 (Fig. 4a, b, c and d; Tables 1 and S1).

Figure 4
figure 4

Germination ratios of buried-seeds (a: WR1-F3, b: WR2-F3) and ratios of tiller regeneration (c: WR1-F1 and WR1-F2, d: WR2-F1 and WR2-F2) in transgenic (dark grey columns) and non-transgenic (white columns) crop-wild hybrid lineages. The comparisons were made between transgenic and non-transgenic hybrid lineages based on independent t-test (N = 3 for seed germination ratio; N = 6 for tiller regeneration ratio). Bars represent standard error. * or ** indicates significances at the levels of P < 0.05 or P < 0.01, respectively.

Full size image

Two-way ANOVAs showed significant effect of transgene (T) on the ratios of seed germination (after being buried for 20, 40, and 60 days) and tiller regeneration. Wild parent (WR) had significant effect on ratios of buried-seed germination. Significant effect was detected for interaction between T and WR on the ratios of buried-seed germination (only for 40 and 60 days) and tiller regeneration (Table 1). No significant differences in germination ratios were detected between transgenic and non-transgenic hybrid lineages before seed burial. However, transgenic hybrid lineages (from WR1) showed 26%, 45%, and 38% higher seed germination ratio than their non-transgenic counterparts, 20, 40, and 60 after days after burial (Fig. 4a). A similar trend was also observed in hybrid descendants derived from WR2 (Fig. 4b). Transgenic F1 and F2 hybrid lineages derived from WR1 showed 55% and 275% higher tiller regeneration ratios compared to their non-transgenic counterparts (Fig. 4c). However, no significant differences were detected between hybrid lineages derived from WR2 (Fig. 4d), indicating apparent maternal influences.

Differences in fitness-related traits affected by endogenous EPSPS level of transgene recipients

The content of endogenous EPSPS was varied significantly among different transgene recipients (WR, WDR, and cultivated rice parents) at different growth stages (Fig. 5). The degree of fitness differences (as indicated by the ratios of increased panicles and seeds per plant) between transgenic and non-transgenic hybrid lineages varied substantially among different transgene recipients. A weak negative correlation was observed between the fitness differences caused by the incorporation of the epsps transgene and the content of endogenous EPSPS of different types of the transgene recipients (Fig. 5).

Figure 5
figure 5

Correlation between the ratios of increased panicles (a, b, and c) or seeds (d, e, and f) and the content of endogenous EPSPS proteins in different transgene recipients (parents) at 60 (a and d), 100 (b and e), and 160 (c and f) days after seed germination. Solid diamonds: wild rice (WR1); empty diamonds: wild rice (WR2); empty circles: weedy rice (WRD1); solid circles: weedy rice (WRD2), and solid triangles: cultivated rice (Minghui-86).

Full size image

The two WDR populations expressed a low level of endogenous EPSPS (0.05~0.33%) at different growth stages. Accordingly, their transgenic lineages showed a more substantial increase in the number of panicles (17~34%) and filled seeds (55~57%) per plant (Fig. 5). In contrast, the two WR populations expressed a relatively high level of endogenous EPSPS (0.23~0.72%), and their transgenic counterparts showed a less increase in the number of panicles (9~13%) and filled seeds (19~27%) (Fig. 5). The parental rice line (Minghui-86) showed a moderate level of endogenous EPSPS and fecundity change (Fig. 5) compared to the WR and WDR populations. These results suggested the possible association, although not strong, between the levels of endogenous EPSPS expression and differences in fitness caused by the epsps transgene.

Discussion

We found evidently altered phenological characteristics (e.g., higher tillering rates and earlier flowering), increased fecundity (more seeds and higher seed-set ratios), and enhanced ability of overwinter survival for stocks/tillers in the crop-wild rice hybrid lineages containing an epsps glyphosate-resistance transgene, based on our three-year common-garden experiments. These results suggest that the transgene over-expressing epsps can change the life-cycle characteristics and increase fitness of crop-wild rice hybrid descendants in the glyphosate-free environment, similar with that reported by Wang et al.16 in a study with the same epsps transgene in crop-weed hybrid descendants. In addition, we also detected differences in the base-line expression level of the EPSPS protein encoded by the endogenous epsps gene in transgenic recipients (parents), which influenced the fitness effect of the epsps transgene in their corresponding crop-wild/weed hybrid lineages. This result suggests that the genetic background of transgene recipient populations can considerably affect fitness of a transgene, which is similar with that in a study involving two insect-resistance transgenes (Bt, Bt/CpTI)1, 3, 43, 58. Therefore, it is necessary to design proper strategies to effectively assess and manage the potential risks caused by introgression of transgenes that can increase weediness and invasiveness of wild relative populations.

Materials and Methods

Production of crop-wild hybrid lineages

An herbicide-resistant transgenic rice (Oryza sativa L.) line (EP3), its non-transgenic rice parent (Minghui-86), and two WR populations (WR1, WR2) were used to generate F1-F3 crop-wild hybrid descendants. The EP3 transgenic line containing a GE endogenous epsps gene from rice was produced via the Agrobacterium-mediated transformation from Minghui-8616, 59. The EP3 line used for hybridization was a T5 generation homozygous for the epsps transgene that was resistant to glyphosate34. The non-transgenic Minghui-86 is a widely-used rice variety in China. The two O. rufipogon populations were collected from Dongxiang in Jiangxi province (WR1) and Suixi in Guangdong province (WR2), China. To avoid the possible gene insertion effects at different loci on phenotypes, we analyzed glyphosate resistance and two fecundity-related traits (number of panicles and filled seeds per plant) of two independent transgenic rice lines (EP3 and EP4), using their non-transgenic parent (Minghui-86) as a control.

For creating crop-wild hybrid lineages, we produced F1 transgenic and non-transgenic hybrids by hand pollination of WR plants (more than 20 plants per population) with EP3 and Minghui-86 rice lines in the designated Biosafety Assessment Centers in Fuzhou, Fujian Province. We selfed the F1 transgenic hybrids (EP3 × WR) to produce F2 and F3 hybrid lineages that segregated for the presence and absence of the target transgene to estimate the fitness effect of epsps transgene under the same genetic background in advanced generations of crop-wild hybrid descendants. The F2 transgenic hybrid plants used in the field experiment contained either hemizygous or homozygous epsps transgene, whereas the F3 transgenic hybrid plants only contained homozygous epsps transgene (Fig. 6).

Figure 6
figure 6

Schematic illustration of the pedigrees to produce F1-F3 crop-wild hybrid lineages. T: transgenic; + +, + −, and − −: transgene homozygous, transgene heterozygous, and non-transgenic, respectively. F1 and F2 hybrid lineages were used to test differences in fitness; the F3 hybrid lineages were used to test differences in seed germination after being buried in soils.

Full size image

The identification of transgene status of crop-weed hybrid descendants in F1 and F2 generations was achieved by specific molecular markers16. For the F3 hybrids, we obtained transgene homozygous and non-transgene homozygous F3 hybrid lineages by randomly screening more than 15 seeds harvested from each F2 plant for the presence or absence of transgene. The above homozygous lineages were preserved for future F3 experiment. Thus, the plant materials used in the experiments included WR parents (WR1, WR2), F1 hybrids (WR1-F1+ and WR2-F1+ with presence of epsps transgene; WR1-F1- and WR2-F1- with absence of epsps transgene), F2 hybrid lineages (WR1-F2+ and WR2-F2+ with presence of epsps transgene; WR1-F2- and WR2-F2- with absence of epsps transgene), and F3 hybrid lineages (WR1-F3+ and WR2-F3+ with presence of epsps transgene; WR1-F3- and WR2-F3- with absence of epsps transgene).

Field experiment design

Common garden field experiments were carried out in the designated Biosafety Assessment Centers in Fuzhou, Fujian Province, to estimate the effects of epsps transgene on vegetative growth, phenology, fecundity, and overwintering regeneration. Six sets of materials were included in the experiments: transgenic F1 or F2 hybrid lineages and their non-transgenic F1 or F2 counterparts derived from WR1 and WR2, as well as the two wild parents. Two cultivation modes, pure planting of transgenic, non-transgenic hybrid lineages, or the parents and mix planting of transgenic and non-transgenic hybrid lineages alternately at different densities, were designed to estimate the competitive abilities between the transgenic and non-transgenic hybrid lineages. For each treatment, six replicates (plots) were included. In pure planting, each plot included 36 plants in 6 × 6 grid with 50 cm spacing for the F1 and F2 experiments. In mix planting, each plot also included 36 plants in 6 × 6 grid but with 30 cm spacing for the F1 and F2 experiments, and 40 cm and 50 cm spacing only for the F1 experiment. Consequently, a total of 72 (in F1) or 48 (in F2) plots were included for the field experiments. The field layout of all experimental plots followed a complete randomized design.

Seed burial experiments were carried out in the confined experimental blocks of Fudan University campus in Shanghai to estimate the germination ability of hybrid seeds after being buried in soils. Four groups of F3 hybrid seeds were included for the seed burial experiments: WR1-F3+, WR1-F3-, WR2-F3+ and WR2-F3-. These hybrid seeds were treated at 50 °C for 7 days to break the seed dormancy. The treated seeds were buried in the soil of a rice field after rice harvesting, for 0, 20, 40, and 60 days before seed germination. Consequently, a total of 16 treatments with three replicates (bags) and 48 bags were included in the experiments. Each nylon bag contained 50 seeds that was randomly buried in 10 cm depth of soils from December to next-year February. The buried seeds were moved out at the different days after burial and germinated on the moist filter papers in a petri dish at 30 °C to examine the seed germination (see detail in Table S5).

Correlation between endogenous EPSPS contents in parental plants and fitness change caused by the transgene

To study relationships between endogenous EPSPS protein contents and fitness changes, we measure EPSPS contents in transgene recipient parents, including WR (WR1 and WR2), WDR (WDR1 and WDR2), and cultivated rice (Minghui-86), using ELISA (enzyme linked immunosorbent assay), in addition to the increased panicles and seeds per plant. The EPSPS protein content was measured as the ratio (%) between the amount of EPSPS protein and the total amount of soluble proteins. Pooled leaf tissues from three plants were collected as one sample (replicate) and nine samples from each type of parental plants were included at the vegetative (60 days), reproductive (100 days), and ripening stages (160 days). The Quantiplate kit (Envirologix, Portland, OR, USA) was used for the detection of the EPSPS proteins following the ELISA manufacturer’s protocols. We set the wavelength of the microtiter plate reader to 450 nanometers (nm) using Plate Reader (Bio-Rad Laboratories, Inc., Hercules, CA, USA). The MICROPLATE MANAGER (MPM) software ver. 6 (Bio-Rad Laboratories, Inc.) was used to summarize the results. The ratios of increased number of panicles and seeds per plant were estimated between transgenic and non-transgenic crop-wild or crop-weed hybrid lineages in the F2 generation, and between EP3 and Minghui-83. Data of crop-weed hybrid lineages used in this study were from Wang et al. (2014).

Data collection and analysis

The methods for data collection follows the description in Table S5. Two-way ANOVAs were carried out to analyze the effects of transgene (transgenic vs. non-transgenic), wild parent (WR1 vs. WR2), and their interaction on fitness in pure-planting plots. Independent and paired t-tests were used to determine differences between transgenic and non-transgenic hybrid lineages for fitness-related traits in pure-planting pots and mix-planting plots, respectively. Independent t-tests was used to detect differences in endogenous EPSPS contents between WR1 and WR2 based on the ELISA experiment. The correlation between endogenous EPSPS protein contents and fitness changes was calculated based on Pearson Correlation Coefficient. All statistical analyses were performed using the software IBM SPSS Statistics ver. 22.0 for Windows (SPSS Inc., IBM Company Chicago, IL, USA, 2010).

Data Availability

The datasets generated during and/or analysed during the current study are available from the corresponding author on reasonable request.