Comment on ‘A novel 5-enolpyruvoylshikimate-3-phosphate (EPSP) synthase transgene for glyphosate resistance stimulates growth and fecundity in weedy rice (Oryza sativa) without herbicide’ by Wang et al. (2014)


We wish to point out the weaknesses in the experimental design and in the interpretation of the results by the authors in the paper of Wang et al. (2014; this issue of New Phytologist, pp. 679–683) entitled ‘A novel 5-enolpyruvoylshikimate-3-phosphate (EPSP) synthase transgene for glyphosate resistance stimulates growth and fecundity in weedy rice (Oryza sativa) without herbicide’ published online on 1 August 2013.

Sloppy research, wrong conclusions

In their manuscript Wang et al. wanted to analyse the effect of the EPSPS-mediated herbicide tolerance trait on the fitness of crop–weed hybrids in the absence of the herbicide glyphosate. They used one herbicide-tolerant genetically engineered (GE) rice line (EP3) and its nonGE rice parent Minghui-86 as control, and crossed them with four weedy rice accessions.

The F1 progeny of the GE (EP3 × weed) and nonGE (Minghui-86 × weed) crop–weed hybrids were analysed in a field experiment in 2009. The authors observed significant increases in the number of tillers and flowering panicles per plant in the GE progeny compared with the nonGE controls in each crop–weed hybrid lineage. However, if one has a close look at the original Supporting Information Table S1 of the Wang et al. paper (here partly reproduced in Table 1), it is clear that the GE rice line EP3 produces significantly more tillers compared to the nonGE parental variety Minghui-86. This means that (1) the Minghui-86 line is not the isogenic line or (2) the EP3 line differs from the Minghui-86 line in more aspects than the transgene expression alone. If option (2) were true – which is most likely – this could, for example, be the result of the insertion of the transgene (Alonso et al., 2003). The authors did not investigate this possibility and thus incorrectly concluded that the observed effects are because of EPSPS expression. The only conclusion that can be drawn is that the parent GE line already differs from the control nonGE line.

Table 1. Mean plant height and number of tillers for the transgenic EP3 and nontransgenic rice (Oryza sativa) Minghui-86 lines
Plant materialPlant heightNo. of tillers
  1. Reproduced from Wang et al. (2014; Supporting Information Table S1). Means were calculated from averages taken from 36 plants in each of four replicate plots (N = 4). Numbers in parentheses following the means indicate standard errors (*P < 0.05, from independent t-tests). Note that the EP3 transgenic line produced significantly more tillers per plant than the nontransgenic control.

EP3 (Minghui-86 with epsps transgene)119.4 (1.5)19.0 (1.1)*
Minghui-86 (nontransgenic)118.8 (1.0)14.3 (0.9)

In a field experiment in 2011, the F2 progeny was analysed. The authors used a different approach compared to the F1 analyses. The GE F1 (EP3 × weed) crop–weed hybrids were selfed to obtain a F2 population. From that F2 population, plants were selected with the transgene (designated as GE F2 crop–weed hybrids) and without the transgene (designated as nonGE F2 crop–weed hybrids; Fig. 1). To evaluate the effect of the EPSPS transgene in the absence of glyphosate both groups were compared to each other. Wang et al. observed that the F2 plants from the four lineages of weedy rice produced 48–57% more seeds per plant than the nonGE controls in monoculture plots and 85–125% more seeds per plant in mixed plots where direct competition favoured the GE plants at the expense of the nonGE plants. Again the authors erroneously conclude that the EPSPS transgene is causing this difference because they assume they solely selected for the transgene. However, they selected for the genomic region which contains the transgene. Given that the authors used only one EPSPS-mediated glyphosate-tolerant line – that is EP3 – and that they did not perform a molecular characterisation of the EP3 line, they can exclude neither any insertion effects of the transgene nor the possible linkage with neighbouring sequences.

Figure 1.

Schematic representation of the transmission of the transgene genomic region from the genetically engineered (GE) rice in F1 and F2 hybrid populations. The grey region represents the 5-enolpyruvoylshikimate-3-phosphate synthase (EPSPS) transgene, the black region marked with ‘?’ illustrates a sequence that putatively stimulates tiller formation and the dotted region symbolizes sequences from weedy rice. As illustrated in the scheme, in the absence of crossing-over selection for the transgene implicates selection for the ‘?’ sequence.

The insertion effect may indeed be one of the explanations of the results because the Wang et al. data clearly show that the GE line EP3 already had more tillers than its isogenic nonGE Minghui-86 line (Table 1). Selecting for the transgene in F2 implicates a selection for the insertion effect. Similarly, in a F2 population obtained from a selfed F1 crop–weed cross the transgene will still be linked to closely neighbouring sequences as in the original EP3 line. If one of the neighbouring sequences would stimulate the number of tillers, then selecting for the transgene in F2 would also mean selecting for an increased number of tillers (Fig. 1).

To exclude these alternative explanations of the results, the authors could have used two independent EPSPS-mediated glyphosate-tolerant rice lines, or at least sequenced the right and left border regions of the transgene insertion to look for genes involved in tiller formation or to check whether the insertion induced expression changes of the neighbouring genes. A more time-consuming but also more accurate approach would be to backcross the crop–weed hybrid multiple times with the respective weedy rice accession. Multiple backcrosses not only isolate the transgene from linked neighbouring sequences, but also give the opportunity to create both GE and nonGE populations with a similar genetic background, and thus the opportunity to analyse the effect of the transgene itself. The bottom line is that Wang et al. studied the effect of a transgene on weedy hybrids without first investigating the impact of the transformation on the parental line.

Contradicting years of experience

The EPSPS-mediated glyphosate tolerance trait is not new. It is incorporated in several major crops which have been commercially grown in the field since 1996. The herbicide tolerance trait has been reported to make farming these crops more productive than nontolerant equivalents because it goes along with efficient weed management, lower inputs, easier use of no-tillage, etc. (Dill, 2005). However, the herbicide tolerance trait does not influence yield per se (Brookes & Barfoot, 2006). Also during the deregulation process phenotypical comparisons between the GE plant and its isogenic line are obligatory. Any effects of additional EPSPS expression should have already been noticed. It is, however, useful to note that Wang et al. used overexpression of the native rice EPSPS gene while the majority of the glyphosate-tolerant crops in the field express the EPSPS gene of Agrobacterium sp. CP4. Nonetheless, the recommendation of Wang et al. that herbicide tolerance increases fitness of plants is completely in contrast to 20 yr of experimental results. Because of the shortcomings in the experimental setup of Wang et al. (already described), their observations are likely to be unrelated to the transgene.

Unnecessary negative influence on the genetically modified organism (GMO) debate

Wang et al. explicitly express their concern on the impact of their ‘finding’ on the use of herbicide-tolerant crops in agriculture worldwide. The conclusion of Wang et al. could only be substantiated if full lifecycle studies indicated an increase in relative growth rates under conditions of competition with the nonGE isogenic controls and in the absence of the herbicide. However, the authors did not perform these experiments and thus do not provide this evidence. Yet, they unnecessarily harm the sensitive debate on GE crops.

Conflict of interest

The authors declare that VIB has no interest in the use of herbicides or in herbicide-tolerant crops. However, VIB is a world authority in plant research that uses genetically modified plants as a research resource. New knowledge that VIB gathers in this way might, in some cases, contribute to the development of genetically modified crops. For this reason, VIB considers it its social and scientific duty to thoroughly examine new information about the possible health and environmental effects of genetically modified plants.