ALAN GRAY

The good sense that pervades earlier contributions to this debate (the notion of relative risk, the need for case-by-case assessment, weighing benefits and costs, and so on) expresses a belief in the scientific method. It reaffirms the view that science, and especially ecological science, has a sufficient toolkit to inform and improve the risk assessment process.

Whilst approving that general premise, I am becoming increasingly alarmed by the direction being taken by the wider debate about the environmental risks of genetically modified food crops. Two strands detectable in recent arguments are particularly worrying.

First is the assumption that by quantifying the risk we will somehow solve the risk assessment problem. This can be illustrated by reference to research on gene flow between crops and their wild relatives.

For example, in an influential study in Denmark¹, plants of the agricultural weed species wild turnip (Brassica rapa) grown experimentally in fields of oilseed rape (B. napus) between 9% and 93% of seeds produced were hybrids of the two plants. When occurring naturally in oilseed rape fields, turnip weeds were found to produce up to 60% hybrid seeds.

Similar levels of hybridisation, and the persistence of the first (F₁) and later generations of turnip/rape hybrids, have confirmed high rates of gene flow between these species.

By contrast, a recent study in the UK of wild turnip populations in a semi-natural environment (along the banks of the River Thames) close to oilseed rape fields within a broadly agricultural landscape, found flowering F₁ turnip/rape hybrids made up less than 7% of the populations. These produced between 0.4% and 1.5% hybrid seeds and less than 2% of their seedlings survived^2*.

The interesting question is not "Which estimate of gene flow is correct?", both are. Between them these studies simply confirm what ecologists might have predicted about rates of gene flow in arable agricultural situations as opposed to patchy natural populations. The interesting question is "Which number would make you change your mind about the risks of growing transgenic oilseed rape - 50%, 10%, or even 1%?".

Scientists rarely address that question explicitly, but we should be under no illusions that others do. Both opponents and supporters of the technology seize upon such numbers to wield as sticks in the argument for or against genetically modified organisms (GMOs).

Quantifying the risk may help us to make more informed judgements, but they remain judgements. Whether the forecast tells us we may expect some showers or that there is a 20% chance of rain, we still have to decide whether to risk taking a picnic.

Second, and more disturbing, is a trend probably caused by the asymmetric nature of risk assessment discussions, in that one can never prove that something is completely risk-free. Each new scientific finding is eagerly seized and exploited (often with astonishing leaps of logic unacceptable in other scientific debates), to call for a ban, or a moratorium, on the growth of transgenic crops.

One recent example is a laboratory study in which lacewing (Chrysoperia carnea) larvae were fed on prey which had been reared on transgenic Bt maize. The lacewings experienced significantly greater mortality and reduced development time than those whose prey fed on non-transgenic maize³. Another example is a field experiment in which wild thale cress (Arabidopsis thaliana) was more frequently fertilised by pollen from plants genetically transformed to confer resistance to a herbicide (chlorsulfuron) than by pollen from plants in which the same resistance allele had arisen by mutation⁴.

The authors of both studies legitimately claim to be offering contributions to the general debate about genetically modified crops. Both sets of authors, in admitting they do not know the underlying mechanisms involved, state they are unclear about the relevance to field situations or the generality of their results. This will not prevent the results being used to claim that transgenic crops are universally harmful to beneficial, non-target insects, or that transgenic plants are more outbreeding ("promiscuous") than conventional plants.

The second study in particular, the number of unknown, or unreported, features ("What is the variation in out-crossing rates across the species as a whole?" "How does one account for the variance in outcrossing rates (1.2% as opposed to 10.8%) in two of the transgenic lines?" "Could the mutant be disabled naturally or during incorporation into the wild-type?"), hardly allow us to assert conclusively that transgenic plants are more promiscuous.

Whether wittingly involved in these leaps of reason or not ecologists, who more than most scientists ought to be aware of the World's complexity, have a responsibility to avoid such 'sound bite' science.

It is difficult, but we must ask ourselves whether we would venture to publish the results of a single, unrepeated, and frankly mystifying experiment, if the debate was not so hot and the knowledge gaps so wide. Publishers of such results must be aware that they could be used to define an unacceptable risk, and even strangle a potentially beneficial technology in its pram.

In trying to enliven the debate, I have left myself little space to discuss what I believe to be the way forward. However, the results of the oilseed rape/wild turnip gene flow research discussed earlier provide a convenient starting point.

If the a genetic construct of concern produces, say, herbicide-tolerance, then for oilseed rape and other crops where hybridization with weedy relatives has been demonstrated unequivocally (for example sugar beet in Europe⁵), farmers and those who produce and sell seed to farmers, could have a problem if they fail to manage these crops effectively. Agronomic practices may reduce or eliminate risk, and that is where risk assessments should be focused. Similar problems, involving different crop/weed combinations, can be expected around the world.

The low rates of gene flow and establishment of turnip/rape hybrids in natural populations (hybrids incidentally known for sufficiently long in the British flora to have been taxonomically dignified with the name B. X harmsiana) alert us to the probability that genes from transgenic oilseed rape can 'escape' by hybridization into the wider countryside. And we must presume they have done so for many years from conventional crops.

This only addresses the first part of the classic risk evaluation equation ('risk' = 'exposure' x 'hazard'). In this context 'exposure' is the frequency/probability of a gene's escape, which can be quantified from the gene flow and distribution estimates becoming available for many crops. This part of the equation will remain largely constant for any given crop in a particular agricultural region.

Where I believe ecologists' attention should be focused is on 'hazard'. Here, 'hazard' is the impact of the inserted gene on the biology of the crop's wild relative and on the dynamics of the population of that plant.

These issues lie on the critical path of risk assessment and so research is urgently needed on problems we can as yet, only see on the horizon. For example, pest- and disease-resistance genes, especially resistance to viruses; combinations of genetically modified traits; stress-tolerance genes of various sorts⁶; ecosystem scale effects in agriculture; and so on.

Where possible, this research should be done in those countries and on those wild relatives where gene flow has been identified as an issue⁷. We have come a long way with oilseed rape and are now targeting research on those stages of the lifecycle that could most affect population growth rates(for example refs 8, 9).

Ecologists are now in a position to frame the questions which should be answered, by experiments where necessary, in the risk assessments that must accompany applications to release novel genes. The growing feeling that the future of genetic modification in agriculture will be settled in the political arena, and not by good science, should not lead us to abrogate that responsibility.

Alan Gray
Institute of Terrestrial Ecology, Furzebrook Research Station, Wareham, Dorset, BH20 5AS

1. Jorgensen, R.B., Andersen, B., Landbo, L. & Mikkelsen, T.R. Spontaneous hybridisation between oilseed rape (Brassica napus) and weedy relatives. Acta Horticulturae 407, 193-200 (1998).

2. Scott, S.E. & Wilkinson, M.J. Transgene risk is low. Nature 393, 320 (1998).

3. Hilbeck, A., Baumgartner, M., Fried, P.M. & Bigler, F. Effects of transgenic Bacillus thuringiensis corn fed prey on the mortality and development time of immature Chrysoperia carnea (Neuroptera, Chrysopidae). Environ. Entomol. 17, 480-487 (1998).

4. Bergelson, J., Burrington, C.B. & Wichmann, G. Promiscuity in transgenic plants. Nature 395, 25 (1998).

5. Boudry, P., Mörchen, M., Saumitou-Laprade, P., Vernet, P. & Van Dijk, H. The origin and evolution of weed beets: consequences for the breeding and release of herbicide-resistant transgenic sugar beets. Theoret. Appl. Genet. 87, 471-478 (1993).

6. Cooper, J.I. & Raybould, A.F. Transgenes for stress tolerance: consequences for weed evolution 1997 Brighton Crop Protection Conference - Weeds, 265-272 (1997).

7. Raybould, A.F. & Gray, A.J. Genetically modified crops and hybridisation with wild relatives: a UK perspective. J. Appl. Ecol. 30, 199-219 (1993).

8. Raybould, A.F., Moyes, C.L., Maskell, L.C., Elmes, G.W., Wardlaw, J.C., Randle, Z., Rispin, W.E., Cooper, J.I., Edwards, M.-L., McCall, D. & Gray, A.J. Predicting the ecological impacts of pest- and disease-resistant genetically modified crops. Annual Report of the Institute of Terrestrial Ecology 1996-97, 41-43 (1997).

9. Linder, R.C. & Schmitt, J.. Potential persistence of escaped transgenes: performance of transgenic, oil-modified Brassica seeds and seedlings. Ecol. Appl. 5, 1056-1068 (1995).

[Moderator's note: there is a minor error relating to the results of the wild turnip study although this does not affect the author's arguments. The work cited2 found that hybrids would form in less than 7% of B. rapa populations and within these populations, a maximum of 0.4-1.5% of seeds set would be hybrids. Only 2% of all seedlings germinating in such populations survive to flowering.]