| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Department of Plant Sciences, University of Manitoba, 222 Agriculture Building, Winnipeg, MB, Canada R3T 2N2
* (christian_willenborg{at}umanitoba.ca)
Dear Editor:
The introduction of genetically engineered (GE) crops has resulted in a dramatic increase in the number of studies providing estimations and predictions of gene flow at various temporal and spatial scales. Before the work of Gustafson et al. (2005), no model existed concerning pollen-mediated gene flow (PMGF) in wheat. This led the authors to develop an empirical regression model that permitted generalized predictions regarding isolation distances, harvest blending, and the effects of field size on PMGF in wheat. However, we are compelled to write this letter because we have serious concerns about the restrictive assumptions the authors used, the robustness of the data sets utilized to derive the model, and, subsequently, the validity of the conclusions drawn from this model.
Any model, whether empirical, mechanistic, or stochastic is inherently limited in its predictive accuracy and precision by the quality of the data and validity of the assumptions on which the model is built. The "general wheat model" (GWM) developed by Gustafson et al. (2005) was an empirical model (negative exponential) fit to data from five studies [see Table 1 in Gustafson et al. (2005)]. Two problems arise from the inclusion of these data sets: first, the largest pollinator size was a 50- x 50-m block (Matus-Cádiz et al., 2004) and second, only one data set (Matus-Cádiz et al., 2004) included any actual measure of long distance PMGF. Gene flow is a stochastic process and adequate sampling over space and time are critical to reveal the range of variation in gene flow (e.g., Rieger et al., 2002). Although sampling in Matus-Cádiz et al. (2004) extended to 2.7 km in some cases, the study was limited by the small pollinator source (50- x 50-m, 100 plants m–2), which not surprisingly, resulted in virtually no gene flow detected past 100 m.
The limited basis for true estimation of long distance PMGF in the model of Gustafson et al. (2005) is problematic because long distance PMGF (the tail of the gene flow curve) is the most important aspect of gene flow due to its effects on metapopulation structure and plant population dynamics (Austerlitz et al., 2004). We find it concerning that no consideration was given to describing the tail of the gene flow curve in the manuscript given that the most fundamental task in studying dispersal is describing the dispersal pattern (Nathan, 2001; Levin et al., 2003). Moreover, the authors conclude that the GWM is conservative or "high end" in its estimates of PMGF. This is misleading because this conclusion is an artifact of fitting a negative exponential model to data from numerous studies examining gene flow at scales of less than 35 m (in fact three of the five studies included examined gene flow at less than 1 m). As a result, the exponential model will overestimate but better describe gene flow at short distances and underestimate and poorly describe gene flow at long distances (Newstrom et al., 2003). By omitting or overlooking the importance of the tail of the gene flow curve, Gustafson et al. (2005) reach the erroneous conclusion that pollinator source size has little effect on PMGF. This conclusion is counterintuitive and contradicts previous findings where model derivation was based on much more robust data sets (Giddings, 2000; Ingram, 2000; Eastham and Sweet, 2002). The error in the conclusions drawn by Gustafson et al. (2005) results from the fact that the literature, and even the work of one of the study's authors (Hucl, 1996; Hucl and Matus-Cádiz, 2001; Matus-Cádiz et al., 2004), shows that as pollinator source size increases, so too does long distance pollen dispersal and hence gene flow. By fitting a PMGF model without adequate estimation of the tail the authors essentially ensure the conclusion that pollinator source size will have little effect on PMGF.
Without robust data sets for estimating long distance PMGF, the predictive accuracy and precision of any gene flow model will be compromised. Consequently, it is important for readers to be able to assess, especially in the absence of the use of robust data sets, the reliability of a given model. The authors presented no coefficient of determination or standard errors of parameter estimates on which to judge model performance. It would have been very helpful to readers if Gustafson et al. (2005) provided such measures of model performance.
There are further conclusions drawn by Gustafson et al. (2005) that are based on rather unusual assumptions. For example, they state on p. 1287 that "Seed containing detectible genetic material would not obviously result in the sustained presence of the gene(s) in subsequent generations if it is consumed as grain rather than planted or left in the ground as seed. Therefore, the PMGF values presented in this work should be treated as upper-bound estimates in subsequent forms of population modeling requiring PMGF as an input parameter." These assumptions are illogical and misleading given that in western Canada, for example, volunteer wheat persists on at least 18% of fields (Leeson et al., 2005). Moreover, harvest losses in wheat are known to be as high as 4%, resulting in approximately 400 to 450 seeds m–2 being returned to the seed bank (Clarke, 1985). Clearly volunteer wheat populations can be numerous and contribute to both gene flow and sustaining the presence of a gene for multiple generations. In fact, volunteer wheat has been shown empirically to persist in western Canada for up to 5 yr (Beckie et al., 2001). According to Gustafson et al.'s (2005) model, close-proximity gene flow between the volunteers and wheat crop could be as high as 1%, ultimately contributing to a high level of genetic admixture. No mention or incorporation of the contribution of volunteers is made within the manuscript, which is surprising considering that volunteers are sexually compatible with and grow within GE wheat crops. Furthermore, in western Canada approximately 70% of farmers plant farm-saved wheat seed (Van Acker et al., 2004), and therefore even the small levels of gene flow predicted by Gustafson et al.'s (2005) model will not only be concerning but could lead to the sustained presence of a gene. In addition, if the gene confers a fitness advantage such as insect, herbicide, or drought resistance, selection pressure could allow for the rapid accumulation of the gene in both cropped and volunteer wheat populations.
While we recognize the inherent difficulties in modeling PMGF, consideration must be given to the cumulative effects of the factors outlined in this letter before making generalized conclusions regarding PMGF in wheat such as those made by Gustafson et al. (2005). It also important to recognize that agricultural landscapes are a mosaic of arable and wild lands composed of randomly distributed plant communities, populations, and metapopulations, and it is within this context that the potential for PMGF must be evaluated.
Received for publication June 15, 2005.
REFERENCES
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| The SCI Journals | Agronomy Journal | Vadose Zone Journal | |||
| Journal of Natural Resources and Life Sciences Education |
Soil Science Society of America Journal | ||||
| Journal of Plant Registrations | Journal of Environmental Quality |
The Plant Genome | |||
