Published online 8 September 2006
Published in Crop Sci 46:2133-2140 (2006)
© 2006 Crop Science Society of America
677 S. Segoe Rd., Madison, WI 53711 USA
PLANT GENETIC RESOURCES
Empirical Modeling of Genetically Modified Maize Grain Production Practices to Achieve European Union Labeling Thresholds
D. I. Gustafsona,*,
I. O. Brantsb,
M. J. Horaka,
K. M. Remunda,
E. W. Rosenbauma and
J. K. Soteresa
a Monsanto Company, 800 North Lindbergh Blvd, St. Louis, MO 63167 USA
b Monsanto International Sarl, Morges, Switzerland
* Corresponding author (david.i.gustafson{at}monsanto.com)
 |
ABSTRACT
|
|---|
An empirical approach is given for specifying coexistence requirements for genetically modified (GM) maize (Zea mays L.) production, to ensure compliance with the 0.9% labeling threshold for food and feed in the European Union. Field data were considered in which pollen-mediated gene flow (PMGF) was measured within maize receptor fields at a series of distances from source fields having a marker. An empirical model is presented that fits the observed decrease of gene flow with distance. The model was parameterized to provide both reasonable worst case and expected case predictions of gene flow for various combinations of isolation distance, use of non-GM border rows in the GM field and/or separately harvested border rows in the receptor field. Based on the data assessed, the model is used to show that the effect of scale is minimal for source fields of surface area 4 ha and greater. Combinations of isolation distance and border rows of 20 m or more are predicted to result in gene flow of less than 0.9%, as a blended average for receptor fields 1 ha or larger. Lesser requirements are necessary when the source field is much smaller than the receptor, and an extension to the model is provided to estimate such effects.
Abbreviations: AP, adventitious presence of a GM-trait in seed of the receptor field • Bt, Bacillus thuringiensis, the soil bacterium source of Cry toxins • EXC, expected case gene flow predictions • FGM, fraction of pollen assumed to contain a detectable trait of interest • GM, genetically modified via transgenic biotechnology techniques • PMGF, pollen-mediated gene flow • RWC, reasonable worst case gene flow predictions
|
INTRODUCTION
|
|---|
POLLEN-MEDIATED GENE flow (PMGF) is the transfer and incorporation of genetic information between distinct, sexually compatible plant populations via cross-pollination. Pollen dispersal is a natural biological process that occurs to some degree in all flowering plant species, including most major crops (Levin and Kerster, 1974). However, PMGF takes place only when incoming, viable pollen successfully out-competes locally produced pollen to form viable seed. PMGF is not unique to GM crops; however, the process has received renewed attention with the need to ensure commercial coexistence of non-GM and GM grain production (Timmons et al., 1996; Klinger and Ellstrand, 1999). In this paper we describe an empirical approach for deriving border row and isolation distance specifications to achieve compliance with the 0.9% EU threshold for labeling GM food and feed (Official Journal of the European Union, 2003).
The extent of pollen dispersal and PMGF in maize (corn) as a function of distance from a pollinator source has been measured in a number of field studies (Jones and Brooks, 1950; Haskell and Dow, 1951; Raynor et al., 1972; Sears and Stanley-Horn, 2000; Luna et al., 2001; Benetrix and Bloc, 2003; Henry et al., 2003; Ma et al., 2004; Halsey et al., 2005; Rosenbaum et al., 2005). These studies fall into two broad classes based on study design (see Fig. 1
). In Class 1, gene flow is measured at varying distances into a contiguous receptor field, where there would be full pollen competition between the incoming foreign pollen and a dense cloud of native pollen. In Class 2, gene flow has been measured in small receptor plots positioned at varying distances from the source, with the intervening distance generally left fallow and where pollen competition is minimal. This second class of field studies is considered separately in the section labeled "Isolation Distance" below.
View larger version (21K):
[in this window]
[in a new window]
|
Fig. 1. Schematic diagram showing the two classes of pollen-mediated gene flow field studies as defined in this work.
|
|
Several of the first class of field studies have been published (Benetrix and Bloc, 2003; Henry et al., 2003; Ma et al., 2004; Halsey et al., 2005; Rosenbaum et al., 2005). Varying degrees of cross-pollination in maize occur depending on biological, agronomic, and environmental factors. The complexity of these interactions restricts attempts to develop a mechanistic model; therefore, we pursued an entirely empirical modeling approach. A mechanistic approach would be one that attempts to model PMGF using complex equations intended to model pollen movement as influenced by environmental conditions. Several papers have been published which have attempted to do this in maize (e.g., Klein et al., 2003), but as pointed out by the authors these models are limited in their application. The empirical approach defined here is one where we examined several available datasets and used them to make general predictions about the effectiveness of border rows and isolation distance for reducing PMGF in a neighboring field below 0.9%.
|
MATERIALS AND METHODS
|
|---|
The 56 Class 1 datasets used in the analysis are listed in Table 1, all of which have source and receptor fields planted on the same date. For each dataset, we determined the distance into a receptor field at which observed PMGF fell and remained below 0.9%. We also determined the first 5-m harvester pass (approximately equal to six maize rows on 0.75-m centers) at which all subsequent passes would have an average PMGF below 0.9%. For example, 30 m means that the sixth harvester pass would give PMGF below the threshold. For cases where the observed data did not permit direct determination, we fit an empirical model to the observed data using one of two techniques involving transformation and subsequent linear regression: (i) using base-10 logarithms of both percentage of PMGF and distance; or (ii) using the logarithm of percentage of PMGF and the square root of distance, as described recently for gene flow in wheat (Gustafson et al., 2005). The empirical model fit having the higher r2 value was used for the extrapolation. The size of the source fields in the various studies analyzed ranged from 0.07 to 6.48 ha.
In addition to these calculations, the number of separately harvested border rows in a receptor field required to achieve 0.9% PMGF in a receptor field 100 m wide (corresponding to a square-shaped, 1-ha field) was determined directly from each field data set. This is the blended gene flow assuming uniform mixing or sampling to achieve a representative sample, as would be the intention of testing. A receptor field of 1 ha was chosen because this approximates one truck (or lorry) load. For larger fields, it could be considered the truck load with the highest PMGF. These calculations were performed for three values of FGM, defined as the fraction of pollen assumed to contain a detectable trait of interest. For current Bacillus thuringiensis (Bt) maize varieties in Europe, where the trait is hemizygous, FGM is 0.5, but FGM would be 0.75 for two-trait products formed by parents each having one trait. The other FGM value used in the calculations was 1, the maximum possible value, that is, all pollen contains a detectable trait. To maintain conservatism, the calculations summarized in Table 1 further assumed that the seed used to plant the receptor field already carried the monitored trait at a level of 0.3% due to the unavoidable adventitious presence (AP) of approved traits in the commercial biotech environment. This model assumes that AP remains constant in the absence of pollen-mediated gene flow, which is a reasonable assumption for a cross-pollinating crop like maize if the majority of AP in the seed of the receptor is hemizygous (Charlton et al., 2005).
Most of the trials listed in Table 1 were conducted using source and recipient fields with synchronous flowering ("nick"). Further details on flowering synchrony and the climatic conditions prevailing during these studies are available in the original publications, but they are broadly representative of the conditions encountered in EU maize production. Studies on this topic (e.g., Halsey et al., 2005; Rosenbaum et al., 2005) demonstrate that synchronization can have a significant impact. By keeping full synchronization as a basic assumption, we help to ensure that the recommendations made based on the model will be maximally robust and conservative.
|
RESULTS AND DISCUSSION
|
|---|
Results are presented and discussed for two options to manage gene flow from GM fields: (i) excluding border rows in the receptor field at harvest; or (ii) planting non-GM border rows within the source field. In both cases, the border rows can be reduced when the source and receptor fields are increasingly isolated from each other. The effect of the size of the GM source field is also considered.
Separately Harvesting Border Rows in the Receptor
According to the data listed in Table 1, separately harvesting border rows of 10 m in the receptor field would be sufficient to reduce blended average gene flow below 0.9% for square, 1-ha receptor fields corresponding to all but two of the 56 datasets, at the FGM value of 0.5 that is appropriate for Bt maize. Separately harvested border rows of 20 m would have been sufficient for the remaining trials. These conclusions hold regardless of isolation distance, which ranged from 0 to 18 m for these datasets.
Isolation Distance
Gene flow or exogenous pollination in maize at the upwind edge of small, isolated receptor plots (the Class 2 studies of Fig. 1) has been measured in a number of studies (Jones and Brooks, 1950; Sears and Stanley-Horn, 2000; Rosenbaum et al., 2005). As shown in Fig. 2
, an empirical model of the same form previously reported (Gustafson et al., 2005) provides a reasonable fit to the upper bound of these data. The model fit to these Class 2 data is:
 | [1] |
in which the following terms are defined:- Fo, fraction of exogenous pollination at the leading edge of the receptor field
- ID, isolation distance (m) from edge of source to receptor field
This equation suggests that an isolation distance of 50 m gives a fivefold reduction in gene flow at the leading edge of the receptor field and an isolation distance of 100 m gives a 10-fold reduction.
In light of the above data, it is possible to construct a single empirical model that combines the effects of adventitious presence, isolation distance, and border rows on gene flow into a receptor field. The model assumes that the percent GM in the receptor field grain is a simple sum of the contributions from AP and incoming pollen-mediated gene flow, which is in turn expressed as the product of four terms: (i) the fraction GM presence in the incoming pollen; (ii) an empirically determined maximum gene flow at the upwind edge; (ii) a term expressing the effect of isolation distance (from Eq. [1]); and (iv) a final term expressing the combined effect of decreases in gene flow with within the receptor field or border rows. The mathematical form of the relationship between gene flow and distance is the same within the receptor or across a fallow isolation buffer: a linear decline in the logarithm of gene flow with the square root of distance. The only difference is the proportionality constant (0.2), which states that pollen competition within the receptor field or border rows is equally effective in reducing gene flow, and twice as effective as an unplanted isolation buffer. This coefficient (0.2) represents a typical value when fitting individual field trial data. The resulting model is then:
 | [2] |
in which the following new terms are defined:- PMGF, pollen-mediated gene flow (%) at a particular point in the receptor field
- AP, adventitious presence (%) of the monitored trait in seed of the receptor field
- FGM, fraction pollen containing the monitored trait
- Po, percentage of gene flow at edge of the receptor field for FGM = 1 and in the absence of border rows, isolation distance, and adventitious presence in the planted seed
- BR, width of non-GM border rows (m) to be planted between source and receptor fields
- x, distance (m) from edge of receptor field nearest to the source
The more conservative predictions of this model for AP = 0.3%, FGM = 0.5, ID = 0.75 m (a standard planter row width), BR = 0 m (no border rows), and Po = 44.8% are shown in Fig. 3
, where it is denoted "RWC Model" for reasonable worst case. The less stringent predictions are labeled as "EXC Model" for expected case, and are based on Po = 5.9%, with other parameters unchanged. The decision to calibrate the model based on Po only was based on an observation that the slope factors in Eq. [2] (0.1 and 0.2) do not appear to change much in going from one set of field data to another, but rather most of the variation was due to changes in the y intercept of individual sets of field data. All field data have been adjusted to AP = 0.3%, FGM = 0.5, ID = 0.75 m, and BR = 0 m to make the comparison between the model and the data as relevant as possible. The two Po values were selected to encompass at least 50% (EXC) or 90% (RWC) of the field data. In other words, just more than half (50.3%) of the observed data lie below the EXC curve, and 90.1% of the data lie below the RWC curve. The RWC curve also gives a similar slope but considerably higher predictions than a regression model (denoted "Defra Model") recently published by the UK Department for Environment Food and Rural Affairs (Henry et al., 2003). An Excel spreadsheet version of the Monsanto model has been prepared and is available from the authors on request.
View larger version (20K):
[in this window]
[in a new window]
|
Fig. 3. Comparison of given empirical model predictions to observed gene flow data in maize (adjusted for isolation distance and FGM). EXC, expected case gene flow predictions; PMGF, pollen-mediated gene flow; RWC, reasonable worst case gene flow predictions.
|
|
The data in both Fig. 2 and 3 reflect a high degree of variability in the amount of gene flow that is observed at any particular distance. This is not surprising given the wide variation in the factors affecting PMGF (Gustafson et al., 2005). Among these are climatic variables, such as wind speed and direction, temperature, and relative humidity, each of which impacts the distance pollen can move and how long it will remain viable. Also, even in a crop such as maize, which has a relatively high out-crossing potential, there may be pollen compatibility issues that serve to reduce the amount of cross-pollination between two particular maize lines. The degree of flowering synchrony and agronomic practices such as fertilization can play a role as well. The net effect of these many competing forces is to cause a very wide range in observed gene flow percentages across the field studies.
The model has also been further enhanced to treat the special case of source fields that are very small relative to the size of the receptor field. It has, heretofore, been assumed that the gene flow may be adequately simulated as being one dimensional, thereby ignoring lateral variations in gene flow across the receptor field. This assumption seems reasonable for sources and receptors of roughly the same size (as shown in Fig. 1), but the assumption breaks down when the receptor field is much larger. A simple approach to generalizing the model has been taken as shown in Fig. 4
, where the source field is a square of side length, s, and the receptor is of width, r, in the direction of the assumed gene flow. It is then assumed that the simple one-dimensional model is valid for sub-receptor0, a rectangular portion of the receptor field closest to the source and of the same width as the source, s. The remainder of the receptor field is then divided into congruent, discrete, rectangular portions numbered as shown. By symmetry, we assume the gene flow into sub-receptorj is equal to that in the corresponding rectangle on the other side of the field, sub-receptor–j. It is further assumed that the effective isolation distance for sub-receptorj may be estimated by rotating the source and sub-receptor such that they come into alignment about a line segment connecting each of their centroids, as shown in for sub-receptor1. A general expression for the effective isolation distance of each sub-receptor may then be calculated by application of the Pythagorean Theorem:
 | [3] |
The RWC model predictions were used to study the influence of isolation distance on the border row requirements to achieve gene flow of less than 0.9% in a 1 ha receptor field is as shown in Table 2 for Scenario A. (The EXC model predictions are for gene flow to remain no higher than 0.7% regardless of isolation distance and border row practice.) The RWC model predicts that the sum of the isolation distance and border rows must be about 20 m to achieve less than 0.9% gene flow: for example, 15-m isolation distance plus 3 m of border rows (total 18 m for Scenario A in Table 2). The model predicts that border rows are more effective than fallow isolation buffers at reducing gene flow, but this advantage is of no practical consequence so long as the fields are at least 20 m apart.
View this table:
[in this window]
[in a new window]
|
Table 2. Combinations of isolation distances and border rows (m) predicted necessary to achieve pollen-mediated gene flow (PMGF) < 0.9%.
|
|
Scale Effect
For pollen sources smaller than 4 ha, a scale effect is present for the data presented in Table 1, with the smaller pollen sources tending to show less gene flow (degree of statistical significance, P = 0.0124, based on a simple t test comparison of mean gene flow between the two groups). However, based on these datasets, the scale effect is minimal for fields larger than 4 ha, as previously reported for gene flow in wheat (Gustafson et al., 2005). Application of the RWC model for maize confirms this, as indicated in Fig. 5
. This conclusion would be consistent with the concept that the basic mechanism of adventitious gene flow is directly related to pollen competition, e.g., the amount of viable foreign pollen relative to the amount of viable indigenous pollen present in the area of viable silks within a maize canopy.
View larger version (9K):
[in this window]
[in a new window]
|
Fig. 5. Predicted effect of source size on observed gene flow according to the given empirical model. PMGF, pollen-mediated gene flow.
|
|
For the special case of source fields much smaller than the receptor, Eq. [3] may be used to estimate gene flow as a function of source size. The predicted gene flow for a large receptor field of 16 ha and a series of successively smaller sources is shown in Table 3, for the case of FGM = 0.5 and AP = 0. Consistent with the above data, very little difference in gene flow is seen when the source is nearly the same size as the receptor, but gene flow decreases more quickly as the source becomes much smaller than the receptor.
View this table:
[in this window]
[in a new window]
|
Table 3. Predicted gene flow to an immediately adjacent, large (16 ha) receptor field as a function of the size of the source field (with no adventitious presence of a genetically modified trait in seed of the receptor field in the receptor and half the source pollen bearing the trait).
|
|
Planting Border Rows within the Source
The datasets and models presented above can be used to directly establish guidelines relative to gene flow when non-GM border rows are included within a GM source with or without isolation between the source and receptor (cf. scenario B in Fig. 6
). This is justified on the basis that PMGF is a function of viable pollen competition and the use of border rows around a source serves as a trap to dilute the viable GM pollen. However, without direct empirical data we cannot eliminate the potential for higher than calculated levels of PMGF to be found in the first row of the receptor field. With this being possible, we recalculated the border row and isolation distance requirements with a lower 0.8% threshold to account for this possibility. The results of these calculations are shown in Table 2 and suggest that the width of border rows would need to be increased by 3 to 5 m relative to Scenario A, depending on isolation distance.
|
CONCLUSIONS
|
|---|
Detailed examination of 56 individual field datasets demonstrated that blended average PMGF to an adjacent 1-ha maize field could have been kept below 0.9% by separately harvesting the first 10-m border rows of the receptor in all but 1 case, for which 20 m of border rows would have been necessary. The empirical model provides both high-end and typical gene flow predictions as a function of isolation distance, use of non-GM border rows in the GM field and/or separately harvested border rows in the receptor field. Any combination of isolation distance and border rows of 20 m or more is predicted to result in gene flow of less than 0.9%, as a blended average for receptor fields 1 ha or larger. These model predictions are for large source fields of surface area 4 ha and greater. Gene flow from smaller sources is predicted to be even less, and a new extension to the empirical model is provided to handle such cases.
Received for publication March 10, 2006.
|
REFERENCES
|
|---|
- Benetrix, F., and D. Bloc. 2003. Possible coexistence. (In French.) Perspectives Agricoles 294:14–15.
- Charlton, S., R. Ram, G. Wandrey, C. Wiltse and D. Wright. 2005. Relationship between transgenic adventitious presence levels in conventional seed and its progeny. Am. Seed Trade Assoc., Alexandria, VA.
- Gustafson, D.I., M.J. Horak, C.B. Rempel, S.G. Metz, D.R. Gigax, and P. Hucl. 2005. An empirical model for pollen-mediated gene flow in wheat. Crop Sci. 45:1286–1294.[Abstract/Free Full Text]
- Halsey, M.E., K.M. Remund, C.A. Davis, M. Qualls, P.J. Eppard, and S.A. Berberich. 2005. Isolation of maize from pollen-mediated gene flow by time and distance. Crop Sci. 45:2172–2185.[Abstract/Free Full Text]
- Haskell, G., and P. Dow. 1951. Studies with sweet corn 5: Seed-settings with distances from pollen source. Emp. J. Exp. Agric. 19:45–50.
- Henry, C., D. Morgan, R. Weekes, R. Daniels, and C. Boffey. 2003. Farm scale evaluations of GM crops: Monitoring gene flow from GM crops to non-GM equivalent crops in the vicinity (contract reference EPG 1/5/138). Part I: Forage maize. Final Report, 2000/2003. Dep. for Environment, Food and Rural Affairs, London.
- Jones, M.D., and J.S. Brooks. 1950. Effectiveness of distance and border rows in preventing outcrossing in corn. Oklahoma Agric. Exp. Stn. Tech. Bull. No. T-38. Oklahoma State Univ., Stillwater, OK.
- Klein, E.K., C. Lavigne, X. Foueillassar, P.-H. Gouyon, and C. Laredo. 2003. Corn pollen dispersal: Quasi-mechanistic models and field experiments. Ecol. Monogr. 73:131–150.
- Klinger, T., and N.C. Ellstrand. 1999. Transgenic movement via gene flow: Recommendations for improved biosafety assessment. p. 129–140. In K. Ammann, Y. Jacot, V. Simonsen, and G. Kjellsson (ed.) Methods for risk assessment of transgenic plants. III. Ecological risks and prospects of transgenic plants. Birkhäuser Verlag, Basel.
- Levin, D.A., and H.W. Kerster. 1974. Gene flow in seed plants. Evol. Biol. 7:139–220.
- Luna S.V., J. Figueroa M., B. Baltazar M., R. Gomez K., R. Townsend, and J.B. Schoper. 2001. Maize pollen longevity and distance isolation requirements for effective pollen control. Crop Sci. 41:1551–1557.[Abstract/Free Full Text]
- Ma, B.L., K.D. Subedi, and L.M. Reid. 2004. Extent of cross-fertilization in maize by pollen from neighboring transgenic hybrids. Crop Sci. 44:1273–1282.[Abstract/Free Full Text]
- Official Journal of the European Union. 2003. Regulation (EC) No. 1830/2003 of the European Parliament and the Council (September 22, 2003) concerning the traceability and labeling of genetically modified organisms and the traceability of food and feed products from genetically modified organisms and amending Directive 2001/18/EC, L268/P.24, 18.10.2003.
- Raynor, G.S., E.C. Ogden, and J.V. Hayes. 1972. Dispersion and deposition of corn pollen from experimental sources. Agron. J. 64:420–427.[Abstract/Free Full Text]
- Rosenbaum, E.W., M.J. Horak, T.A. Pester, and T.E. Nickson. 2005. Evaluation of temporal isolation on frequency and distance of pollen-mediated gene flow in corn measured at two interfield spacings. p. 132. In Proc. of the North Central Weed Science Society Annual Meeting, Kansas City, MO. 12–15 Dec. 2005. North Central Weed Sci. Soc., Champaign, IL.
- Sears, M.K., and D. Stanley-Horn. 2000. Impact of Bt corn pollen on monarch butterfly populations. In C. Fairbairn, G. Scoles, and A. McHughen (ed.) Proc. of the 6th Int. Symp. on The Biosafety of Genetically Modified Organisms, Saskatoon, SK. Univ. of Saskatchewan, Saskatoon.
- Timmons, A.M., Y. Charters, J.W. Crawford, D. Burn, S. Scott, S.J. Dubbels, N.J. Wilson, A. Robertson, E.T. O'Brien, G.R. Squire, and M.J. Wilkinson. 1996. Risks from transgenic crops. Nature 380:487.[CrossRef][Medline]
TOP
ABSTRACT
INTRODUCTION
