An empirically derived model of field-scale gene flow in winter wheat.

Abstract

The potential introduction of wheat (Triticum aestivum L.) cultivars with transgenic traits has generated increased interest in pollen-mediated gene flow (PMGF). The objectives of this study were to estimate wheat PMGF between commercial fields across multiple years and locations, and to compare estimates from large fields to those from smaller experimental plots. The study was conducted in a total of 56 commercial field locations in eastern Colorado in 2003, 2004, and 2005. We measured PMGF by tracking the movement of an imidazolinone herbicide resistance gene from resistant to susceptible cultivars, sampled at distances of 0.23 to 61 m. At least one sample from all 56 fields and from all 18 evaluated cultivars had detectable PMGF. The highest observed PMGF was 5.3% at 0.23 m. The farthest distance at which PMGF was detected was 61 m and the highest PMGF at that distance was 0.25%. Higher levels and greater distances of PMGF were detected in commercial fields than in experimental plots. Based on estimates from a generalized linear mixed model with a random location effect, the distance required to ensure 95% confidence that 95% of locations have PMGF less than 0.9% is 41.1 m for cultivars heading earlier than the pollen source and 0.7 m for cultivars heading later than the pollen source. These confidence limits should represent the highest levels of PMGF expected to occur in winter wheat in the west-central Great Plains and will be useful for wheat biotechnology regulation.

Full text

Page 1

Reproduced from Crop Science. Published by Crop Science Society of America. All copyrights reserved.
2308
CROP SCIENCE, VOL. 47, NOVEMBER–DECEMBER 2007
RESEARCH
G
plants through pollen dispersal. Gene fl ow in wheat has histori-
cally been a concern only for seed production, as wheat is consid-
ered a self-pollinating species with generally less than one to two
percent outcrossing (Poehlman and Sleper, 1995). More recently,
wheat cultivars and experimental lines with novel non-transgenic
traits such as herbicide resistance (Newhouse et al., 1992) or trans-
genic traits such as herbicide and pathogen resistance (Zhou et
al., 2003; Schlaich et al., 2006) have generated increased interest
in gene fl ow between wheat fi elds. Understanding gene fl ow via
pollen movement will be critical to ensure coexistence of trans-
genic and non-transgenic crops (Messean et al., 2006). Although
no transgenic wheat cultivars have been commercialized to date,
ene flow, defi ned as movement between populations that
results in genetic exchange (Hedrick, 2005), can occur in
An Empirically Derived Model of
Field-Scale Gene Flow in Winter Wheat
Todd A. Gaines, Patrick F. Byrne,* Philip Westra, Scott J. Nissen,
W. Brien Henry, Dale L. Shaner, and Phillip L. Chapman
ABSTRACT
The potential introduction of wheat (Triticum
aestivum L.) cultivars with transgenic traits has
generated increased interest in pollen-medi-
ated gene fl ow (PMGF). The objectives of this
study were to estimate wheat PMGF between
commercial fi elds across multiple years and
locations, and to compare estimates from large
fi elds to those from smaller experimental plots.
The study was conducted in a total of 56 com-
mercial fi eld locations in eastern Colorado in
2003, 2004, and 2005. We measured PMGF
by tracking the movement of an imidazolinone
herbicide resistance gene from resistant to sus-
ceptible cultivars, sampled at distances of 0.23
to 61 m. At least one sample from all 56 fi elds
and from all 18 evaluated cultivars had detect-
able PMGF. The highest observed PMGF was
5.3% at 0.23 m. The farthest distance at which
PMGF was detected was 61 m and the highest
PMGF at that distance was 0.25%. Higher levels
and greater distances of PMGF were detected
in commercial fi elds than in experimental plots.
Based on estimates from a generalized linear
mixed model with a random location effect, the
distance required to ensure 95% confi dence that
95% of locations have PMGF less than 0.9% is
41.1 m for cultivars heading earlier than the pol-
len source and 0.7 m for cultivars heading later
than the pollen source. These confi dence lim-
its should represent the highest levels of PMGF
expected to occur in winter wheat in the west-
central Great Plains and will be useful for wheat
biotechnology regulation.
T.A. Gaines, P.F. Byrne, Colorado State Univ., Dep. of Soil and Crop
Sciences, 1170 Campus Delivery, Fort Collins, CO 80523; P. Westra,
S.J. Nissen, Dep. of Bioagricultural Sciences and Pest Management,
1177 Campus Delivery, Fort Collins, CO 80523; W.B. Henry, USDA/
ARS Central Great Plains Research Station, Akron, CO 80720, pres-
ent address, Room 326 Dorman Hall, Stone Blvd., Mississippi State, MS
39762; D.L. Shaner, USDA/ARS Water Management Research, 2150
Centre Avenue, Fort Collins, CO 80526; P.L. Chapman, Dep. of Statis-
tics, 1877 Campus Delivery, Fort Collins, CO 80523. Received 23 July
2007. *Corresponding author e-mail: Patrick.Byrne@colostate.edu.
Abbreviations: ALS, acetolactate synthase; AIC, Aikake information
criterion; GLMM, generalized linear mixed model; GWM, general
wheat model; IR, imidazolinone-resistant; IS, imidazolinone-suscep-
tible; PCR, polymerase chain reaction; PMGF, pollen-mediated gene
fl ow; RHC, relative heading class.
Published in Crop Sci. 47:2308–2316 (2007).
doi: 10.2135/cropsci2007.05.0248
© Crop Science Society of America
677 S. Segoe Rd., Madison, WI 53711 USA
All rights reserved. No part of this periodical may be reproduced or transmitted in any
form or by any means, electronic or mechanical, including photocopying, recording,
or any information storage and retrieval system, without permission in writing from
the publisher. Permission for printing and for reprinting the material contained herein
has been obtained by the publisher.

Page 2

Reproduced from Crop Science. Published by Crop Science Society of America. All copyrights reserved.
CROP SCIENCE, VOL. 47, NOVEMBER–DECEMBER 2007
WWW.CROPS.ORG 2309
biotechnology regulatory agencies require robust esti-
mates of cross-pollination frequency between wheat fi elds
to determine isolation distances for fi eld evaluation of
experimental transgenic lines.
Wheat seed production guidelines address gene fl ow
between fi elds. In Colorado, isolation distances of 3 m
are required for certifi ed seed production (Anonymous,
2003). Several authors have suggested that to ensure
adequate purity in wheat seed production, a larger isola-
tion distance is needed, such as 30 m (Matus-Cadiz et al.,
2004) or 45 m (Hanson et al., 2005). Gene fl ow in wheat
has been detected at distances as great as 2.75 km, though
at the very low level of 0.01% (Matus-Cadiz et al., 2007).
Pollen movement is the fi rst step required for gene
fl ow to occur. Synchrony in fl owering between fi elds is
necessary, because viable pollen must successfully enter
fl orets, germinate on stigmas, and fertilize ovules (Waines
and Hegde, 2003). Anthesis in wheat can last up to 10 d
(De Vries, 1973) and stigmas can be receptive for a period
of four to 13 days (De Vries, 1971). Cross-pollination in
wheat can be aff ected by both genetic and environmental
factors. Cultivars are known to vary in receptivity to gene
fl ow under controlled conditions (Griffi n, 1987; Hucl,
1996; Hucl and Matus-Cadiz, 2001; Lawrie et al., 2006;
Martin, 1990). The optimal temperature for pollen pro-
duction is 21°C (Porter and Gawith, 1999), but freezing
or near-freezing temperatures at the boot, heading, and
anthesis stages can kill anthers, resulting in male steril-
ity (Shroyer et al., 1995) that leaves the surviving stigmas
more receptive to foreign pollen. High temperatures or
drought can increase cross-pollination frequency (Briggs
et al., 1999; Waines and Hegde, 2003) by increasing glume
opening, which is associated with higher outcrossing (De
Vries, 1971; Hucl, 1996; Waines and Hegde, 2003).
Pollen source size is known to infl uence outcross-
ing in recipient wheat (De Vries, 1974). Several studies
in wheat have used single row pollen sources to measure
gene fl ow via pollen over close distance (Griffi n, 1987;
Hucl, 1996; Martin, 1990). Other studies have used pollen
source blocks ranging in size from 5 m2 (Hucl and Matus-
Cadiz, 2001) to 50 m2 (Matus-Cadiz et al., 2004) to 1640
m2 (Hanson et al., 2005). Matus-Cadiz et al. (2004) sug-
gested that further research at the commercial fi eld scale
was necessary, because small pollen sources may underes-
timate gene fl ow in wheat.
Realistic measurements of gene fl ow require adequate
sampling over space and time to reveal the range of variation
that may occur (Rieger et al., 2002). The release in 2001
of a non-transgenic herbicide-resistant winter wheat culti-
var in Colorado presented a unique opportunity to moni-
tor gene fl ow at the commercial fi eld scale. This cultivar,
Above (Haley et al., 2003), has single-gene imidazolinone
herbicide resistance due to a mutant allele at the acetolac-
tate synthase (ALS) locus on the D genome (Anderson et
al., 2004). Above and several subsequently released cultivars
were developed through induced mutagenesis followed by
conventional plant breeding (Newhouse et al., 1992; Tan
et al., 2005). Above was fi rst planted by commercial grain
producers in fall of 2002 and the fi rst summer during which
PMGF could have occurred on a large scale was 2003. In
this system, gene fl ow is identifi ed by the presence of het-
erozygous imidazolinone-resistant (IR) plants in the prog-
eny of imidazolinone-susceptible (IS) plants.
Ensuring coexistence of transgenic and non-trans-
genic wheat, including appropriate biotechnology regula-
tions for transgenic wheat fi eld trials and appropriate seed
production standards, will require knowledge of commer-
cial fi eld-scale gene fl ow variation. A comparison between
gene fl ow in commercial fi elds and in experiments with
small pollen sources would also aid in evaluating previous
wheat gene fl ow studies. This study utilized the novel IR
trait to (i) estimate gene fl ow in wheat between commer-
cial fi elds across years and locations and (ii) compare gene
fl ow estimates from commercial fi eld sampling to those
obtained from small experimental plots.
MATERIALS AND METHODS
Commercial Fields
Sample Collection
Field sites were located in eastern Colorado where IR winter
wheat cultivars Above or Bond CL (Haley et al., 2006) were
grown bordering IS winter wheat during 2003, 2004, and
2005. Commercial fi eld sites were either a single IS cultivar
planted adjacent to an IR cultivar or cultivar strip trials, where
multiple IS cultivars were planted in strips parallel to IR wheat.
Dates of heading, when approximately 50% of heads were vis-
ible above the fl ag leaf, were recorded for both pollen source
and recipient cultivars. Heading typically occurs two to three
days before anthesis.
Commercial fi eld samples were collected by hand harvest-
ing single rows of the IS cultivar along a transect perpendicular
to the fi eld border at distances from 0.23 to 61 m from the IR
border. Each sample was a composite of at least 20 subsamples of
all wheat heads in a 1 m length of row, spaced at least 3 m apart,
in a single row at each distance from the IR pollen source. For
the strip trials, a single cultivar was represented by one to three
composite samples at diff erent distances from the IR cultivar.
Samples were grouped by distance and threshed individually
with a stationary thresher, starting with samples from the far-
thest distances and proceeding to the closest distances to mini-
mize potential contamination between samples.
A total of 455 samples were collected from 56 locations in
eastern Colorado (Table 1). The samples included 18 commer-
cial wheat cultivars, representing six relative heading classes
(RHC) for eastern Colorado (Table 2). Relative heading
classes range from 1 to 8, with 1 being earliest and adjacent
classes diff ering by approximately 1.5 d (S. Haley, personal
communication, 2006). At all locations IS wheat was sampled
next to Above, with the exception of one location where IS
wheat was sampled next to Bond CL. Above is classifi ed as an

Page 3

Reproduced from Crop Science. Published by Crop Science Society of America. All copyrights reserved.
2310
WWW.CROPS.ORG
CROP SCIENCE, VOL. 47, NOVEMBER–DECEMBER 2007
none herbicide) at 44 g ha–1, 0.25% (v/v) non-ionic surfactant
(Activator 90, Loveland Industries Inc., Greeley, CO), and 2.5%
(v/v) urea ammonium nitrate applied in a spray volume of 187
L ha–1 at 206 kPa. A second imazamox application of 35 g ha–1
was applied in early May to ensure adequate diff erentiation
among homozygous resistant, homozygous susceptible, and
heterozygous phenotypes. In this system, most plants derived
from IS cultivar seeds are expected to be killed by the herbicide
treatment, and only plants derived from cross-pollinated seeds
are expected to survive.
The number of surviving plants was determined in late
May. A heterozygous phenotype, indicative of gene fl ow, was
identifi ed by mild foliar chlorosis, stunting, increased tiller-
ing, and twisted spikes. This phenotype was consistent with
greenhouse observations of imidazolinone-treated heterozy-
gous plants (Pozniak and Hucl, 2004). We estimated the total
number of emerged plants in each plot by counting subsamples.
The subsample plant population was used to estimate the aver-
age number of plants per meter of row, and that number was
multiplied by the number of meters of row in a plot to estimate
the total number of emerged plants.
Survivor Verifi cation
Resistance to ALS-inhibiting herbicides in plants can occur via
spontaneous mutation (Saari et al., 1994). To verify that plants
scored as heterozygous carried one copy of the ALS mutant
allele from Above and one copy of the wild-type ALS allele
from the susceptible parent, we conducted polymerase chain
reaction (PCR) analysis on a subsample of plants. Proprietary
PCR-based protocols and primers specifi c to the D genome
ALS mutant allele from Above and the wild-type D genome
ALS allele were provided by the BASF Corp. (Research Tri-
angle Park, NC). Reactions were performed on a PTC-100
thermal cycler (Bio-Rad Laboratories, Hercules, CA). Amplifi -
cation products were separated on 2% agarose gels, stained with
ethidium bromide, and visualized under ultraviolet light.
Data Analysis
Percent PMGF was calculated as (number of verifi ed survi-
vors/total number of plants) × 100. Analysis of variance for
PMGF was performed in SAS Proc GLM (SAS, 2004) on veri-
fi ed PMGF results from samples taken within 6 m of the pollen
source. Means separation was conducted using Fisher’s Least
Signifi cant Diff erence (LSD) at α = 0.05.
Experimental Plots
In a separate experiment, a Nelder wheel design (Nelder, 1962) was
used to measure gene fl ow from smaller pollen sources. Plots were
planted in fall 2003 and 2004 and sampled the following summer.
Alternating strips of the IS cultivars Prairie Red and Halt were
planted around a central 10-by-10 m block of Above. Dates of 50%
heading for the three cultivars were recorded. Wind speed and
direction data were obtained from a permanent USDA weather
station 400 m from the site. Samples were collected by hand har-
vesting all wheat heads within a 1 m2 area at 1, 3, 7, 11, 15, 23,
31, and 39 m from the edge of the pollen source along eight tran-
sects in north, northeast, east, southeast, south, southwest, west,
and northwest directions. In 2004, two additional transects were
RHC 3 cultivar, and Bond CL is considered to be in RHC
5. All locations had some degree of heading date overlap
between IS and IR cultivars suffi cient to provide fl owering
synchrony. Sampling sites included 25 cultivar strip trials (177
samples) and 36 large commercial fi elds (278 samples). Some
locations included both commercial fi eld transects and culti-
var strip trials. Commercial fi eld locations ranged in size from
32 to 130 ha for pollen source and recipient fi elds, and border
length varied from 200 to 800 m. Cultivar trial pollen source
and recipient plots ranged from 6- to 24-m wide with a 200-
to 400-m long border.
Field Screening Method
To detect cross-pollination, samples were screened for herbi-
cide resistance using a fi eld-based method. Field plots consisted
of 17-m long plots with six rows (2003) or four rows (2004 and
2005). The design had four replicates and 15,000 total seeds
planted per sample (2003) or three replicates and 11,250 total
seeds planted per sample (2004 and 2005). Seed numbers were
estimated based on 200-kernel mass for each sample. Known
homozygous IR plants (Above) were included as checks in sep-
arate plots. Plots were planted in October and treated in April
at the three to fi ve leaf stage with imazamox (an imidazoli-
Table 1. Number of commercial wheat fi eld sample locations
and total plants screened.
YearLocationsSamplesPlants screened
× 106
1.06
1.41
1.54
4.01
2003
2004
2005
Total
17
17
22
56
123
167
165
455
Table 2. Wheat cultivars sampled in commercial fi elds.
Relative heading
class†
1
1
2
3
5
5
5
5
5
5
5
5
5
6
6
6
6
8
Cultivar
Number of
locations
11
1
8
2
11
3
21
16
8
2
13
2
2
1
1
1
19
3
Number of
samples
72
7
57
7
45
16
50
29
18
14
41
9
4
8
1
5
52
20
Prairie Red
TAM 107
Jagger
Halt
Akron
Alliance
Ankor
Avalanche
Enhancer
Hatcher
Jagalene
Yuma
Yumar
Ike
Millennium
Platte
Trego
Prowers 99
†Relative rating for eastern Colorado on a scale of 1 to 9 with 1 being earliest (S.
Haley, personal communication, 2006).

Page 4

Reproduced from Crop Science. Published by Crop Science Society of America. All copyrights reserved.
CROP SCIENCE, VOL. 47, NOVEMBER–DECEMBER 2007
WWW.CROPS.ORG 2311
sampled on a north by northwest and a west by northwest orienta-
tion. Farther distances were sampled where the size of the recipient
wheat plot permitted, up to 70 m from the pollen source. These
samples were processed separately from commercial fi eld samples
and evaluated using a greenhouse screening method.
Greenhouse Screening Method
Samples from the 2004 Nelder wheel were screened by plant-
ing 12 rows of 15 seeds each in rectangular fl ats with commer-
cial potting media (Sunshine Mix #3, SunGro Horticulture,
Bellevue, WA). Each sample was planted in two fl ats in each
of two replications. Flats were placed in a greenhouse under
natural light supplemented with 400 W sodium halide lamps
to provide a 14 h daylength and watered as required. Plants
were sprayed at the two- to three-leaf stage with imazamox
at 35 g ha–1, 0.25% (v/v) non-ionic surfactant, and 1.0% (v/v)
urea ammonium nitrate in a research track sprayer (DeVries
Manufacturing Corp., Hollandale, MN) calibrated to apply
187 L ha–1 at 206 kPa. Two days after treatment, plants were
clipped to approximately 1 cm above the newest emerging leaf.
Plants that re-grew and displayed an injured phenotype were
identifi ed as heterozygous survivors. The number of survivors
divided by the number of emerged plants in each sample was
used to calculate PMGF.
Samples from the 2005 Nelder wheel experiment were
screened with a more effi cient method of soaking seeds in ima-
zamox, described by Gaines et al. (2007). A foliar application
of imazamox as described for the 2004 samples was applied 10
to 14 d after emergence to eliminate any susceptible plants that
escaped imazamox treatment during seed soaking. The number
of survivors divided by the expected number of emerged plants
in each sample was used to calculate PMGF. The methods used
in 2004 and 2005 were compared by correlating PMGF esti-
mates from a set of commercial fi eld samples taken in 2003.
Empirical Model
As a starting equation for this empirical modeling approach, we
used the following “General Wheat Model” (GWM) proposed
by Gustafson et al. (2005):
(0.2
−
)
10
x
PMGF
The GWM can be rewritten as
−
= ×10
Ya

where Y is PMGF, a, b, and c are model parameters describing
the height, steepness, and curvature of the line, respectively,
and x is the distance in m between pollen source and recipient
plants. Gustafson et al. (2005) fi xed a at 1, b at 0.2, and c at 0.5
based on graphing data sets from published studies and adjust-
ing model parameters to obtain a line that exceeded 95% of
available data and had conservative (high-end) predictions of
gene fl ow at all distances. Because heading date in the receiving
fi eld relative to the pollen source likely infl uences receptivity to
gene fl ow (Waines and Hegde, 2003), an additional term was
added to the model
f
c
bxdz
Ya

where d is a model parameter, z is RHC, and f is the exponent
in a power transformation of z. Based on preliminary fi tting,
=

[1]
c
bx
[2]
−+
= ×10[3]
both c and f were fi xed. By fi xing these exponential values,
the model becomes a linear combination of parameters, thereby
substantially reducing the complexity of calculating standard
errors for the value of interest (PMGF). A generalized linear
mixed model (GLMM) was estimated using the maximum
likelihood method in SAS Proc Nlmixed (SAS, 2004). The
GLMM was fi t separately to commercial fi eld and experimental
plot data. Model fi t was judged using the Aikake information
criterion (AIC) (Burnham and Anderson, 2002).
Equation [3] was modifi ed for additional analysis in Proc
Nlmixed. For a given location, Hi (number of heterozygous plants)
is binomially distributed with ni trials and a probability given by
f
c
bxdzl
a

where li is a random, normal location eff ect with mean zero and
standard deviation σ. Three lines were computed based on the
model parameter estimates. The fi rst line is a model of the median
response, which represents a typical location where l = 0,
f
c
bxdz
Ha

The second line represents an estimate of the response for a
location at the upper 95th percentile of locations, 10 ˆ χ, where
ˆ χ = log10(ˆa) – ˆbxc + ˆdzf + 1.645ˆσ
and where 1.645 is the upper 95th percentile of a standard nor-
mal distribution.
The third line represents an upper 95% confi dence limit on
10χ and requires a standard error of ˆ χ calculated using estab-
lished methods for variances of linear combinations of param-
eter estimates. An interpretation of this line is that we have 95%
confi dence that 95% of locations in a given RHC have gene
fl ow less than this bound at a given distance.
−++
×10
i
[4]
−+
= ×
ˆ 10
ˆˆ
[5]
[6]
RESULTS
Commercial Fields
Resistance to an imidazolinone herbicide proved to be a
reliable and easily implemented trait for estimating gene
fl ow. To verify our phenotypic scoring, a total of 92 sus-
pected heterozygous plants were sampled for PCR analysis
over the three years of the study and all were confi rmed as
having both mutant and wild-type alleles at the D genome
ALS locus. In 2004 and 2005, we tested two single plants
that were considered ambiguous as to whether they were
homozygous or heterozygous for imidazolinone resistance.
Both plants were scored homozygous mutant by PCR and
were not included in the PMGF calculations.
The highest observed gene fl ow in a sample was 5.3%
in ‘Jagger’ at 0.23 m, while no gene fl ow was detectable in
125 (27%) of the samples at distances ranging from 0.3 to
61 m from the pollen source. Gene fl ow was detected in at
least one sample from every location and from each of 18
cultivars tested. Across all three years, 33% of samples were
taken within 6 m of the pollen source and 67% of total
observed gene fl ow occurred within that distance; 77%
of samples were taken within 30 m and 92% of observed
gene fl ow occurred within that distance. The farthest

Page 5

Reproduced from Crop Science. Published by Crop Science Society of America. All copyrights reserved.
2312
WWW.CROPS.ORG
CROP SCIENCE, VOL. 47, NOVEMBER–DECEMBER 2007
distance at which gene fl ow was detected was 61 m, our
furthest sampling point, and the maximum outcrossing at
that distance was 0.25% in a Prairie Red sample.
Analysis of variance for gene fl ow in samples taken
within 6 m of the pollen source indicated that cultivar,
year, and a cultivar-by-year interaction were all signifi -
cant (P < 0.01, data not shown). Cultivars with
mean PMGF > 1% in at least one year were Jagger,
Prairie Red, and TAM 107 (Table 3), all of which
are early heading cultivars (RHC 1 or 2, Table 2).
Ankor (RHC 5) also had relatively high cross-pol-
lination rates in two of the three years.
Experimental Plots
A total of 191 samples was analyzed from the
Akron Nelder wheel sites in 2004 and 2005 (Table
4). The Pearson correlation coeffi cient between
results of the 2004 and 2005 screening methods
on a common set of commercial fi eld samples from
2003 was 0.90 (n = 28, P < 0.0001), indicating that
the two methods gave comparable results. In 2004
an average of 675 plants was screened per sample.
Ninety-one of the 98 samples had no detectable
gene fl ow. In 2005 an average of 5700 plants was
screened per sample. Seventy-six of the 93 samples
had no detectable gene fl ow. Maximum gene fl ow
observations were greater in 2004 than in 2005,
while mean and maximum gene fl ow observations
were higher for Prairie Red than for Halt (Table
4). The farthest PMGF was observed in 2005 in a
Prairie Red sample (0.02%) at 31 m from the pol-
len source (Table 4). A subsample of 13 putative
heterozygous IR plants from 2004 and 2005 was
verifi ed heterozygous by PCR.
Empirical Model
Model Parameters
The Nlmixed procedure was used to fi t values of c in Eq.
[5]. The value of 0.53 was determined to provide the best
fi t based on AIC. We fi xed c at 0.5, that is, a square root
transformation of distance, because
it was very close to our fi tted value
and is supported by the literature
(Gustafson et al., 2005). For relative
maturity, the response of the param-
eter d fi t for each value of z (RHC)
was not linear, contrary to our ini-
tial expectations. Instead, fi tted val-
ues of d appeared to be much higher
for low values of z (RHC 1 and 2)
and similar for values of z from 3 to
8. Testing various exponential trans-
formations of z using Proc Nlmixed
to fi t values of f in Eq. [5] indicated
that f = −1.4 provided the best fi t
based on AIC. According to USDA
weather station data, no single pre-
vailing wind direction occurred
during morning hours of the head-
ing and anthesis periods from 2003,
Table 3. Pollen-mediated gene fl ow in wheat samples collected at dis-
tances ≤ 6 m from the pollen source, by cultivar and year.
Year
2004
Number of
samples
2003
Number of
samples
2005
Number of
samples
CultivarMeanMeanMean
––– % –––
0.04 b†
0.14 b
0.42 b
0.15 b
0.12 b
––– % –––
0.12 bc
0.24 abc
0.57 a
––– % –––
0.10 b
Akron
Alliance
Ankor
Avalanche
Enhancer
Halt
Hatcher
Ike
Jagalene
Jagger
Platte
Prairie Red
Prowers 99
TAM 107
Trego
Yuma
Yumar
Mean
8
1
4
2
5
7
4
6
3
0.07 b 6
0.02 b
0.05 b
2
6
0.13 b4
0.03 c
0.21 abc
4
8
0.30 b
0.28 b
0.40 b
0.56 b
0.02 b
1.66 a
0.07 b
8
9
1
12
9
3
6
2.47 a5
1.14 b20.46 ab9
0.09 b
0.03 b
0.09 b
0.46
1.32
4
4
2
0.18 abc3
0.07 bc
0.27
0.41
2
0.29
0.61
LSD (α = 0.05)
†Means with the same letter within a column are not signifi cantly different at α = 0.05.
Table 4. Pollen-mediated gene fl ow (PMGF) results from Nelder wheel experimental
plots (EP) and commercial fi elds (CF) for cultivars Halt, Prairie Red, and TAM 107.
Samples
With
PMGF
6
13
7
24
36
7
1
4
0
1
PMGF
Cultivar
Site
type
EP
Year Analyzed
Distance
range (m)Mean†Maximum†Maximum
1–42 0.04
1–690.01
0.15–370.26
0.3–460.24
0.3–610.33
0.3–610.85
1–55 0.003
1–70 0.004
14 0.0
3–61 0.01
distance‡
11 (0.15)
31 (0.02)
37 (0.01)
46 (0.10)
61 (0.25)
61 (0.10)
15 (0.15)
7 (0.02)
–
3 (0.03)
Mean
1–3 m§
0.11 (9)
0.04 (10)
1.14 (2)
0.55 (6)
0.63 (7)
2.15 (2)
0.00 (10)
0.01 (10)
–
0.03 (1)
Prairie Red 2004
2005
2003
2004
2005
2005
2004
2005
2004
2005
42
50
10
26
36
7
56
43
1
6
0.74
0.09
1.15
1.06
1.08
3.71
0.15
0.06
–
0.03
CF
TAM 107¶
Halt
CF
EP
CF
†Percent PMGF, defi ned as (number of resistant survivors/total plants screened) × 100.
‡Maximum distance in m at which PMGF was observed, with percent PMGF from the sample in parentheses.
§Percent PMGF in samples collected 1 to 3 m from the pollen source, with the number of samples in parentheses.
¶Prairie Red was derived via backcrossing with TAM 107 as a recurrent parent and is expected to be > 95% geneti-
cally similar to TAM 107.