Influence of intergenotypic competition on multigenerational persistence of abiotic stress resistance transgenes in populations of Arabidopsis thaliana

Abstract Reducing crop losses due to abiotic stresses is a major target of agricultural biotechnology that will increase with climate change and global population growth. Concerns, however, have been raised about potential ecological impacts if transgenes become established in wild populations and cause increased competitiveness of weedy or invasive species. Potential risks will be a function of transgene movement, population sizes, and fitness effects on the recipient population. While key components influencing gene flow have been extensively investigated, there have been few studies on factors subsequent to transgene movement that can influence persistence and competitiveness. Here, we performed multiyear, multigenerational, assessment to examine fitness effects and persistence of three mechanistically different abiotic stress tolerance genes: C‐repeat binding factor 3/drought responsive element binding factor 1a (CBF3/DREB1a); Salt overly sensitive 1 (SOS1); and Mannose‐6‐phosphate reductase (M6PR). Transgenic Arabidopsis thaliana overexpressing these genes were grown in pure populations and in competition with wild‐type (WT) parents for six generations spanning a range of field environment conditions. Growth, development, biomass, seed production, and transgene frequency were measured at each generation. Seed planted for each generation was obtained from the previous generation as would occur during establishment of a new genotype in the environment. The three transgenes exhibited different fitness effects and followed different establishment trajectories. In comparison with pure populations, CBF3 lines exhibited reduced dry weight, seed yield, and viable seed yield, relative to WT background. In contrast, overexpression of SOS1 and M6PR did not significantly impact productivity measures in pure populations. In competition with WT, negative fitness effects were magnified. Transgene frequencies were significantly reduced for CBF3 and SOS1 while frequencies of M6PR appeared to be subject to genetic drift. These studies demonstrate the importance of fitness effects and intergenotype competition in influencing persistence of transgenes conferring complex traits.


| INTRODUC TI ON
A major target of crop improvement through agricultural biotechnology is reduction in losses due to abiotic stresses (Liang, Prins, van de Wiel, & Kok, 2014;Mittler & Blumwald, 2010). Concerns, however, have been raised about potential ecological impacts if introduced stress tolerance genes become established in wild populations and cause increased competitiveness of weedy or invasive species (e.g., Ding et al., 2014;Hails & Morley, 2005;Lu & Snow, 2005;Nickson, 2008). While presence of transgenes per se does not pose environmental risk, potential for ecological harm could occur if the transgene causes release from ecological constraints leading to population increase or expansion into new geographic regions (Hails & Morley, 2005;Lu & Snow, 2005;Lu, Yang, & Ellstrand, 2016;Warwick, Beckie, & Hall, 2009).
A prerequisite for ecological impacts via population increase or range expansion is establishment of self-sustaining populations containing the transgene (Guitierrez, Cantamutto, & Poverene, 2011;Hooftman et al., 2011;Kost, Alexander, Emry, & Mercer, 2015;Mercer et al., 2014). Transgenic plants could become feral and form a self-sustaining population outside of agricultural plantings, perhaps first along field margins, and later by expansion into natural areas. For example, persistent populations of herbicide-resistant canola resulting from roadside seed spillage have been observed in Australia and Canada, although persistence was limited when plants moved into natural environments (Beckie & Warwick, 2010;Busi & Powles, 2016). Alternatively, the transgene could enter a wild compatible relative through hybridization and introgression.
Evidence of naturally occurring crop-wild hybrids has long been documented for several crops such as such as lettuce, sunflower, radish, oilseed rape and rice; however, concerns about transgenes have led to heightened awareness about the potential for gene transfer from crops to wild populations (Campbell, Snow, Sweeny, & Ketner, 2009;Ellstrand et al., 2013;Hooftman et al., 2011;Lu et al., 2016).
Several experimental systems have studied performance of hybrids between transgenic crops and their weedy relatives [e.g., oilseed rape (Halfhill et al., 2005;Rose et al., 2009), rice ]. In addition, unintended transfer of transgenes has been documented in distant populations of bentgrass in western USA (Zapiola, Campbell, Butler, & Mallory-Smith, 2008;Zapiola & Mallory-Smith, 2017), and introgression of herbicide resistance was observed in wild Brassica rapa in natural environments over several years, although gene frequency decreased rapidly over time (Warwick, Legere, Simard, & James, 2008).
The key variables influencing gene movement via hybridization, such as mode of pollination, occurrence of compatible relatives, cropping system, population size, and spatial and temporal proximity, have been extensively studied, and several models have been developed to predict extent of gene flow from transgenic crops into recipient populations (e.g., Baker & Preston, 2003;Colbach, Clermont-Dauphin, & Meynard, 2001;Weekes et al., 2005). However, as emphasized by Ellstrand and Rieseberg (2016), there have been very few multigenerational field experiments tracking transgene persistence. While the extent of pollen-mediated gene flow is generally independent of the specific transgene introduced, potential gene establishment, and ultimately ecological impact such as weediness or invasiveness, will depend on biology of the crop and recipient wild relatives, abiotic and biotic characteristics of the recipient environment, and importantly, fitness effects of traits conferred by the transgene and surrounding genomic regions (Grumet, Wolfenbarger, & Ferenczi, 2011;Hails & Morley, 2005;Hartman et al., 2014;Keese, Robold, Myers, Weisman, & Smith, 2014; Thompson et al., 2003;Warwick et al., 2009;Xia et al., 2016).
The first wave of commercialized transgenic crops, dominated by Bt-mediated insect resistance and herbicide tolerance, generally had minimal effects on fitness, except under the selective conditions (insect pressure, herbicide application) for which the transgenic crops were developed (e.g., Beckie et al., 2006;Snow et al., 2003;Warwick et al., 2009). These traits are conferred by genes whose protein product is directly responsible for the desired trait: Bt proteins are toxic to the target insect; herbicide resistance genes encode proteins that prevent binding of the herbicide or otherwise inactivate the herbicide. Thus, these genes and their gene products are largely inert with respect to other cellular functions as evidenced by results of global transcriptome, proteome, and metabolome studies revealing that transgene-induced differences were smaller than the differences that currently exist among conventionally bred varieties (e.g., Cheng et al., 2008;Coll et al., 2008;Ruebelt et al., 2006). These physiologically simple traits may be contrasted with engineered resistances to abiotic stresses which are likely to have complex effects on gene expression, physiology, and growth responses, and may provide variable selective advantages or disadvantages, depending on the environment.
Indeed, studies of clinal gradients and reciprocal transplants have found associations between physiological characteristics that influence response to the environment with adaptation to local environmental conditions (e.g., Baxter et al., 2010;Debieu et al., 2013;Wolfe & Tonsor, 2014), and experiments combining common garden and genomic analyses have revealed potential for a small number of quantitative trait loci or effective alleles to modulate local adaptations (e.g., Agren, Oakley, McKay, Lovell, & Schemske, 2013;Hancock et al., 2011). These observations suggest that a small number of genetic changes can influence fitness, depending on the environment.
Research over the past two decades has identified a wide variety of genes with potential for enhancing abiotic stress tolerance such as chaperone proteins, membrane stabilization proteins, metabolic and detoxification enzymes, stress signaling pathway genes, and transcriptional activators (Bhatnagar-Mathur, Vadez, & Sharma, 2008;. Given the variety of underlying mechanisms associated with such genes, it is anticipated that not all stress tolerance genes would have equivalent effects on fitness and competitiveness of recipient plants. In addition to the selective advantages that may result from the primary, intended effect of the transgene, it is also possible that variable pleiotropic effects resulting from transgene-induced physiological modifications may also influence fitness, either positively or negatively (Little, Grumet, & Hancock, 2009;Warwick et al., 2009). In addition, extensive cross talk among stress responses, including both abiotic and biotic stresses, has been widely documented (Krasensky & Jonak, 2012;Mittler & Blumwald, 2010).
Mannitol can act as a compatible solute and osmoprotectant, counterbalancing the osmotic and toxic effects of sodium and chloride ions, and as reactive oxygen quencher, reducing damage caused by free radicals produced under salinity stress or in response to pathogen attack Sickler, Edwards, Kiirats, Gao, & Loescher, 2007;Zhifang & Loescher, 2003). The SOS1 gene encodes a plasma membrane Na+/H+ antiporter that shuttles toxic sodium ions away from the cytoplasm, influencing intracellular ionic balance (Shi et al., 2003).
Results from the growth chamber-grown plants indicate potential for complex fitness effects that could ultimately influence transgene establishment (Chan et al., , 2012. Several studies have indicated the importance of evaluating fitness based on performance throughout the life cycle, in a range of environments, and under competitive conditions (Agren et al., 2013;Bhatnagar-Mathur et al., 2008;Hartman et al., 2014;Mercer et al., 2014;Mittler & Blumwald, 2010). Therefore, in these experiments, we sought to examine fitness impacts of overexpression of the three mechanis-
Each of the abiotic stress resistance transgenes was expressed constitutively using the Cauliflower mosaic virus (CaMV) 35S promoter.
The lines provided were previously demonstrated to confer salt stress resistance in growth chamber studies (Gilmour et al., 2000;Shi et al., 2003;Zhifang & Loescher, 2003). With the exception of CBF3 A30, all also conferred resistance to 100 mM salt stress in our growth chamber experiments relative to WT parents as measured by reduced salt injury, greater dry weight, and increased seed yield (Chan et al., 2012). Each line also contains the neomycin phosphotransferase II (NPTII) gene for kanamycin resistance as a selectable marker under control of the nopaline synthase (NOS) promoter. The WT parents [Col, Col(gl) and WS] are all highly homozygous, rapidcycling A. thaliana ecotypes. For each transgenic line, the initial seed provided was generation T 2 or T 3 , derived by self-pollination from the original molecularly verified transgenic plant. Plants grown from the seed provided were verified to contain and express the specified transgene by PCR, Southern and Northern blot analysis ( Figure   S1) and verified to be homozygous by kanamycin screening of progeny seed. Verified plants were used for seed amplification by selfpollination in the glasshouse to provide sufficient starting material for the first generation of the field experiments.

| Field experiment design
The experimental design for the multigenerational field study was adapted from a glasshouse experiment examining long-term fitness effects of Arabidopsis mutations under competitive conditions (Roux, Camilleri, Bérard, & Reboud, 2005). Planting density (2600 seeds/m 2 ) [yielding ~180 (±20%) seedlings per 26 × 26 cm tray] was selected to ensure a naturally high level of interplant competition for each population as was previously utilized by Roux et al. (2005).
Fourteen replicate intergenotypic competitive populations (one tray = one population = ~180 plants) were established for each transgenic line and background WT mix with a starting allelic frequency of 50% transgenic and 50% WT. This level of replication was chosen to ensure that fitness impacts from transgene expression ≥5% would be detectable after three generations by exceeding the 95% confidence intervals attributable to genetic drift based on theoretical distributions using the following formula, V qt = q 0 *p 0 *(1-(1-1/(2N e )) t ) (Falconer & Mackay, 1996). The initial frequencies of transgenic and wild-type plants at the t generation are q 0 and p 0 , respectively, and N e is the effective population size. Due to the highly selfing nature of Arabidopsis thaliana, 98%-99% (Snape & Lawrence, 1971), the effective population size was calculated based on the following formula, (Caballero, 1994) where β is the proportion of selfing (0.98) and N is the observed population size (180) with regard to contained movement of materials between laboratory, glasshouse, and field; autoclaving and/or disposal of containers; location and security of the field site; containment of plants and seeds during the field season, including isolation distance from native Arabidopsis populations and maintenance of an Arabidopsisfree zone surrounding the test plots; removal of plants to prevent dissemination of seeds; harvesting and storage of seeds to assure containment; treatment of residual plant material after removal of flower stalks; treatment of the field location with herbicide; and monitoring of the field location for the volunteers in the following year. The populations were arranged in a computer generated, completely randomized design. A tray-in-flat potting method was developed to allowing for subsoil irrigation via trickle hose and provide secure anchoring to the ground ( Figure S2a). Plants were grown in the field until approximately 75% of siliques had begun to senesce.
The trays were then returned to the glasshouse for final maturation and dry down prior to harvest. This procedure minimized seed loss prior to harvest, assuring seed confinement and increasing accuracy of seed yield measurements.
At harvest, all aboveground plant material was harvested from the tray, transferred to paper bags, and allowed to dry to ambient humidity levels. The dried plant materials were cleaned of seed. Seed derived from each population (pure line or mixed) was maintained separately. Seed viability was tested by percent germination following 48-to 60-hr stratification at 4°C (three batches of 100 seeds were tested for each population at each generation). Calibration counts (relative to the custom plate hole size) were performed for each population at each generation to compensate for possible differences in seed size due to genotypic and/or environmental effects.
The number of seed planted for each population in each generation was adjusted as needed for seed size and viability to provide approximately 180 seedlings per population to initiate the subsequent generation.
Climatic data were recorded at the HTRC weather station, located less than 400 m from the field site. The weather station recorded air temperature, precipitation, solar radiation, potential evapotranspiration, and wind speed across all six field growing seasons. Information about mean daily maximum and minimum temperatures and mean daily total solar flux for each of the six seasons is provided in Figure S3.

| Developmental and productivity data
Developmental data were collected on all pure populations on a per tray basis as days postsowing when ~75% of the population reached the following developmental stages: germination; two true leaves emerged; rosette formation (5-6 true leaves); bolting; flowering; siliques mature. Time of first bolting and time of first flowering also were recorded. At harvest, all aboveground plant material from each replicate population at each generation, from both pure and mixed populations, was removed and allowed to dry to ambient humidity levels. The dried plant materials were cleaned of seed, and total plant dry weight and seed mass were recorded. Seed viability (percent germination) and seed size were tested as described above for

| Genotyping field grown progeny via kanamycin screening
Transgene frequency was monitored in each replicate population of the mixed populations at each generation by kanamycin screening of progeny seed to determine relative gene frequency of the NPTII selectable marker gene. Sterilized seed were plated via 20μl pipette, onto ½ MS media +1% agar containing 100 mg/L kanamycin (with the exception of lines A40 and S1-1 which were plated onto media containing 75 mg/L kanamycin due to a lower level of resistance than the other transgenic lines). Kanamycin screenings for the mixed populations were performed in triplicate for each population and generation with 100 scored seedlings/replicate plate. Each plate also included an aliquot of seed from the respective WT as negative controls to verify kanamycin effectiveness. WT seedlings showed complete bleaching and did not proceed past the cotyledon stage, while transgenic seedlings grew normally; resistant and susceptible individuals were also easily distinguished from seeds that failed to germinate ( Figure S2b).
The overall mean germination rate for all populations was 92.62 ± 0.33%, indicating minimal impact of the sterilization procedure or the presence of kanamycin in the media on seed viability and germination. Median germination percentages of pure populations ranged from 89% to 97% ( Figure S4). Continued presence and expression of the transgenes in pure transgenic populations, and absence in WT lines, also were verified by kanamycin resistance screening after each of the six generations.

| qPCR (quantitative polymerase chain reaction) verification of transgene frequency
Progeny seed was sterilized as described above and plated via pipette on to ½ MS media +1% agar, and grown to the two-leaf stage.
A set of standards was created by producing mixes of confirmed transgenic or WT seedlings to result in batches of 100 seedlings with 0% transgenic, 10%, 20%, 50%, and 100% transgenic plants.  Seasonal differences in environmental factors such as temperature, day length, light intensity, rainfall, and humidity reflect the range of environmental conditions experienced by natural populations ( Figure S3). Significant variation in plant development and productivity was observed across the six growing seasons as measured in the pure populations (Tables 1 and S2) Table S2). Seasonal effects did not uniformly affect all populations, indicating significant genotype by environment (GxE) interactions ( Figure 1, Table 1).   Figure S4).

| Performance of pure populations
The CBF3 lines exhibited reduced dry weight, average seed yield, and viable seed yield, relative to the WT WS background (Table 2). lines and their respective WT parents (p < .001; Table S3B,C), there were not significant transgene or season X transgene interactions for seed yield.
The reduced viable seed yield of CBF3 lines resulted in reduced fitness relative to WT WS (Table 2). Although the relative fitness of the M6PR lines averaged over the six seasons was 1.33 and 1.23 (Table 2), there was variation among the seasons. For example, while yield was greatly reduced for both parents and M6PR lines in the highly unfavorable seasons 2 and 5, the M6PR lines were less badly affected than the WT parent in the cool season 2, producing approximately twice as much seed. However, in the very wet season 5 the situation was reversed. Consistent with these observations, analysis of variance indicated a significant effect of season on fitness values for M6PR (Table S4C; p < .05). There was not significant variation among seasons for fitness effects of CBF3 or SOS1 (Table S4A,B).

| Performance in competition with WT
Each of the transgenic lines was also grown in competition with their  *,**,Value is significantly different from WT within transgene group, paired t test (by season) (df = 5), p < .05, 0.01, respectively. c Each value is the mean of fitness estimates calculated for each of the six seasons. Each fitness estimate within a season was calculated as the mean viable seed yield of five replicate transgenic populations/mean seed yield of five replicate WT populations for that season. d Mean fitness is significantly different from WT (t test, n = 6, H 0 : relative fitness =1). frequency relative to expected frequency of 50%, if fitness were equal) that can be compared to that obtained from fitness estimates of pure populations. Comparison of fitness estimates from mixed populations at the end of generation 6 to the mean of fitness estimates obtained from pure populations over the six generations, showed exacerbation of negative fitness effects when in competition with WT parents (Figure 4). Although CBF3 lines showed decreased fitness in pure stands relative to WT WS, fitness in competition with WT was even more reduced. Fitness of SOS1 line 1-1 was comparable to WT based on pure populations, but when in competition with WT, it was significantly lower. M6PR estimates were not significantly different between pure line and mixed population estimates.

| D ISCUSS I ON
Mounting challenges due to climate change and global population growth will necessitate development of crops with increased resistance to abiotic stresses. While transgenic approaches may contribute to more resilient crops, deployment of transgenic crops is met with concerns about potential environmental impacts. Studies showing relationship between variation in key genes conferring stress-related traits and geographic distribution suggest that such genes play a key role in adaptation to adverse environments (e.g., Baxter et al., 2010;Busoms et al., 2015). Thus, introgression of genes enhancing stress tolerance could provide a fitness advantage allowing for increased weediness or invasiveness of recipient populations in cases where the environmental stress is a limiting factor. On the other hand, adaptations favoring success in the face of abiotic and biotic stresses are frequently accompanied by reduced growth rate, leading to trade-offs between competitiveness and survival that could influence potential for transgene establishment in native populations (Anderson, Willis, & Mitchell-Olds, 2011;Claeys & Inze, 2013;Karabourniotis, Liakopoulos, Nikolopoulos, Bresta, & Sumbele, 2014;Trontin, Tisne, Bach, & Loudet, 2011). The objective of this study was to examine the fitness effects and persistence of three mechanistically diverse abiotic stress tolerance transgenes as observed over multiple generations and environmental conditions.

| Potential for pleiotropy and genotype X environment interaction varies with the transgene
Multiple fitness components, such as growth rate, survival, and reproduction, as influenced by variable environmental conditions, ultimately drive individual and population success (Laughlin & Messier, 2015;Voille et al., 2007). While our APHIS field release permits did not allow us to test fitness effects at all stages of development in field conditions (e.g., germination, silique dehiscence, and seed dispersal), during their time in the field over the six seasons, the plants were exposed to a range of conditions that markedly influenced both F I G U R E 3 Transgene frequency within mixed populations. (a, b) wild-type (WT) WS and CBF3 overexpression lines, A40 (a) and A30 (b); (c, d) WT Col(gl) and SOS1 lines, S1-1 (c) and S7-6 (d); and (e, f) WT Col and M6PR lines, M2-1 (e) and M5-1 (f), as determined by selectable marker screening for growth on kanamycin. Each value for each population is the mean of three replicate kanamycin screening plates/ generation with 100 seedlings/plate. All populations began at 50% starting frequency (FG0) and were maintained separately in subsequent generations. Gray lines-transgene frequency in each of the 14 replicate mixed populations; solid black line-mean of the 14 replicate populations. Dashed lines indicate the 95% confidence intervals for populations undergoing solely genetic drift. Reductions in transgene frequency were confirmed by qPCR analysis at generation 6 (i.e., were not an artifact due to gene silencing) (Table S3)  WT when grown in the absence of salt stress (Chan et al., 2012).
In contrast, CBF3 plants exhibited markedly reduced fitness.
The negative fitness effect of CBF3 is likely due to an associated dwarfing phenotype, as has been previously observed in growth chamber and glasshouse studies (Gilmour et al., 2000;Jackson, Stinchcombe, Korves, & Schmitt, 2004), and delayed development as was observed in these field experiments. Inferred negative fitness effects of CBF genes in climates that are not subject to extensive cold stress also have been observed in naturally occurring Arabidopsis populations. Populations from warmer climates exhibited reduced induction of CBF regulon genes in response to cold, and occurrence of nonsynonymous or frame-shift alleles influencing functionality of the CBF genes was observed in association with populations from warmer climates (Gehan et al., 2015;Monroe et al., 2016;Oakley, Agren, Atchison, & Schemske, 2014).
Although the plants experienced a range of growing conditions, the abiotic stresses for which these genes have been previously shown to improve tolerance in the growth chamber were not specifically applied [e.g., salinity stress for SOS1, CBF3, and M6PR; freezing and drought for CBF3 (Gilmour et al., 2000;Kasuga, Liu, Miura, Yamaguchi-Shinozaki, & Shinozaki, 1999;Shi et al., 2003;Zhifang & Loescher, 2003)]. Thus, the observed transgene effects, both positive and negative, indicate pleiotropic effects and GxE interactions.
Given the function of the introduced transgene products (transcription factor, sodium antiporter, compatible solute and possible pathogen signal molecule) and thus their capacity to modify a range of cellular functions, their extensive influence on the transcriptome (Chan et al., 2012), and the known interplay among plant responses to multiple abiotic and biotic stresses (Krasensky & Jonak, 2012;Mittler and Blumwald 2010), it was anticipated that the transgenes would cause pleiotropic effects, but that the extent of pleiotropy could vary with the transgene, as was observed in these experiments. Of course, the fitness effects observed with the CBF3 lines in the higher yielding environments in these experiments were quite extreme; cultivars aimed at increasing crop production would not be anticipated to have strong negative effects on productivity.

| Relative performance is influenced by intergenotypic competition
Processes of transgene introgression and establishment in native populations will require successful competition with plants within the population that do not possess the transgene. In the case of a crop and its interfertile wild relatives, a range of differences may impact fitness of the resulting hybrids, including some, such as domestication traits, that may have negative effects on the progeny (Campbell et al., 2009;Hails & Morley, 2005;Hooftman et al., 2011;Lu et al., 2016). The intent of these experiments was examine the effect of the transgene per se, isolated from other potentially confounding background genetic differences. Therefore, the transgenic lines were planted in direct competition with their respective WT parents. The seed planted for each generation was a subset of the progeny seed produced in the previous generation, allowing the impacts of one season and transgene combination to influence the transgene frequency in the next season. This method follows the natural processes of selection and drift that would occur dur-

_______ _______ _______
The few studies that have compared effects of transgenes in the presence and absence of intergenotypic competition, also have shown differences in fitness estimates from the two methods. For example, insect-resistant (Bt) rice exceeded yield of WT under insect pressure in pure populations, but lost that advantage in competition (Yang, Wang, Su, & Lu, 2012). Transgenic insect-resistant canola produced more seed in pure populations than WT, but did not show increased seed yield at all sites when grown in 1:1 competitive populations with WT canola (Ramachandran, Buntin, All, Raymer, & Stewart, 2000). An inverse effect was observed for transgenic rice expressing a Bt gene for insect resistance (Liu, Ge, Liang, Wu, & Li, 2015). In the absence of competition with WT, the transgenic rice did not produce as much seed as WT, both in the absence and presence of insect pressure; however, in most cases, the disadvantage was reduced or lost when planted in mixed populations with WT.
While the above competition studies were replicated in locations and/or years, to our knowledge, this is the first study that provides longitudinal data following populations over multiple generations.
In addition to direct differences resulting from intergenotypic competition versus performance in pure lines, the estimates obtained from the mixed populations in this study reflect cumulative effects of the fitness impacts in the prior generation. Thus, as would occur during introgression into native populations, the effects of relative fitness differences would be compounded from one generation to the next. Depending on the nature of the genetic differences and selective environment, intraspecific competition can result in additive, antagonistic, or neutral effects (Weis & Hochberg, 2000).
In the case of additive competition, a fitness disadvantage is amplified by increased competition from more rapidly growing plants.
This appears to be the case in the intergenotype populations with CBF3 plants, which were overgrown and inhibited from reaching full development by their WT neighbors, resulting in more rapid decline in representation in the population than was predicted based on the pure populations. However, despite somewhat earlier flowering for the M6PR plants, there did not appear to be effects of additive competition in intergenotypic populations under the conditions tested in these experiments. This is likely due to the relatively modest differences in rate of development. The M6PR plants did not exhibit increased vegetative growth as measured by aboveground dry weight at harvest, and so likely did not cause increased competition for resources such as light, water, and nutrients that can contribute to additive competitive effects. In several cases (CBF3 A30 and A40, Clearly competition experienced by plants within natural settings will not be limited to intraspecific competition. Plants encounter competition from other plant species as well as biotic stresses such as herbivory (Mercer et al., 2014;Weis & Hochberg, 2000). While it is not possible to account for all possible factors, a few studies directly comparing inter-and intraspecific competition have found that intraspecific competition, and especially population density, plays a predominant role in relative fitness.
For example, although Brassica rapa and B. napus both suffered when in competition with Lolium perenne, plant density, regardless of species, had a greater impact on performance, even in the presence of insect herbivore pressure (Damgaard & Kjaer, 2009).
Similarly, conspecific plant density had a greater influence on fitness parameters of sunflower and crop-wild hybrids than did the presence of interspecific competition from other weed species (Mercer et al., 2014).

| CON CLUS ION
A series of multiyear, multigenerational experiments found that three mechanistically different abiotic stress resistance transgenes exhibited different fitness effects in Arabidopsis thaliana, likely reflecting diverse effects on gene expression, plant development, and productivity when exposed to varying environmental conditions. Thus potential impacts following transgene release are likely to be highly dependent on the specific nature of the transgene and its physiological effects, necessitating case- Hammar for help with progeny seed screening. We also thank Drs.
Karen Hokanson and Carolyn Malmstrom for helpful comments on the manuscript.

CO N FLI C T O F I NTE R E S T
None declared.