There is growing evidence that many interspecific hybrids between crops and weeds, including transgenic hybrids, have inferior performance compared with their weedy parent (reviewed in ), although a few exceptions have been reported [13, 38]. There are many possible explanations for these observations, most of which focus on the fitness penalties associated with hybridization and backcrossing. A variety of genetic mechanisms can affect hybrid fitness . Heterosis or hybrid vigor could be enhanced by the segregation of additive genetic traits or optimal environmental conditions. In cases where the crop and weedy or wild relative are the same species, there should be fewer genetic barriers to hybrid fitness and transgene persistence [37, 40, 41]. Unless the trait or transgene is deleterious, as in the case of TM, hybridization and introgression might easily progress, thereby potentially allowing transgene persistence. This is far more likely to occur with related weeds existing and competing in the same agro-ecosystem as the crop than with wild species, residing in non-agricultural areas where a transgenic trait may have little or no value. The performance of new genotypes introduced via transgenesis and subsequent hybridization are likely also subject to genotype × environment (G × E) interactions [3–7, 41–43].
When distantly related taxa hybridize, recombination between homologous or homoeologous chromosomes can result in irregular chromosome pairing. The latter can lead to reduced growth and fertility , thus decreasing potential for introgression. In addition, the dilution of weed alleles by crop alleles, i.e., genetic load, might also account for slower introgression regardless of environment [2, 6]. Our results showed a positive correlation between genetic load (estimated by B. napus-specific AFLP marker numbers) and plant productivity in only one of our transgenic events, GT1, a result congruent with a prior study . No correlation was observed for the other transgenic events suggesting a strong transgenic event or genotype effect. The experimental design allowed the detection of a surprising effect--a counter genetic load finding (i.e. genetic load, conferred by crop alleles, was not detrimental to hybrid performance as originally hypothesized). In our experiments, the number of B. napus crop-specific AFLP markers was associated with an increase in hybrid performance (as opposed to a decrease in performance) in the GT1 hybrids in the absence of interspecific competition with wheat and weeds. Conversely, in the presence of interspecific competition with wheat, there was a strong correlation between B. rapa weed-specific AFLP markers and productivity, again, only for the GT1 line. These results indicate a strong G × E interaction and that inherited crop alleles, introgressed as a result of hybridization, can be preferentially selected to ensure maximum growth potential and productivity under crop-favorable conditions (as seen in the absence of interspecific competition). Conversely, inherited weedy alleles can be preferentially selected under more "weed-like" conditions (i.e. under interspecific competitive conditions) . We anticipate that this will continue to occur as the lines continue to backcross.
This evidence of local and situational adaptation demonstrated in our study as well as other studies indicates that many alleles are beneficial in some but not all environments [40–44]. As sessile organisms, plants must be able to maximize resource acquisition; this can be accomplished with adapted genomes. In agriculture and in our study, environmental genomic adaptation could translate to favoring some crop alleles in crops grown in monocultures where weeds are controlled; weed alleles might be favored in weeds in competitive environments . Dilution in either direction could be maladaptive. Environmental adaptation and trait selection are further demonstrated in a case-study performed on recombinant inbred sunflower lines . Baack et al.  showed that crop traits (earlier flowering time, larger stem diameter, and larger flowering disk diameter) were preferentially favored in one site, while weedy traits (smaller ray and seed size) were favored at another location . These studies indicate that, with regards to genetic load and trait selection, environmentally-dependent selection must be taken into consideration for every transgene, species, and situation studied [40, 42–44].
Transgenic event effects and genetic load
Since transgene insertion is random, the flanking host DNA can differ significantly for individual transgenic events; hence the analysis of multiple events in the present study. Zhu et al.  examined the transgene segregation ratios for each of our lines using controlled crosses and progeny analysis. The GT1 and GT5 lines did not deviate from expected Mendelian segregation, indicating homeologous recombination of the transgene locus on an A-genome chromosome (the A genome is shared by the two parental taxa, B napus AACC- and B. rapa AA genomes), and hence the transgene would have a decreased chance of being lost in subsequent backcrossing as a result of genomic incompatibility. Thus, linkage effects could likely cause decreased introgression and genetic load in the GT1 line. Transgenes located on the C-genome of B. napus might be subject to greater genomic incompatibility since B. rapa does not have the C-genome of B. napus. Thus C-genome-localized transgenic events could be an important investigative avenue for decreasing introgression [2, 27], although recent research found that homeologous recombination occurs at the same rate as homologous recombination in the B. rapa × B. napus system [29, 46], suggesting that "safe-integration" sites in B. napus are unlikely .
Engineered vs. randon transgenic event genetic load
TM B. napus plants containing the dwarfing gene Δgai were previously analyzed under shade-house conditions [31–34]. Homozygous TM dwarf B. napus plants [TM(H)] grown alone (in the absence of competition) had a much higher seed yield (P < 0.01), and double the shoot and root biomass compared to non-transgenic counterparts (P ≤ 0.01). The TM(H) line also produced more leaves than tall non-transgenic plants when grown alone at 2.5-cm spacing (P ≤ 0.05) and 10-cm spacing (P ≤ 0.05). However, when grown in competition with tall non-transgenic cohorts, these same plants were exceedingly unfit . The reproductive fitness of TM(H) plants, when grown in competition, was less than 12%, and the harvest index (grain-to-straw ratio) was less than 40% of that of the non-transgenic B. napus competitors. When grown in the absence of competition, the TM(H) line produced the greatest amount of seeds per plant . From these studies, Al-Ahmad et al.  concluded that the Δgai gene greatly enhanced seed and biomass yield in a weed-free transgenic crop. However, if dwarfed plants emerged as volunteer weeds and competed with non-transgenic cohorts (and presumably other species), dwarfed plants would be eliminated from poor competitive ability, especially if selective herbicides were not used.
Our field results were congruent with the previous TM research [31–35]. When the homozygous TM line [TM(H)] was grown under agronomic non-competitive conditions, it had equivalent performance as the non-transgenic B. napus line with regards to seed yield and biomass. However, under competitive conditions, the TM(H) line produced the least seed and biomass and, hence, performed the poorest of all the lines that were analyzed. In terms of transgene persistence, the transgenic progeny from the backcrossed TM(B) line approached a 50- to 85-fold reduction, under competitive and non-competitive conditions, compared with the 3-fold reduction in transgene persistence with the non-TM (GT) lines. It must be noted that the final portion of the transgene persistence studies; i.e., germination of seed collected from field grown plants were carried out under optimal growth chamber conditions. This environment did not take into consideration the effects of over-wintering survivorship, dormancy issues or seedling dynamics. Consequently, when environmental factors are considered, one could potentially expect a decrease in performance (germination rates, etc) under realistic field conditions. Hence, ex situ analysis of transgene persistence could appear higher than it actually would be in the field . Regardless, these data do indicate that TM constructs were effective in severely limiting the impact of gene flow from transgenic crops to their wild relatives and transgene persistence. The TM approach could further be enhanced by stacking other weed mitigating transgenes, such as those that prevent seedpods from shattering or those that prevent secondary dormancy. In addition to making plants shorter the dwarfing gene could also confer pleiotropic effects such as altered flowering traits. These could conceivably also have G × E interactions and effect fitness . Depending on the trait, the pleiotropic effect could increase or decrease fitness. In this particular instance, dwarfed plants were very late flowering and also had delayed seed maturation and decreased germination. In this case, the pleiotropic effects would almost certainly decrease introgression.
Some researchers point out that dwarfing in TM is a trait that could foreseeably be selected against in the hemizygous state; e.g., in semi-dwarfed plants [45, 46]. Consequently, the dwarfing allele might not be expressed as highly in first-generation hybrids because of the presence of the dominant GAI allele from the weedy parent genome. Therefore, hybrids containing the TM construct might not have sufficient dwarfing to be selected against in competition, yet harbor the fitness-enhancing gene from the TM construct. In subsequent generations, the deleterious allele would only be expressed in homozygous individuals, which would strongly reduce its capability to decrease fitness. Moreover, if the hybrids are fertile (as is the case for Brassica TM hybrids) , this strategy would not prevent them acting as a genetic bridge and pollinating the wild parent .
Reuter et al.  challenged the theory that dwarfed volunteers or transgenic escapees would be less competitive in feral environments. They noted that feral non-transgenic B. napus populations, growing in rural and urban area in northern Germany, were, on average, 41% shorter than cultivated non-transgenic B. napus. They attributed this height difference to phenotypic adaptation to local environments. Reuter et al.  concluded that, under certain environmental and ecological conditions, the proposed mitigation approach (dwarfing) could actually increase escape and persistence of transgenic varieties rather than reducing them, but their study did not consider genotype differences, local adaptation, and environmental effects such as differences in nutrient availability between agronomic and feral conditions. Our transgene persistence data indicate that even if the dwarfing trait is in the semi-dominant state, it is still effective in limiting transgene persistence. Our results indicate that when dwarfing is utilized for a containment strategy, the plants are not able to compete as well and hence persist in subsequent generations. These observations hold true regardless of whether the plants were grown in the presence or absence of interspecific competition. Hence, under the conditions and situations analyzed in this study, hybrids containing TM constructs were not effective weeds.
Different species will likely require various mitigators, and possible also for different environments . Anti-shattering genes will be appropriate for many row crops that have seed shattering weedy relatives, and anti-secondary dormancy genes would be appropriate where weeds possess seed-bank longevity as part of their survival mechanisms. Anti-bolting genes (e.g. anti-kaurene oxidase preventing gibberellic acid biosynthesis) would be appropriate mitigators for biennial crops (carrots, beets) or storage crops (onions, radishes, cabbages) that have weedy relatives. Sterility genes are appropriate for vegetatively propagated species (potatoes, poplars). The waxy transgene in maize would be a good mitigator for pharmaceutical genes expressed in maize embryos, as seeds from self or cross pollination would be shrunken, and unable to survive over-winter in most soils . Thus, to choose an appropriate mitigator for each crop and environment, the researcher must seek out genes that are positive or neutral for the crop in that environment and would be detrimental to the related weedy or wild species.