GIS assessment of the risk of gene flow from Brassica napus to its wild relatives in China

Environ Monit Assess (2018) 190:405 https://doi.org/10.1007/s10661-018-6753-9 GIS assessment of the risk of gene flow from Brassica napus to its wild relatives in China Jing-jing Dong & Ming-gang Zhang & Wei Wei Ke-ping Ma & Ying-hao Wang & Received: 6 December 2017 /Accepted: 25 May 2018 # Springer International Publishing AG, part of Springer Nature 2018 Abstract Risk of gene flow from canola (Brassica napus) to species of wild relatives was used as an example to evaluate the risk of gene flow of transgenic crops. B. juncea and B. rapa were the most common weedy Brassica species in China, which were both sexually compatible with canola. Data on canola cultivation in China were collected and analyzed using geographic information system (GIS), and the distribution of its wild relatives was predicted by MaxEnt species distribution model. Based on biological and phenological evidence, our results showed that gene flow risk exists in most parts of the country, especially in places with higher richness of wild Brassica species. However, risk in dominant canola cultivation regions is relatively low owing to the reduced distribution density of wild species in these regions. Three regions of higher risk of gene flow had been identified. Risk of gene flow is Electronic supplementary material The online version of this article (https://doi.org/10.1007/s10661-018-6753-9) contains supplementary material, which is available to authorized users. J.<j. Dong : W. Wei (*) : K.<p. Ma : Y.<h. Wang State Key Laboratory of Vegetation and Environmental Change, Institute of Botany, Chinese Academy of Sciences, Beijing, China e-mail: weiwei@ibcas.ac.cn M.<g. Zhang Institute of Loess Plateau, Shanxi University, Taiyuan, China Present Address: J.-. Dong Shandong Agriculture and Engineering University, Jinan, China relatively high in certain areas. China has been assumed to be the original center of B. juncea and B. rapa, and gene flow may lead to negative effects on the conservation of biodiversity of local species. Strategies had been proposed to reduce the possibility of gene flow either by monitoring introgression from crops to wild relatives in the areas of high adoption of the crop or by taking measures to limit the releasing of new crops or varieties in the areas with abundant wild relatives. Keywords Biosafety. Canola. Species distribution model. Transgenic crops. Weedy species Introduction Gene flow from transgenic crops to their wild relatives and its effect on these species is a heavily deputed issue in biosafety risk assessment, especially when the transgenic crops have been commercially released. Oilseed rape (Brassica napus, canola) is a very important oilseed crop, providing oil and protein (Raymer 2002). However, it has also been defined as the most risky crop in terms of gene flow (Eastham and Sweet 2002), because it has a high outcrossing rate, high seed and pollen production, and long dispersal distances. Canola occupies approximately 40% of the total oil crop cultivation area in China and accounts for approximately one third of total oil crop production. Canola is also an important transgenic material, and many transgenic cultivars have been produced, incorporating features such as herbicide and pest resistance and high yield (Cardoza

405 Page 2 of 12 Environ Monit Assess (2018) 190:405 and Stewart 2007). Since 1996, transgenic canola has been commercially released in several countries, which account for most of the canola production. In 2016, approximately 24% of the canola cultivation in the world used transgenic canola, which was resistant to herbicides (James 2016). In other countries, transgenic canola is not allowed to be grown commercially yet, although production conditions are optional. In China, no transgenic canola is cultivated commercially; however, more than 90% of the imported rapeseed is transgenic (Li et al. 2011). In addition, breeding is in progress with the goal of developing transgenic oilseed, and several cultivars have been released for field testing (Guan 2005). B. napus has several closely related and reproductively compatible wild relatives growing near agricultural fields. The weedy B. juncea is widely distributed in China near roadsides and farmlands, and the occurrence of the weedy B. rapa is reported in southwestern China (Wang et al. 2006). These two species and B. napus all belong to the Brassicaceae family and have similarities in their primary gene pools. B. napus (AACC, 2n =38) and B. juncea (AABB, 2n = 36) are allotetraploids derived from three diploid species B. nigra (BB, 2n =16), B. rapa (AA, 2n =20),andB. oleracea (CC, 2n =18)(U 1935). To assess hybridization and gene flow between B. napus and its wild relatives, several prerequisites should be considered, including geographic proximity (sympatry), species density, overlap of flowering phenology, and reproductive compatibility (hybridization rate). Researchers have reported that gene flow between B. napus (including the transgenic canola) and its wild relatives occurred easily via pollination. Breeding of interspecific hybrids was also successful (Bing et al. 1996; Kerlan et al. 1992; Metzetal.1997; Scheffler and Dale 1994), which indicated a high environmental risk of transgene escape in nature. Among these species, B. rapa was ranked highest in terms of its ability to hybridize with B. napus (Scheffler and Dale 1994); its hybridization rate in the field ranged from 0 to 69% (Jørgensen et al. 2004). The spontaneous hybridization rate between B. napus and B. juncea ranged from 0.3 to 2.3% (Jørgensen et al. 1998), ranking second after the hybridization rate between B. napus and B. rapa (Scheffler and Dale 1994). In addition to frequent natural crosses among B. napus and other Brassica species, transgene flow from B. napus to other species may also occur through pollination, and selectively advantageous traits could be transferred into wild populations. For example, a wild relative B. rapa could acquire transgenic traits such as herbicide resistance via spontaneous crop wild species hybridization, without any significant costs associated with the transgene. In addition, transgenic resistance to glufosinate is capable of persisting, even in the absence of selection of herbicide application (Snow et al. 1999). At present, no compelling data suggests that these hybridizations are inherently risky, although the actual consequences might be trait-dependent or event-specific (Liu et al. 2013; Warwick et al. 2008). Some traits might increase weediness or invasiveness, especially when environmental positive selection exists. However, even in the absence of selection pressures, the persistence of transgenic hybrids could result in the persistence of transgenic traits over time, as was observed with B. rapa in Québec (Warwick et al. 2008). Because hybridization of B. napus and its wild relatives is possible, assessing their sympatry is the first step in the risk analysis of gene flow. In southern England, Davenport et al. (2000) used satellite images to identify localities where gene flow was most likely to occur, by searching for oilseed rape fields adjacent to regions where the wild parent plants occurred. After visiting these localities and screening for hybrids, a single hybrid between B. napus and B. rapa was observed. Gene flow from B. napus to its wild relatives might not only introduce difficult-to-control weeds in agricultural systems but also impact the native populations in the non-agricultural habitats when the introgressed traits confer fitness benefits (Liu et al. 2015). There has been no research, at the regional or national level, to assess the risk of gene flow from crops to wild species based on distribution. In this study, information on the distribution of canola cultivation at the county level, the distribution of wild canola relatives from herbarium specimens and previous investigations throughout the country, and flowering period of each species, which is variable according to their geographic locations, was collected and analyzed to evaluate the risk of gene flow at the national level, as well as identify regions of high risk. These results will increase our understanding of the risk of gene flow from B. napus to its relatives and will be helpful in further studies and management for ecological safety.

Environ Monit Assess (2018) 190:405 Page 3 of 12 405 Materials and methods Area of canola cultivation The distribution of canola fields at the county level in China was mapped by ArcGIS (ESRI, Inc.) using agricultural statistical data from statistical yearbooks. According to the China Administrative Divisions (ADM), ADM1 was defined at provincial level and consisted of over 300 ADM2 regions; ADM2 included cities and consisted of over 3000 ADM3 subregions. The ADM3, generally corresponding to county, was the basic unit in which gene flow was assessed in our study. Data on canola cultivation areas were collected from statistical yearbooks from each ADM1 during 2008. Unfortunately, canola cultivation areas could not be obtained directly from the statistical yearbooks for all of the ADM3 subregions and further calculations were conducted. No canola was cultivated in Beijing, Tianjin, Jilin, Hainan, Hong Kong, or Macau, according to the statistical yearbooks at the ADM1 level. Apart from these, there were three ways in which data were collected for the cultivated area of canola in the remaining 28 ADM1s in China: (1) area of canola cultivation in each ADM3 was listed directly in the yearbooks of Shanxi, Xinjiang, and Taiwan; (2) canola yield of each ADM3 was given, which was used to estimate the cultivated area according to the local canola yield per unit area in Zhejiang, Hunan, Guizhou, Yunnan, and Xizang; (3) total area of oil crop cultivation or production in each ADM3 was given and these data were used to estimate the cultivated area of canola by determining the proportion of canola in the total oil crop using the yearbooks for the rest 20 ADM1s (Hebei, Neimenggu, Heilongjiang, Jiangsu, Anhui, Fujian, Shandong, Hubei, Guangdong, Guangxi, Chongqing, Shaanxi, Liaoning, Jiangxi, Henan, Gansu, Qinghai, Ningxia, Sichuan, and Shanghai). Wild Brassica species Distributions of wild Brassica species were mapped according to specimen records in herbariums and distribution data in published literatures. Through reviewing these data sources, we found that B. nigra and B. elongata were rarely reported in China, and thus were excluded from this study. The main wild species were B. juncea and B. rapa. More than 400 records of wild Brassica plants were collected, including the name of the species, location, habitat, and flowering period. The data resources included: (1) the online database through the Chinese Virtual Herbarium (http://www.cvh.org.cn, an open online portal allowing access to herbarium specimen information and other botanical knowledge with collaboration among more than 20 major herbaria in China), China Germplasm Bank of Wild Species and Herbarium of National Taiwan University; (2) published literature; and (3) national and local floras. A list of herbariums and floras used as data sources in this study is given in the Supplementary Information (S1 and S2). Published literatures and some of the specimens had detailed GPS coordinates. As for those without GPS data (173 out of 431 records), we georeferenced them using location descriptions. Prediction of species distributions Species distribution models (SDMs) are numerical tools that combine observations of species occurrence or abundance with environmental estimates via statistical or machine-learning procedures. Data collected from herbariums and literatures only described known occurrences of species and were always isolated. This type of data has been used in many investigations, being justified by the lack of systematic survey data (Elith et al. 2006; Elith and Leathwick 2009). We also used SDMs to predict the possible distribution of wild Brassicas plants, using the function MaxEnt (ver. 3.3.1) (Phillips et al. 2006). Because no absence data for wild Brassica species was available, the area under the receiver operating characteristic curve (AUC) was the only indicator of usefulness of the model. The AUC ranged from 0 to 1, where a score of 1 indicated perfect discrimination between sites in regard to species presence or absence, a score of 0.5 implied a prediction equivalent to random, and < 0.5 indicated discrimination less accurate than random (Elith et al. 2006). Environmental factors for MaxEnt included elevation and climate data at 2.5 arc-minutes resolution, which were available at the WorldClim website (http://www. worldclim.org/). There were 19 bioclimatic variables (BIO01 BIO19) in the original climate dataset, which were selected to avoid the SDM overfitting problem (Graham 2003; Pearson et al. 2007). Spearman s rank correlation test was used to select the BIO variables, and when two variables were highly correlated (Spearman s

405 Page 4 of 12 Environ Monit Assess (2018) 190:405 r > 0.75), the variable with greater potential ecological implications was retained and the other was removed (Zhang et al. 2013). Seven BIO variables remained for MaxEnt, including (1) BIO01, annual mean temperature; (2) BIO02, mean diurnal temperature range; (3) BIO04, temperature seasonality; (4) BIO07, annual temperature range; (5) BIO12, annual precipitation; (6) BIO14, precipitation of driest month; and (7) BIO15, precipitation seasonality. Gene flow risk index An index was designed to quantify the risk of gene flow from cultivated B. napus to its wild relatives at the ADM3 level. Hybridization is a prerequisite for gene flow, and it may happen when certain circumstances are all achieved: (1) sympatry of occurrence, (2) overlapping of flowering phenology, (3) pollen and seed dispersal, and (4) successful fertilization and viable offspring. The successful fertilization via pollen and seed dispersal, that resulted in viable offspring, between Brassica napus crop and its wild relatives (B. juncea and B. rapa) has been confirmed in current literatures (e.g., Liu et al. 2013) once the first and second circumstances were met. Therefore, the risk evaluation of gene flow in this paper focused on the first and second circumstances stated above. Sympatric occurrence was evaluated by mapping canola cultivation areas and the distribution of its wild relatives. In each ADM3, the area of cultivation was available but its specific distribution was unknown. Therefore, we improved the cultivation factor in the risk index by using the ratio (Area cultivation /Area non-habitat ), where Area cultivation equaled the area of canola cultivation and Area non-habitat was the area that was not suitable for canola or its wild relatives. Land cover data BGlobCover2009^ (v2.3) from the ESA website (http://due.esrin.esa.int/globcover/) was used to identify non-habitat regions. Among all the wild Brassica records used in this analysis from herbariums and published materials, 179 records contained environmental description of where the plants were found. Those descriptions were used to determine appropriate habitats for wild Brassica, and we concluded that areas including as deserts, forests, water bodies, and artificial surfaces were non-habitat areas for Brassica (described in S3). The ratio Area cultivation /Area non-habitat was divided into 5 ranks according to the minimum (0%) and maximum (15.8%) value and assigned a score of 1 to 5 (score 1 contained 18.5% of the cells with a ratio value of 0; score 2 contained 71.2% of the cells with a ratio value > 0 but 0.5%; score 3 contained 4.4% of the cells with a ratio value > 0.5 but 1.0%; score 4 contained 3.7% of the cells with ratio values > 1.0 but 2.0%; and score 5 contained 2.2% of the cells with ratio values > 2.0%). The possibility of the presence of wild relatives, or their distribution, was predicted by MaxEnt at a resolution of 2.5 arc-minutes, and prediction results ranged from 0 to 1. These values were also equally divided into 5 ranks (i.e., break values were 0.2, 0.4, 0.6, and 0.8) and assigned a score from 1 to 5. Overlap of flowering periods was defined as the number of days during which both B. napus and its wild relatives were flowering. Three ranks were designated. Rank 1 was assigned when the overlap of flowering periods was 10 days; rank 2 was assigned when the overlap was > 10 days but 1 month; and rank 3 was assigned when the overlap was > 1 month. The risk index was defined as the sum of the three scores (the ratio of Area cultivation /Area non-habitat [score 1 to 5], possibility of wild relative presence [score 1 to 5], and overlap of flowering period [score 1 to 3]). Thus, the minimum risk index was 3 and the maximum was 13, and the higher the risk index the more possible gene flow became. In order to adjust the risk assessment, rank 1(value 0.2) of the MaxEnt prediction was used as a control for areas of low risk. Results Canola cultivation in China Data regarding the area of canola cultivation were collected at the ADM3 level. According to the statistic yearbooks, canola cultivation was widespread in country, regardless of the longitude and latitude. In regions along the Changjiang River, the cultivation density was relatively high, and these places are also the dominant canola cultivation regions in China. We assumed that locations with more non-habitat regions would increase the difficulty of gene flow between canola and its wild relatives via either limitation of pollen or seed dispersal. It was obvious that locations with larger areas of canola cultivation usually covered smaller non-habitat areas, but few of them changed their ranking when dividing the cultivation area (Area cultivation )byareaofnon-habitat regions (Area non-habitat )(Fig.1a, b). For example,

Environ Monit Assess (2018) 190:405 Page 5 of 12 405 Fig. 1 Brassica napus crop cultivation of China in 2008. a The original canola cultivation proportion: ratio of canola area to AMD3 area. b The canola area ratio rank, according to the ratio of canola area to non-habitat area

405 Page 6 of 12 Environ Monit Assess (2018) 190:405 Fig. 2 Distribution of wild Brassica species in China. Counts refer to the number of records found in each ADM1 Chunhua and Qianxian counties of the Shaanxi Province, because of a low percentage of non-habitat (< 1%), had a higher score for canola cultivation compared to other ADM3 areas with a similar level of Area cultivation / Area ADM3t, whereas Pengze and Anyi counties in the Jiangxi Province had a lower score (b-value) compared to similar counties, which was due to a higher percentage of non-habitat (> 30%). Distribution of wild Brassicas Surveying the herbarium database (59%) and literature (41%) resulted in a total of 431 records of wild Brassica specimens. Among these, there was only one record of B. elongata and one record of B. nigra, which were not used in further analysis; in addition, records without detailed species information or location were excluded. The remaining data consisted of 325 records identified as B. juncea, and 46 records identified as B. rapa, which were only found in three ADM1s (Xizang, Yunnan and Taiwan). The number of wild Brassica specimens found in each ADM1 demonstrated that western China had a greater abundance of wild Brassica,whereas the density of wild Brassica in the dominant canola production regions was relatively low, including areas such as Hubei and Hunan provinces (Fig. 2). Gene flow risk based on SDM prediction The potential distribution of wild Brassica plants (i.e., B. juncea and B. rapa) was successfully predicted by MaxEnt, and the training AUC was 0.913 and 0.984 for

Environ Monit Assess (2018) 190:405 Page 7 of 12 405 the two species, respectively. The main area of the predicted distribution for B. juncea covered more than half of China, except for a few regions in southern, northeastern, and northwestern China (Fig. 3a). The predicted distribution of B. rapa was mainly in southwestern China, including Yunnan, Sichuan, and Xizang (Fig. 3b). The gene flow risk index was calculated for each ADM3, and we defined the top one third of the high index values as high risk. For B. juncea, the risk index ranged from 3 to 13, and the index values of 10 13 indicated high risk (the orange-red and red area in the Fig. 4a). Areas having a high-risk index appeared in region A located in some areas of Jiangsu, Anhui, Henan, and Hubei provinces, region B located to the east of Sichuan, and region C in the middle of Shaanxi (Fig. 4a). High-risk areas of gene flow for B. rapa were areas with an index level > 7, as the maximum risk index for B. rapa was 10 because of the lower canola cultivation ratio in B. rapa habitat. These high-risk areas, presented in orange-red and red in Fig. 4b, were mainly located in southwestern China (i.e., Yunnan, western Sichuan, southeastern Xizang, and a small region of southern Qinghai province). From the establishment of the risk index, two types of high-risk regions could be identified: (1) regions that had wild relatives and a very high canola cultivation ratio; (2) regions that had canola cultivation and a large amount of habitat for wild relatives. Discussion Risk of gene flow from canola to its wild relatives in China was addressed in this study. Gene flow requires the co-occurrence of both canola and wild relatives. Area of canola cultivation was addressed rather than production, because the geographic distribution directly affected gene flow, especially when the traditional farmlands were relatively small and highly dispersed, as is usually observed in southern China (Zhu et al. 2012). On the other hand, in modern industrial agriculture, there may be enormous areas in cultivation. Thus, weed populations hundreds of meters away may have gene flow with cultivars (Ellstrand et al. 1999). Based on field investigations of wild Brassica species and the prediction of their distribution using SDM MaxEnt, wild relatives of B. napus occurs, or could survive, in most parts of China. The presence and proximity of wild relatives are the necessary conditions for gene flow. Usually, the risk of gene flow between one species and its wild relatives has been found to be negatively correlated with the spatial distance between them (Liu et al. 2013), but no clear cutoff distance where levels of outcrossing reached zero could be determined (Hüsken and Dietz-Pfeilstetter 2007). The farther species were from each other, the less likely gene flow would take place. On the contrary, if they grow in similar habitats or very close in space, the possibility of gene flow is increased. In this paper, data on the area of canola cultivation were collected or estimated from statistical yearbooks. However, their detailed distributions within ADM3s were unknown, which resulted in uncertainty in the assessment of distance between cultivated canola and wild Brassica species. We introduced the area of non-habitat to decrease this uncertainty. The area data for canola cultivation in this paper were from the same year (2008) for data integrity and consistency, because the area of canola cultivation may vary among years. For example, the area of canola cultivation in the Xinjiang province in 2009 was twice that of 2008. Furthermore, Toksun County of Xinjiang cultivated about 430 ha of canola in 2009, whereas no canola cultivation occurred during 2008. Thus, regions without canola cultivation might have had or will have canola cultivation in the past or in the future, respectively. Along with this, seed shattering and seed dormancy are other factors that should not be neglected. These are oilseed rape seed traits, which may enhance gene flow (Zhu et al. 2012). Up to 30% of canola seed may be lost each year by shattering during harvest (Gulden et al. 2003), and seed spill during transport is also a likely mechanism for seed escape (Schafer et al. 2011). Spilled seeds might remain dormant for up to 10 years in unplowed land and 5 years in plowed land (Pekrun et al. 1998). The seed shattering and dormancy properties lead to a soil seed bank. Considering effects of soil seed banks and the transport of oilseed rape products, high-risk regions might extend beyond our estimation. Online databases, floras, and literature were the main sources of data for wild Brassica distribution. The distribution data were not all investigated by systematically organized field surveys, and therefore, the occurrence of localities might be biased and sampling intensity and methods often varied widely (Anderson et al. 2003; Reddy and Dávalos 2003). The MaxEnt species distribution model was used to improve this deficiency of presence-only data (Phillips et al. 2006). Although

405 Page 8 of 12 Environ Monit Assess (2018) 190:405 Fig. 3 Predicted distribution of wild Brassica species B. juncea (a) and B. rapa (b) in China by MaxEnt

Environ Monit Assess (2018) 190:405 Fig. 4 Gene flow risk index from Brassica napus to B. juncea (a) and B. rapa (b) Page 9 of 12 405

405 Page 10 of 12 Environ Monit Assess (2018) 190:405 MaxEnt prediction accuracy, i.e., AUC, was larger than 0.9 based on presence-only data, the bias of specimen collection might result in an invalid assumption. However, regarding the aspect of existence of risk, gene flow risk does certainly exist and this information will assist in field management of transgenic oilseed rape. A risk index was designed to quantify gene flow risk, for which three factors were included. Days of flowering overlap is a very interesting factor. For example, during a systematic field investigation conducted from 1999 to 2004 in the Xizang Autonomous Region (Wang et al. 2006), an abundance of wild Brassica species was found, including B. juncea and B. rapa. Despite the abundance of wild species and the high hybridization rate between B. rapa and canola, the fact that few days of flowering overlap occurred resulted in a lower risk assessment as calculated by the risk index. Researchers found that the flowering period of the same oilseed rape species, even seeds from the same source, differed significantly between cultivars and feral plants (Wang et al. 2006). Thus, increasing days of flowering overlap increased the gene flow risk index, but absence of overlap may not eliminate the risk. In addition, as there are no records of wild B. rapa in most parts of China, the predicted distribution was limited to the southwest where both lower and higher gene flow risk indices were calculated. This finding may change when new areas for the distribution of the wild species are reported. Generally, gene flow between crops and their wild relatives may have two potentially harmful consequences: the evolution of increased weediness and the increased likelihood of extinction of wild rare relatives (Ellstrand et al. 1999). In regard to canola, we might focus on the problem of weediness rather than rare species extinction. Evidence showed that gene flow from canola to its wild species did exist, and it was beneficial for the wild species. Traditionally improved cultivars and gene-modified cultivars likely confer a selective advantage for weed populations. The F 1 hybrids and hybrids that backcrossed with wild plants were fertile, and did not suffer reduced fitness, and some traits exhibited positive revolution according to field experiments (Hauser and Shaw 1998; Liu et al. 2015). After the commercial release of transgenic canola in America, Schafer et al. (2011) found that widespread distribution of feral canola was established outside of the cultivated areas both near and far from the cultivated fields, with and without obvious selection pressure (e.g., glyphosate herbicide). About 36% (231/634) of these feral canola samples expressed at least one transgene. Huangfu et al. (2007) conducted investigations in nine provinces of China and found that wild B. juncea in the Jiangsu province had evolved a very high resistance to glyphosate herbicide. As no transgenic canola was approved for commercial cultivation in China, this resistance to glyphosate could be a result of evolution responding to glyphosate selection pressure rather than transgene introgression. However, gene flow from herbicide resistant canola after its environmental release would increase the risk of resistance. Along with weediness, the wild populations with new traits of selective advantages, such as herbicide resistance, would improve fitness of noxious weeds and might exacerbate the problem of weeds control and crop failure. Conversely, the transgene introgression could also raise competition with native species and threaten local wild species biodiversity (Ellstrand et al. 1999; Jørgensen et al. 2004). Different circumstances require different management for safety. Our results suggest that, in order to reduce or manage the risk of gene flow, different strategies should be employed. For example, in regions having high proportion of canola cultivation, such as Anhui and Hubei, it is necessary to monitor functional traits of wild relatives that may have ecological impacts, such as herbicide and insect resistance, which were probably introgressed from transgenic canola or weedy resistant population through gene flow. In regions having abundant wild relatives, such as Gansu and Qinghai, the introduction of a new crop or a new cultivar should be done with caution. Enhancement of the management of agricultural areas could be the better option, such as building separation barriers in space and time to prevent gene flow. The risk of gene flow from canola to its wild relatives was assessed herein according to the biological and ecological properties (blooming period, hybridization rate, pollen spread, etc.), regardless of industrial factors and transmission of products. Although transgenic canola has not been commercially released in China yet, related transgenic canola seeds and products are legally imported from abroad every year. As Schafer et al. (2011) also pointed out the movements by transport were likely to explain the distribution of feral canola populations with transgene(s). If seed spillage during transportation occurred in China, the risk of gene flow from canola to wild relatives would have been

Environ Monit Assess (2018) 190:405 Page 11 of 12 405 higher and wider spread, as would the possibility of transgene spread in field populations. Acknowledgements This work was funded by a National Natural Science Foundation of China (NSFC) grant (31370357) and a Commonweal Scientific Programme of the Ministry of Environmental Protection of China (201109028). References Anderson, R. P., Lew, D., & Peterson, A. T. (2003). Evaluating predictive models of species distributions: Criteria for selecting optimal models. Ecological Modelling, 162, 211 232. Bing, D. J., Downey, R. K., & Rakow, G. F. W. (1996). Hybridizations among Brassica napus, B. rapa and B. juncea and their two weedy relatives B. nigra and Sinapis arvensis under open pollination conditions in the field. Plant Breeding, 115, 470 473. Cardoza, V., & Stewart Jr., C. N. (2007). Rapeseed biotechnology. Advances in Botanical Research, 45, 435 449. Davenport, I. J., Wilkinson, M. J., Mason, D. C., Charters, Y. M., Jones, A. E., Allainguillaume, J., Butler, H. T., & Raybould, A. F. (2000). Quantifying gene movement from oilseed rape to its wild relatives using remote sensing. International Journal of Remote Sensing, 21, 3567 3573. Eastham, K., & Sweet, J. (2002). Genetically modified organisms (GMOs): The significance of gene flow through pollen transfer. Denmark: European Environment Agency Copenhagen. Elith, J., Graham, C. H., Anderson, R. P., et al. (2006). Novel methods improve prediction of species distributions from occurrence data. Ecography, 29, 129 151. Elith, J., & Leathwick, J. R. (2009). Species distribution models: Ecological explanation and prediction across space and time. Annual Review of Ecology, Evolution, and Systematics, 40, 677 697. Ellstrand, N. C., Prentice, H. C., & Hancock, J. F. (1999). Gene flow and introgression from domesticated plants into their wild relatives. Annual Review of Ecology, Evolution, and Systematics, 30, 539 563. Graham, M. H. (2003). Confronting multicollinearity in ecological multipie regression. Ecology, 84, 809 2815. Guan, C. Y. (2005). Breeding of rapeseed by genetic transformation. Chinese Journal of Oilseed Crop Sciences, 27, 97 103 (in Chinese with English Abstract). Gulden, R. H., Shirtliffe, S. J., & Thomas, A. G. (2003). Harvest losses of canola (Brassica napus) cause large seedbank inputs. Weed Science, 51, 83 86. Hüsken, A., & Dietz-Pfeilstetter, A. (2007). Pollen-mediated intraspecific gene flow from herbicide resistant oilseed rape (Brassica napus L.). Transgenic Research, 16, 557 569. Hauser, T. P., & Shaw, R. G. (1998). Fitness of F1 hybrids between weedy Brassica rapa and oilseed rape (B. napus). Heredity, 81, 429 435. Huangfu, C. H., Song, X. L., Qiang, S., & Zhang, H. J. (2007). Response of wild Brassica juncea populations to glyphosate. Pest Management Science, 63, 1133 1140. Jørgensen, R., Andersen, B., Hauser, T. P., Landbo, L., Mikkelsen, T. R., & Østergård, H. (1998). Introgression of crop genes from oilseed rape (Brassica napus) to related wild species an avenue for the escape of engineered genes. Acta Horticulturae, (459), 211 217. Jørgensen, R. B., Ammitzbøll, H., Hansen, L. B., Johannessen, M., Andersen, B., & Hauser, T. P. (2004). Gene introgression and consequences in Brassica. In H. C. M. D. Nijs, D. Bartsch, & J. Sweet (Eds.), Introgression from genetically modified plants into wild relatives (pp. 253 262). Cambridge: CABI Publishing. James, C. (2016). Global status of commercialized biotech/gm crops: 2016. ISAAA Brief No. 52. Ithaca, NY: ISAAA. Kerlan, M. C., Chèvre, A. M., Eber, F., Baranger, A., & Renard, M. (1992). Risk assessment of outcrossing of transgenic rapessed to related species. Euphytica, 62, 145 153. Li, X., Pan, L. W., Li, J. Y., Lü, R., Zhang, S. Y., Liu, Y. M., & Gao, Q. (2011). Identification of imported genetically modified rapeseeds. Chinese Journal of Oilseed Crop Sciences, 33, 77 82 (in Chinese with English Abstract). Liu, Y. B., Darmency, H., Stewart Jr., C. N., Wei, W., Tang, Z. X., & Ma, K. P. (2015). The effect of Bt-transgene introgression on plant growth and reproduction in wild Brassica juncea. Transgenic Research, 24, 537 547. Liu, Y. B., Wei, W., Ma, K. P., Li, J. S., Liang, Y. Y., & Darmency, H. (2013). Consequences of gene flow between oilseed rape (Brassica napus) and its relatives. Plant Science, 211, 42 51. Metz, P. L. J., Jacobsen, E., Nap, J. P., Pereira, A., & Stiekema, W. J. (1997). The impact on biosafety of the phosphinothricintolerance transgene in inter-specific B. rapa B. napus hybrids and their successive backcrosses. Theoretical and Applied Genetics, 95, 442 450. Pearson, R. G., Raxworthy, C. J., Nakamura, M., & Peterson, T. A. (2007). Predicting species distributions from small numbers of occurrence records: A test case using cryptic geckos in Madagascar. Journal of Biogeography, 34, 102 117. Pekrun, C., Hewitt, J. D. J., & Lutman, P. J. W. (1998). Cultural control of volunteer oilseed rape (Brassica napus). Journal of Agricultural Science, 130, 155 163. Phillips, S. J., Anderson, R. P., & Schapire, R. E. (2006). Maximum entropy modeling of species geographic distributions. Ecological Modelling, 190, 231 259. Raymer, P. L. (2002). Canola: An emerging oilseed crop. In J. Janick & A. Whipkey (Eds.), Trends in new crops and new uses: Strength in diversity (pp. 122 126). Alexandria, VA: ASHS Press. Reddy, S., & Dávalos, L. M. (2003). Geographical sampling bias and its implications for conservation priorities in Africa. Journal of Biogeography, 30, 1719 1727. Schafer, M. G., Ross, A. A., Londo, J. P., Burdick, C. A., Lee, E. H., Travers, S. E., van de Water, P. K., & Sagers, C. L. (2011). The establishment of genetically engineered canola populations in the U.S. PLoS One, 6, e25736. Scheffler, J. A., & Dale, P. J. (1994). Opportunities for gene transfer from transgenic oilseed rape (Brassica napus) to related species. Transgenic Research, 3, 263 278. Snow, A. A., Andersen, B., & Jørgensen, R. B. (1999). Costs of transgenic herbicide resistance introgressed from Brassica napus into weedy B. rapa. Molecular Ecology, 8, 605 615.

405 Page 12 of 12 Environ Monit Assess (2018) 190:405 U. N. (1935). Genimic analysis in Brassica with special reference to the experimental formation of B. napus and peculiar mode of fertilization. Journal of Japanese Botany, 7, 389 452. Wang, J. L., Luan, Y. F., Daci, Z. G., & Chang, T. J. (2006). Geographical distribution and biological characters of wild rapeseed in Tibet. Chinese Journal of Oilseed Crop Sciences, 28, 134 137 (in Chinese with English Abstract). Warwick, S. I., Légère, A., Simard, M. J., & James, T. (2008). Do escaped transgenes persist in nature? The case of an herbicide resistance transgene in a weedy Brassica rapa population. Molecular Ecology, 17, 1387 1395. Zhang, M. G., Zhou, Z. K., Chen, W. Y., Cannon, C. H., Raes, N., & Slik, J. (2013). Major declines of woody plant species ranges under climate change in Yunnan, China. Diversity and Distribution, 20, 405 415. Zhu, Y. M., Li, Y. D., Colbach, N., Ma, K. P., Wei, W., & Mi, X. C. (2012). Seed losses at harvest and seed persistence of oilseed rape (Brassica napus) in different cultural conditions in Chinese farming systems. Weed Research, 52, 317 326.