Introduction
Many domesticated plants have closely related wild and weedy types (Jarvis and Hodgkin, Reference Jarvis and Hodgkin1999) from which they were derived (e.g. de Wet and Huckabay, Reference de Wet and Huckabay1967) and in many cases, both occur in overlapping regions (Hooftman et al., Reference Hooftman, De Jong, Oostermeijer and Den Nijs2007; Adugna and Bekele, Reference Adugna and Bekele2013). This coexistence may result in the formation of viable and fertile hybrids (Chapman and Burke, Reference Chapman and Burke2006). Wild and weedy sorghum populations exhibit great diversity (Adugna et al., Reference Adugna, Snow, Sweeney, Bekele and Mutegi2012) and may hybridize among themselves as well as with cultivated sorghum populations (de Wet, Reference de Wet1978; Adugna et al., Reference Adugna, Sweeney and Bekele2013).
Hybrids between wild and transgenic cultivated plants could be more weedy and more invasive than either parent (Paterson et al., Reference Paterson, Schertz, Lin, Liu and Chang1995; Snow et al., Reference Snow, Pilson, Rieseberg, Paulsen, Pleskac, Reagon, Wolf and Selbo2003; Warwick et al., Reference Warwick, Légère, Simard and James2008); however, certain genetically modified traits might confer a fitness cost to crop–wild hybrids (Chen et al., Reference Chen, Snow, Wang and Lu2006). For example, if a transferred gene was to lead to an enhancement in the fitness of a recipient wild relative, then this might lead to the extinction of certain wild crop relatives (Small, Reference Small and Grant1984) mainly due to outbreeding depression and swamping (Ellstrand et al., Reference Ellstrand, Prentice and Hancock1999). Consequently, the escape of transgenes through the hybridization of crops with their wild relatives in their centres of origin has been cited as one of the potential environmental risks of transgenic crops (Hails, Reference Hails2000) because the centres of origin and diversity harbour highly diversified crop landraces, wild progenitors and related wild species (Gepts and Papa, Reference Gepts and Papa2003). Ethiopia lies within the broad geographic range where sorghum is believed to have been domesticated, yet little has been done to conserve wild relatives. To our knowledge, this is the first empirical study to investigate the risk of crop–wild gene flow in Ethiopia.
Genetic transformation of sorghum has been underway (e.g. Zhao et al., Reference Zhao, Cai, Tagliani, Miller, Wang, Pang, Rudert, Schroeder, Hondred, Seltzer and Pierce2000, Reference Zhao, Glassman, Sewalt, Wang, Miller, Chang, Thompson, Catron, Wu, Bidney, Kedebe, Jung and Vasil2003; Liang and Gao, Reference Liang and Gao2001), and transgenic sorghum is expected to be released in the near future (Zhao, Reference Zhao and Xu2007). Therefore, it is timely to undertake biosafety risk assessment studies case by case in each country (Ejeta and Grenier, Reference Ejeta, Grenier and Gressel2005). Furthermore, although wild/weedy types of sorghum populations are sources of genes for crop improvement, as weeds, they have direct negative consequences on sorghum production through competition and harbouring of insect pests and disease pathogens (Adugna and Bekele, Reference Adugna and Bekele2013). Hence, in order to provide better weed control methods, it is important to perform studies to learn about the fitness of wild–crop hybrids that will make their way into weed populations. The objective of this study was to assess the fitness components of wild–cultivated sorghum F1 hybrids in comparison with their parents, to gain a better understanding of their population ecology in Ethiopia.
Materials and methods
Plant materials
The list of the parents used in the study is presented in Table 1. Twenty-three wild sorghum accessions and two released cultivated sorghum varieties were involved in the initial crossing. Crossing was made in 2010 during the main rainy season at Melkassa Agricultural Research Center (MARC) located in the central Rift Valley of Ethiopia (39°21′E, 8°24′N, altitude 1550 m) using hand emasculation techniques. The wild sorghum lines served as female parents and the pollinators were two improved sorghum varieties, 76T1#23 and WSV 387 (Melkam), as this was the natural pattern of gene flow intended to be investigated. The crop varieties belong to the race caudatum and are early maturing, recommended for production in lowlands of Ethiopia similar to the places of collection of most of the wild sorghum parents used in this study. Three of the wild parents were acquired from the International Crops Research Institute for the Semi-Arid Tropics (ICRISAT) in India. Two of the ICRISAT wild parents belong to subsp. verticilliflorum race aethiopicum (originated from Sudan) and one belongs to subsp. verticilliflorum race virgatum from the USA. The remaining parents were Ethiopian collections in November 2008 from the Hararghe and Tigray areas belonging to subsp. verticilliflorum race arundinaceum and from the Wello area which had unknown grouping, but most probably belonging to subsp. drummondii. The seeds of all of the experimental parents were multiplied under the same management at the same field where crossing of hybrids was made. The harvested seeds were kept in the laboratory at room temperature until the beginning of the experiment.
Table 1 List of the parents involved in the experiment

Laboratory and field experiments
Germination test was carried out just after sowing of the field experiment for those entries that had an adequate amount of remnant seeds. Before the test, the glumes were removed with fine sandpaper to facilitate germination. In each of the four replications, 25 seeds of the hybrids and their parents were germinated on 115 mm Petri dishes lined with number one Whatman® filter paper moistened with distilled water. Each replication was represented by a compartment in the incubator. The Petri dishes were kept at a constant temperature of 30 ± 3°C in the dark. Distilled water was added as needed.
The field experiment was carried out at MARC during the main rainy season of 2011 (from 29 June to 29 October 2011). The amount of rainfall during the season was 589.7 mm. The majority of this amount was distributed in the first 3 months (June–July 131.2 mm, August 208.8 mm and September 197.5 mm). There was no rainfall in October. The soil at the experimental site was silty clay loam Andosol with a pH of 7.8. Out of a pool of crosses made in 2010, seven F1 hybrids were selected for this study based on the availability of adequate seed. Hence, a total of 16 entries including hybrids, their cultivated and wild parents were involved. The experimental design was a randomized complete block (RCBD) with five replications. The plot size was a single row of length 3 m. The inter-row spacing was 0.75 m and the space between the plants was 15 cm. The space between the blocks was 1.5 m. Two seeds were dropped by hand per hole and later, the seedlings were thinned to a single plant. The thinned seedlings were transplanted wherever there were some missing holes. In order to avoid transplanting shock, transplanting was done while it was raining. Because the wild parents and some of the hybrids tended to lodge, they were braced with stakes. All management practices were done as per the recommendation for sorghum in the lowlands where the wild sorghums are better adapted in Ethiopia. Accordingly, diammonium phosphate (DAP) fertilizer was applied at the rate of 100 kg/ha during planting. Moreover, urea was applied at the rate of 50 kg/ha when the plants reached 30–40 cm in height. To control sorghum shoot fly and stem borers, Karate 5% EC was applied at the rate of 320 ml/ha. The field was kept free of weeds throughout the duration of the experiment.
Data recording and statistical analysis
Data were recorded at three main life-history stages following the procedure of Burke et al. (Reference Burke, Carney and Arnold1998). Germination percentage and days to emergence were considered to be juvenile survival traits (Table 2). Days to flowering, plant height, number of productive tillers, flag leaf length and width, leaf number, head length and head width were regarded as adult (growth) characters. Likewise, seed yield per plant (female fecundity) was regarded as fertility (reproductive) traits. Data on germination percentage were taken in the laboratory every day, starting from the day 3 to day 15, and germinated seeds were removed after the count. Data on days to 50% seedling emergence and flowering were recorded on a plot basis. Plant height, flag leaf length and width, leaf number, head length and head width were recorded at flowering on an individual plant basis from six to ten plants that were randomly selected and permanently tagged in each plot throughout the field experiment. The number of productive tillers was counted at maturity. To avoid shattering of seeds, panicles were covered using waterproof Lawson® paper bags when flowering was completed and started producing seeds. Bagging continued until all tillers completed flowering. The number of seeds per plant (female fecundity) was estimated from grain yield and the weight of 1000 grains counted using an electronic grain counter (Wagtech International, Tachtam, Berkshire, UK). Grain yield per plant was determined in each plot on the tagged plants. For estimating grain yield per plant, all heads from the six to ten main plants and from all productive tillers were threshed together, cleaned carefully and weighed, and the total weight was divided by the number of plants. Count data were log-transformed and percentage data were arcsine-transformed. To assess the variation among the parents and crosses, all measured variables were subjected to analysis of variance and mean separation by Duncan's multiple range test using SPSS Release 17.0 software. Two methods were followed to estimate the level of fitness components: mid-parent heterosis and direct comparison of relative fitness. Mid-parent heterosis (MPH) of the hybrids was computed using the following formula:

Table 2 Description of the morphological and phenological characters measured for estimating fitness
GERM, seed germination; DTE, days to 50% seedling emergence; DTF, days to 50% flowering; PHT, plant height; LL, flag leaf length; LW, flag leaf width; LN, leaf number; TILL, number of tillers per plant; HL, head length; HW, head width; NSPP, number of seeds per panicle.
where F 1 is the mean value of the wild × crop hybrid and MP is the mean values of the parents involved in the cross. Significance of mid-parent heterosis was tested using a t test following the procedure of Wynne et al. (Reference Wynne, Emery and Rice1970) as follows:

Relative fitness of the hybrids and their parents was computed following the procedure of Song et al. (Reference Song, Lu, Wang and Chen2004) and also as described in Templeton (Reference Templeton2006). Accordingly, the mean values of two parents and the hybrid for each trait were arranged in ascending order with the highest mean value receiving a relative fitness of 1.0. The remaining values were computed from that value as a reference.
Results
Morphology and relative fitness of hybrids and parents
Analysis of variance showed significant differences for all the measured juvenile survival, growth and fertility traits (Table S1, available online). Laboratory seed germination and field emergence of crop–wild hybrids were compared with that in the parents as juvenile survival traits. Early emergence was regarded as a better fitness component. There were significant differences between the parents and hybrids in laboratory seed germination and field emergence. Three of the five hybrids compared had better mean germination than their wild parents (Table 3). In one of the crosses, the hybrids did not significantly perform better than both of their parents in terms of seed germination percentage. However, the hybrids showed relatively later emergence than their wild parents.
Table 3 Mean of the life-history traits measured on wild×crop F1 sorghum hybrids and their parents at Melkassa in 2011
GERM, germination percentage; DTE, days to 50% seedling emergence; DTF, days to flowering; PHT, plant height; TILL, number of tillers per plant; HL, head length; HW, head width; LL, leaf length; LW, leaf width; LN, leaf number; NSPP, number of seeds per panicle; WP, wild parent; NA, not available (seed was not adequate); CP, crop parent.
Values with unlike superscript letters are significantly different (P< 0.05).
Most of the hybrids assumed intermediate magnitude for most of the growth traits, perhaps because their parents were extremely variable for some characters. Four of the seven hybrids showed earlier flowering than their cultivar parents. The remaining three hybrids had flowering times not significantly different from both of their parents. All hybrids except W5-20 × 76T1#23 were significantly taller than their cultivar parents. Moreover, three were significantly taller than both parents.
All hybrids except one had a significantly lower number of tillers per plant than their wild parents, but significantly a higher number of tillers than their cultivated parents. Five of the seven hybrids had significantly longer flag leaves than their wild parents, three of which had also a significantly shorter flag leaf length than their crop parents. Two hybrids (W5-20 × 76T1#23 and IS18804 × 76T1#23) had a leaf length not significantly different from their cultivar parents. All of the hybrids had significantly wider leaves than their wild parents, but had significantly narrower leaves than their cultivar parents. All of the hybrids had a significantly more number of leaves than their wild parents and a less number of leaves than their cultivar parents except IS18804 × 76T1#23, which had leaf traits at par with its wild parent (Table 3). The average number of seeds per plant was in the range of 2249 (76T1#23) to 12,363 (H2-16 × WSV 387). Virtually, all of the hybrids had higher fecundity than both of their parents, but it was relatively lower in W5-20 × 76T1#23. Cultivar parents showed the highest juvenile survival and adult/growth fitness components. The hybrids also showed juvenile survival fitness components comparable with their cultivar parents, but had adult/growth fitness components intermediate between their cultivar and wild parents. They showed the highest fertility fitness component. The wild parents generally had the lowest juvenile survival and adult growth fitness components, but had a better reproductive fitness component than the cultivar parents. Generally, cultivar parents had lower fecundity than wild parents and hybrids. The hybrids showed more composite fitness (0.85) than their wild (0.69) and cultivar (0.63) parents (Table S2, available online).
Mid-parent heterosis of wild×crop hybrids
Three of the five hybrids, which had data, had significant mid-parent heterosis for germination (Table 4). None of the hybrids showed significant mid-parent heterosis for days to 50% seedling emergence and number of tillers per plant. Most of the hybrids exhibited earlier flowering (10–20%) than their mid-parents, but three of the seven hybrids had significant mid-parent heterosis. All of the hybrids except one had significantly high mid-parent heterosis for plant height. Three hybrids showed significant mid-parent heterosis for head length and two others had significant mid-parent heterosis for head width. Most of the hybrids had panicle shapes intermediate between their parents (for an example see Fig. 1). Four hybrids had significant mid-parent heterosis for leaf number, four hybrids for leaf width and one for leaf length. Four hybrids exhibited significantly higher mid-parent heterosis for number of seeds per panicle. None of the crosses showed consistently superior mid-parent heterosis for all the characters. However, two crosses, T1-1 × WSV 387 and IS18822 × WSV 387, exhibited significant mid-parent heterosis for most of the traits. On the contrary, one cross was not significantly different from the mid-parent for all the characters except for leaf width.
Table 4 Mid-parent heterosis (%) for different life-history traits measured in wild×crop hybrids

GERM, germination percentage; DTE, days to 50% seedling emergence; DTF, days to flowering; PHT, plant height; TILL, number of tillers per plant; HL, head length; HW, head width; LN, leaf number; LL, leaf length; LW, leaf width; NSPP, number of seeds per panicle; NA, not available (seed was not adequate).
* Significantly different (P< 0.05).
** Significantly different (P< 0.01).
Fig. 1 (colour online) Comparison of panicle shape differences among the wild (left) and cultivated (right) sorghum parents and their hybrid (middle).
Discussion
Fertility and relative fitness
Studying the fitness/adaptive value of crop–wild/weed hybrids is crucial to predicting the future fate of rare hybrids on the environment, as it determines the effectiveness of initial gene flow and its introgression into populations in cases where alien transgenes might be included in the hybrids and cause biosafety concerns (Snow et al., Reference Snow, Pilson, Rieseberg, Paulsen, Pleskac, Reagon, Wolf and Selbo2003; Groot et al., 2003; Song et al., Reference Song, Lu, Wang and Chen2004). The present study dealt with comparing crop–wild sorghum hybrids and their parents for various fitness-related traits (components): juvenile survival; adult growth; reproductive/fertility. These traits are believed to have a significant contribution to competition and reproductive potentials in crop–wild hybrids and thus are useful to estimate the adaptive value. Usually, the low fertility of early-generation hybrids has been considered to be the major hindrance for introgression to take place (e.g. Weis, Reference Weis, Poppy and Wilkinson2005). For instance, F1 hybrids from certain crop–wild crosses of rice showed reduced fertility (Ellstrand et al., Reference Ellstrand, Prentice and Hancock1999). However, during crossing in 2010, we did not observe any form of barrier for hybridization between the wild and cultivated sorghum populations.
In agreement with the present study, some empirical fitness assessment studies have been carried out in crop–wild hybrids in different species with or without transgenes, but the results were quite variable. Some field experiments showed that different fitness parameters exhibited by non-modified hybrids could be as high as, even higher than, those of the weedy parent (Arriola and Ellstrand, Reference Arriola and Ellstrand1997; Hauser et al., Reference Hauser, Shaw and Østergård1998). For instance, F1 fitness traits from crosses between wild and cultivated rice were generally high (Ellstrand et al., Reference Ellstrand, Prentice and Hancock1999). Di et al. (Reference Di, Stewart, Wei, Shen, Tanga and Ma2009) also reported an intermediate composite fitness and the lowest reproductive fitness of crop–wild hybrids formed between transgenic oilseed rape [Brassica napus (L.)] and wild brown mustard [B. juncea (L.) Czern & Coss.]. Moreover, Arriola and Ellstrand (Reference Arriola and Ellstrand1997) detected interspecific hybridization between the tetraploid (Sorghum halepense) and diploid (Sorghum bicolor) sorghum using allozyme progeny analysis, and found that the resulting hybrids showed no fitness differences from the weedy parent under field conditions. Muraya et al. (Reference Muraya, Geiger, Sagnard, Toure, Traore, Togola, de Villiers and Parzies2012) studied the fitness of three generations (F1, F2 and F3) resulting from wild × cultivated crosses of S. bicolor in Kenya, and found that some of the crosses were as fit as their wild parents, showed no penalties for adaptive traits and there were no consistent differences among the three generations.
Intraspecific gene flow and introgression occur more readily than interspecific ones since cross compatibility is high and improved offspring fitness is common (Jørgensen and Wilkinson, Reference Jørgensen, Wilkinson, Poppy and Wilkinson2005). To this end, cultivated sorghum and its closest wild relatives belong to the same biological species [S. bicolor (L.) Moench] and can freely hybridize (de Wet, Reference de Wet1978). Due to this reason, such penalties of fertility were not evident in the present study involving wild–crop sorghum hybrids. Moreover, all of the hybrids did not show reduced fitness components for the measured phenological and morphological traits. In some cases, even the hybrids exhibited significant mid-parent heterosis for some of these traits. Introgression of crop genes into the wild relatives is also likely to occur as the two forms often coexist in sympatric ranges and have overlapping flowering windows.
Survivorship and reproduction of early-generation hybrids in a field environment determines gene flow between crops and weedy relatives (Raybould and Cooper, Reference Raybould and Cooper2005). In the present study, the better fitness components of the hybrids at the juvenile and fertility (fecundity) stages than their wild parents may indicate that crop genes in the hybrids are working in two different directions to recombine the parental characters. First, unlike the wild parents, dormancy could be removed in the hybrids and second, unlike the crop parents, fecundity could be enhanced. This recombination could be beneficial to the survival, introgression and spread of crop genes (including transgenes) into the wild gene pool.
Mid-parent heterosis
Some levels of mid-parent heterosis were observed at all developmental stages in different wild × crop cross combinations, but were not evident for emergence and number of tillers per plant. Heterosis was negative for days to flowering, which is considered advantageous in crop improvement. This can also help the hybrids as an adaptation to shatter their seeds earlier during the season to escape selective removal by farmers. With the exception of a single hybrid, all had overlapping flowering times with their wild parents. This is in agreement with the results of Snow et al. (Reference Snow, Moran-Palma, Rieseberg, Wszelaki and Seiler1998) and will help the hybrids to backcross with their wild parents and may ensure introgression. Furthermore, while plants in the cultivated sorghum plots flowered in synchrony, the tillers in the wild plants continued flowering for more than 15 d. This behaviour of asynchronous and wider range of flowering in the wild sorghum is in agreement with Jenczewski et al. (Reference Jenczewski, Ronfort and Chèvre2003), and this opens overlapping windows with the cultivated sorghum for possible natural outcrossing.
Heterosis was also negative for leaf traits (number, length and width) because the domesticated parents have values approaching to a positive maximum for these traits. However, it was positive for germination, plant height, head length and width, and number of seeds per plant (fecundity). The number of seeds per panicle was lower in the hybrids and wild parents than the cultivars, but superiority of most hybrids in fecundity over their cultivar parents was attributed to the large number of productive tillers. Fecundity itself was considered to substitute fitness, and all other things being equal, plants producing more seeds reportedly make a greater proportional increase to the next generation (Hails and Raymond, Reference Hails, Raymond, den Nijs, Bartsch and Sweet2004). A similar heterosis pattern in crop–wild crosses has also been observed in sorghum in Kenya (Muraya et al., Reference Muraya, Geiger, Sagnard, Toure, Traore, Togola, de Villiers and Parzies2012) and in other crop species (Hauser et al., Reference Hauser, Bjorn, Magnussen, Shim, den Nijs, Bartsch and Sweet2004; Guadagnuolo et al., Reference Guadagnuolo, Clegg and Ellstrand2006). The observed heterosis in the present crosses could be degraded in subsequent generations. Nevertheless, based on two races of subsp. verticilliflorum (i.e. verticilliflorum and arundinaceum), Muraya et al. (Reference Muraya, Geiger, Sagnard, Toure, Traore, Togola, de Villiers and Parzies2012) studied the fitness of wild × cultivated sorghum in F1, F2 and F3 generations in Kenya and Mali and found no consistent differences among the three generations. However, they did not observe fitness differences in seed yield (fecundity) of all the generations.
Implications for survival and introgression of transgenes
Because of the fear that transgenes might increase the fitness and competitiveness of wild and weedy relatives and as the consequences of releasing transgenic crop plants cannot be predicted (Fredshavn and Poulsen, Reference Fredshavn and Poulsen1996), gene flow from transgenic crops to their wild relatives remains an area of concern (Stewart et al., Reference Stewart, Halfhill and Warwick2003). The economic and environmental impact of enhanced weediness, however, depends on how much fitness increases and under what environmental conditions (Ellstrand et al., Reference Ellstrand, Prentice and Hancock1999). The substantial degree of fitness components in the wild × crop sorghum F1 hybrids in the present experiment without the addition of alien transgene(s) into the cultivated parents may have implications for introgression and persistence of crop genes in the wild pool. However, the addition of transgenes may alter fitness and the competitiveness cost to either direction depending on the behaviour of the inserted gene and the environment. Unfortunately, the effect of the environment was not investigated in this study due to the lack of seeds for undertaking multi-location trials. Usually, for weed control purposes, transgenes that have fitness costs to the weedy species or tightly linked to a disadvantageous gene are considered to be more preferable for transfer (Andow and Zwahlen, Reference Andow and Zwahlen2006). So far, it has been suggested that hybridization between wild and cultivated parents may result in a novel combination of genes, which may enhance fitness relative to wild plants (Hauser et al., Reference Hauser, Shaw and Østergård1998; Kiær et al., Reference Kiær, Philipp, Jorgensen and Hauser2007), and the results of the present study also support this idea with evidence based on morphological observations.
Supplementary material
To view supplementary material for this article, please visit http://dx.doi.org/10.1017/S1479262113000129
Acknowledgements
This study was supported by a grant from the Biotechnology and Biodiversity Interface (BBI) programme of the United States Agency for International Development (USAID) to Prof. Allison A. Snow, Department of Evolution, Ecology and Organismal Biology, Ohio State University, Columbus, Ohio. We extend our sincere thanks to ICRISAT for providing the seeds of some of the wild sorghum accessions used for the study.