Cross compatibility in intraspecific and interspecific hybridization in yam (Dioscorea spp.)

Yam (Dioscorea spp.) is a staple crop for millions of people in the tropics and subtropics. Its genetic improvement through breeding is being challenged by pre-zygotic and post-zygotic cross-compatibility barriers within and among species. Studies dissecting hybridization barriers on yam for improving the crossability rates are limited. This study aimed to assess the cross-compatibility, which yielded fruit set, viable seeds and progeny plants in an extensive intraspecific and interspecific crossing combinations in a yam genetic improvement effort to understand the internal and exogenous factors influencing pollination success. Cross-compatability was analyzed at the individual genotype or family level using historical data from crossing blocks and seedling nurseries from 2010 to 2020 at the International Institute of Tropical Agriculture (IITA). The average crossability rate (ACR) was lower in interspecific crossing combinations (6.1%) than intraspecific ones (27.6%). The seed production efficiency (SPE) values were 1.1 and 9.3% for interspecific and intraspecific crosses, respectively. Weather conditions and pollinator's skills are the main contributors to the low success rate in the intraspecific cross combinations in yam breeding. At the same time, genetic distance and heterozygosity played little role. Interspecific cross barriers were both pre-zygotic and post-zygotic, resulting from the evolutionary divergence among the yam species. Dioscorea rotundata had higher interspecific cross-compatibility indices than D. alata. Distant parents produced intraspecific crossbred seeds with higher germination rates compared to closest parents (r = 0.21, p = 0.033). This work provided important insights into interspecific and intraspecific cross-compatibility in yam and suggested actions for improving hybridization practices in yam breeding programs.


Results
Weather conditions at IITA breeding sites. At the Ibadan station, the temperature was stable across years while the rainfall was erratic and significantly fluctuated from year to year (Fig. S1a). The year 2010 had the highest rainfall (1926.3 mm) while 2020 had the lowest (1074.0 mm). June (246 mm, 15.5 days) and September (245 mm, 17.5 days) were the wettest months of the year. During the yam flowering window, August (130 mm, 13.9 days) and November (31.9 mm, 3.3 days) were the driest months. The mean minimum and maximum temperatures were 22.4 and 31.6 °C, respectively. Ibadan received a mean annual rainfall of 1546.9 mm from 116 rainy days, mean sunshine of 5.8 h, and mean maximum and minimum relative humidity of 92.8 and 50.25%, respectively (Fig. S1b). As for Ibadan, Abuja station had stable temperatures across years (min. 21.12 -21.77 °C, max. 32.21 -33.48 °C) while there was high inter-annual rainfall variability (945 mm in 2015 -2105.88 mm in 2019) (Fig. S1c). At Abuja, the flowering window had well distributed monthly rainfall amounts (217 mm, 20 rainy days from June-October) with moderate temperatures (21.92 °C min, 29.93 °C max). August was the wettest month (276 mm, 25 rainy days), while November was the driest (11 mm, 5 days) within the yam flowering window. Abuja minimum and maximum relative humidity values were 48.84 and 90.97%, respectively. This station experienced a wind speed of 1.89 km h −1 and solar radiation of 14.94 MJ m −2 day −1 (Fig. S1d). The two stations were significantly different for maximum temperatures (t = − 4.62, p = 0.006), minimum temperatures (t = 6.54, p = 0.001), and wind speed (t = 7.90, p < 0.001). All other parameters showed no differences between the two stations.

Exogenous and genetic factors on yam cross-pollination success. Exogenous factors.
Year, location, and pollinator effects significantly influenced the supervised cross-pollination success in yam (Tables 3,  4, 5, Table S5). The cross-pollination success indices such as the ACR and SPE were consistently higher at the Ibadan site than the Abuja on both D. rotundata and D. alata crossing blocks ( Table 4). The highest ACR values, 27.7% on D. rotundata and 69.4% on D. alata crossing blocks were recorded in 2019. The lowest 10.7% on D. rotundata and 12.3% on D. alata crossing blocks were in 2017 and 2020, respectively ( Table 5).
The annual weather data did not fully explain the inter-annual variability in yam cross success. Instead, weekly variability in rainfall, temperature, relative humidity, sunshine, and the number of rainy days within the yam flowering window (July to November) significantly influenced the pollination outcomes in either the D. Table 2. Interspecific cross-compatibility indices among yam species (Dioscorea spp.) The first species in the pedigree served as the female parent in the interspecific cross while the second species provided the male parents, ACR = average cross-compatibility/crossability rate, SPE = seed production efficiency. Dioscorea bulbifera wild refers to individuals of the species from the forest/wild environments. SD = standard deviation. ACR and SPE are presented by means ± standard deviation.   Table S6). In August, a lower success rate was recorded in the D. rotundata crossing block, while this happened in the D. alata crossing block in November. Crosses made after 14 November had no fruit setting regardless of the cross-combinations in D. alata (Fig. 3). Technicians involved in yam hand pollination also played a significant role in the cross-pollination success rates (Table S5). Some technicians achieved up to 5 times more pollination success rates than others. When assessed on the same sets of parents and cross-combinations, the pollinator's efficiency ranged from 6.2-30.3% and 27.9-67.4% in D. alata and D. rotundata pollination blocks, respectively. Genetic relatedness and cross-pollination success. Genetic distance among D. alata parental clones ranged from 0.02 to 0.84 (Fig. 4a, Supplementary data 1). There were no significant correlations between the parental genetic distance and the ACR (r = 0.046, p = 0.52) and between the parental genetic distance and the SPE (r = 0.007, p = 0.92)( Fig. 4b and 4c).
The D. rotundata clones used in crosses had a genetic distance of 0.01-0.96 (Fig. 5a, Supplementary data 2). No relationship was established between parental genetic distances and the ACR (r = 0.09, p = 0.2) or the parental genetic distance and the SPE (r = 0.052, p = 0.46) for D. rotundata cross-combinations ( Fig. 5b and 5c).

Discussion
This study used 11-year historical data to investigate intra-and interspecific cross-compatibility and seed germination rates in yam. We ascertained the generally low cross-compatibility rates among and within yam species reported by several studies 2,7,12,16,23-26 . Our retrospective analysis in yam crossing blocks at the IITA breeding Among the climatic factors, weekly differences in rainfall, temperature, relative humidity, number of rainy days, and sunshine during the pollination window were the main contributors to success rates. Our results revealed that making crosses in August for D. rotundata and November for D. alata could result in low pollination success due to the suboptimal weather conditions. This suggests the need for supplemental irrigation in yam crossing blocks to reduce the water deficit's adverse effects on yam reproductive phases during these months. Our results supported those of Abraham and Nair 27 , which linked successful pollination in yam with high relative humidity and moderate atmospheric temperatures.
Examining technicians' performance on the same cross-combinations simultaneously, we found highly significant differences among technicians in making successful pollination (p < 0.001). The anther removal from the male flower and its transfer to the female stigma is delicate. It requires special skills that might be lacking in many short-term pollinator staff involved in yam hand pollination operations. Pollinator's skill and weather conditions partly explained the significantly lower pollination success recorded at the IITA Abuja site than at Ibadan. Abuja site was characterized by lower relative humidity and higher temperatures compared to Ibadan. Hence, selecting conducive crossing sites and enhancing technicians' skills are highly relevant to improve the cross-pollination success rates in yam breeding programs.
Genotype genetic relatedness had no significant effect on pollination success rates in both D. rotundata and D. alata intraspecific crosses. The high or low success rates were recorded in crosses irrespective of the genetic distance between parents. This wide genetic compatibility range implies that breeding programs could exploit desirable genetic variant gene transfer within closely related and distant parents within the yam species. Although admitting that inbreeding depression could occur with low genetic distances, Willi and Van Buskirk 28 found no single peak of optimal genetic outbreeding in outcrossing plants. Since the genetic distance had no significant effect on yam cross-compatibility rates, it might be wise to cross genetically distant parents to exploit heterosis in yam breeding. Past studies supported that tuber yield is positively correlated with heterozygosity in yam 16,20,21 . It is, however, noteworthy that outbreeding over more considerable genetic distances could lead to a decline in the clonal component of fitness, as warned by Willi and Van Buskirk 28 .
Although no significant effect of genetic relatedness and heterozygosity of crossed parents was established on cross-compatility indices, there was an influence of parent combinations on ACR and SPE for both major yam species. Using the same D. alata plant materials as in this study, genome-wide association studies (GWAS) showed that ACR could be controlled by loci on chromosomes 1, 6 and 17 12 . That study, therefore, opened an avenue for developing genomic tools for predicting hybridization success in yam breeding programs.
As previously suggested by Mondo et al. 7 , several other internal and exogenous factors seem to influence yam pollination successes and should be given attention in future studies. Factors such as the optimum pollination time, floral morphology, and biology, especially for anthesis detection, pollen parent viability, etc., should also be investigated to identify additional causes of low intraspecific pollination success in yam at IITA. To complement this study conclusions, we recommend a more refined experiment on ploidy status among and within species to elucidate its effects on intra and interspecific cross-pollination success.
Pre-breeding involving interspecific hybridizations is valuable for broadening the cultivated yam species' genetic base by introducing desirable genes. Past efforts of broadening the genetic bases of white (D. rotundata) and water (D. alata) yams involved six relative species at IITA. Those relative yam species served as valuable   (Table S9). Although some natural/spontaneous interspecific hybridizations between D. rotundata and its wild relatives has been reported in West Africa 22,29 , yam is only propaged using tubers by farmers and botanical seed is only used for breeding purposes. Spontaneous crosses would require parental reconstruction to identify the pollen parent before advancing progenies. Developing bi-parental crosses through hand pollination is thus more precise. Average crossability rates, through hand pollination, were highest when crossing D. rotundata with its wild relatives (D. burkilliana and D. hirtiflora). The same trend was observed on SPE as the highest values were on D. rotundata × D. burkilliana cross-combinations. This relatively higher cross-compatibility rate could be attributed to high genetic relatedness among these species, as reported by previous studies 22,25,[29][30][31] . However, interspecific crosses involving D. rotundata with D. cayenensis had generally low success rates due to expected differences in ploidy level. Diocorea cayenensis is mostly triploid with male sex flowers, while diploid plants dominate D. rotundata with dioecious flowering patterns 32,33 . It is noteworthy that successful interspecific crosses were dependent on the cross direction such that reciprocals seldom ensured satisfactory results. This demonstrated the presence of unilateral incompatibility in yam species translated by rejection, in one crossing direction, of pollen from one species by pistils of a related species 19,34 . Therefore, when designing an interspecific hybridization program, male and female parents should be selected carefully to optimize the pollination success rates.
Our results showed that yam species experience both pre-zygotic and post-zygotic compatibility barriers. The pre-zygotic barriers prevented successful fertilization even when viable pollen and receptive stigmatic parents were used. The pre-zygotic barriers seemed much stronger among yam species as the crossability rate (translated by fruit set) was rare in interspecific crosses. Luque et al. 19 associated such barriers to the degree of domestication. Besides, that pre-zygotic barrier, also referred to as pre-fertilization barrier, could be a result of the halfway arrest of the elongation of the pollen tube in the stylar canals 18 . On the other hand, post-zygotic barriers prevented embryo development and seed formation even though the fertilization was successful. Such post-zygotic barriers could be mainly attributed to differences in ploidy levels 19 or by degradation of hybrid embryos and/or endosperm in incongruous crossings 18 . For successful interspecific yam hybridizations, both barriers need to be overcome. Direct in vitro pollination is among the techniques for overcoming the pre-zygotic incompatibility barriers in yams 7 . Besides, hybrid genotypes acting as a bridge between related incompatible yam species could be developed 19 or to experiment the cut-style pollination technique which had been successful in other crops 18 . The post-zygotic barriers resulting in seedless fruits can be controlled by biotechnology techniques such as the embryo rescue to prevent post-zygotic abortions 17 . Such techniques have been fully developed and implemented by CIRAD-Guadeloupe, France, and could be transferred to other programs to exploit interspecific and interploidy crosses in yam breeding 20,21 . We have suggested a scheme illustrating how a successful interspecific hybridization can be designed in yam breeding programs (Fig. 8).
Dioscorea rotundata had higher seed germination under unsupervised open pollination than controlled hand pollination. Pollen donors of higher quality might have been involved in open pollination than those used in hand pollination 35 . This study showed also that distant parents produced highly viable crossbred seeds than closely related ones. This could be associated with the hybrid vigor of distant parents' progenies. Fischer et al. 35 argued that low genetic distance among crossed individuals reduces cross-compatibility and offspring fitness in species with a self-incompatibility system. This reduced offspring fitness could have translated into low seed germination and poor seedling vigor due to inbreeding depression. We also found seed germination rate improvement across years as a result of growth medium optimization at IITA since 2014. Most seedling nurseries have been using carbonized rice husk under proper management in screenhouse conditions 36 . Therefore, the seed germination rate is not only under genetic control but also influenced by growing media and conditions.

Conclusions
This study ascertained the generally-reported low crossability and seed setting rates among and within yam species. Low success rates in yam crossing blocks were mainly attributed to the suboptimal weather conditions and technicians' skill rather than the genetic relatedness and parents' heterozygosity levels, as previously assumed. However, the seed germination rate was highest for crossbred seeds from genetically distant parents. Many other factors seem to influence the yam pollination success and should be investigated in future studies.

Materials and methods
Plant materials and weather conditions. In this study, 11-year (2010-2020) historical data generated from the IITA Yam Breeding Unit's crossing blocks and seedling nurseries were explored. Information on intra-and interspecific cross-compatibility and seed germination rates was gathered irrespective of the breeding objectives. Eight yam species, sourced from IITA core collection, were used for inter-and intraspecific cross- Cross-compatibility analysis. Historical data on the number of flowers pollinated, and the corresponding fruit and seed sets for the different cross-combinations conducted in IITA yam crossing blocks from 2010 to 2020 were used to calculate the cross-compatibility indices: ACR, PHC, and SPE. The cross-compatibility rate of a cross was calculated using the following formula: The ACR for a parent was calculated as the sum of cross-compatibility rates in specific crosses divided by the number of cross-combinations involving that particular parent: www.nature.com/scientificreports/ The PHC for a parent was calculated as the number of times the cross-compatibility rate exceeded the species overall cross-compatibility divided by the number of cross-combinations in which that parental genotype was involved: The SPE for a cross was calculated as the number of viable seeds divided by six times (the expected number of seeds in a yam fruit is six) the number of pollinated flowers multiplied by 100: Analysis of genetic relatedness and heterozygosity level. To assess the influence of the genetic distance between crossed parents on ACR and SPE, Jaccard pairwise dissimilarity matrices were generated using philentropy package implemented in R 37 . This genetic relatedness analysis was performed using single nucleotide polymorphism (SNP) markers. This analysis involved 77 D. alata genotypes previously sequenced using Diversity Array Technology (DArT-Seq) 12,38 . SNP information of 302 D. rotundata genotypes was extracted from the genotyping by sequencing (GBS) studies reported in Bhattacharjee et al. 39 . The heterozygosity level of the different parental genotypes was then estimated using Tassel 5 40 . It is noteworthy that analyses on genetic relatedness and heterozygosity were only performed on the two economically-important yam species, D. alata and D. rotundata, for which sufficient molecular information was available. Unlike the estimation of cross-compatibility indices which included historical crossing data of all the genotypes (434 D. alata and 712 D. rotundata) maintained in the yam breeding at IITA, analyses of genetic relatedness and heterozygosity level only focused on popular clones (77 D. alata and 302 D. rotundata) for which both sequencing and crossing information were available.
Crossbred seed germination rate assessment. The crossbred seed germination rates were determined using sample data of 2014-2020. The seed germination rate was estimated by dividing the seedling stand count in nurseries by the number of seeds sown multiplied by 100: Seed germination rate was further associated with parents' genetic distance information to assess the causeeffect relationship.
Determination of weather parameters and pollinator effects on pollination success. We used ten weather parameters to assess their influence on the cross-pollination success in yam crossing blocks. The weather parameters assessed were rainfall (mm), evaporation (mm), wind speed (km h −1 ), solar radiation (MJ m −2 day −1 ), minimum and maximum temperatures (°C), minimum and maximum relative humidity (%), sunshine (h) and the number of rainy days (a day was considered rainy when rainfall amount exceeded 0.5 mm). The weather parameters records on a yearly, monthly, weekly, and daily basis were assessed using multiple regression analysis. Differences in meteorological parameters at the two crossing sites, Ibadan and Abuja, were assessed using a t-test after testing the data's normal distribution using the Shapiro-Wilk test.
The pollinator effect was assessed on selected 2020 D. alata and D. rotundata crossing blocks at the Ibadan station using the same cross-combinations, crossing time, and same technicians. A total of eight technicians were monitored per yam species (a separate record was filled for each pollinator without their awereness and data was analysed anonymously). It is noteworthy that the two yam species' flowering windows are spaced in time, June to early October for D. rotundata and the end of September to November for D. alata. Therefore, a different set of pollinators was used for each species since crossing activity is mostly performed by short-time staff.
Data analysis. Historical data on ACR, SPE, and seed germination rate were summarized by means and standard deviations. Pearson's correlation was estimated to assess the relationship between ACR and SPE with internal (genetic relatedness, heterozygosity level) and external (weather conditions) factors. Effects of the year, location, and cross-combination on cross-compatibility indices were determined using the analysis of variance (ANOVA). ANOVA was also used to assess differences in seed germination rates among crossbred seeds from interspecific, intraspecific and open-pollinations. Means were further separated by Fisher's least significant difference (LSD) test at 5% p-value threshold. The relationship of parents' genetic distance with crossbred seed germination rate was also assessed using Pearson's correlation analysis. Heatmap function implemented in package 'qgraph' 41 was used to estimate the genetic relationship among crossed parents. Results were presented as a heatmap with different color gradients based on genetic distance. Regression plots were then drawn to determine the level of significance and influence of genetic distance on ACR, SPE, and seed germination rate. Regression analysis was also used to assess the impact of parents' heterozygosity level on the ACR and PHC values. Pollinating technician's effect on yam pollination success was evaluated using ANOVA and Tukey HSD All-Pairwise Comparisons Test at 5% p-value threshold.