Combining High Yields and Blast Resistance in Rice ( Oryza spp . ) : A Screening under Upland and Lowland Conditions in Benin

The future security of the supply of rice for food in Africa depends on improving the level of local production to achieve self-sufficiency. In order to cope with the existing gap between production and actual demand, combining a high level of rice blast tolerance and a high-yield potential is necessary. The current study was conducted under upland and lowland conditions in Benin to gain insight into the performance of selected blast-resistant accessions along with some currently grown varieties. This study revealed a high phenotypic variability among these accessions. Furthermore, differences in the performance of these accessions under lowland and upland conditions were observed. Principal component analysis showed their grouping in three clusters. The analysis also demonstrated a high yield potential among the blast-resistant rice accessions whether they were Oryza sativa or O. glaberrima. Furthermore, there was a significant correlation between yield and both spikelet fertility and growth cycle duration. In conclusion, the present study identified promising rice accessions for future breeding. High phenotypic variability in combination with interesting traits can help to develop new resilient varieties. Finally, when the traits correlate with yield, they can be used as markers for an early screening method for identifying promising accessions at an early stage.


Introduction
Rice (Oryza spp.) is the most cultivated cereal crop after wheat, and a primary food consumed by millions of people worldwide. Rice provides the bulk of calories and a number of micronutrients (iron, zinc and ß-carotene) for many people in African developing countries [1]. Africa has an abundant supply of natural resources that can support a huge expansion in food, specifically rice production [2]. Indeed, rice can be grown under diverse environments e.g., dryland, rainfed wetland, deepwater and mangrove swamps, and irrigated wetland [2,3]. Africa harvests annually more than 12,503,331 ha of rice to feed many low-income households with limited access to food [4]. However, annual rice production only covers 62% of the actual needs, whereas the demand is growing faster than for any the present study was undertaken to evaluate the grain production of this selected subset under irrigated upland and lowland field ecologies. This will allow determination of whether these rice accessions have better grain yields than currently cultivated varieties. The study can also provide a wide range of accessions with interesting agronomic characteristics that can form the basis for future breeding programs in Benin. Furthermore, scientific knowledge generated might be ideally used to improve ultimately rice production in Africa.

Plant Materials
The germplasm of Oryza spp. included in this study is a subset of 42 rice accessions (5 O. sativa accessions and 37 O. glaberrima) originating from six West African countries (Appendix A Table A1). This subset of rice accessions has been derived from an entire African collection of 350 rice accessions based on geographical origins, pairwise genetic distances (revealed by 77 AFLP markers) and differential reactions to blast disease [31]. Several field blast resistance patterns were observed in this selected germplasm: 26 highly resistant accessions; 9 moderately resistant, 3 moderately susceptible and 4 susceptible.

Field Trials for Agronomic Evaluation
Two field experiments were conducted concurrently under upland and lowland conditions in Cotonou (Benin) at AfricaRice's experimental site during (June 2016-December 2016; 2 • 21 20 E, 6 • 26 54 N). Rainfall starts in mid-March and ends (with an average of 1100 to 1200 mm) in early November with a mid-season dry period from mid-July to mid-August.
The experimental design was a randomized complete block design (RCBD) with three replications. Forty-two rice accessions (5 O. sativa and 37 O. glaberrima) and seven controls (ARICA 4, ARICA 5, CG 14, NERICA 1, NERICA 2, NERICA 4 and Moroberekan), serving as reference of cultivated upland rice, were tested. Direct sowing of 121 seeds per accession was done in a plot measuring 2 × 2 m. Planting was done at a spacing of 20 × 20 cm between and within rows.
The lowland field experiment was also performed in a RCBD with three replications. But, only 37 of 42 rice accessions of the subset were tested (because of insufficient seed) along with the seven lowland refence controls (ARICA 1, ARICA 2, ARICA 3, IR 841, NERICAL 14,NERICAL 19 and TOG 5681). Pre-germinated seeds of each accession were transplanted 21 days after sowing into small plots of 1.60 × 1.60 m at a spacing of 20 × 20 cm between and within rows.
A chemical treatment with mancozeb (80 g/15 L) and deltamethrin (Decis®, 40 mL/15 L) was performed to protect the plants for diseases and pests. The plots were weeded regularly to minimize weed infestation. A pre-planting base application of 200 kg ha −1 of NPK (15-15-15) was done followed by a total of 100 kg ha −1 of urea at panicle initiation (35 kg ha −1 ) and booting stages (65 kg ha −1 ), respectively.

Data Collection
The following 15 agronomic traits were evaluated in both field experiments (lowland and upland): total number of tillers (Tillers_Total), number of fertile tillers (Tillers_Fertile), percentage of fertile tillers (%Fertile_Tillers), panicle length (Length_Pan), plant height at maturity (Plant_height), total number of spikelets (Spikelets_TotalNum), number of filled spikelets (Filled_Spikelets), percentage of filled spikelets (%Fertile_Spikelets) number of primary branching (Ram_Iaire), number of secondary branching (Ram_IIaire), ratio secondary branching/primary branching (Ratio_RamIIRamIaire), total number of panicles per square meter (Panicules_Num), number of days to 80% flowering (CSE), number of days to 80% maturity (CSM) and grain yield (Yield). A number of plants in the middle of the inner two rows of each elementary plot were considered for the data collection using the Standard Evaluation System for rice and wild (IRRI, 2007). The list of data collected, and methodology used were presented in Appendix A Table A2. At 80% crop maturity stage, a quadrat of 1 m × 1 m size was measured and all the plants in each quadrate of each plot were harvested to estimate grain production of rice accessions.

Statistical Analysis
An analysis of variance (ANOVA) was conducted to gain insights into the effect of genotype on several phenotypic traits. Correlations analyses were performed assess the relationship between variables. Principal component analysis (PCA) was used to identify phenotypic traits and use these to identify superior accessions and similarities between accessions. Additionally, principal component regression was adopted to predict the yield based on the linear combinations (PCAs) of the phenotypic traits. A t-test test was performed to compare lowland and upland grain yield characteristics of rice accessions.

Phenotypic Variability for 15 Agronomic Measured Traits
The results of the ANOVA analyses of the 15 agronomic traits evaluated in the lowland and upland ecologies are presented in Table 1. Significant variations in some characteristics were observed between replications for lowland conditions (total number of tillers, number of fertile tillers, plant height at maturity and spikelet fertility) and upland conditions (spikelet fertility). Accessions performances in the lowland significantly differed from the upland for seven traits ( Table 2). The data for the percentage of fertile tillers, total number of spikelets, spikelet fertility, panicle secondary branching, days to 80% heading, days to 80% maturity, and grain yield were the major discriminants between lowland and upland ecologies. A correlation matrix (Table 3) was constructed with lowland data, showing that the number of days to 80% flowering was positively and significantly associated with the number of days to 80% maturity (R = 0.94, P = 0.0001) but, negatively associated with spikelet fertility. The total number of spikelets was positively correlated with secondary branching (R = 0.61, P = 0.004). In upland conditions, a positive association was found between grain yield and spikelet fertility (R = 0.57, P = 0.0001) and both were significantly negatively correlated with the number of days to 80% flowering and maturity. Moreover, the secondary branching had a positive correlation with the total number of spikelets (R = 0.69, P = 0.0001) ( Table 3).

Performance Evaluation of 42 Rice Accessions and 7 Reference Varieties for Yield and Yield Components in Upland Conditions
To reduce data dimensions for a better description of the relationships between accessions, PCA was performed using prior seven identified traits that contributed most to the phenotypic variation. The first two principal components explained 68.82% of phenotypic variability within the 42 rice accessions and 7 reference varieties. The trait contribution revealed by both principal components is presented in Table 4. PCA 1 showed a positive association with yield (0.42) and the spikelet fertility (0.47) whereas the number of days to 80% flowering (−0.53) and the number of days to 80% maturity (−0.52) were negatively linked with PC 1. Table 1. Mean sum of squares for the effect of "Accession (Access.)", "Replication (Rep.) and the residuals (Error) for the 15 agronomic traits for the experiments under lowland and upland conditions. ***, ** or * indicate a significant effect of accession and/or replication on a certain trait at a significance level of α = 0.001, α = 0.01 and α = 0.05, respectively.   Panicle secondary branching (0.69) and total number of spikelets (0.69) were positively correlated with PCA 2. A two-dimensional scatter plot involving all the 42 rice accessions and the 7 controls is presented in Figure 1. Three accessions groups were clearly separated with reference to PCA 1 and PCA 2. The majority of the accessions in cluster 1 were characterized by a short growing cycle and had high grain yields, whereas accessions in cluster 2 were low-yielding and had a long cycle duration. Furthermore, accessions in cluster 1 were characterized by a higher spikelet fertility, total number of spikelets, and secondary panicle branching compared to accessions in cluster 2. Accessions in cluster 3 showed intermediate agronomic performance. Accessions in cluster 2, WAB0032230, WAB0002093, WAB0032394 and WAB0015772 were particularly of a very long growth cycle (146, 151, 170 and 172 days to 80% maturity) and low grain yield (50.62, 98.18, 51.06 and 257.30 t/ha), respectively. Three highly resistant O. glaberrima accessions, namely WAB0002143 and WAB0029182 from cluster 1 and WAB0029194 (cluster 3) out-yielded all the seven reference controls with yields of 540, 573, and 603 t/ha, respectively. Two highly resistant O. sativa accessions, namely WAB0035059 and WAB0035038 from cluster 1 out-yielded all the seven reference controls used in the upland with yields of 669 and 717 t/ha, respectively.

Performance Evaluation of 37 Rice Accessions and 7 Reference Varieties for Yield and Yield Components in Lowland Conditions
A similar PCA analysis was conducted to examine the relationship between rice accessions in lowland ecology. The number of days to 80% heading, and days to 80% maturity were positively correlated with PCA 1 (0.63), whereas a significant negative correlation was found with spikelet fertility (−0.40). The total number of spikelets (0.62) and panicle secondary branching (0.70) were positively correlated with PCA 2, whereas grain yield (−0.58) and percentage of fertile tillers (−0.74) were negatively correlated with PCA 3 (Table 4). The total of 37 rice accessions and 7 reference controls was split into three main clusters relatively to the two first principal components ( Figure 2). The two principal components (PCA 1 and PCA 2) accounted for 57.64% of the total variation among studied germplasm. The first group, cluster 1 A number of eight rice accessions, including WAB0029182 (highly resistant) and WAB0030263 (highly susceptible), matured earlier than all the controls used (less than 108 days). Accession WAB0030263 was found to be the earliest maturating rice of all (94 days).

Grain Yield Performance of the Rice Accessions in Lowland and Upland Agro-Ecology
Grain yield scores were used as selection index to rank accessions from the most to the least important. In the upland, WAB0035059 showed the highest index followed by WAB0029194 and WAB0029182, while the lowest indexes were observed for WAB0032394, WAB0002093 and WAB0024105. In the lowland, WAB0008956 and WAB0029342 possessed the highest selection indexes, whereas the lowest values were scored in WAB0002136 and WAB0030263 ( Table 5). The blast resistant accession, WAB0035055, possessed selection indexes of 4 and 5 in lowland and upland conditions, respectively and might be recommended for farmers' cultivation in both ecologies. Nearly high similar indexes were observed in a susceptible accession, WAB0008937, that can be used for rice yield breeding in Benin. Accessions, WAB0030263 versus WAB0006684, WAB0015772 versus WAB0029315 and WAB0029323 versus WAB0026176 showed a comparatively equal performance in both environments (marked as circle in Appendix A Figure A1). Cluster analysis based on the principal components enabled the identification of cluster 1 that comprised of the majority of lowland and upland reference controls included. Accessions in cluster 1 had similar characteristics of the controls used and were found to be promising adapted rice to the lowland or upland growing conditions in Benin. There were no upland controls in cluster 2 suggesting low-yielding rice accessions that were not adapted to an upland environment. An ANOVA analysis with a post-hoc Tukey honest significant difference (HSD) test was performed to compare the grain yield and number of days to maturity observed between lowland and upland. Results are presented in Table 5. On the

Grain Yield Performance of the Rice Accessions in Lowland and Upland Agro-Ecology
Grain yield scores were used as selection index to rank accessions from the most to the least important. In the upland, WAB0035059 showed the highest index followed by WAB0029194 and WAB0029182, while the lowest indexes were observed for WAB0032394, WAB0002093 and WAB0024105. In the lowland, WAB0008956 and WAB0029342 possessed the highest selection indexes, whereas the lowest values were scored in WAB0002136 and WAB0030263 ( Table 5). The blast resistant accession, WAB0035055, possessed selection indexes of 4 and 5 in lowland and upland conditions, respectively and might be recommended for farmers' cultivation in both ecologies. Nearly high similar indexes were observed in a susceptible accession, WAB0008937, that can be used for rice yield breeding in Benin. Accessions, WAB0030263 versus WAB0006684, WAB0015772 versus WAB0029315 and WAB0029323 versus WAB0026176 showed a comparatively equal performance in both environments (marked as circle in Appendix A Figure A1). Cluster analysis based on the principal components enabled the identification of cluster 1 that comprised of the majority of lowland and upland reference controls included. Accessions in cluster 1 had similar characteristics of the controls used and were found to be promising adapted rice to the lowland or upland growing conditions in Benin. There were no upland controls in cluster 2 suggesting low-yielding rice accessions that were not adapted to an upland environment. An ANOVA analysis with a post-hoc Tukey honest significant difference (HSD) test was performed to compare the grain yield and number of days to maturity observed between lowland and upland. Results are presented in Table 5. On the basis of the ranking scores and the significant differences detected in grain yield and cycle duration, 13 and 6 rice accessions performed better in lowland and upland conditions in Benin, respectively. Nineteen rice accessions were relatively stable in both ecologies and might be suggested for farming in Benin.

Relationships of Yield Components with Blast Resistance and Genetic Population Diversity
A correlation analysis was performed to assess the relationship between yield-related traits and blast-resistance patterns. Results revealed that resistant accessions tend to produce a relatively higher number of secondary branching compared to susceptible ones in the lowland (R = 0.52) and the upland (R = 0.44) growing conditions. However, there was no direct association between blast resistance and grain yield in the lowland (R = 0.08) and the upland (R = 0.26) respectively. There was no blast disease incidence because of fungicide application during both field experiments.
The presence of three genetically distinct populations (population 1, population 2 and population 3) was revealed in the current subset germplasm using 20 SSR polymorphic markers [17]. The population genetic structure was significantly correlated with secondary panicle branching (R = 0.67) and primary panicle branching (R = 0.41) of rice accessions in upland growing conditions. Accessions in population 1 produced a significantly higher number of secondary panicle branches than those in population 2 and population 3. Accessions in population 3 produced a significantly higher number of primary panicle branches than those in population 1 and population 2.
In the lowland, we found significant correlations between population genetic structure and the following agronomic traits: total number of tillers (R = 0.41), percentage of fertile tillers (R = 0.45), secondary panicle branching (R = 0.70) and ratio of secondary branching to primary branching (R = 0.73). The majority of accessions in population 2 and population 3 tend to develop a better tiller ability, whereas accessions in population 1 showed a higher number of secondary panicle branching and higher ratio of the secondary branching to primary branching.

Discussion
Significant differences were observed between rice accessions for the 15 agronomic traits evaluated under lowland and upland ecologies. This attests to the existence of high genetic variability in the studied rice germplasm [32]. This variability can be exploited for further yield improvement of rice. In fact, the presence of high variable germplasm can help plant breeders to properly select parental lines to use in breeding programs [33].
Furthermore, it was seen that rice accessions responded differently across the three repetitions in the upland for spikelets fertility and in the lowland, for the total number of tillers, number of fertile tillers, plant height and spikelets fertility. Plant height and tillers are in general sensitive to environmental conditions (water level), especially when there is standing water, as experienced during the experiment in the lowland. Although there was no significant difference in grain yield between repetitions, the observed changes in spikelets fertility in both ecologies could be due to bird damage. Rice crop is highly susceptible to bird damage during grain maturation stages (milk to hard-dough stages) [34,35]. The performances of accessions differed significantly for seven agronomic traits between upland and lowland. Both experiments were simultaneously conducted on AfricaRice's site in similar physico-chemical conditions. This indicates that the significant differences observed between the lowland and the upland might be more attributed to the hydrological conditions [36]. This subset of seven traits identified should thus be given a greater priority for the selection of suitable lowland and upland rice accessions in Benin.
Concerning the relationship between yield and other traits, it was concluded that secondary panicle branching and total number of spikelets were strongly correlated and could have contributed to the grain yield performance of the accessions under lowland conditions. Zhao et al. (2016) have recently identified a SNP locus (G/C) that substantially affects both the total number of spikelets per panicle and the number of primary and secondary branches in some high-yielding japonica rice varieties [37]. According to Ashikari et al. (2005) grain yield is mostly determined by total number of spikelets per panicle [38]. The upland experiment revealed that spikelet fertility and growth cycle duration (the number of days to 80% flowering and 80% maturity) were the most important grain yield contributors. The shorter the cycle duration, the more spikelets were fertile and the higher the grain yield under upland conditions. Mokuwa et al. (2013) have also pointed out a negative relationship between grain yield and number of days to 50% flowering in O. glaberrima and O. sativa accessions [39]. A short rice growing season generally contributes more to stable harvest indexes than a late growing season especially under less favorable conditions [40]. Several authors [39,41,42] have suggested that grain yield and maturity duration are the most important characteristics used by farmers to select varieties. A recent participatory ethnobotanical survey indicated that Beninese farmers give a particular emphasis to high-productive and early-maturing varieties for selecting varieties [24]. Three O. glaberrima rice accessions (WAB0008956, WAB0029342 and WAB0015043) showed a higher grain productivity than O. sativa rice accessions in the lowland. Previous work demonstrated that O. glaberrima is the species mostly adapted to African adverse environmental conditions compared to O. sativa [9,39,[43][44][45]. But a significant reduction in grain yield is usually observed in this species due to the grain shattering and susceptibility to lodging [36,45]. The present study demonstrated that some O. glaberrima accessions clustered in cluster 1 (lowland and upland) achieved a good performance for the total number of spikelets and panicle secondary branching, although O. glaberrima is described as generally low compared to that of O. sativa [43].
Long-held assumptions about O. glaberrima agronomic traits need to be reconsidered for a better valorization of this rice species. Better agronomic performance was particularly observed in previously identified blast-resistant rice accessions compared to the varieties currently grown in Benin. This means that in case severe attacks would occur in the field, these accessions would yield better than the commonly used susceptible varieties. In fact, the findings of  are among the first giving any insight on the resistance/susceptibility of each rice accession across two different environments in Benin. The study clearly demonstrated that rice germplasm exhibiting high blast disease resistance is potentially resistant to all isolates/races of the pathogen prevalent in those two environments [31]. Odjo et al. (2011) reported areas in country where severe blast attacks frequently occur (hotspots) that can be recommended for germplasm evaluation [46]. Higher secondary panicle branching was positively correlated with blast resistance in the lowland (R = 0.27) and the upland (R = 0.19). Association of secondary branching and blast resistance also suggests there are more opportunities for potential rice breeding.
Since blast disease is the most harmful biotic threat to Beninese rice production, the combination of high-yield potential of rice accessions with their resistance to the disease can help minimize yield losses and thus reduce chemical pesticide applications [47,48]. Genetic structure analysis of the 42 selected rice accessions revealed the presence of three genetically distinct populations with a significant level of gene flow between O. sativa and O. glaberrima accessions across the population 1 [49]. Valuable information on the relationship between population structure and agronomic characteristics was highlighted in this paper and can be integrated in breeding for attaining higher yield potentials [50]. It was also shown in this study that nearly all O. glaberrima accessions in population 1 yielded at least 4 tons ha −1 in lowland ecology except WAB0030263. However, WAB0030263 was the most early maturing accession out of the total number of germplasms studied, which is one of the farmers' preference for rice traits. Gene flow might then strongly contribute to these yield-related traits of the accessions in population 1.

Conclusions
This study revealed a high phenotypic variability among the screened rice accessions, which is highly valuable for breeding. The differential performance in upland and lowland conditions for several traits indicates that these traits are substantially influenced by environmental factors. In addition, significant correlations between yield and several phenotypic traits were observed, which are important and can be used as markers for early screening for identifying promising high-yielding varieties. Results of the present study highlight the potential of the identified core selection of the African rice germplasm collection for developing new blast-resilient rice varieties in general and especially for Beninese growing conditions. The genetic relationship between agronomic traits associated with blast resistance and genetic structure shown for these rice accessions will globally help breeders improve rice productivity and especially for Benin. Multi-year trials at multiple locations are required, however, to ensure the performance of the varieties.