The authors have declared that no competing interests exist.
Conceived and designed the experiments: WM PAR JDE TJC ID SM OB NC. Performed the experiments: WM JDE ID SM NC OB NND. Analyzed the data: WM PAR. Wrote the paper: WM PAR JDE TJC.
Current address: Plant Systems Biology, Biosciences Division, Oak Ridge National Laboratory, Oak Ridge, Tennessee, United States of America
Current address: Generation Challenge Program, International Maize and Wheat Improvement Center, Tescoco, Mexico
Current address: Bill and Melinda Gates Foundation, Seattle, Washington, United States of America
The stay-green phenomenon is a key plant trait with wide usage in managing crop production under limited water conditions. This trait enhances delayed senescence, biomass, and grain yield under drought stress. In this study we sought to identify QTLs in cowpea (
Delayed senescence, grain yield, and biomass yield under drought stress rank among the most important traits targeted for improvement in crop plants
The recent advent of cost-effective high-throughput genotyping technologies which enable construction of dense genetic maps and rapid genotyping of large germplasm collections has facilitated genome-level understanding of these traits and is helping to revolutionize marker-assisted breeding
The stay-green trait, a phenomenon where QTLs for delayed senescence, biomass, and grain yield are co-located has been described in several crop species and is considered to be an important trait for improvement of yield under drought stress for cereals such as sorghum and maize
In cowpea, an important legume crop for subsistence farmers of semi-arid regions of sub-Saharan Africa, South America, and Asia, improvement of grain yield under drought stress is a key objective in breeding programs. Gwathmey et al.
Despite the critical role of cowpea in sustaining the food systems of many parts of the developing world, until recently very limited genomic resources have been developed. However, a high-throughput SNP genotyping platform has been implemented for cowpea, enabling the construction of a high density consensus genetic linkage map
In this study, a 1536-multiplex SNP genotyping platform was applied to cowpea germplasm originating from diverse geographical locations and breeding programs together with the consensus genetic map developed from six constituent RIL populations to (i) determine the extent of linkage disequilibrium in cowpea and (ii) detect QTL related to early and late-season delayed senescence, grain and biomass yield. Further, we compared map locations for QTL identified previously for seedling drought tolerance
Based on the Illumina GoldenGate assay 1,080 SNP markers had robust performance across the entire panel of 383 genotypes. Of these 1,054 SNPs had call rates ≥99% and only 14 SNPs had missing data between 5 and 10%. From the collection of diverse genotypes evaluated for these SNPs, 98.5% had call rates ≥99%. The remaining genotypes had call rates between 91.4 and 96% for the 1,080 SNPs. Genotyping results for the RIL population were reported previously
For inter-chromosomal LD, 722 of the 281,731 pairwise correlations between markers exceeded the observed threshold of r2 = 0.25. This represented 0.26% of all possible correlations. Of these, only 30 exceeded r2 = 0.4 representing 0.01% of possible correlation and only two correlations were above r2 = 0.5 (r2 = 0.5001 and 0.561963). Therefore, 99.7% of r2 values for unlinked markers fell below the set threshold. These statistics excluded SNP marker 1_0069 that mapped at the 39.2 cM-locus on linkage group (VuLG11) of the consensus map. This marker had high r2 values with markers on VuLG3 (up to 0.70) but had negligible correlation with neighboring markers, which suggested a possible map error given that there was also >3 cM distance between the marker and adjacent markers on the map. Therefore, marker 1_0069 was excluded from all subsequent analyses.
With a few exceptions, intra-chromosomal r2 values for LGs 1, 5, 6, 8, 9, 10, and 11 decayed completely within 10 cM. LD for VuLG4 also decayed within 10 cM with the exception of 8 correlations that were above the 0.25 threshold up to distances of 50 cM. For VuLG7, LD decayed within 7 cM with the exception of 5 correlations that exceeded the threshold up to a distance of 17 cM. VuLGs 2 and 3 exhibited the highest level of LD persistence with 19 and 24 pairwise r2 values, respectively, exceeding the 0.25 threshold at distances greater than 10 cM. In total, 74 pairwise correlations had r2 values greater than the 0.25 threshold over 10 cM, representing 0.46% of all cases for all 11 LGs. The r2 values plotted against map distance for each linkage group are illustrated in
Based on LOESS curve and r2 threshold intercept, LD decayed between 0 and 2 cM for VuLGs 1, 2, 3, 4, 6, 8, and 10. For VuLGs 5, 7, 9, and 11 there was no intercept between the LOESS curve and the 0.25 threshold, suggesting that LD generally decayed within sub-cM distances for these four LGs (
Although the number of markers between16 and 612 generally resulted in the same profiles of mean ln likelihood values plotted against K, 187 markers were chosen for the population structure analysis based on optimal allocation of genotypes to individual subpopulations. Using the model and run parameters described in methods and plotting the resulting mean ln likelihood versus K values, there was no significant change in ln likelihood values beyond K = 24 (
In all, 41 separate observations were recorded from 13 experiments encompassing grain yield, grain yield components, biomass yield, seedling-stage senescence, early vegetative senescence and post-flowering senescence (
PFS | Grain yield | Biomass | Seed weight | |
Nigeria 2007 (number of genotypes = 339) | ||||
Grain yield | 0.85 |
|||
Biomass | 0.86 |
0.94 |
||
Pod weight | 0.87 |
0.98 |
0.94 |
|
Senegal 2008 (number of genotypes = 155) | ||||
Grain yield | 0.34 |
|||
Biomass | 0.34 |
0.05ns | ||
100-seed weight | −0.05ns | 0.13ns | 0.10ns | |
Burkina Faso (number of genotypes = 187) | ||||
Grain yield | 0.49 |
|||
Biomass | −0.26 |
0.11ns | ||
100-seed weight | 0.53 |
0.79 |
0.08ns | |
Pod number | 0.58 |
0.82 |
0.07ns | 0.84 |
ns = not significant at 0.05 level,
= p≤0.001,
= p≤0.0001. PFS = post-flowering senescence.
In the RIL population (IT93K-503-1 × CB46), 6 QTL for grain yield were identified across locations, reaching the more stringent 0.005 Kruskal-Wallis significance level in at least one experiment (
Vertical broken lines represent significance thresholds based on 1000 permutations at the 0.05 significance level. Delayed senescence LOD profile was based on reanalyzed data from a greenhouse experiment reported in Muchero et al. (2009a) study, grain yield data was derived from Senegal 2008 (red) and USA 2007B (magenta) experiments.
Vertical broken lines represent significance thresholds based on 1000 permutations at the 0.05 significance level.
QTL name | Consensus map position:LG (cM range) | IT93K-503-1 × CB46 mapposition: LG (cM range) | Kruskal-Wallis significance level | |||||||
Burkina Faso | Senegal | USA 2007A | USA 2007B | USA 2007C | USA 2008A | |||||
GY |
BY |
GY | BY | GY | GY | GY | GY | |||
1 (12.2 – 33.1) | 0.005 | 0.0005 | 0.05 | – | 0.05 | 0.01 | 0.1 | 0.001 | ||
3 (60.5–70.8) | 0.05 | – | 0.001 | – | 0.05 | – | – | 0.05 | ||
4 (34.3–59.3) | 0.0005 | 0.005 | 0.01 | – | 0.05 | 0.05 | 0.05 | 0.05 | ||
7 (8.8–34.5) | 0.05 | 0.001 | 0.001 | – | 0.005 | 0.01 | 0.05 | 0.005 | ||
8(33.7–48.6) | 0.001 | 0.05 | 0.0001 | – | 0.001 | 0.005 | 0.005 | – | ||
11 (2.6–43.0) | 0.05 | – | 0.005 | 0.05 | – | – | 0.05 | – |
Linkage groups (LG) in brackets correspond to the AFLP-only map (Muchero et al. 2009a); LGs in bold correspond to the SNP-only map (Muchero et a l. 2009b) and the cM range of the delayed senescence QTL remapped on the SNP-only map.
GY = Grain yield; BY = Biomass yield.
Significance thresholds (αG) calculated for the MLM analysis using the Simple
Trait_Experiment | Marker | LG | Position(cM) | P-value | r2 | QTL | ||
Seed weight per plant_BF 2008B | 1_0029 | 1 | 18.420 | 1.62E-04 | 0.050 | 5_39290868 (5.7e-29) | 15_1274309 (1.3e-8) | |
Pod weight per plant_BF 2009C | 1_0029 | 1 | 18.420 | 2.90E-03 | 0.018 | |||
100-seed weight_BF 2009A | 1_0029 | 1 | 18.420 | 4.70E-03 | 0.023 | |||
Biomass_BF 2009C | 1_0029 | 1 | 18.420 | 4.90E-03 | 0.026 | |||
Seed weight per plant_BF 2009C | 1_0029 | 1 | 18.420 | 5.90E-03 | 0.016 | |||
Biomass_BF 2008 | 1_0589 | 2 | 17.046 | 2.21E-04 | 0.049 | 7_47769161 (6.9e-47) | 10_32269550 (8.1e-43) | |
Grain yield_BF 2008 | 1_0589 | 2 | 17.046 | 1.40E-03 | 0.044 | |||
Senescence_Greenhouse 2 | 1_0067 | 3 | 10.556 | 1.59E-04 | 0.057 | 2_43425949 (2.4e-46) | 13_41927008 (2.6e-36) | |
Senescence_Greenhouse 1 | 1_0067 | 3 | 10.556 | 5.00E-03 | 0.019 | |||
Senescence_Greenhouse 1 | 1_0206 | 4 | 0.009 | 7.71E-05 | 0.039 | 5_516531 (1.5e-42) | 4_12399615 (1.8e-38) | |
Biomass_Senegal 2008 | 1_0296 | 4 | 0.009 | 4.10E-03 | 0.038 | 3_43338952 (5.7e-48) | 5_3321676 (5.4e-45) | |
Seed weight per plant_BF2009A | 1_0888 | 4 | 0.455 | 4.04E-04 | 0.056 | 5_794965 (1.0e-6) | 5_33610435 (4.19e-2) | |
Seed weight per plant_BF2009C | 1_0888 | 4 | 0.455 | 8.42E-04 | 0.052 | |||
Pod weight per plant_BF2009C | 1_0888 | 4 | 0.455 | 2.10E-03 | 0.019 | |||
Grain yield_USA 2007C | 1_0049 | 5 | 9.456 | 1.92E-04 | 0.095 | 1_859379 (1.3e-43) | 17_41145050 (2.6e-36) | |
Grain yield_USA 2007A | 1_0049 | 5 | 9.456 | 5.70E-03 | 0.059 | |||
Senescence_BF 2008 | 1_0108 | 7 | 15.844 | 6.76E-04 | 0.023 | 2_7436099 (2.1e-15) | 9_39304124 (2.3e-5) | |
Biomass_BF 2008 | 1_1150 | 7 | 15.844 | 3.60E-03 | 0.033 | 2_7409687 (2.1e-34) | 2_3284032 (4.5e-8) | |
Grain yield_BF 2008 | 1_1150 | 7 | 15.844 | 7.90E-03 | 0.031 | |||
Senescence_Greenhouse 1 | 1_0279 | 7 | 18.213 | 2.70E-03 | 0.019 | 2_27942808 (1.2e-37) | 9_39955414 (1.7e-32) | |
Seed weight per plant_BF 2009B | 1_0279 | 7 | 18.213 | 7.30E-03 | 0.025 | |||
Senescence_USA 2009 | 1_0983 | 7 | 19.081 | 9.12E-05 | 0.070 | 2_27089201 (6.5e-41) | 9_40434989 (1.2e-21) | |
Senescence_Greenhouse 1 | 1_0983 | 7 | 19.081 | 2.40E-03 | 0.022 | |||
Grain yield_Nigeria 2007 | 1_0140 | 10 | 34.896 | 4.52E-04 | 0.015 | 10_30910809 (3.1e-51) | 1_46091028 (2.2e-37) | |
Senescence_Nigeria 2007 | 1_0140 | 10 | 34.896 | 9.00E-03 | 0.008 | |||
Seed number per pod_BF 2009C | 1_0759 | 10 | 35.578 | 8.20E-03 | 0.033 | 10_29466207 (1.3e-43) | 1_45401915 (7.6e-18) | |
Grain yield_Nigeria 2007 | 1_0759 | 10 | 35.578 | 9.40E-03 | 0.009 | |||
Senescence_USA 2009 | 1_1405 | 10 | 37.373 | 2.14E-04 | 0.063 | 10_16853286 (3.4e-38) | 1_44037621 (2.3e-24) | |
Pod number per plant_BF 2009A | 1_1405 | 10 | 37.373 | 5.70E-03 | 0.033 |
At the phenotypic level, pleiotropy between the three traits (stay-green phenomenon) was suggested by significant correlations among mean phenotypic values for delayed senescence, grain yield, grain yield components, and biomass yield in experiments conducted in Burkina Faso, Nigeria, and Senegal (
Allele effects were scaled on a percent basis relative to the highest value for each phenotypic parameter.
Significant positive correlations of allele effects obtained from the MLM analysis for each of the 794 mapped and 62 unmapped SNPs suggested that the predominant pleiotropic effect was positive in which delayed senescence conferred higher biomass and grain yield, although a single instance of significant negative correlation was observed between delayed senescence and biomass in Burkina Faso (
Fourteen ESTs harboring SNPs with phenotypic associations had significant BLAST matches in common bean and (or) soybean genome assemblies (
The general decay of LD within distances less than or equal to 2 cM in cowpea suggested that genome-wide association mapping is practical in cowpea with the potential to provide QTL resolution up to 10× greater than bi-parental QTL mapping. Given the total length of the cowpea genetic map distance (680 cM) and the conservative 1 cM distance for complete LD decay, the cowpea consensus map should provide sufficient coverage for QTL detection. Therefore, the cowpea consensus map and its recently improved version
Using the genome-wide scan with 856 SNP markers, we were able to identify seven loci with significant marker-trait associations. Of these, five were associated with more than one trait and three of them mapped in QTL intervals that were associated with delayed senescence at the seedling stage as well grain and biomass yield in the RIL population. The co-location of QTLs for at least two of the three traits is suggestive of common genetic determinants as well as the existence of the pleiotropic stay-green phenomenon in cowpea. This was further supported by positively correlated allele effects, and largely correlated mean phenotypic values for delayed senescence, grain yield, grain yield components and biomass yield in experiments conducted in Burkina Faso, Nigeria, and Senegal. However, there was also evidence suggesting genotype × environment interaction for this trait with the Senegal location exhibiting the lowest levels of correlation between delayed senescence, biomass, and grain yield. These findings suggest the importance of the stay-green trait in cowpea as an adaptive trait in drought-prone environments. The stay-green trait must have been favored in the arid and semi-arid environments of sub-Saharan Africa where cowpea was domesticated and has been cultivated for thousands of years. The coupling of mechanisms conferring drought tolerance, grain yield, and biomass yield would confer fitness while minimizing energy usage by utilizing the same genetic pathways for survival and productivity.
Further, the fact that these QTLs could be identified across different agro-ecological environments in both West Africa and North America indicates that these loci represent viable targets for the genetic improvement of cowpea production in arid and semi-arid regions. The apparent positive pleiotropy observed here suggests that genomic selection-based approaches to introgress more than one locus at a time
While recognizing that the ‘stay-green’ trait does not encompass all potential drought tolerance mechanisms contributing to drought adaptation in cowpea, this study demonstrated a robust association of expression of the stay-green trait with grain yield and biomass yield under drought stress. These findings indicate that assessments of delayed senescence could be useful for indirect selection of grain yield and biomass yield under drought, especially for cost-effective preliminary assessments of very large samples of germplasm or early generation screening of breeding lines in breeding programs targeting improved drought tolerance. Such a cost-saving approach would be important for breeding programs where the time and resource requirements for grain yield phenotyping in drought environments are currently high relative to seedling or early vegetative stage high-throughput phenotyping assays. These findings in cowpea are distinct from crops such as sorghum, maize, and rice where phenotyping for this trait has only been done at the post-flowering stage under field conditions
From practical and basic genetics considerations, cowpea offers an important opportunity for the dissection of the stay-green trait in legume crops and perhaps more broadly as a model for indeterminate, non-cereal crops. With its relatively simple, small, diploid genome and short generation time, rapid advances can be made in understanding this trait including synteny and translational genomics-based analysis relevant to other important crops. Notably, the close evolutionary relationship of cowpea to soybean and common bean and the demonstrated high level of synteny between their genomes
Our results also showed that a RIL population as small as 48 lines together with a diverse germplasm set of 96 genotypes provided enough resolution to detect the co-location of delayed senescence, grain yield and (or) biomass yield QTLs. This finding corroborates previous results where we demonstrated that subsets of 57 and 70 RILs provided nearly the same mapping resolution for the delayed senescence trait as populations of 124 and 127 RILs
In conclusion, isolating and characterizing the genetic determinants of the stay-green trait hold tremendous opportunity for the improvement of crop productivity in light of contemporary challenges that include global warming, rapidly increasing population, the need to reduce inorganic nitrogen applications in farming systems, and increasingly constrained agricultural water resources, as well as the need to develop sustainable renewable energy sources
A selection of 383 inbred cowpea genotypes (
Each of the 383 genotypes was SNP-genotyped using the Illumina GoldenGate 1536-SNP assay and data for allele calls at each SNP processed as described by Muchero et al.
LD decay along linkage groups was analyzed using the Graphical Genotyping (GGT2.0) software
Population structure was evaluated using the software STRUCTURE 2.1
The delayed drought-induced senescence trait (early vegetative delayed senescence) was evaluated on 205 genotypes in two greenhouse (Senescence_Greenhouse 1 and Senescence_Greenhouse 2) and one field (Senescence_USA 2009) experiments in the USA. Four replicates with one plant of each genotype were evaluated in greenhouse experiments. Four replicates of 1-meter plots with ten plants each were used to evaluate genotypes under field conditions. In greenhouse experiments, ratings were taken for visual-based senescence and maintenance of stem greenness on 2–6 week-old cowpea plants as described by Muchero et al.
Experiments in all locations were conducted under rainless or limited rain conditions to induce drought stress. Climatic conditions related to maximum and minimum temperatures, evapo-transpiration, and total rainfall during the course of each experiment are summarized in
Briefly, 96 cowpea genotypes and 57 RILs from the IT93K503-1 × CB46 population were evaluated for grain yield at the University of California-Riverside CVARS station under three treatments: fully irrigated (USA 2007A); irrigation withheld for 15 days from the day of planting (USA 2007B); and irrigation withheld for 30 days from the day of planting (USA 2007C). Normal irrigation cycle was resumed after stress treatment. In 2008, 201 diverse genotypes and 84 RILs were evaluated for grain yield at CVARS (USA 2008).
In 2007, 339 cowpea genotypes were planted at the International Institute of Tropical Agriculture (IITA) Kano station in Nigeria (Nigeria 2007), and evaluated for biomass, grain yield, 100-seed weight, and post-flowering delayed senescence. Three replicates of 3 m×75 cm plots were planted with 15 seeds of each genotype. Pre-planting irrigation was applied to soil saturation and two additional irrigation cycles were applied at 2 and 4 weeks after planting after which no additional irrigation was applied. A total of 4 mm rainfall was recorded over the course of the experiment (
In Burkina Faso, 201 genotypes and 49 RILs were planted at the Institute of Environmental and Agricultural Research (INERA) Kamboinse experiment station (BF 2008) and were evaluated for biomass, grain yield, and delayed senescence. The same set of material excluding the RIL population was evaluated in Burkina Faso in 3 additional experiments, BF 2009A and BF 2009B at the Pobe-Mengao experiment station, and BF_2009C at the Kamboinse experiment station. In each experiment, 2–2 m-plots replicated three times were planted with a total of 40 seeds for each genotype. Experiment BF_2008 in Kamboinse was conducted entirely under natural rainfall conditions with 123 mm and 28 mm being received for last two months of the experiment, September and October, respectively. Similarly, experiment BF_2009C was conducted under natural rainfall conditions. Planting followed a 260 mm rainfall event in late August with 98 mm being received in the last six weeks of the experiment For experiments BF_2009A and BF_2009B at Pobe-Mengao, planting was carried out after a 58-mm rainfall at the end of July for BF_2009A, and a 43-mm rainfall event for BF_2009B. A total of 324 mm of rainfall was received within 4 weeks from the beginning of August until the end of September, and 20 mm was received during the whole month of October. No additional rainfall or supplemental irrigation was received or applied for the remainder of the experiment. Delayed senescence was scored as described above and ten plants per replicate were harvested for biomass, grain yield, and grain yield components evaluation (
163 genotypes and 51 RILs were evaluated at the Senegalese Institute of Agricultural Research (ISRA-CNRA) Bambey station in Senegal (Senegal 2008). Three replicates, each with 2–4 m long rows were planted with a total of 68 seeds. Plots were planted after a 22.7-mm rainfall event that was followed by a total of 102.3 mm rainfall during the course of the experiment (
The Kruskal-Wallis and Multiple-Model QTL Mapping (MQM) packages of MapQTL 4.0
QTLs detected for the early vegetative delayed senescence trait in the RIL population had been described by Muchero et al.
TASSEL 2.1 software (
Allele effects were determined concurrently with marker-trait associations using the MLM analysis in the TASSEL 2.1 software.
To avoid spurious association due to LD persistence, LD was determined for all loci with significant association. Two or more QTLs were considered the same when (i) the markers next to the adjacent QTLs had r2 greater than the threshold or (ii) unmapped markers had high pair-wise correlations with mapped loci harboring a detected QTL. For non-adjacent loci as well as loci on different linkage groups, the locus showing the highest significance was accepted as the QTL peak.
Pleiotropic effects between delayed senescence, grain yield, and biomass yield were accepted only when (1) QTLs coincided on the same map position, (2) statistically significant Spearman rank correlation of p-values, and (3) statistically significant Spearman rank correlation of allele effect of each marker. For unmapped markers, co-location was accepted when the same marker gave significant associations with at least two of the three traits. Further, remapping the QTLs for delayed senescence at the seedling stage was used to assess the incidence of markers within these QTL regions that also showed statistically significant association with grain yield and biomass in the association mapping study. Co-location was visualized by plotting –log (p-value) against map distance along individual linkage groups of the cowpea consensus map for each of the three traits. To rule out spurious correlations, correlation analysis was carried out using MLM-derived p-values of association for the three traits and flowering from the Nigeria 2007 dataset.
The Simple
For the FDR method, the p-values from the MLM analysis were subjected to the FDR method based on a false discovery rate, q* = 0.20 as described by Benjamini and Yekutieli
Analysis of Variance (ANOVA) and Spearman rank correlations were calculated using Statistix 9 software
Cowpea ESTs harboring SNPs with significant associations to phenotypes were used in BLAST analysis to infer positions on the common bean and soybean reference genomes (
(PDF)
(TIF)
(XLSX)
(XLSX)
(XLSX)
(DOCX)
(DOCX)
The authors thank Dr. Dominique Dumet, Germplasm Curator, International Institute of Tropical Agriculture, for supplying many of the cowpea breeding lines used in the present study.