Root and shoot variation in relation to potential intermittent drought adaptation of Mesoamerican wild common bean (Phaseolus vulgaris L.)
Data files
Mar 13, 2020 version files 106.85 KB
-
Tube_Experiment_data.xlsx
Abstract
• Background Wild crop relatives have been potentially subjected to stresses on an evolutionary time scale prior to domestication. Among these stresses, drought is one of the main factors limiting crop productivity and its impact is likely to increase under current scenarios of global climate change. We sought to determine to what extent wild common bean (Phaseolus vulgaris) exhibited adaptation to drought stress, whether this potential adaptation is dependent on the climatic conditions of the location of origin of individual populations, and to what extent domesticated common bean reflects potential drought adaptation.
• Methods An extensive and diverse set of wild beans from across Mesoamerica, along with a set of reference Mesoamerican domesticated cultivars, were evaluated for root and shoot traits related to drought adaptation. A water deficit experiment was conducted by growing each genotype in a long transparent tube in greenhouse conditions so that root growth, in addition to shoot growth, could be monitored.
• Results Phenotypic and landscape genomic analyses, based on single-nucleotide polymorphisms, suggested that beans originating from central and north-west Mexico and Oaxaca, in the driest parts of their distribution, produced more biomass and were deeper-rooted. Nevertheless, deeper rooting was correlated with less root biomass production relative to total biomass. Compared with wild types, domesticated types showed a stronger reduction and delay in growth and development in response to drought stress. Specific genomic regions were associated with root depth, biomass productivity and drought response, some of which showed signals of selection and were previously related to productivity and drought tolerance.
• Conclusions The drought tolerance of wild beans consists in its stronger ability, compared with domesticated types, to continue growth in spite of water-limited conditions. This study is the first to relate bean response to drought to environment of origin for a diverse selection of wild beans. It provides information that needs to be corroborated in crosses between wild and domesticated beans to make it applicable to breeding programmes.
Methods
Experimental setup To observe root growth during the experiment, a screening method (Polanía et al., 2009) was used with some modifications in a climate-controlled greenhouse (19–30 °C diurnal range, ~1000 µmol photons m−2 s−1 PAR at midday) in the Orchard Park Greenhouse facilities at the University of California-Davis (Fig. 3A). We used transparent tubes of 7.6 cm diameter and 1.2 m length made from polyethylene terephthalate (Fig. 3B; www.cleartecpackaging.com). The tubes were filled with a 2:1 sand:topsoil mix (bulk density of 1.70 g cm−3) up to 1 m high, with a similar amount of soil volume and compaction. A 1-cm layer of perlite was placed on the surface to reduce water loss through evaporation (Fig. 3B). The tubes were arranged in three blocks, each consisting of a wire grid structure 9.6 m long by 1.2 m wide, for a plant density of 24.3 plants m−2. The structure was covered in plastic (white exterior, black interior; Fig. 3A, B) to avoid root exposure to light and overheating. A 1.2-m long bamboo culm was placed at the centre of each tube for plant support.
The experimental design was a randomized complete block design, with two treatments and three replicates per treatment, and one plant per block–treatment combination. Blocks were planted on consecutive days. Two or three mechanically scarified seeds per genotype were planted in the tubes, one or two being later discarded to leave one plant after germination. The treatments consisted of full irrigation and irrigation withdrawal. For irrigation, the tubes were watered to field capacity at the time of planting and initial weight was recorded. Further irrigation to field capacity was given to both treatments until the expanded first trifoliate leaf stage (between V2 and V3), ~16 d after planting. The irrigation was continued in the irrigated treatment until the end of the experiment, while the drought treatment received no further irrigation. Plants were submitted to water stress for ~18 d. The drought treatment was applied to the vegetative stage only, immediately before reproductive initiation. This was performed to account for the non-synchronism in flowering initiation, especially among wild genotypes. Moreover, this vegetative stress represents a common intermittent water deficit in tropical and subtropical regions of Mexico and central and northern South America.
Plant traits After seeding, all tubes were monitored on a daily basis for time for emergence (the shoot was visible over the perlite layer; stage V1). The number of days to reach V3 stage was also recorded when the first trifoliate leaf was completely expanded. The number of days to reach each developmental stage was counted based on the day that the seed was placed in each block. Visual root depth and plant height were measured at an interval of 3–5 d (following the sequence used to plant each block). The relative amount of chlorophyll was measured 28 d after planting using a SPAD-502Plus Chlorophyll Meter (Konica-Minolta). At the end of the experiment, the above-ground biomass was collected, separating the leaves and stems. Leaf area was determined with Easy Leaf Area (Easlon and Bloom, 2014). Specific leaf area was calculated using six leaf punches with a known area (2.83 cm2 ) of the most recently expanded mature leaves (Fig. 3C), dried in a 50 °C oven for a week (Cornelissen et al., 2003). At the end of the experiment, the roots were washed (Fig. 3E). Nodules and root whorl number were then counted and the roots were measured for their final length and dried. The environmental variables [Priestley–Taylor α coefficient (PTAC), annual temperature and soil bulk density] were extracted from the geographical coordinates of the origin of each population using the package raster (Hijmans and van Etten, 2016). The PTAC is the ratio of actual to potential evapotranspiration, and integrates soil water availability assuming similar soil characteristics (Priestley and Taylor, 1972). The PTAC was obtained from the CGIAR-CSI database (Trabucco and Zomer 2010) (www.cgiar-csi.org), annual mean temperature from WorldClim (Hijmans et al., 2005) (www.worldclim.org) and soil bulk density from SoilGrids (Hengl et al., 2014) (www. soilgrids1km.isric.org). All variables were at 1-km resolution.
Statistical analyses The data were analysed using linear mixed models in R (R Development Core Team, 2017). Genotype, treatment (irrigated, drought) and their interaction were fixed effects. Blocks and their interaction with genotype and drought treatment were random effects. Statistical analyses were performed with the lme4 package (Bates et al., 2015) and lmerTest to determine the significance of effects and to calculate least-squares means, using type-III hypothesis testing with Satterthwaite approximation for degrees of freedom. We determined R2 with the piecewiseSEM package (Nakagawa and Schielzeth, 2013). Marginal R2 is the variance explained by the fixed factors, while conditional R2 includes fixed and random factors. The coefficient of variation was calculated with sjstats (Lüdecke, 2016). Broad-sense heritability was estimated with REML (Holland et al., 2003) as h2 =σG 2 /(σG 2 +σGT 2 +σE 2 ), where σG 2 is the genetic variance, σGT 2 is the variance of the genotype × treatment interaction and σE 2 is the variance of the experimental error. The correlation between traits was calculated only among wild samples and plotted with the corrplot package (Wei and Simko, 2016). Environmental associations. A Bayesian network analysis was performed with bnlearn (Scutari, 2012) to jointly analyse the phenotypic and environmental variables and the treatment effect. All continuous variables were jointly discretized to preserve the dependence structure while bypassing normality assumptions (Nagarajan et al., 2013; Scutari and Denis, 2014). The continuous variables were discretized as multinomials with three levels using the hartemik method, and the treatment was left as a binomial. Structure learning was performed with a score base structure using the TABU greedy search. Average bootstrapping (10 000 iterations) with the Bayesian Dirichlet equivalent (bde) was used to obtain a consensus network. The nodes from phenotype towards environmental variables and treatment were blacklisted to improve stability. The trait ‘root whorl number’ was excluded due to the small phenotypic range, which did not allow proper discretization of the rest of the traits. To evaluate the variation between genetic groups, the phenotypic and ecological data were analysed using linear mixed models in R (R Development Core Team, 2017). Genetic group (n = 3), treatment (irrigated, drought) and their interaction were considered fixed effects. The accessions were considered to belong to the group with the highest ancestry coefficient among the three groups. Statistical analyses were performed in the lme4 package (Bates et al., 2015). The package lmerTest (Kuznetsova et al., 2015) was used to estimate the effect significance using type-III hypothesis testing with Satterthwaite approximation for degrees of freedom and least-squares mean calculation. Genotyping. DNA was extracted from a random individual of each accession using a modified ammonium acetate-based protocol (Pallotta et al., 2003). The 112 selected accessions were genotyped with 5398 single-nucleotide polymorphism (SNP) markers from the BARCBean6K_3 BeadChip SNP chip platform (Song et al., 2015) at the USDA-ARS Soybean Genomics Improvement Laboratory, Beltsville, MD, USA. After filtering in GenomeStudio Module v1.8.4 (Illumina, San Diego, CA, USA), the SNP calling was performed automatically and with subsequent manual adjustments; after filtering for quality control and a 0.15 Gencall score cutoff, 5186 SNPs remained. The same accessions were also subjected to genotyping by sequencing (GBS) based on the CviAII enzyme (Ariani et al., 2016). The reads were aligned to the G19833 reference genome version 2.1 [https://phytozome.jgi.doe.gov/pz/portal. html#!info?alias=Org_Pvulgaris (S. Jackson, P. McClean, J. Schmutz, unpubl. res.)] with BWA (Li and Durbin, 2009) and filtered with SAMtools (Li et al., 2009) for minimum mapping quality of 10 and mean and maximum read depths of 5 and 1000, respectively. The joint dataset of the GBS and SNP chips consisted of 11 447 SNP markers.