Phenotypic and genome-wide association analysis of spike ethylene in diverse wheat genotypes under heat stress
Summary
- The gaseous phytohormone ethylene plays an important role in spike development in wheat (Triticum aestivum). However, the genotypic variation and the genomic regions governing spike ethylene (SET) production in wheat under long-term heat stress remain unexplored.
- We investigated genotypic variation in the production of SET and its relationship with spike dry weight (SDW) in 130 diverse wheat elite lines and landraces under heat-stressed field conditions. We employed an Illumina iSelect 90K single nucleotide polymorphism (SNP) genotyping array to identify the genetic loci for SET and SDW through a genome-wide association study (GWAS) in a subset of the Wheat Association Mapping Initiative (WAMI) panel.
- The SET and SDW exhibited appreciable genotypic variation among wheat genotypes at the anthesis stage. There was a strong negative correlation between SET and SDW. The GWAS uncovered five and 32 significant SNPs for SET, and 22 and 142 significant SNPs for SDW, in glasshouse and field conditions, respectively. Some of these SNPs closely localized to the SNPs for plant height, suggesting close associations between plant height and spike-related traits.
- The phenotypic and genetic elucidation of SET and its relationship with SDW supports future efforts toward gene discovery and breeding wheat cultivars with reduced ethylene effects on yield under heat stress.
Introduction
Ambient temperatures, on a global scale (IPCC, 2007) as well as at the on-farm level (Battisti & Naylor, 2009), have been increasing since the beginning of the century and this trend is predicted to continue into the future (IPCC, 2013). Field studies (White et al., 2011) and controlled environment experiments (Prasad & Djanaguiraman, 2014) have reported a substantial negative impact of heat stress (HT) on wheat (Triticum aestivum) yield. Further, short episodes of extreme heat are expected to become more frequent, which could have an adverse impact on yield components, and thus grain yield (Prasad et al., 2006; Jagadish et al., 2007). Such negative impacts are more likely to occur at lower latitudes, where c. 100 million ha of wheat is cultivated producing c. 280 million tons of grain, although some positive benefits at higher latitudes can be expected (Cossani & Reynolds, 2012). Intensified efforts employing a diverse set of wheat genotypes can help to better understand the physiological and genetic responses to HT and this understanding potentially can aid in enhancing future wheat yields.
Under heat stress, many plants produce extra ethylene (ET) (Hays et al., 2007), a colorless gas that has been extensively used in commercial agriculture to force the ripening of fruits. Ethylene has been linked to yield penalty under heat stress in various crops including wheat (Hays et al., 2007; Huberman et al., 2013), which mainly is attributed to a reduction in spike fertility (Cheng & Lur, 1996) and final grain weight (Yang et al., 2004). These trait responses are attributable partly to reduced carbohydrate export to pollen grains, accompanied by poor pollen germination and accelerated senescence (Torre et al., 2006; Feng et al., 2011). Further, accelerated stigma and ovule development and abnormal ovary development result in reduced pollen tube growth and grain set when heat stress-induced ET coincides with meiosis (Barnabás et al., 2008). From these studies, it is apparent that ET physiology plays a crucial role in the regulation of spike fertility and kernel abortion under heat stress, although ET-induced kernel abortion and premature maturation can be seen as a strategic survival trait in order to maintain optimal progeny load in warm, dry climates.
The effects of heat stress vary among species and between individuals in the same species and are highly dependent on the timing of the high-temperature challenge (Prasad et al., 2006). Previous studies have followed various approaches to reduce the negative impact of heat-induced ET production on crop yield. Typically, the application of various chemicals that act as antagonists of ET biosynthesis or sensitivity of response has been used in practice, which has significantly increased grain yield in wheat (Hays et al., 2007; Huberman et al., 2013) and rice (Oryza sativa) under conditions where ET might otherwise be limited by heat stress (Tamaki et al., 2015). At the molecular level, biotechnological interventions have been employed to reduce ET biosynthesis and its sensitivity in crop plants. Down-regulation of one or more metabolic components of ET signaling (e.g. ACS6 (1-Aminocyclopropane-1-carboxylic acid Synthase 6 ) (Young et al., 2004; Habben et al., 2014) and ETR2 (Ethylene Response 2) (Wuriyanghan et al., 2009)) has been shown to decrease ET biosynthesis, and sensitivity to this hormone in maize (Zea mays) (Young et al., 2004; Habben et al., 2014) and rice (Wuriyanghan et al., 2009). Such manipulations contributed to an increased grain yield under both optimal growing conditions (Wuriyanghan et al., 2009) and field-drought conditions (Habben et al., 2014). These studies suggest that manipulating ET signaling may have a positive impact on crop yield under stress.
A longer term, robust and cost-effective approach to reduce the negative effect of ET on crop yield could be through exploiting natural diversity in ET production. A substantial genetic diversity in pollen viability and spike fertility under heat stress has been identified in rice (Prasad et al., 2006; Jagadish et al., 2007), cotton (Gossypium hirsutum) (Burke & Chen, 2015), maize (Herrero & Johnson, 1980), and Brachypodium (Harsant et al., 2013), while in wheat, variation in spikelet fertility in response to various agronomic treatments has been reported (Cantrell & Haro-Arias, 1986; Guo & Schnurbusch, 2015). Further, wheat genotypes differ in heat stress tolerance (Fu et al., 2015) and exhibit large intraspecific variation in grain yield (Lopes et al., 2015). However, there is a significant lack of understanding of genotypic variation in spike dry weight (SDW), a proxy for grain number and grain weight (Fischer, 1985; González et al., 2011), and its association with spike ethylene (SET), in a diverse set of wheat genotypes on exposure to heat stress under field conditions.
Plant height is an important agronomic trait closely linked to yield and biomass. In heat- and drought-stressed environments, a considerable yield variation between taller and shorter phenotypes has been reported (Richards, 1992; Buttler et al., 2005). Further, taller genotypes have been shown to maintain cooler canopies, although only when surrounded by shorter plots, in heat-stressed environments (Miller et al., 1981; Pask et al., 2014) as a result partly of a greater coupling of the plant boundary layer with the environment. We hypothesize that taller genotypes may produce lower SET compared with their shorter counterparts, particularly in a heat-stressed environment. Nevertheless, this hypothesis has never been explicitly tested.
To elucidate the impact of heat stress and other environmental stresses, proteomic and transcriptomic analyses routinely have been performed in various crops including wheat (Li et al., 2013; Wang et al., 2015a,b). However, there have been few studies, limited to only a few crop species, dissecting the genetic basis of fertility dynamics under heat stress, for instance, in rice (Ye et al., 2015), cotton and tobacco (Nicotiana tabacum) (Burke & Chen, 2015). Genome-wide association analysis offers the high-resolution genetic capacity to capture insights into the genetic architecture of complex traits such as SET. To our knowledge, there are no reports of the use of genome-wide association mapping to elucidate the basis of heat stress responses of SET and SDW in wheat.
In this study, we first (Exp 1) examined a diverse set of 130 spring wheat genotypes that consisted of elite material (EM) and landraces (LR) to examine the genotypic variation of SET and SDW in the presence of an ET response inhibitor under heat-stressed field conditions. We hypothesized that different wheat genotypes differ for SDW sensitivity to heat stress particularly through their response to SET (Hays et al., 2007). Second, a random subset of 30 genotypes (Exps 2 and 3) from the initial 130 genotypes was used to dissect the responses of, and the tradeoffs between, spike-related traits, plant height, yield and yield attributes in response to SET. Finally, in order to dissect the genetic basis of SET and SDW, a genome-wide association study (GWAS) was performed on a separate subset of the Wheat Association Mapping Initiative (WAMI; Lopez et al., 2012) panel grown under controlled environment (Exp 4) and field conditions (Exp 5).
Materials and Methods
Plant material
A diverse set of spring wheat (Triticum aestivum L.) genotypes (130 genotypes; Supporting Information Table S1) consisting of EM (85) and LR (45) was chosen for the initial genetic diversity study (Exp 1). Subsequently, a random subset (30 genotypes: EM, 13; LR, 17; Exps 2 and 3; Table S2) was chosen for further studies. In addition, a separate diverse set of the spring wheat elite WAMI panel (190 genotypes; Exps 4 and 5) was used for genetic studies.
Experimental design
Four field experiments (Exps 1–3 and 5) were carried out during 2013–2015 in the Yaqui Valley at CIMMYT's Experimental Station in Obregon (20°27′N, 109°54′W; 38 m above sea level), Mexico. One glasshouse study (Exp 4) was carried out during 2014–2015 at the Lancaster Environment Center, Lancaster University, UK.
All field experiments (Exps 1–3 and 5) were sown during mid-to-late March in each year (24, 29 and 19 March in 2013, 2014 and 2015, respectively). The plots were arranged in a randomized lattice design with two replications in 2-m-long and 0.8-m-wide plots consisting of one raised bed with two rows per bed (0.3 m between rows). The soil type is a sandy-clay, low in organic matter (< 1%) and slightly alkaline (pH 7.5), with a plant available water-holding capacity of c. 200 mm. Gravity-based flooding irrigation was used with a frequency of 2–3 wk to avoid water deficit. Similar fertilizer regimes and appropriate cultural practices such as weed, disease, and pest control were implemented to avoid any yield limitations.
Sowing in the month of March generally exposes the crop to a heat-stress environment compared with the normal planting time (November) of yield potential trials. Hence, the crop plants encountered a natural high-radiation environment. The average air temperatures during the preanthesis period were 32.8°C, 33.8°C and 34°C (maximum: 37.5°C, 41.0°C and 40.4°C; minimum: 28.1°C, 26.8°C and 27.4°C) with an average air relative humidity of 65%, 59% and 60% in 2013, 2014 and 2015, respectively. The average temperatures of the crop canopy throughout the preanthesis period, however, were 29.5°C, 31.2°C and 31.9°C in 2013, 2014 and 2015, respectively. Thus, the crop experienced heat stress (average maximum temperature (Tmax) > 32°C; total crop water supply > 700 mm) during the preanthesis period.
For genetic analysis, in addition to one field study (Exp 5, sown on 19 March 2015), one controlled glasshouse study (Exp 4) was carried out using the same subset of the WAMI panel. Briefly, seeds of a randomly selected subset of WAMI lines (190) were germinated on wet filter paper placed in Petri dishes at room temperature. Seven-day-old seedlings were transplanted into the plastic pots (20 × 15 cm) containing a well-prepared mixture of soil-based compost (John Innes No. 2; John Innes Manufacturers Association, Berkshire, UK). The plants were grown in the glasshouse which was set to a 14-h light regime (c. 900 μmol m−2 s−1 photosynthetically active radiation (PAR)) with day : night temperatures of 35°C : 23°C, respectively, at a relative humidity of 70–75%. All plants were well watered daily and a half-strength Hoagland solution was supplied as a source of nutrients on alternate days. Two pots per genotype, and two plants per pot were maintained. Samples for SET and SDW were collected at the appropriate developmental stage.
Silver nitrate spray in field experiments
In Exps 1–3, two treatment conditions were maintained: heat stress without silver nitrate spray (control; HT) and heat stress with silver nitrate spray (HSN). Both the HT and HSN conditions were imposed in an area of a 1-m length of plot containing two rows. These plots were specifically marked at an early seedling stage to eliminate border effects. When the plants reached an early booting stage (ranging from 50 to 70 d after sowing (DAS) with a range of heading dates of 61–78 DAS), leaves of all plants in the premarked area were sprayed with either silver nitrate (Sigma-Aldrich, Toluca, Mexico) as described previously (Kumar et al., 2009) or water (control plots). As a consequence of its high water solubility and a lack of phytotoxicity at effective concentrations (Beyer, 1976), silver nitrate has been used extensively for both agronomic and research purposes to inhibit the ET response (Rodrı́guez et al., 1999). Silver nitrate was dissolved in water (0.05 mM) containing 0.01% (v/v) Tween 20 (Sigma-Aldrich) and ethanol (0.1%), and the solution was foliar-sprayed onto the leaves of whole plants until runoff. The spraying always took place between 06:00 and 08:00 h and spraying of each plant was repeated on each of three successive days. The control plots were sprayed with water containing 0.01% (v/v) Tween 20 and ethanol (0.1%) only. During spraying, a specially made wind protector was used to cover the spray area. As the genotypes reached an early booting stage at different time-points, a staggered spray schedule was followed to avoid the confounding effects of spraying at different developmental stages.
Spike ethylene measurement
In all the experiments, spike sampling was performed at the anthesis stage. Endogenous ET production from spikes was measured using a laser-based ET detector (ETD-300; Sensor Sense BV, Nijmegen, the Netherlands) in combination with a gas handling system (VC-6; Sensor Sense BV) (Valluru et al., 2016). Five tagged spikes (2–2.5 g FW) on the main tillers (in Exp 4, only four spikes) were sampled and placed in paper bags, and the bags were closed and transported to the lab with similar transport times (< 10 min). Each spike was then weighed and placed in a glass tube. Previous studies have shown that ET evolution from excised leaves and other shoot tissues can be steady for c. 15 min and then increase, presumably as a consequence of wounding (Sharp et al., 2000; Yang et al., 2006). We therefore left the glass tubes open for 60 min to allow any wound-induced ET to subside (Yang et al., 2006). Spikes were then transferred into 50-ml glass vials containing moist filter paper, which were tightly capped with a double-bent rubber stopper, and were further incubated for 2 h under an artificial light setting in the lab. A thermometer was used to maintain a constant temperature regime of c. 35°C (± 1°C) during incubation for all spikes. Using a 5-ml syringe, 4 ml of gas was extracted through the rubber stopper and stored in 4-ml sealed glass vials. These vials were connected to inlet and outlet cuvettes of the VC-6 system, which allowed measurement of six cuvettes at once. Vials were continuously flushed with air at a constant flow of 4 l h−1. ET released from each vial was successively connected to the ET detector in sample mode for 10 min. To remove any traces of external ET or other hydrocarbons, the air flow was passed through a platinum-based catalyzer before entering the cuvettes. A scrubber with KOH and CaCl2 was placed before the ET detector to reduce the CO2 and water content in the gas flow, respectively. The SET production was corrected for incubation period and SDW to determine ET production rate. After the SET measurement, SDW was determined after oven drying at 70°C for 72 h.
Plant height and grain yield sampling
Plant height (cm) was measured as the distance from the soil surface to the tip of the spike excluding the awns at the anthesis stage. At harvest, the HT and HSN plots were harvested separately for yield analysis whereby grain yield (GY; m−2) and thousand-grain weight (TGW; g) were estimated.
Statistical analyses
All statistical analyses were performed using the software R v.3.1.3 (R Core Team, 2015). The SDW and SET data were subjected to one-way and two-way ANOVA, and a two-sample t-test with Bonferroni-adjustment alpha level was used for post hoc comparison of treatments, genotype categories and their interaction. In order to analyze the response of SDW to changes in SET, an exponential model was fitted between SDW and SET. Pearson's correlations were performed between SET, SDW and plant height. Hierarchical clustering (hclust function in R) was used to categorize all genotypes into two SET groups, the low-ET group (LETG) and the high-ET group (HETG), while a quantile approach was used to categorize genotypes into two extreme plant height groups (short plant height group (SPHT), < 10% quantile value; and taller plant height group (TPHT), > 90% quantile value).
DNA extraction and genotyping
The methods of DNA isolation and genotyping are reported elsewhere (Sukumaran et al., 2015a,b). Briefly, each genomic DNA sample of WAMI lines was isolated using a modified cetyltrimethyl ammonium bromide (CTAB) protocol (Saghai-Maroof et al., 1984). The DNA samples were SNP genotyped at the United States Department of Agriculture, Agricultural Research Service Small Grain Genotyping Center, Fargo (http://wheat.pw.usda.gov/GenotypingLabs) using an Illumina iSelect 90K SNP array (Wang et al., 2014). SNP allele clustering and genotype calling were performed with Genome Studio software v.2011.1 (Illumina Inc., San Diego, CA, USA). We used the default clustering algorithm implemented in Genome Studio to identify distinct clusters corresponding to the AA, AB and BB genotypes. The accuracy of SNP clustering was validated visually.
Genotype selection, linkage disequilibrium and population structure
In order to detect genotype outliers between field and glasshouse data sets, we used a Bonferroni outlierTest (car package in R; Williams, 1987). The Studentized genotype residuals with significant Bonferroni P-values (< 0.05) were identified (also confirmed using Cook's distance measure; Cook & Weisberg, 1982) and eliminated, and all subsequent analyses were then performed on the remaining 163 lines.
Linkage disequilibrium (LD) among SNPs was calculated using the full matrix and sliding window options in Tassel 4.0 (Bradbury et al., 2007) using 18 704 SNPs. Only SNPs with minor allele frequency > 5% were considered. Using the squared allele frequency correlations, pair-wise LD was measured according to Weir (1996). LD decay and the percentage of SNP pairs below and above the critical LD were determined (Sukumaran et al., 2015a,b).
Patterns of population structure detected using the principal component analysis (PCA) approach were also confirmed using the model-based clustering approach implemented in Structure (Pritchard et al., 2000). Briefly, we ran Structure using the default model parameters and varying the assumed number of genetic groups (K ) from 1 to 15, using an admixture model consisting of 20 000 burn-in iterations and 5000 replicates. Finally, we used the SPAGeDi 1.3 program (Hardy & Vekemans, 2002) to calculate the genetic relationship matrix among the 163 lines. The broad-sense heritability for SET, SDW and their responses to HSN was calculated as 
, where 
is the genetic variance, 
is the genotype by environment (G × E) variance, and 
is the residual variance. These parameters were obtained using the linear mixed model, treating genotype and G × E as random effects.
GWAS
We used the Tassel algorithm population parameters previously determined (P3D; Zhang et al., 2010) to perform GWAS. To control for the familial relatedness and population structure, we incorporated a kinship matrix as the random component and the principal components (PCs) as fixed effects. We compared different models to select the best model based on the deviation of the observed P-values from the expected P-values for each trait (Q–Q plots) for detecting SNP–trait associations following unified mixed model association mapping (Sukumaran et al., 2012). The threshold for defining an SNP to be significant was taken as 10−03, considering the number of SNPs and the deviation of the observed F-test statistics from the expected F-test distribution in the quantile–quantile (QQ) plots (Sukumaran et al., 2012).
Results
Genotypic variation for spike dry weight at anthesis stage among 130 wheat lines
SDW under HT conditions ranged from 0.31 to 1.68 g, with a mean of 0.94 g, and a broad-sense heritability of 61% (Fig. 1a). The SDW increased after the plants were sprayed with silver nitrate solution (HSN) with values ranging from 0.36 to 1.67 g, with a mean of 1.18 g (Fig. 1a), that is, an average c. 25% higher than without HSN (P < 0.05; Fig. 1b). The response of SDW to HSN among lines varied by up to 3-fold with an average SDW response of 1.2-fold, and with a broad-sense heritability of 57%. Genotypes (P < 0.001), treatments (P < 0.001) and their interaction (P < 0.001) had a significant effect on SDW (Table S3a). It is noticeable that, in a few genotypes, particularly those that had higher SDW under HT, SDW decreased in response to HSN.
The EM and LR differed significantly for SDW (Fig. 1c,d). EM had a significantly higher SDW compared with LR across the two conditions, the HT conditions (+ 14%) and the HSN conditions (+ 11%). Two-way ANOVA indicated a significant main effect of genotype category on SDW (P < 0.001); however, its interaction with treatment was not significant (Table S3a). These results suggest that there exists large genotypic variation for SDW among the wheat genotypes studied and silver nitrate spray at the early booting stage contributed to an increased SDW at the anthesis stage.
Genotypic variation for spike ethylene at anthesis stage among 130 wheat lines
SET under HT conditions (Fig. 2a) ranged from 0.017 to 0.155 nl g DW−1 h−1, with a mean of 0.077 nl g DW−1 h−1, and a broad-sense heritability of 25%. In response to HSN, SET decreased noticeably among genotypes, ranging from 0.01 to 0.145 nl g DW−1 h−1, with a mean of 0.063 nl g DW−1 h−1, that is, an average c. 18% lower than without HSN (Fig. 2b). The response of SET to HSN had a broad-sense heritability of 39%. Genotypes that had high SET under HT exhibited more reduction in SET compared with low-SET genotypes. There were significant main effects of genotype (P < 0.001), treatment (P < 0.001), and their interaction (P < 0.001) on SET (Table S4b).
In response to HSN, both EM and LR exhibited a reduction in SET, by 18% and 17%, respectively (Fig. 2c,d). Across both HT and HSN conditions, EM genotypes had a significantly lower SET (−13%; P < 0.001; Fig. 2c) compared with LR. Genotype category had a significant (P < 0.001) effect on SET, but its interaction with treatment was not significant (Table S3b). Overall, these findings suggest that there exists large genotypic variation for SET among the wheat genotypes studied and SET decreased significantly in response to HSN.
Spike ethylene was negatively correlated with spike dry weight
We used an exponential model to fit the relationship between SDW and SET. SDW decreased exponentially with increasing SET across all the genotypes (Fig. 3a). The relationship between SDW and SET was not altered by HT and HSN conditions (for HT: r2 = 0.73; P < 0.001; for HSN: r2 = 0.62; P < 0.001; Fig. 3a) or by the genotype category (EM HT: r2 = 0.79; P < 0.001; EM HSN: r2 = 0.67; P < 0.001; LR HT: r2 = 0.71; P < 0.001; LR HSN: r2 = 0.56; P < 0.001) (Fig. 3b,c). On average, 68 ± 3% of the variance of the SDW decay rate across HT and HSN conditions can be explained by an exponential regression of the mean rate on the SET.
HT and HSN condition-specific exponential models between SDW and SET were similar to the global model of HT+HSN conditions across all the genotypes (r2 = 0.50; P < 0.001; Fig. S1a), within the EM genotypes (r2 = 0.56; P < 0.001; Fig. S1b) or within LR (r2 = 0.45; P < 0.001; Fig. S1c). We also checked the relationship between SDW and SET in a relatively small subset of genotypes (30 genotypes; Exps 2 and 3), which also revealed a significant negative relationship between SET and SDW (Fig. S2a–c).
We further examined the tradeoffs between SET and SDW using a hierarchical clustering by establishing two ET groups: LETG and HETG. The LETG had a significantly lower SET (−41.5%; P < 0.001; Fig. S3a) compared with the HETG. By contrast, the LETG had a significantly higher SDW (+41%; P < 0.001; Fig. S3b) compared with its counterpart. Taken together, these results suggest that there was a consistent negative relationship between SET and SDW, exhibiting differential SDW sensitivity to increasing levels of SET, whereby higher SET profiles appeared to be detrimental to SDW.
Correlations between spike-related traits and plant height
We generated Pearson's correlations between SDW, SET and plant height using phenology as a covariate (Fig. 4). Plant height had a negative correlation with SET under HT conditions (r2 = 0.66; P < 0.001) and HSN conditions (r2 = 0.10) (Fig. 4a). By contrast, plant height showed a positive correlation with SDW (r2 = 0.36; P < 0.05; and r2 = 0.01) under both HT and HSN conditions, respectively (Fig. 4b). However, these correlations were not significant under HSN conditions. These correlations were present across all genotypes (Fig. 4a,b), within EM genotypes (Fig. S4a,e), and within LR (Fig. S4c,g).
To further dissect the association between SET, SDW and plant height, we used a quantile approach to develop two extreme plant height groups based on plant height (genotypes below the 10% quantile were assigned to the shorter plant height group (SPHT) while genotypes above the 90% quantile were assigned to the taller plant height group (TPHT)). The taller group had a significantly lower SET across genotypes (−53%; P < 0.001; Fig. 4b), within EM genotypes (−50%; P < 0.05; Fig. S4b) and within LR (−70%; P < 0.001; Fig. S4d). By contrast, the taller group had a significantly higher SDW across all genotypes (+19%; P < 0.001; Fig. 4d), within EM genotypes (+36%; P < 0.05; Fig. S4f) and within LR (+42%; P < 0.05; Fig. S4h) compared with the shorter group. These results indicate that the SET and SDW changes identified by this arbitrarily chosen limit of 10% reveal close associations of these traits with plant height.
Lower spike-ethylene dynamics were strongly associated with higher grain yield
In order to examine the effects of SET on grain yield, grain yield was regressed against SET, which revealed negative correlations between SET and grain yield under HT conditions (r2 = 0.37; P < 0.001) and HSN conditions (r2 = 0.17; P < 0.05) (Fig. 5a) as well as among ET groups (low-ET group: r2 = 0.42; P < 0.001; high-ET group: r2 = 0.23; P < 0.01; Fig. 5b). These results suggest that there is a yield penalty associated with higher SET profiles. Similar results were observed between SET and TGW, although they were not significant (Fig. S5).
Further yield analyses indicated that, across all genotypes, HSN conditions had a yield gain of 23% over HT conditions (P < 0.05; Fig. 6b). A similar yield gain (22%) was observed in the low-ET group compared with the high-ET group (P < 0.01; Fig. 6a). When different combinations of ET groups, treatments, and genotype categories were compared for yield gain (Fig. S6), in all cases, the low-ET group had a higher yield gain compared with the high-ET group, ranging from 12% to 45%, with an average yield gain of 26% (Fig. S6). When only significant comparisons, that is, five out of eight combinations (P < 0.05), were considered, the mean yield gain rose to 34%. The yield gain of the low-ET group over the high-ET group was higher for EM genotypes (42%) compared with LR (12%; P < 0.001).
TGW differed significantly between ET groups (P < 0.01; Fig. 6d), treatments (P < 0.05; Fig. 6e) and different combinations of ET groups and treatments (Fig. 6c–e) but not between genotype categories. HSN conditions as well as the low-ET group had a TGW gain of 5.5% and 9%, respectively, while their combination varied from 3% to 7% (Fig. S6c,d). Taken together, these results suggested that changes in SET were closely but negatively associated with changes in grain yield and TGW.
Genome-wide association for spike ethylene and spike dry weight
We performed a GWAS on SET and SDW for the glasshouse and field data separately using 18 704 SNPs. We detected a number of candidate SNPs (five and 32 SNPs in the glasshouse and field, respectively, with P-values < 10−3). These SNPs are located on chromosomes 1A, 1B, 1D, 3B, 5B and 7B (Fig. 7; Table 1) and explained 5.7–10.5% of the variation in SET. A major SNP (kukri-c15603-1116) at 69 cM on chromosome 3B explained 10.5% while a major SNP (wsnp-JD-c38123-27754848) localized at 40 cM on chromosome 5B explained 8.8% of total SET variation under field conditions. Interestingly, two distinct SNPs (kukri-c29668-338 and BS00009866-51 in glasshouse and field conditions, respectively) are localized at the same position at 70 cM on chromosome 1A (Fig. 7a,b; Table 1). The SNP BS00009866-51 is predicted to be a putative membrane steroid-binding protein (MSBP1; Balmer et al., 2006). For the SNP kukri-c29668-338, 149 accessions had genotype AA and 10 accessions had genotype GG at the locus, while based on the SNP BS00009866-51, 10 accessions had genotype AA and 148 accessions had genotype GG at the locus. The allelic diversity for these two SNPs indicates that the AA genotype had significantly lower and higher SET (38% and 25% for SNPs kukri-c29668-338 and BS00009866-51, respectively) compared with the GG genotype (Fig. 8a,b).
| Trait | Marker | Chromosome | cM | P | PVE (%) | Allele | MAF |
|---|---|---|---|---|---|---|---|
| SET, lab | Tdurum-contig32033-125 | 1B | 68 | 9.00E-04 | 6.8 | A/C | A (0.05) |
| Kukri-c29668-338 | 1A | 70 | 2.44E-03 | 5.7 | A/G | G (0.05) | |
| SET, field | Kukri-c15603-1116 | 3B | 69 | 4.63E-05 | 10.5 | C/T | C (0.34) |
| wsnp-JD-c38123-27754848 | 5B | 40 | 1.98E-04 | 8.8 | G/T | G (0.32) | |
| BS00009866-51 | 1A | 70 | 6.85E-04 | 7.1 | A/G | A (0.07) | |
| RAC875-c37183-331 | 1D | 3 | 7.10E-04 | 7.0 | A/G | G (0.92) | |
| wsnp-CAP11-rep-c4076-1926235 | 7B | 102 | 9.01E-04 | 6.8 | A/G | A (0.19) | |
| SDW, lab | wsnp-Ku-c5693-10079278 | 7A | 208 | 2.80E-05 | 10.3 | A/C | A (0.06) |
| RAC875-rep-c109658-211 | 5B | 131 | 1.33E-04 | 8.7 | C/T | T (0.82) | |
| Tdurum-contig49576-75 | 5A | 82 | 3.30E-04 | 7.7 | A/C | A (0.11) | |
| RAC875-c31299-1302 | 6B | 109 | 4.27E-04 | 7.6 | A/G | A (0.29) | |
| Kukri-c32958-390 | 4B | 81 | 7.38E-04 | 6.8 | C/T | C (0.06) | |
| Kukri-c657-1139 | 2B | 141 | 7.57E-04 | 6.8 | C/T | C (0.49) | |
| TA003922-1105 | 3B | 72 | 7.76E-04 | 6.8 | A/G | A (0.08) | |
| SDW, field | BS00055584-51 | 7B | 140 | 1.01E-04 | 14.9 | C/T | T (0.54) |
| wsnp-Ku-c5693-10079278 | 7A | 208 | 6.51E-06 | 12.3 | A/C | A (0.06) | |
| Kukri-c657-1139 | 2B | 141 | 6.27E-05 | 9.9 | C/T | C (0.49) | |
| Ku-c24961-1176 | 4B | 71 | 6.43E-05 | 9.8 | A/G | A (0.26) | |
| Tdurum-contig49576-75 | 5A | 82 | 7.81E-05 | 9.5 | A/C | A (0.11) | |
| RAC875-rep-c109658-211 | 5B | 131 | 9.90E-05 | 9.3 | C/T | T (0.82) | |
| RAC875-c31299-1302 | 6B | 109 | 1.12E-04 | 9.3 | A/G | A (0.29) | |
| Kukri-c32958-390 | 4B | 81 | 1.41E-04 | 8.8 | C/T | C (0.06) | |
| wsnp-Ra-c12086-19452422 | 6A | 91 | 1.72E-04 | 8.9 | C/T | C (0.46) | |
| Kukri-c11327-977 | 2A | 101 | 1.94E-04 | 8.9 | G/T | T (0.79) | |
| wsnp-Ku-rep-c102901-89769309 | 6A | 91 | 2.24E-04 | 8.7 | C/T | C (0.45) | |
| Ku-c14907-456 | 6A | 72 | 3.95E-04 | 7.9 | A/C | A (0.42) | |
| wsnp-Ex-rep-c69816-68774416 | 3A | 68 | 4.08E-04 | 7.9 | A/C | A (0.45) | |
| Kukri-c43208-335 | 3D | 67 | 4.39E-04 | 8.0 | A/G | G (0.65) |
- A subset (163) of the wheat association mapping initiate (WAMI) panel accessions were grown in experimental glasshouse conditions (2014) and field conditions (2015) for scoring SET and SDW of wheat, and 18 704 single nucleotide polymorphism (SNP) markers (MAF > 5%) of the wheat Illumina iSelect 90K SNP array were applied for genotyping. PVE, the percentage of phenotypic variation explained by each marker. P-values indicate significance levels. MAF, minor allele frequency.
The association mapping for SDW identified a number of candidate SNPs in glasshouse and field conditions (22 and 142 SNPs, respectively, with P-values < 10−3). These SNPs were localized on chromosomes 2A, 2B, 3A, 3B, 3D, 4B, 5A, 5B, 6A, 6B, 7A and 7B, and explained 6.8% to 14.9% of SDW variation (Fig. 7c,d; Table 1). A major SNP, BS00055584-51, which explained c. 15% of total SDW variation, was detected at 140 cM on chromosome 7B under field conditions (Fig. 7c,d; Table 1). Further, we detected six SNPs (wsnp_Ku_c5693_10079278, RAC875_rep_c109658_211, Tdurum_contig49576_75, RAC875_c31299_1302, Kukri_c32958_390 and Kukri_c657_1139) on chromosomes 7A, 5B, 5A, 6B, 4B, and 2B, respectively, that were common across environments. Among 163 accessions, two groups could be established based on two SNPs (wsnp_Ku_c5693_10079278 and Kukri_c657_1139) whose allelic diversity (AA vs CC and CC vs TT, respectively) exhibited a significant effect on SDW. For the SNP wsnp_Ku_c5693_10079278, eight accessions had the AA genotype, and this had a significantly higher SDW (15%) compared with the CC genotype, which was found in 150 accessions, while for the SNP Kukri_c657_1139, 77 accessions had the CC genotype, and this had a significantly higher SDW (10%) compared with the TT genotype, which was found in 80 accessions (Fig. 8c,d). The SNP Kukri_c657_1139 is predicted to be an E3 ubiquitin-protein ligase (He et al., 2013).
Discussion
Plants synthesize ET in response to numerous environmental stimuli. ET has long been thought to act mainly as a growth-inhibiting phytohormone, although positive organ growth at lower concentrations has been widely reported in several species (Ku et al., 1970; Valluru et al., 2016). To date, only a limited number of studies have explored the possibility of the ‘ET trait’ being used as a means of crop improvement under stress. Further, previous studies focused only on a few individual genotypes (Yang et al., 2006; Hays et al., 2007). This study used as a base a population showing wide genetic diversity, and found substantial genotypic variation for SET and SDW at the anthesis stage (Figs 1, 2). There was a strong negative correlation between SET and SDW (Figs 3, S1, S2). An application of an ET response inhibitor significantly decreased SET (Fig. 2; Hays et al., 2007; Huberman et al., 2013), and subsequently contributed to increased SDW and grain yield under heat stress (Figs 1, 6, S6). Finally, we detected numerous putative SNPs explaining genotypic variation of SET and SDW (Fig. 7; Table 1), suggesting that both SET and SDW are accessible for genetic manipulation in wheat.
Conditions without silver nitrate spray during an entire vegetative period caused a significant decline in SDW compared with silver nitrate spray conditions (Fig. 1), with the result being a genotype-dependent decrease in grain yield (Fig. 6). This decrease in SDW and grain yield was partly rescued by silver nitrate spray (Figs 1, 6), a phenomenon previously observed (Hays et al., 2007; Huberman et al., 2013). Taking this into account, our results suggest that an 18% decline in SET contributed to increases in SDW and grain yield of 14% and 22%, respectively, which are not inconsistent with increases in yield gains (1–45%) reported previously in various crops in response to the use of inhibitors of the biosynthesis of, or the response to, ET (Young et al., 2004; Hays et al., 2007; Wuriyanghan et al., 2009; Huberman et al., 2013; Tamaki et al., 2015). Notably, a moderate heritability of the responses of SET and SDW to silver nitrate spray (39% and 57%, respectively) further suggests the possibility of leverage of decent genetic gains through these traits. These results support the view that SET is an invaluable ‘candidate’ for partial mitigation of heat stress effects on yield. In this regard, silver nitrate spray can be seen as a cost-effective field approach to enhancing on-farm yields under heat stress. Importantly, silver nitrate spray decreased SDW in a few genotypes that already had lower SET profiles, warranting a genotype-specific tailoring of SET, as a drastic reduction of ET signaling may not be beneficial to maximizing grain yields (Guo et al., 2014). Such responses may, however, vary depending on the stress conditions.
In an agricultural context, an exploration of the potential natural variation for SET seemed to be warranted. We found large genotypic variation for SET with a low broad-sense heritability, suggesting that SET may be conditioned by a large number of genes of small effect, reflecting a high nonadditive genetic variance and strong genotype by environment crosstalk. Further, EM lines had a significantly lower mean SET profile compared with LR (Fig. 2), indicating that SET might have been inadvertently selected against in EM lines, probably as a consequence of targeted selection for higher yield potential. Substantial genotypic variation, albeit in a small set of genotypes, has been previously reported for numerous phytohormones such as ET in sorghum (Sorghum bicolor) (Zhang & Kirkham, 1990), and abscisic acid (ABA), gibberellic acid (GA) and indole acetic acid in wheat (Sridhar, 2003). Further, ABA concentrations differ markedly among different ploidy levels of wheat (Iehisa & Takumi, 2012), and also between landraces and hybrids of tomato (Solanum lycopersicum) (Ruiz et al., 2006). Such a wide genetic diversity provides an excellent base for uncovering the physiological and genetic underpinning of ET insensitivity. For instance, SDW seemed to be less sensitive to SET in higher SET profile genotypes (Fig. 3). It is attractive to speculate that these genotypes or spikes may have metabolic signals that rendered them less sensitive to ET. These genotypes would therefore be useful for developing new cultivars with enhanced yield, although breeding for favorable SET expression would be very challenging, particularly in field settings.
ET-based strategies with the aim of improving crop yields have so far been limited, partly as a consequence of the fact that comprehensive understanding of ET regulation at the whole-plant level is limited (Sharp et al., 2000; Sharp & LeNoble, 2002), as most studies have focused largely on the cell and tissue levels (Takatsuka & Umeda, 2014). Moreover, an inconsistent pattern of gene expression and protein activity has been reported for certain ET biosynthetic genes (e.g. ACO1 (1-Aminocyclopropane-1-Carboxylate Oxidase 1); Van de Poel et al., 2014), reflecting the complex nature of ET regulation in plants. The lack of a high-throughput phenotyping tool to quantify ET production, especially in the field setting, is another bottleneck. In view of this, SDW can serve as a high-throughput proxy with which to rapidly estimate SET, as it is strongly negatively correlated with SET (Fig. 3); it represents the degree of spike fertility at the anthesis stage (González et al., 2011), and also determines grain weight and grain number at maturity (Fischer, 1985; González et al., 2011). Notably, SDW can easily be rapidly measured and hence crops can be screened for SDW on a large scale in the field. Therefore, use of SDW can be widely applied to high-yield breeding in wheat.
Genetic knowledge of ET in crops has been lacking partly because of the fact that ET is a complex trait. Through a GWAS, we detected five and 32 SNPs for SET in glasshouse and field conditions, respectively (Fig. 7; Table 1), suggesting that some are environment-specific. A limited SNP overlap and a lower heritability (25%) suggest that SET may indeed be governed by a large number of genes of small effect that may not be possible to detect consistently across environments. However, two distinct SNPs, kukri-c29668-338 and BS00009866-51, were consistently detected at the same position on chromosome 1, indicating that this genomic region may be an important regulator of SET. Further, many SNPs are localized close to previously detected markers controlling several agronomic traits such as peduncle length, maturity and plant density adaptation (Sukumaran et al., 2015a,b).
Two distinct SNPs (kukri-c29668-338 and BS00009866-51) of SET were detected at the same position (at 70 cM on chromosome 1A). The latter SNP, a homolog of MSBP1 in Arabidopsis, is implicated in diverse hormone signaling (Balmer et al., 2006) and negatively regulates brassinosteroid (BR) signaling (Song et al., 2009). In young panicles of rice, a putative serine carboxypeptidase-like protein, OsGS5 (Oryzae sativa Grain Size 5), which is implicated in natural variation of grain size (Li et al., 2011), has been shown to competitively inhibit MSBP1, and thereby promote BR signaling and grain size (Xu et al., 2015). The OsGS5 homolog in wheat, TaGS5, has been recently cloned and shown to be associated with grain weight (Wang et al., 2015a,b). Whether TaGS5 and its promoter polymorphism cause MSBP1 inhibition and promote BR signaling and spike growth in wheat remains to be investigated.
The spike-related traits are important components of wheat yield that are under complex genetic control and are more highly heritable (61% in this study) than yield per se (Ma et al., 2006). We detected 22 and 142 SNPs for SDW in glasshouse and field conditions, respectively (Fig. 7; Table 1). Of these, six SNPs were common across the two conditions. The SNP wsnp_Ku_c5693_10079278 on chromosome 7A controls the stem water-soluble carbohydrate (WSC) concentration, while another SNP, RAC875_rep_c109658_211 on chromosome 5B, is localized at 31 cM, close to the marker that controls stem WSC content, as reported in the double-haploid population of wheat (Rebetzke et al., 2008). The allelic diversity of the former SNP also revealed a significant effect on SDW (Fig. 8c). In addition, an SNP on chromosome 6B, related to stem WSC remobilization efficiency, is localized at 82 cM, close to a fructan catabolic gene, 1-fructan exohydrolase (1-FEH w3), which triggers stem fructan-based WSC release that is indeed used to buffer grain-filling in wheat, and its expression is regulated by hormones including ET (Zhang et al., 2014). These findings emphasize a key role for the contribution of stem WSC to SDW. A spatiotemporal analysis of WSC and hormones (Valluru, 2015) in different spike organs is therefore required if we are to track spike development under heat stress.
In addition, a vernalization gene, vrn-A1 on chromosome 5A, controlling days to heading, has been detected across two conditions (Lopes et al., 2015). This indicates a key role for phenological phases in modifying SDW. It has long been known that an extended stem elongation period greatly improves SDW in wheat (Miralles et al., 2000). An SNP associated with maturity on chromosome 2B (Sukumaran et al., 2015a,b) is 4 cM from the SNP detected in this study, whose allelic diversity (CC vs TT) had a significant effect on SDW (Fig. 8d). This SNP is predicted to be an E3-ubiquitin ligase (He et al., 2013), which is involved in the regulation of ET action by targeting ubiquitination of ACS proteins, particularly types 2 and 3 (Lee & Seo, 2015). Of the 873 E3-ligase genes identified so far in wheat, 173 genes are expressed differentially during grain development (Capron et al., 2012). Some of them are closely linked to ET signaling, and may underlie the process of balancing growth and stress tolerance (Lee & Seo, 2015). Previously, E3 APC/C ligases have been shown to be positively associated with grain yield in maize (Fu et al., 2010). These findings emphasize a crucial role for ET signaling in the regulation of SDW.
Interestingly, two of the SET SNPs, kukri-c29668-338 and BS00009866-51 on chromosome 1A, and one of the SDW SNPs, Kukri-c32958-390 on chromosome 4B (Table 1), are localized close to the previously detected markers of plant height (Lopes et al., 2015; Sukumaran et al., 2015b), establishing a causal genetic link between SET and plant height. Subsequent analysis suggests that allelic variation in SET SNPs (AA vs GG) had appreciable effects on SET and height, with the effect on SET being opposite to that on height. By contrast, allelic variation (CC vs TT) in an SDW SNP had effects on SDW and height in the same direction. The height difference between two distinct alleles of these SNPs was, however, not significant (data not shown), as expected, because the WAMI panel is corrected for height variation (Lopes et al., 2015). Nevertheless, the observed directionality of the allelic effects suggests that genetic pleiotropy may act on these traits. In support of this conclusion, physiological data show significant negative and positive correlations of SET and SDW with plant height, respectively (Figs 4b–e, S4a–h). These findings call into question the physiological benefits of using dwarf plants to enhance yield in future high-temperature scenarios, although a stem-specific ET manipulation would be useful to alter plant height, as shown recently in poplar (Populus alba) (Plett et al., 2014).
Acknowledgements
The authors thank the field operational staff of the wheat physiology group at CIMMYT's Experimental Station in Obregon for assistance with field experiments. The authors also thank Guofeng Li for assistance with the glasshouse experiment. This work was financially supported by the Consultative Group for International Agricultural Research (CGIAR) Research Program for Wheat (CRP-Wheat, SI6: Heat and Drought) through CIMMYT. The authors thank two reviewers for their constructive comments on a previous version of the manuscript.
Author contributions
R.V., M.P.R. and W.J.D. devised the experimental design; R.V. carried out the experimental work and data analysis; R.V., and S.S. performed the GWAS analysis; R.V. wrote the manuscript and all authors reviewed and commented on the manuscript.
