Detection of quantitative trait loci associated with drought tolerance in St. Augustinegrass

St. Augustinegrass (Stenotaphrum secundatum) is a warm-season grass species commonly utilized as turf in the southeastern US. Improvement in the drought tolerance of St. Augustinegrass has significant value within the turfgrass industry. Detecting quantitative trait loci (QTL) associated with drought tolerance will allow for advanced breeding strategies to identify St. Augustinegrass germplasm with improved performance for this trait. A multi-year and multi-environment study was performed to identify QTL in a ‘Raleigh’ x ‘Seville’ mapping population segregating for phenotypic traits associated with drought tolerance. Phenotypic data was collected from a field trial and a two-year greenhouse study, which included relative water content (RWC), chlorophyll content (CHC), leaf firing (LF), leaf wilting (LW), green cover (GC) and normalized difference vegetative index (NDVI). Significant phenotypic variance was observed and a total of 70 QTL were detected for all traits. A genomic region on linkage group R6 simultaneously harbored QTL for RWC, LF and LW in different experiments. In addition, overlapping QTL for GC, LF, LW and NDVI were found on linkage groups R1, R5, R7 and S2. Sequence alignment analysis revealed several drought response genes within these regions. The QTL identified in this study have potential to be used in the future to identify genes associated with drought tolerance and for use in marker-assisted breeding.


Introduction
St. Augustinegrass (Stenotaphrum secundatum [Walt.] Kuntze) (2n = 2X = 18) is a warm-season turfgrass that is well adapted to tropical and subtropical regions of the world [1]. It has been a popular turfgrass in the southern United States for its broad leaf blades and rapid stolon elongation, which makes the grass well-suited for sod production [2]. However, one of the greatest challenges that turfgrass industry facing is the limited availability and reduced quality of water for irrigating turfgrass areas. To address these concerns, there is a need to develop St. Augustinegrass cultivars with the improved levels of drought tolerance. Considerable efforts have been devoted to determining the drought response of St. Augustinegrass, including turf quality, leaf firing, percent green cover, canopy temperature and root characteristics [3,4]. PLOS  and on 2 January 2018 for GH18 until soil water content reached approximately 5%, which occurred on 10 th day for GH17 and 14 th day for GH18. Soil water content was monitored using a TDR 100 Soil Moisture Meter (Spectrum Technologies, Inc. IL). At the end of the drought stress treatment, leaf wilting (LW) and leaf firing (LF) were visually scored on a scale of 1 (no wilting/firing) to 5 (severe wilting/firing) according to National Turfgrass Evaluation Program (NTEP) guidelines (Fig 1) [16]. Leaf relative water content (RWC) was determined according to the formula: RWC = (FW-DW)/(TW-DW) × 100, where FW is fresh weight, DW is dry weight and TW is turgid weight. Relative chlorophyll content (CHC) in leaf was measured using MultispeQ v2.0 (PhotosynQ Inc, East Lansing, MI). In addition, percent green cover (GC) was calculated using digital image analysis with ImageJ software on third day after recovering from stress [17].

Field experiments
A field trail was planted in June 2016 at the Sandhills Research Station. Plots were 0.91 m × 0.91 m in size. Plots were irrigated and fertilized according to recommended practices for NC and mowed weekly at a 6.35 cm height. Drought stress was applied during the annually recurring drought period for the region. Data were not collected in summer of 2017, as turfgrass were not stressed due to short continuous drought period. After drought stress started on 12 August 2018 and lasted 18 days, visual ratings were taken for leaf firing (LF) using a scale of 1 (no firing) to 5 (severe firing) according to NTEP guidelines [16]. Normal Difference Vegetation Index (NDVI) were assessed using a TCM 500 NDVI Turf Color Meter (Spectrum Technologies, Inc, Aurora, IL).

Experimental design and statistical analysis
Both greenhouse and field experiments were arranged in a randomized complete block design with three replicates. Analysis of variance (ANOVA) was performed using SAS PROC GLM procedure (SAS Institute, Cary, NC) with replication as random effect. Correlation analysis between traits and trails was performed using SAS PROC CORR procedure (SAS Institute, Cary, NC).

QTL detection
QTL analysis was performed using the integrated two-way pseudo-testcross approach with MapQTL 6.0 [18]. This approach was applied by analyzing data for each parental meiosis separately. Kruskal-Wallis (KW) analysis, interval mapping (IM) and multiple QTL method (MQM) analysis were performed to detect significant associations between markers and phenotypic traits using a regression approach. Genome-wide LOD thresholds (p < 0.05) were determined for each trait using a permutation test with 10,000 iterations. Regions with a LOD score above threshold values were considered as potential QTL intervals. Lastly, the sequences flanking SNP markers that fall within the identified regions of interest were searched against the NCBI NR database using blastn/blastp tools to obtain their orthologs. Gene annotation was conducted using the UniProt database to predict gene function in the QTL regions.

Phenotypic trait variation and correlation under drought stress
The hybrids in the 'Raleigh x Seville' population showed a wide range of phenotypical variation in all evaluated traits and all independent trails (Table 1, S1  (Table 1 and S1 Table). No significant correlations among traits were observed between greenhouse and field experiments. However, significant correlations were found between traits under the same conditions. In experiment SRS18, a positive correlation was identified between NDVI and LF ( Table 2). In greenhouse evaluations, significant positive correlations (p-value < 0.01) were observed among GC, RWC, LF and LW (Table 2). CHC only showed positive correlations with LW (p-value < 0.01) and LF (p-value < 0.05), but did not with RWC and GC ( Table 2). In addition, the correlation analysis between GH17 and GH18 showed all traits have positive correlation between two experiments except LW (S2 Table).

QTL detection for drought-related phenotypic traits
Using the parental linkage maps developed for the 'Raleigh x Seville' population [14], a total of 70 QTL associated with six drought-related traits were identified in three experiments    (Table 3). Among them, 46 QTL were mapped on linkage groups for Raleigh, while 24 QTL were found on the Seville map. These QTL were identified on all of the 18 linkage groups with the exception of S6 and S7. Twenty-one QTL for GC were detected on linkage groups R1, R3, R4, R5, R6, R7, R8, R9, S2, S3, S4, S8, with explained variance from 3.1 to 10.7% (Table 3). Ten of them were identified in experiment GH17, while 11 of them were found in experiment GH18. A total of 18 QTL were found to be associated with LF, with nine in SRS18, six in GH17 and three in GH18. These QTL explained between 5.1 and 20.4% of the phenotypic variance, with LF-S2. 2 showing the highest explained variance and LOD value (Table 3). Forty QTL for LW were identified in the two greenhouse experiments, which explained 5.36 to 17.6% of the phenotypic variance (Table 3). Six and two QTL for RWC were detected in GH17 and GH18, respectively, with only one of them (RWC-S3.1) mapping on Seville's map (Table 3). There were only two QTL (CHC-R4.1 and CHC-R4.2) mapped for CHC and both of them were located on linkage group R4 (Table 3). A total of seven QTL were identified for NDVI in the field evaluation. These QTL were spread across linkage groups R7, S2 and S9 (Table 3).
Sequences flanking SNP markers in putative QTL were used to search for candidate genes for drought tolerance. Nine genes were found to be related to drought tolerance in these regions, including ZHD and WRKY transcription factors, ethylene-insensitive protein, coldresponsive protein kinase, OBERON-like protein, light-harvesting complex-like protein (OHP2), Magnesium-chelatase subunit(ChlD), Osmotin-like protein and LRR receptor-like serine/threonine-protein kinase (GSO1) ( Table 4).

Field
Greenhouse †

Discussion
QTL analysis has been a primary method used to identify molecular markers linked to drought tolerance related traits, which have the potential to improve selection efficiency for desired traits in breeding [10]. However, molecular markers linked to drought tolerance in turfgrass remain largely unexplored [12]. Genotyping and phenotyping have been the bottlenecks for establishing marker-trait associations of drought stress in turfgrass. This is mostly due to the limited genomic information available for turfgrass species and also to the multigene nature of the genetic control for drought tolerance. In recent years, with the adoption of genotype by sequencing approaches, high density genetic maps have been developed for several turfgrass species [11,12,13], including St. Augustinegrass [14]. The genetic maps developed from the 'Raleigh x Seville' mapping population contain nine linkage groups for each parent and consist of 2871 SNP markers [14]. These maps provide a powerful foundation for association analysis of traits of interest in St. Augustinegrass. In terms of phenotyping, morphological and physiological traits including GC, LW, LF, NDVI, RWC and CHC have been used as indicators of drought tolerance in the evaluation of turfgrasses [8,9,19]. A wide range of phenotypic variance for these traits was found in the 'Raleigh x Seville' population evaluated in this study under drought stress conditions under both greenhouse and field conditions ( Table 1, Fig 2). However, mean values for each trait were not correlated between different environments, which might be largely due to the enormous environmental difference present between greenhouse and field conditions. Greenhouse experiments provide a controlled, stable environment and reproducible evaluations, but are limited in the size of plots that can be used for evaluation. On the other hand, field experiments reflect 'real-world' responses under water deficit stress, but these can also be affected by complicated environmental factors [20]. Given  consideration of the environmental difference, both greenhouse and field phenotypic data were collected for QTL analysis in this study, which could enable us to detect drought response QTL in various environments. Significant associations between drought tolerance traits and SNP markers were identified in both greenhouse and field conditions ( Table 3, Fig 3). In the future, high-throughput phenotyping platforms might be used to improve the efficiency and precision of drought tolerance evaluation, which could lead to higher reliability in QTL mapping.
A total of 70 QTL associated with six drought-related traits were detected. Among them, 46 QTL were mapped on all nine linkage groups for Raleigh, while 24 QTL were mapped on seven of nine linkage groups for Seville (Table 3). These results suggested that two parents showed different patterns of genetic response to drought stress. Although most of the QTL were found only for a specific trait in one experiment, co-localization of some QTL for different traits and in different experiments was still found in this study (Fig 3). There were significant correlations among most traits under both greenhouse and field conditions. Thus, as expected, QTL for GC overlapped with QTL for LW, as did QTL for LF and NDVI. Likewise, QTL for LF and LW overlapped with QTL for RWC, suggesting that visual rating of LW and LF are reliable approaches for drought evaluation in turfgrass. Responses of LW and LF under drought stress have also been studied on other turfgrass species including zoysiagrass, bermudagrass and tall fescue [21,22]. Merewitz et al. also found the co-localization of QTL for chlorophyll content, RWC and NDVI in a creeping bentgrass population under drought stress [8]. Identification of QTL associated with multiple traits suggested there might be genes acting epistatically to promote drought tolerance. These findings highlight the value of these regions in gene discovery and function studies. In addition, some co-localized QTL attracted more attention as they were detected in different experiments, specifically under both greenhouse and field conditions. These QTL might be involved in drought response regardless of environmental effects. Beyond that, there were only two QTL identified for chlorophyll content (CHC) in this study, and these QTL were not co-localized with any other traits (Table 3, Fig 3). These results might be due to inadequate genetic variation for CHC among the population for QTL detection.
Through comparative genomic analysis, sequences of SNP markers in St. Augustinegrass were aligned to the genomes of Setaria italica, Sorghum bicolor and Oryza sativa [14]. Several markers closely linked to the QTL identified in this study could be annotated with genes of known function, especially those with functions in drought stress response (Table 4). It is well known that drought resistance in turfgrass may be affected by maintenance of an extensive and deep root system [23]. SNP58543 within QTL RWC-S3.1 had significant sequence match with gene GSO1, encoding an LRR receptor-like serine/threonine-protein kinase in Setaria italica. In Arabidopisis, GSO1 together with GSO2 were found to control primary root growth by modulating sucrose response after germination [24]. In addition, QTL LF-R8.1 was matched to gene OBERON, which was also reported to regulate root development through the auxin pathway [25]. Thus, these QTL markers could be related to root system development during drought stress. Marker SNP31882 showed high identity with OHP2, a gene encoding a lightharvesting complex-like protein, which has been reported to play a photoprotective role within photosystem I in response to light stress [26]. The sequence of SNP47240 was aligned to gene CHLD encoding a magnesium-chelatase subunit, which is involved in chlorophyll biosynthesis [27]. It is possible that these markers could respond to drought stress through photosynthetic regulation. However, further research is needed to determine gene function during drought stress in St. Augustinegrass and whether these SNPs could be used for marker-assisted selection.

Conclusions
Overall, significant phenotypic variance during drought stress was observed among the 115 hybrids in the 'Raleigh x Seville' mapping population. A total of 70 QTL associated with six drought tolerance traits were detected, with several QTL co-localizing in the same genomic region. Further candidate gene identification and gene function studies within these QTL regions will contribute to our understanding of the genetic control of drought tolerance in St. Augustinegrass. These QTL have potential value to be used to develop St. Augustinegrass cultivars with improved drought tolerance through marker-assisted breeding.