Detection of molecular markers associated with yield and yield components in durum wheat (Triticum turgidum L. var. durum Desf.) under drought conditions

Durum wheat is one of the most important staple food crops grown mainly in the Mediterranean region where its productivity is drastically affected by drought. Detecting the important traits and molecular markers associated with drought and grain yield traits will aid breeders to understand the genetic mechanisms of drought tolerance of durum wheat that will assist in the development of drought tolerant varieties through the use of marker-assisted selection (MAS). The study objectives were to identify putative chromosomes regions affecting grain yield expression under drought conditions. A population of 114 F8 recombinant inbred lines (RILs), derived by single-seed descent from a cross between "Omrabi 5" (drought tolerant variety) and "Belikh 2" (less drought tolerant), was used for this purpose. Phenotypic data of the RILs and parental lines were measured for 16 traits at Maru Agricultural Research Station (Jordan) for two consecutive seasons; 2006 to 2007 (annual rainfall 268.9 mm) and 2007 to 2008 (annual rainfall 230.6 mm). Marker-trait associations were investigated using a mixed linear model (MLM). A total of 32 simple sequence repeat (SSR) loci were significantly associated with these traits. For phonological traits, 18 markers were associated with morpho-phonological, while 31 markers were associated with yield and its components. Marker alleles from "Omrabi 5" were associated with a positive effect for the majority of traits for yield and its components. Six markers were closely linked with grain yield and these markers could be used for MAS in durum wheat breeding under drought conditions. 
 
   
 
 Key words: Durum wheat, marker-assisted selection, trait analysis by association, evolution, and linkage (TASSEL), recombinant inbred lines (RILs), simple sequence repeat (SSR), transgressive segregation.


INTRODUCTION
Wheat is the most important and strategic cereal crop for the majority of the world's population. It is the most important staple food crop of about 40% of the world population, occupying 17% of crop acreage worldwide, and providing nearly 55% of the carbohydrates and 20% of the food calories consumed globally (FAO, 2003;Gupta et al., 2008). Developing cultivars with high grain yield for drought-prone environments has been the principal aim of wheat breeding programs worldwide. In the Mediterranean region, it is of special interest because of the low and erratic distribution of rainfall, which has explained as much as 75% of the variation in wheat yield (Blum and Pnuel, 1990). Durum wheat (Triticum turgidum L. var. durum Desf.) is a tetraploid made up of A and B genomes (AABB); it is the main source of semolina for the production of pasta, couscous, burghul and other Mediterranean local end-products (Nachit, 1992). Detecting the important traits and the molecular markers associated with drought tolerance and grain yield traits will aid breeders to understand the genetic mechanisms of drought tolerance of durum wheat that will assist in the development of drought tolerant varieties through the use of marker-assisted selection (MAS).
Association mapping (AM) is a method used to identify association between genotype and phenotype (Yu et al., 2006), and facilitates the identification of agronomically valuable alleles using a reasonably low number of markers to reveal significant marker-trait associations (Breseghello and Sorrells, 2006). However, computational methods have been developed to account for population structure (Pritchard et al., 2000) and implemented in AM studies in a number of crop species including barley (Rostoks et al., 2006), wheat (Parker et al., 1998(Parker et al., , 1999Breseghello and Sorrells, 2006;Crossa et al., 2007;Tommasini et al., 2007) and rice (Agrama et al., 2007).
Most AM studies in wheat have focused on specific chromosomes where major quantitative trait locus (QTL) have previously been reported and there is still a lack of published literature on genome-wide AM studies of important target traits in wheat. Although AM is intrinsically more powerful than genetic linkage mapping because it scrutinizes the results of thousands of generations of recombination and selection (Syvänen, 2005;Cattivelli et al., 2008), most of the data available up to date, on drought tolerance, are based on segregation mapping and QTL analysis. Future studies of AM in crop plants will also elucidate further the structures of plant genomes and will facilitate the use of MAS and map based cloning of genes for difficult traits (Gupta et al., 2005). Therefore, the aim of this study was to determine the number of markers linked with grain yield and its related traits in durum wheat under dry conditions.

Plant material
The plant material used is a population stemmed from a cross between "Omrabi 5" and "Belikh 2". The population is constituted of 114 single seed descent recombinant inbred lines (RILs) developed by Nachit in 2005 at ICARDA durum wheat breeding program (Nachit, Personal Communication). The "Omrabi 5" durum cultivar is a cross between the landrace "Haurani" and the improved cultivar "Jori-C69"; "Omrabi 5" and "Belikh 2" were developed for the Mediterranean dryland conditions (Nachit, 1998). "Omrabi 5" is released in Jordan, Turkey, Algeria, Iran, and Iraq for commercial production; it combines drought tolerance with high and stable yields. "Belikh 2" is released in Algeria, Lebanon, and Syria; it was bred at ICARDA (Crane/Stork) and developed for rain-fed areas. It is early heading and maturing and shows good protein quality for pasta processing.

Testing environments
The RILs trial was grown at Maru experimental station (NCARE-Jordan) for two consecutive seasons: 2006 to 2007 and 2007 to 2008. Maru station is located in Jordan at 32° 36 N latitude, 35° 40 E longitude and 620 m above sea level, with silt clay soil with an average annual precipitation of 425 mm. The experiment was arranged in an augmented complete block design (Federer, 1956) in six blocks, with 1.5 m between them. Each block consisted of two sub-blocks, with 50 cm in between. Twelve genotypes, grown each in four rows (2.5 m long rows, spaced 0.35 m apart) were tested per each sub-block. Five checks per each block were used: "Omrabi 5", "Belikh 2", "Cham 1", "Korifla", and "Haurani Nawawi". Seeding rate was 120 kgha -1 . Fertilizers were applied at sowing date at a rate of 60 kgha -1 nitrogen units (NH4 + ; ammonium form) and 40 kgha -1 H2PO4 for both growing seasons and weeds were mechanically controlled.

Data collection
In addition to grain yield, information on adult-plant characters was taken for the two consecutive seasons at Maru experimental station. Days to heading (DH) was recorded as the number of days from emergence, to the day when half of the spikes appeared in 50% of the plants. Days to maturity (DM) was recorded as the number of days from emergence to the day when the peduncle was completely discolored in 90% of the plants. Plant height (PH) was measured at harvest maturity from the ground level to the top of the spikes excluding awns. Peduncle length (PL) was measured from the node to the ligule of the flag leaf as the mean of 10 random spikes. Spike length (SL) was measured from the base to the top of the spike excluding the awns as the mean of 10 random spikes. Awns length (AL) was measured from the top of the spike to the top of awns as the mean of 10 random spikes. Number of tillers (NT) was accounted in two 2.5 m row segments, 35 cm apart and converted to square meters. Number of fertile tillers (NFT) was calculated from the number of grain-bearing culms in two 2.5 m row segments, 35 cm apart and converted to square meters. Main spike weight (SW), number of grains per spike (NGS) and weight of grains per spike (GWS), and number of spikelets per spike (NSS) were determined from the mean of 10 random spikes. 1000 grain weight (TGW) was measured by weighing 200 grains taken randomly from the harvested seeds and converted to the weight of 1000 grains. Biological yield (BY) was measured as the weight of above-ground dry matter (straw + grain) harvested from the inner two 2.5 m row segments, 35 cm apart and converted to kgha -1 . Grain yield was measured as the weight of grain harvested from the inner two 2.5 m row segments, 35 cm apart and converted to kgha -1 . Straw yield was calculated as the difference between biological yield and grain yield.

DNA extraction
Leaves from seedlings grown in the field were used for DNA extraction. The DNA was extracted following the protocol developed at ICARDA durum wheat MAS laboratory (Nachit et al., 2001) and quantified by the spectrophotometer (S2100 Diode Array Spectrophotometer).

Microsatellites (simple sequence repeat, SSRs)
Wheat microsatellites consortium (WMC) and Beltsville Agriculture Research Center (BARC) were used as described by Nachit et al. (2001). Molecular analysis was conducted at ICARDA durum wheat MAS laboratory. The parents were screened for polymorphism using 300 SSR primer pairs, out of which 96 SSR gave clear, reproducible and polymorphic amplification products. The polymerase chain reaction (PCR) amplification was carried out in Eppendorff thermal cycle, in a 7.5 μl reaction mixture. Each reaction contained 1X Taq polymerase buffer, 200 μM of each dNTPs, 0.5 μΜ of each of the two primers, 1U Taq polymerase, and 20 ng of genomic DNA as a template. Touchdown PCR cycling was performed as follows: an initial denaturation was performed at 94°C for 5 min, followed by 35 cycles of denaturation at 94°C for 1 min, annealing in 1°C incremental steps between 63 and 56°C for 1 min, extension at 72°C for1 min), with a final extension at 72°C for 5 min. PCR products were mixed with loading buffer, 5 to 10 μl of mixture was denaturated and loaded into wells in 0.4 mm thick 15% acrylamide gel (20 × 20 cm) resolved at constant power (30 w) in 1X TBE running buffer for 15 min to 1 h depending on size of the primer pairs of SSRs. Bands were visualized by a silver-staining method as described by Nachit et al. (2001). Dried gels were visually scored and scanned for records.

Statistical analysis and AM
The statistical analysis was performed using the MIXED procedure of the SAS statistical package (SAS, 1998). The broad-sense heritability values (H 2 ) were calculated as ratios of genetic and phenotypic variances (σ 2 g/σ 2 p), where σ 2 p is the phenotypic variance and was estimated as σ 2 g + σ 2 e, σ 2 g is the genetic variance and σ 2 e is the residual variance. Pearson's correlations between phenotypic traits and phenotypic frequency distribution were performed using SPSS 17.0 statistical software.
Because of the low number of molecular markers probed in this study to construct a genetic map, we have opted for AM between molecular markers and morpho-phonological traits under investigation. We used a mixed linear model (MLM) which is implemented in the software package version 2.0.1 (http://www.maizegenetics.net) where the marker was considered as a fixed-effect factor and the lines of the population considered as a random-effect factor (Kennedy et al., 1992). Significance of associations between marker loci and traits under investigation was based on an F-test, at a level αc corresponding to α corrected for multiple testing. Corrected significance levels αc were computed by 1000 permutations. The locations of these markers were identified following the genetic maps carried out at ICARDA durum wheat MAS laboratory (Nachit, Personal Communication).
The least significant difference (LSD), coefficient of variation (CV), and additive effect of the marker were estimated using Genstat (Version 11). In general and with few exceptions, the alleles of "Omrabi 5" (Mrb) had a positive additive effect on increasing trait value, whereas "Belikh 2" (Blk) had a negative one.

Weather conditions
The annual rainfall and its monthly distribution differed from year to year at Maru station, Jordan. The total amount of rainfall was 268.9 and 230.6 mm for Maru in the 2006 to 2007 and 2007 to 2008 growing seasons, respectively (Table 1). In general, rainfall was lower than the long-term average (425 mm) in both growing seasons. In both growing seasons, during the post-anthesis developmental stage, rainfall in the months of April and May was low and much less than the crop requirements. This low rainfall and its poor distribution in both growing seasons affected crop performance; it subjected the crop to severe drought stress, particularly during the grain filling period. However, the months of April and May

Phenotypic evaluation
A total of 114 lines and their parents ("Omrabi 5" and "Belikh 2") were investigated under drought conditions. The grand means, standard deviation, and ranges of measured 16 agronomic traits for the parents and the RIL population are presented in Table 2. The two parents showed a significant difference in all 16 traits. "Omrabi 5" was earlier in DH, DM and taller than "Belikh 2". Also, "Omrabi 5" showed larger/greater NFT, NGS, GWS, PL, TGW, and yield (grain, biological, and straw) than "Belikh 2". In contrast, "Belikh 2" showed larger NSS, SL, and AL than "Omrabi 5". The largest ranges were shown for PH, NFT, PL, AL, NGS, TGW, BY, and GY ( Table 2). The phenotypic distributions of all examined traits for the lines displayed a continuous normal pattern. Obviously, these traits were quantitatively inherited. In addition, transgressive segregation in both directions was observed for all traits (Figure 1), and the parent "Omrabi 5" transmitted favorable alleles for the majority of the traits. The combined mean, LSD at 5% level, CV (%), and heritability (H 2 ) for all studied traits are shown in Table 3.
High heritability values were shown for DH, DM, PH, PL, SL, AL, NFT, GWS, NGS, SW, NSS, and TGW, moderate values for NT, and BY; and low values fo r GY, and SY (Table 3). Significant positive correlations (P < 0.01) were observed between GY and PH, PL, NT, NFT, GWS, NGS, TGW, BY, and SY; and negative correlations between GY and DH, and DM. The strongest positive correlation was between BY and SY (r = 0.79), whereas the correlation between GY and SL, AL, NSS, and SW was not significant (Table 4).

Marker-trait association
The trait analysis by association, evolution, and linkage (TASSEL) program was used to detect associations between markers and phenotypic data by using a MLM of the combined values for the two growing seasons. A total of 32 SSR markers showed significant associations with 16 agronomic traits on all 14 chromosomes of durum wheat ( Table 5). The significantly associated markers with these traits are identified as putative QTLs. The detected QTL individually accounted for 2.41 to 7.42% of the phenotypic variation. The positive effects of the linked markers originated from the parent "Omrabi 5" (increasing effects of QTL); whereas the markers with negative effects originated from the parent "Belikh 2". A list of the putative markers loci along with their df, F-test, ρ-values association, and additive effects are presented in Table 5. Most of the significantly associated markers with the examined traits were mapped on Chromosome 2A. The results indicated that this region of Chromosome 2A is important for drought tolerance in durum wheat. It may be assumed that there is a QTL cluster for drought tolerance in the region of Chromosome 2A (Table 5), and thus the region may be used as an important target for improving drought tolerance of durum wheat.

Phenological traits
Three markers showed significant association with DH. The Xwmc177 (P < 0.06) located on Chromosome 2A accounted for 5.98% of the phenotypic variation. In addition, the marker Xwmc388 (P < 0.012) accounted for 2.54% and the marker Xwmc322 (P < 0.093) accounted for 4.90%; both markers are located on 3B. The added effects of these 3 markers have accounted for 13.45% of the total variation for DH. The significantly associated markers are defined as QTLs and all these QTLs have alleles from "Omrabi 5" (  worked on bread wheat. Five significantly associated with DM were identified; these markers are localized on Chromosomes 2A, 3A, 5A, 5B, 6A, and 6B. Their contributions ranged from 2.54 to 5.75% of the total variation. All of these markers, except for Xbarc309, had alleles from "Omrabi 5" ( Table 5). The results of the present study agree with previous study of Merza et al. (2006) for the location of QTLs for DM on Chromosomes 2A and 6B, except for the QTL on 1B which was not detected in our study.

Morphological traits
For PH, two significantly associated markers were detected. The marker Xwmc177 (P < 0.095) located on Chromosome 2A accounted for 2.49% of the phenotypic variation. In addition, the marker, Xwmc24 (P < 0.098) located on Chromosome 1A accounted for 2.44% of the phenotypic variation. The total contribution of both markers was 4.93%. The positive alleles for this trait are from "Omrabi 5" (Table 5) Five putative markers were significantly associated with PL. These markers were detected on Chromosomes 1B, 2B, 3A, 3B, 4A, 5A, 6A, 6B, and 7A. Their contributions for the phenotypic variation ranged from 3.05 to 4.95%. Similarly to plant height, "Omrabi 5" alleles have contributed to an increased PL (Table 5).

Yield components
In this study, five significantly associated markers with SL were localized on chromosomes 1B, 4B, and 7B and their contributions ranged from 2.42 to 3.37%. All of these markers are from "Omrabi 5", except, Xwmc415 and Xwmc617 are from "Belikh 2" (Table 5). All QTLs identified in this study were mapped earlier by Merza et al. (2006) who identified the location of QTLs on Chromosomes 1A, 1B, 2B, 3B, 4B, 5B, 7A, and 7B.
The NT was found to be significantly associated with two markers: Xwmc35 (P < 0.013) and Xbarc100 (P < 0.031), located on Chromosomes 2A, and 2B and accounted for 5.83 and 4.1% of the phenotypic variation, respectively. The alleles of these markers, which increased NT, originated from "Omrabi 5" (Table 5). Our results were in disagreement with Kumar et al. (2007) and Kaur et al. (2008) that worked on bread wheat, identified three QTLs for NT on Chromosomes 3A, 7A, and 7B. On the other hand, four different markers for NFT were detected on Chromosomes 4A, 4B, 5A, 5B, and 7A and their contributions ranged from 2.51 to 3.83%. All positive alleles originated from "Omrabi 5" (Table 5). These detected markers were not found in the previous studies (Quarrie et al., 2005;Marza et al., 2006).

DH
For TGW, 3 significantly associated markers were detected. The marker, Xwmc322 (P < 0.016) located on Chromosomes 3A, and 3B, accounting for 5.57% of the phenotypic variation, and its positive alleles came from "Omrabi 5". Another marker, Xwmc475 (P < 0.070) was identified on Chromosomes 5A, and 7B. This marker accounted for 2.99% of the TGW variation, and its positive alleles also came from "Omrabi 5", whereas, the third marker, Xbarc100 (P < 0.027) was detected on Chromosomes 2A, 2B, and 5AL accounting for 4.31% of the phenotypic variation, and its alleles came from "Omrabi 5" ( Table 5). The total contribution of the 3 markers was 12.87%. These markers are located on Chromosomes 2A, 3A, and 7B and confirm the QTLs location detected by Sun et al. (2009), andYao et al. (2009).

Source of alleles for drought tolerance
Drought tolerance genes are located throughout the genome and are genotype dependent. In this study, it was found that for grain yield, 4 significantly associated markers were derived from "Omrabi 5". However, some of the markers were also derived from a drought sensitive parent of the population. Börner et al. (2002) and Huang et al. (2003) found similar results for other traits of interest. These findings further confirm that drought tolerance is a quantitative trait and that apparently sensitive parents may contain alleles for tolerance, which may not be found in the tolerant parents. It can be concluded that the sensitive parent "Belikh 2" may contain some tolerance alleles that when combined with alleles from tolerant parents can result in an increased level of tolerance.

Perspectives for MAS
Molecular markers, closely linked to genes of agronomic important traits, have been identified to be useful markers for an indirect selection in durum wheat breeding programs (Nachit, 1998). MAS is time efficient (Landjeva et al., 2007), non-destructive and depending on associated relationships, characterized by low selection error (Nachit et al., 2000).
Further investigations for drought tolerance will be required to establish the importance of the identified genomic regions in other backgrounds. In addition, field evaluation is required to establish the effectiveness of the drought screening system in modeling water responses and in evaluating the stability of QTLs across environments (Mohan et al., 1997). Our results indicate the existence of markers/genes, which are involved in the control of the phenotypic variation in quantitatively inherited traits related to drought tolerance.
Further research is needed on molecular markers and QTLs mapping to screen potential genotypes for drought tolerance in durum wheat. The current study represents potential targets for MAS of grain yield improvement in durum wheat. These were Xbarc59, Xbarc70, Xbarc100, Xwmc322, Xwmc388, and Xwmc398.

Conclusions
This study has shown the importance of using molecular markers to identify traits and markers for further improving durum performance under drought conditions. AM revealed the presence of a number of significantly associated markers with grain yield and its related traits. Several of these markers overlapped with QTLs mapped in other durum and bread wheat mapping populations. In addition, we are working on developing genetic linkage maps with marker saturated chromosomes.
The comparative analysis carried out, considering the result from QTL analysis obtained so far in durum wheat and bread wheat, provide further supporting evidence for the relative importance of the identified markers, chromosome regions, and the potential value for the application of MAS to the most agronomic traits of interest in durum wheat.