Hexaploid bread wheat (Triticum aestivum L.) is a major commodity for export in Kazakhstan and is grown annually on more than 12 million hectares. The history of wheat cultivation in Kazakhstan shows that most of wheat cultivars have been developed in collaboration with Russian breeders and using Russian wheat genetic resources [1]. Even after the breakup of the USSR, this trend is still in the place as the two countries share their expertise and genetic resources based on bilateral projects and international activities under the CIMMYT umbrella [2, 3]. It is also found that Kazakh-Russian wheat germplasm is genetically close to wheats from the USA [4]. It is hypothesized that the heavy importation of this crop from Russia at the end of the nineteenth century after the successful introduction of Turkish Red Wheat types to the US by Russian Mennonites [5] might be a main reason behind the close genetic relationship of Kazakh and US accessions [4].

Over 80% of the wheat harvesting area in Kazakhstan is grown with spring type and it is cultivated at higher latitudes in parts of the country, including Northern and North-Eastern Kazakhstan. The other important growing regions stretch along the Tian-Shan mountain chain in Southern and South-Eastern parts, where both winter and spring types are grown successfully. The climatic conditions in these regions are very variable, as are the soil types, the temperature during the growing season, the precipitation levels, and the photoperiod length [6]. Therefore, studies of yield performance in different ecological niches are important for strategies in current and future breeding activities across wheat growing regions of the country. It is projected that the improvement of agronomy and use of new breeding methods in this country may lead to the development of new varieties, and, consequently, improve the yield productivity up to 3 tons per hectares [7]. In the past, plant breeders successfully relied on using conventional tools and methodologies. Nowadays, the availability of new genomic tools and resources is leading to new opportunities to dissect the genetic mechanisms of complex traits associated with yield improvement [8].

As costs for high throughput genotyping are decreasing, genome-wide association studies (GWAS) are becoming a powerful approach for the detection of QTL (quantitative trait loci) associated with wheat agronomic traits, with the final goal of accelerating local breeding activities based on the application of marker-assistant selection [9, 10]. The success of GWAS in wheat is largely based on the development of high-density SNP genotyping platforms by Affymetrix [11, 12] and Illumina [9, 13], which are now providing rich resources for high-throughput genotyping data for wheat diversity panels. In GWAS, detection of significant associations relies primarily on genetic marker coverage, the number of individuals studied, and linkage disequilibrium (LD) between causative and linked polymorphisms [14, 15]. Currently, GWAS has been used successful in hexaploid wheat for identification of QTL for yield components [16–18], abiotic stress resistance [19–21] disease resistance [22, 23], and grain quality [24, 25]. A survey of the literature shows that GWAS is actively applied in wheat studies in many different parts of the World, including North America [26], Central America [25], Europe [18, 24], Africa [19], Australia [22], and Asia [20].

Although GWAS has proven to be a very efficient approach for capturing important marker-trait associations (MTA), results reported from studies in different regions of the World are revealing the tendency for a strong influence of the growth environment in which yield QTL are identified with significant genotype x environment interaction revealed (GEI). For instance, results obtained from three different GWAS studies related to identification of QTL for yield performance in Europe [24], India [27], and Mexico [16] showed different responses and QTL for yield components in different parts of the genome. This trend is also confirmed in studies when the same germplasm was tested in different regions of Asia [20]. This outcome is congruent with result reported by Quarrie et al. (2005) from studies using bi-parental mapping populations [28], and can be explained by the sensitivity to environmental factors at crucial growth phases which determines the potential number of grains per ear [29]. Therefore, the success of regional projects may largely depend on separate, local, GWAS experiments using genotyped adapted germplasm. The main goal of this work was GWAS using spring wheat accessions from Kazakhstan, Russia, Europe, and CIMMYT (Mexico) for identification of MTA in field trials in three diverse environments of Kazakhstan. The study is the first attempt to employ GWAS for identification of important QTL and enhancing of spring wheat breeding projects in this county.

The spring wheat panel consisted of 96 commercial and prospective cultivars from Kazakhstan and the Russian Federation, 38 cultivars from Europe, and 60 CIMCOG (CIMMYT Mexico Core Germplasm) lines (CIMMYT, Mexico) (Additional file 1). Currently, 61 cultivars from Kazakhstan and Russia in this genetic panel have been registered through the State Seed Trials Commission of the Republic of Kazakhstan (2015), and are grown officially in Kazakhstan. The panel also included 29 prospective cultivars developed in Kazakhstan and Russia (Additional file 1). The European cultivar collection predominantly comprised accessions originating in the United Kingdom. The CIMCOG lines are a special population developed for studying opportunities for improvements in photosynthesis and biomass [30]. The field trials were conducted in three different latitude regions of Kazakhstan (Additional file 2), specifically at the Karabalyk breeding station (Northern Kazakhstan), Karaganda breeding Research Institute (Central Kazakhstan), and at the Kazakh Rice Research Institute (Southern Kazakhstan). The collection was planted at each site in randomized experiments each of three replicates in the seasons of 2013–2015. The distance between rows was 15 cm and the distance between plants within a row was 5 cm. The experiments in the Northern and Southern regions were conducted in 1 metre blocks, while in the Southern region the accessions were planted in 3 rows per repetition. In total, the data for mean values of 12 agronomic traits of the 194 hexaploid wheat accessions harvested in nine environments were subjected to further statistical analysis. The 12 traits included the following: days to booting (BD), days to heading time (HT), days to maturity (MT), thermal time at heading (TT-H), thermal time at maturity (TT-M), plant height (PH), peduncle length (PL), number of fertile spikes (NFS), number of kernels per spike (NKS), thousand kernel weight (TKW), dry biomass per plant (BPP) and yield per plant (YPP).

DNA samples were extracted and purified from single seeds of individual cultivars using commercial kits (Qiagene, CA, USA). The DNA concentration for each sample was adjusted to 50 ng/μl. Accessions were genotyped using the wheat 90 K Illumina iSelect SNP array as described in [4].

Statistical analyses of data, including multiple factor ANOVA, Pearson’s correlation and t-test were calculated using the software package GraphPad Prism 5.0 [31]. GGE Biplot methods were employed by using the GenStat package (17th release, VSN International, Hertfordshire, UK). The symmetric scaling option of both methods and available field data for all three sites were used in estimations.

GWAS analysis of QTL governing plant growth stages and yield parameters in the set of 194 accessions was performed with the TASSEL 5.0 package [32]. For this, the SNP dataset was filtered using a 10% cutoff for missing data and only markers with a minor allele frequency ≥ 0.10 were considered for GWAS. The STRUCTURE and STRUCTURE HARVESTER [29] programs were used for the development of delta K values (ΔK) and Q-matrix for identified clusters.

To facilitate the discovery of MTAs, three groups of wheat accessions with different breeding origins were studied. In genetic terms, the accessions were clearly separated into the three subgroups, with the first subgroup (Russia/Kazakhstan) being the most diverse, and the third subgroup (CIMCOG) being the least diverse group in the analysis. The separation of accessions into the three subgroups was also congruent with population structure analysis using on STRUCTURE software.

Yield performance during the 3 years was lowest in Northern Kazakhstan. Yield performance in the Southern region was always higher. However, the size of arable land in this region is limited and insignificant in comparison with the wheat growing area in Northern Kazakhstan. Results suggest that accessions in the EU collection can successfully be used in the Central region for obtaining better yield, and selected CIMCOG lines can be efficiently used for improvement of TKW in all three regions. The GGE biplot based on yield data helps to confirm that groups within the three breeding origins have different genetic backgrounds, and local accessions are well adapted to Northern and Southern regions.

Incorporation of the CIMCOG lines in the analysis resulted in a reduction of polymorphic SNP markers available for GWAS, as genotyping data for those lines was restricted to those shared between the Axiom and Illumina SNP arrays. Therefore, only 3245 aligned polymorphic SNPs were used in the GWAS with 1340 SNPs positioned in the A genome, 1448 in the B genome, and 457 in the D genome. The Diversity index was relatively high in the A and B genomes and lower in the D genome, which is well in agreement with previous observations [10, 18]. GWAS was performed separately for nine field trials over 3 years and identified 114 MTAs (Additional file 6). Only 12 of the identified MTA were significant in two environments both within and between regions (Additional file 6), suggesting that stability of the associations was undermined by a strong influence of environmental factors.

Despite similarities in the genetic positions of MTAs identified with other studies, a number of MTAs from this study were missing in GWAS conducted in different geographic regions. This study revealed four MTAs related to NKS, which is one of the major yield components in wheat [28]. Surveys of the literature suggests that the loci identified on chromosomes 3B (56.9 cM) and 5B (144.1 cM) were not identified in previous GWAS studies in other regions of the World [17, 25, 28] nor in Quarrie et al. [28]. Therefore, it was hypothesized that these MTAs, along with other associations shown in Table 3, are new MTAs.

Recently, there have been a number of discussions related to the importance of size and level of genetic variation in diversity panels for the success of GWAS projects [14, 15, 37]. It was pointed out that experiments with less than 384 accessions [14] and large LD blocks [15] may lead to the identification of false positive associations. The study by Turner et al. (2016) indicated that smaller panels may allow the detection of false negative associations that would not have been detected in the larger panels [37]. Results in this study using a relatively small panel (n = 194) are largely in agreement with the study by Turner et al. [37].

The study confirms the efficiency of GWAS for the identification of molecular markers which tag important agronomic traits. In total 114 MTAs for 12 physiological and agronomic traits determined using spring wheat samples from Kazakhstan, Russia, EU and CIMMYT studied in field conditions of three regions of Kazakhstan. Locations of identified MTAs for plant height were similar with genetic positions corresponding to Rht-B1, Rht-D1, and Rht8 genes of wheat. In addition, from field trials it was found that EU and CIMMYT germplasm has high breeding potential to improve yield performance in Central and Southern regions of the country. The use of Principal Coordinate analysis clearly separated the panel into three distinct groups according to their breeding origin. The study identifies a network of key genes that will be further validated for improvement of yield productivity in wheat growing regions of Kazakhstan.