Using two cloning primers, genomic PCR amplifications were conducted, and one desired DNA band was detected in each accession of wild emmer wheat. Cloning the fragments yielded 244 positive clones from 13 accessions (randomly selected from 10 populations), which were subsequently sequenced (data not shown). Only three out of 244 dimeric α-amylase inhibitor genes had a common three bp deletion, and those three genes were obtained from one accession derived from Mt. Hermon, whereas the other cloned fragments had 426 bp long (data not shown). It was predicted that all of the 426-bp sequences would encode functional dimeric α-amylase inhibitors. Alignment of the gene sequences from emmer wheat with sequences from the species of Aegilops section Sitopsis (including Ae. speltoides, Ae. bicornis, Ae. longissima, Ae. searsii, and Ae. sharonensis), Ae. tauschii, einkorn wheats, and common wheat clearly indicated that the emmer wheat sequences were derived from the B genome 27.

The frequency of SNPs in the dimeric α-amylase inhibitor genes in emmer wheat was 1 out of 5.7 bases, which was higher than the SNPs observed for kunitz-type α-amylase inhibitor and α-amylase/subtilisin inhibitor genes in barley and dimeric α-amylase inhibitor genes in common wheat 282930. Among the 426 nucleotides, there were 351 conserved positions and 75 variable positions among the 244 α-amylase inhibitor genes sequenced from 13 accessions.

A total of 74 haplotypes were revealed by sequence analysis (Figure 1); 53 of these were each found in only a single sequence. Haplotype 41 was observed at the highest frequency, i.e., in 38 gene sequences, followed by haplotype 27 in 33 sequences (Figure 1).

The relationship between SNPs and amino acid changes in the α-amylase inhibitor proteins is summarized in Table 1. The 75 SNPs resulted in 38 amino acid substitutions. The position of each SNP in the sequence, whether the predicted change was synonymous (silent) or non-synonymous (replacement), was determined. Forty percent of SNPs were found to occur at the third codon position, and as expected, most of these were synonymous (Table 1). A number of changes were also identified in codon positions 1 and 2, and these accounted for more than 95% of the non-synonymous changes (Table 1). In total, 60% of the SNPs resulted in non-synonymous changes.

Using the information from the 75 SNPs identified in the α-amylase inhibitor genes, 16 primers (combined with the reverse cloning primer, R, as SNP markers) were successfully designed to detect the SNPs in 205 accessions from 18 populations. The primers, with the SNP (bold letters) at the 3' end and an extra mismatched nucleotide (underline) on the third nucleotide from the end are listed in Table 2. A total of 14 SNPs were detected with the 16 SNP markers from position 19 to 288 of the α-amylase inhibitor gene, and the size of the amplified fragments ranged from 158 to 426 bp. The data was then organized in terms of genotypic frequencies ("0" or "1") to assess the population structure.

There were only 5 and 2 accessions from Yehudiyya and Achihood, respectively. Thus, the data for Yehudiyya and Achihood were not used in further analyses. Positive fragment frequency for each primer in the 16 populations is listed Additional file 1.

Some genetic parameters of the 16 populations of wild emmer wheat are summarized in Table 3. The proportion of polymorphic loci P (5%), the expected heterozygosity He, and Shannon's information index of the 16 populations of wild emmer wheat were 0.887, 0.404, and 0.589, respectively. The values of He ranged from 0.182 to 0.437, and the population of Kokhav Hashahar had the highest value of He (0.437), followed by the population of Rosh-Pinna, whereas the population from Daliyya was characterized by the lowest He value of 0.182.

The genetic distances (D) were calculated for comparisons of all 16 populations based on the positive fragment of SNP markers among all population pairs (see Additional file 2). The highest genetic distance (0.263) was obtained between populations of Kokhav Hashahar and Daliyya, whereas the most related populations were Qazzrin and Gamla with a genetic distance of 0.017. However, lower D values (< 0.050) were observed between some populations from different areas, and, for the most part, the estimates of D value were geographically independent. Large genetic distances and sharp genetic differentiation over long geographic distances could be found. For example, Kokhav Hashahar in southern Israel had higher D values with the populations from Gamla (0.221), Nahef (0.247), Beit-Oren (0.215), Daliyya (0.263), and Bat-Shelomo (0.224) in northeast Israel.

To assess if some of the ecological factors are correlated to each other, principle components analysis (PCA) was carried out using 23 ecological factors as variables. A combination of the first four components could give us a high cumulative percent (88.81%) according to the eigenvalues of the correlation matrix (see Additional file 3A; 4), which could be used to explain the ecological associations. The main ecological factors of the first component were Rd and Ev (two water availability factors) that could give 40.60% eigenvalues, and the second component could give 28.66% eigenvalues (see Additional file 3). And from PCA analysis, it was known that the accessions from Mt. Hermon were affected the most by ecological factors (see Additional file 3C).

After analyzed of the factors by projection of the variables on the factor-plane (see Additional file 4) and consulted the correlations of these factors (see Additional file 5), 11 independent ecological factors were chosen. And then, multiple regression analysis was done using these 11 factors to investigate the relationship between environmental variables and SNPs.

The geographical, temperature, water, and solar radiation factors in Table 4, singly or in combination, explained a significant proportion of the diversity in the SNPs (Table 5). The best variable predictors of P, He, and Shannon's information index, significantly explaining 0.264 – 0.355 of the variance, was the water availability factor Hu-an. The combination of three variable predictors accounting for geographic and water availability factors Hu-an, Ev, and Lt (mean annual humidity, mean annual evaporation and latitude) accounted significantly (p < 0.05) for about 0.55 of the genetic diversity. The addition of a fourth variable predictor, Rad (total solar radiation per year), or Sh (mean number of Sharav days) to the first three factors accounted for approximately 0.65 of the diversity (significant at p < 0.05) (Table 5).