The association mapping panel comprised of 130 diverse pearl millet lines including two checks, namely, ICTP 8203Fe and ICMB 98222 for grain iron and zinc content (Table S1). The association panel represents B- lines (seed parents), R- lines (restorers or pollen parents) and advanced breeding lines derived from different parts of India and some introduced selections from Africa.

The experiment was conducted at three diverse geographical locations, representing all the three pearl millet growing agro-climatic zones of India, (i) ICAR-Indian Agricultural Research Institute Research farm, New Delhi (28° 382′ N, 77° 802′ E) representing Zone A receiving more than 400 mm annual rainfall (ii) ICAR-National Bureau of Plant Genetic Resources, Regional Station farm, Jodhpur (26° 252′ N, 72° 992′ E) falling in zone A₁ with annual rainfall less than 400 mm and (iii) ICAR-IARI Regional Centre farm, Dharwad (15° 212′ N, 75° 052′ E) from zone B (covering the southern peninsular India). In all locations, the trial was taken for 2 consecutive years (Rainy Season, 2014 and 2015). The planting of genotypes was taken up in an alpha-lattice design (Patterson and Williams, 1976), with three replications. Each replication comprised of 13 incomplete blocks with 10 entries in each block. Each accession was represented in 2 rows of 4 m length with a spacing of 50 cm between rows and 15 cm from plant to plant. All the experiments were managed as per recommended agronomic practices across the locations and years to raise a normal healthy crop.

Open-pollinated panicles from five representative plants were harvested for each accession at physiological maturity, threshed with a wooden mallet and then cleaned, while taking utmost care to avoid dust or metal contamination of the samples. Further, oven-dried samples were used on duplicate basis to estimate iron and zinc by di-acid digestion (Singh et al., 2005) followed by readings taken on an atomic absorption spectrometer (AAS, ZEE nit 700 tech Analytikjena). Genotypes, which recorded high iron content (>77 ppm, as this is the criterion given by Harvest Plus) were also tested for aluminum content in ICPMS (Nex ION 300X, Perkin Elmer USA) in order to determine, whether higher grain iron content could be due to contamination with soil, dust or is innate (Pfeiffer and McClafferty, 2007).

The data over locations and years were subjected to combined analysis of variance across the six environments using PROC GLM with random statement, considering environments as random and genotypes as fixed effect in SAS v9.2 (http://iasri.res.in/design). The adjusted means thus obtained after analysis were used to generate all the descriptive statistics using IBM SPSS v20. The broad-sense heritability was calculated as per the formula:

where, σG2 is genotypic variance, σGE2 is genotype × environment interaction variance, σe2 is error variance, n is number of environments and r is number of replications.

Datasets for analysis at Delhi during 2014 and 2015 were referred as Del-14 and Del-15. Similarly, data of Dharwad in 2014 and 2015 as DW-14 and DW-15. Likewise, Jodhpur data during 2014 and 2015 as Jod-14 and Jod-15. Mean data over years in a single location as Y14-M and Y15-M. Mean data across locations in a year were referred as Del-M, DW-M, and Jod-M. Mean data over all the six environments as grand mean (GM). Altogether, 12 datasets were included for analysis with TASSEL using both general liner model (GLM) and mixed linear model (MLM), but the results of only MLM are presented here. Further, though STRUCTURE and neighbor-joining (NJ) results were similar, we chose Q matrix files over PCA files due to greater relevance of the former with the pedigree of population for MLM analysis.

Pooled leaf samples from five plants for each genotype were collected at 2 weeks stage and genomic DNA of all genotypes in the association mapping panel was isolated using modified CTAB method (Murray and Thompson, 1980). The PCR was performed in a 10 μl volume consisting of 1.0 μl DNA (25 ng/μl), 1.0 μl 10 X Buffer (Banglore Genei), 0.60 μl dNTP (10 mM, Banglore Genei), 0.20 μl (25 mM MgCl₂, Banglore Genei), 1.0 μl each forward and reverse primer (10 mM), 0.13 μl Taq polymerase (3 U/μl, Banglore Genei) and the final volume is made up to 10 μl with autoclaved distilled water for every reaction. Thermo cycling was carried out in 384 well-blocked thermal cycler machine (GenePro, Bioer Technology Co. Ltd.) with a thermal cycling program having initial denaturation for 5 min at 94°C, followed by 40 cycles of 30 s at 94°C for denaturation, 30 s at temperatures as indicated in Table S2, extension at 70°C for 30 s and finally, 7 min at 72°C with final extension.

For association mapping analysis, a total of 267 polymorphic markers were used of which 250 were SSR markers and 17 were genic markers. To target genes involved in the pathway of iron and zinc uptake from soil to final sequestration into grains by plants, 55 genic primers were screened for polymorphism. Twenty primer pairs were selected from previous studies on maize (Mondal et al., 2014) and rice (Sperotto et al., 2010). Further, orthologous genes were searched in phytozome database of other related crops for cross amplification in pearl millet, since the genome sequencing in this crop (http://ceg.icrisat.org/ipmgsc/) is still underway and not available in public domain. After identification of orthologous genes in foxtail millet and sorghum genome, a total of 35 gene specific primers (foxtail millet: 22 and sorghum: 13) were designed with the help of Primer3Plus web application (http://primer3plus.com/cgi-bin/dev/primer3plus.cgi). Out of 55 genic markers used, 17 were observed to be polymorphic and they were further utilized in the study. TASSEL v3 (Bradbury et al., 2007) with 1,000 permutations was used to generate an LD plot with r² and p-values among all the 267 polymorphic markers. MTA was studied in TASSEL v3 (http://www.maizegenetics.net) with both general linear model (GLM) and mixed linear model (MLM) (Yu et al., 2006) as given below (Tadesse et al., 2015).

where,

y is phenotype vector, a is marker vector with fixed effects, b is a vector with fixed effects, u is a vector with random effects (kinship matrix), e is a residuals vector, X denotes the accessions/genotypes at the marker, Q is the Q-matrix, result of STRUCTURE software, Z is an identity matrix.

A widely adopted threshold for the significance, Bonferroni correction (Bland and Altman, 1995; Li C. Q. et al., 2016) was used uniformly for all traits to observe association between trait and markers. In this context, p < 3.74 × 10⁻³ (where, p = 1/n and n = number of total markers used (267); also −log10 (1/267) = 2.43). Quantile-quantile (QQ) plots, plotted against expected and observed p-values to assess the adequacy of Type I error were generated with TASSEL (Figures S1, S2).

The position of the marker trait associations were determined following the published consensus linkage map of pearl millet by Rajaram et al. (2013).

Further, the significantly associated markers were tested for associations at individual sub-population level using one way ANOVA (http://vassarstats.net/anova1u.html), assuming null hypothesis of no phenotypic difference between different alleles of a marker.

Initially, Breseghello and Sorrells (2006) utilized null alleles (missing and rare alleles were also included) to calculate the phenotypic effect of an allele for identified MTAs. But all SSR markers may not have null alleles, sometimes rare alleles themselves are superior alleles. Hence, population mean was introduced in place of null alleles (Cai et al., 2014). Favorable allele for a marker loci associated with specific trait was identified using the formula (Cai et al., 2014; Li X. et al., 2016).

where,

a_i is ith allele phenotypic affect, x_ij is phenotypic value of ith allele in jth genotype, n_i is number of accessions with ith allele, N_k is phenotypic value over all genotypes and n_k is total number of genotypes. When, a_i > 0, then this allele is said to have positive effect on the trait. When a_i < 0, the allele gives a negative effect (Zhang et al., 2013). Since, positive alleles give an increment, we considered superior positive alleles for planning crosses to pyramid maximum possible number of favorable alleles together for enhancement of grain iron and zinc content.

Four lines having higher mean iron and zinc content (PPMI 1102, PPMI 708, PPMI 683, PPMI 1107) and check (ICTP 8203Fe) along with two lines (5540B and PPMI 1155) with low mean iron and zinc content were selected for validation. Three consistently associated SSR marker primers, IPES 0096, IPES 0180, and PSMP 2261 were used to amplify DNA of all the seven selected genotypes. The amplified products were sequenced by standard Sanger's sequencing method by M/S Chromous Biotech, Bangalore. The sequenced reads were compared using BLAST against recently sequenced pearl millet reference genome,Tift23D₂B1−P1-P5 (unpublished) using MEGA6 (Tamura et al., 2013).

Initially, 114 polymorphic SSR markers, covering the entire genome were used to characterize pearl millet association panel to assess the genetic relationship. A total of 294 alleles were detected with an average number of 2.65 alleles per locus (Table S3). The observed number of alleles (n_a) ranged from 2 to 5. Maximum number of five alleles were detected by genomic SSR loci, Xpsmp 2081. Four alleles per loci were observed in Xipes 200, Xipes 126 and Xpgird 50, and Xpsmp 2086. The mean effective allele number per loci (n_e) for 114 polymorphic markers studied was 1.64. This shows that there exists a difference of 37.89% in observed and expected number of alleles. The discrepancy observed between n_a and n_e indicates that alleles detected were having frequencies <5% (Sehgal et al., 2015). The average major allele frequency was 0.75 with a range of 0.32 (Xpsmp 2066) to 0.98 (Xipes 0079. Measure of gene diversity, Nei's gene diversity (Nei) was calculated across the population. Xipes 0079 had the least value of 0.03 while highest value (0.73) was recorded by Xpsmp 2066 with a mean of 0.35. A frequently used measure to know the polymorphism information of a marker, PIC ranged from 0.03 (Xipes 0079) to 0.68 (Xpsmp 2066). In this study, seven SSRs (Xpsmp 2066, Xpsmp 2081, Xpsmp 2263, Xpsmp 2086, Xipes 0200, Xipes 0163, and Xicmp 3066) had a PIC value more than 0.5 and in total 67 markers were showing moderate PIC values.

The true association is known only after considering population structure (Yu et al., 2006). A Bayesian model based approach, STRUCTURE, assigns individuals to sub-population based on genotyping. Pritchard et al. (2000) was of the opinion that, actual number of sub-populations in a panel may not be easy to identify, but one should consider the smallest K-value which captures the major composition in the population. Evanno et al. (2005) suggested the use of ΔK-value based on standard deviation to find the true value of K. The LnP(D) graph (Figure 1A) depicted with number of sub-populations(K) on x-axis and logarithmic of probability distribution on y-axis did not give a clear picture of the true value of K, so ΔK value was plotted against number of sub-population (Figure 1B) showed the highest peak at K = 3, which indicated that there are three sub-populations in the panel. The genotypes were assigned to individual sub-population keeping in view of highest membership likelihood criterion based on Q-values obtained from STRUCTURE software. The decision point taken here is, a genotype indicating a Q-value more than or equal to 0.6 is assigned to that particular sub group A or B or C and remaining genotypes which did not reach Q = 0.6 were allotted to admixture group (Table S1). In this manner, the first sub-population A is the biggest group containing 50 lines, B had 22, and C had 31, while remaining genotypes (30) were kept in admixture. These three sub populations were represented by three colors in the bar plot showing three sub-populations (Figure 1C). The sub population A is depicted by red, B by green, and C by blue color. Here, sub-population A is said to have fewer admixtures (0.19%) compared to B (0.29%) or C (0.21%) and in overall 23.08% of total population is under admixture. Sub-population A consisted of genotypes from different sources of pedigree, which included B lines while sub-population B comprised of few promising lines derived from a cross between, PPMI 683 and PPMI 627, few lines from 843B × 841B lines and some low iron Jamnagar lines and few other breeding lines. Sub-population C, majorly included genotypes selected for good agronomic score in a cross between PPMI 683 and PPMI 627, few lines derived from a cross between WGI 148 and WGI 52, and few others. Admixture consisted of thick panicle restorer lines, early maturing lines and dual purpose lines.

The dendrogram generated by DARwin using neighbor joining, sub-classified population into three groups. The maximum number of individuals was clustered in group 1 consisting 75 genotypes (Figure 2), whereas group 2 consisted of 29 and group 3 comprised of 26 individuals. Cluster 1 was only 64.0% similar to sub-population A, whereas Cluster 2 had 72.4% similarity with sub-population B while cluster 3 shared 88.5% similarity with sub-population C.

The principal coordinate analysis further supported this clustering pattern, where all the breeding lines were distributed in all four quadrants (Figure 3). Cluster I, similar to sub-population A of STRUCTURE analysis was observed in first quadrant. In same way, cluster II, similar to sub-population B fell in second quadrant and cluster III, similar to sub-population C, was mostly in third and a little of fourth quadrant. Remaining genotypes which belong to admixture in STRUCTURE were dispersed throughout all the quadrants.

Results from AMOVA showed that significant differences exist between sub populations which accounted for 11.8% of total variation. Within sub-population variation was much larger, contributing 82.2% to total variation and within individual variation was 6.0% (Table 1).

Pearl millet grain iron and zinc content were having a wide range of variation at all six environments studied showing significant differences between genotypes in the analysis of variance (Tables S4A–E). Grain iron content (grand mean) ranged from 32.3 to 111.9 ppm whereas grain zinc content ranged from 26.6 to 73.7 ppm (Tables 2A,B, Tables S5A,B). The broad sense heritability estimates were high (>0.8) at all the individual environments (Tables 2A,B). It ranged from 0.84 (Jod-15) to 0.89 (DW-15) for grain zinc content while for grain iron content it was 0.85 (Del-14) to 0.92 (Jod-14). In pooled environments, the highest heritability value was recorded at Dharwad location (0.86) while the lowest during the year 2015 (0.60) across all the three locations for grain zinc content with 0.79 combined over all six environments. Grain iron content exhibited the highest heritability at Delhi location (0.8) while the lowest at Dharwad location (0.65). Hence, these results indicate that heritability was high for both grain iron and zinc content.

The LD measured as squared value of Pearson correlation, R² ranged from 0 to 1 with a mean of 0.017. Highest R²-value was observed between Xpsmp 2063 and Xipes 0145 on LG7, Xipes 0103 and Xipes 0146 of LG1 (Figure 4). Out of 35,511 pair-wise combinations obtained from 267 marker loci, 69 pairs had R² more than 40%. Only 9.53% of the 35,511 pairs were in extent of LD (p < 0.05) across all the genotypes accessed. Before performing the association mapping, correlations between grain iron and zinc content and population structure was computed (Table S6). Grain iron was significantly associated with structure at DW-15, while grain zinc content recorded its significance at DW-14, DW-15, and DW-M. In remaining datasets, there is no correlation of the trait with the structure. The Q+K model holds good in all the cases, but when there is correlation of trait with the structure, Q+K model is a better model which avoids spurious associations (Yu et al., 2006). The threshold for identifying significant associations with a trait kept here was p < 0.0037, which was a Bonferroni corrected value considered by many workers in association mapping studies. Significant associations with grain iron and zinc content were tested across six different environments and their six means in different combinations.

The MLM analysis (Table 3) identified Xipes 0180 and Xsinramp 6 to be significantly associated with the trait explaining 12.4 and 10.1% variation, respectively, in Delhi, 2014. During the year 2015 at Delhi, Xpsmp 2261 recorded significant association with grain iron content. Mean of Delhi over 2 years, indicated Xpsmp 2261, Xipes 0096 and Xsinramp 6 to be highly associated with grain iron with an R² more than 9.0%. Dharwad 2014 dataset analysis indicated significance of only one SSR marker (Xicmp 3092) with grain iron content, while during the year 2015 at Dharwad, Xpsmp 2209 was observed to be associated. No association was observed for grain iron content at Dharwad over mean of 2 years.

At Jodhpur during the year 2014, Xpsmp 2261 and Xipes 0096 were observed to be associated with the trait, where Xpsmp 2261 explained phenotypic variance of 15.7%, while another marker, Xipes 0096 recorded more than 8.0% variation. Xpsmp 2261 showed statistical significance by explaining more than 10% variation in the year 2015 at Jodhpur. Mean iron content of Jodhpur over 2 years revealed only one marker locus, Xpsmp 2261 to associate with the trait by explaining 18.1% variation.

During 2014, the data pooled over locations identified Xpsmp 2261, Xipes 0096, and Xipes 0180 to associate with grain iron content. Similarly, during year 2015, Xpsmp 2261 and Xipes 0096 were observed to be significantly associated with the trait. Likewise, three SSR loci, Xpsmp 2261, Xipes 0096, and Xipes 0180 were recorded to be associated with grain iron content in grand mean data set.

During 2014 at Delhi, Xpsmp 2261 recorded significant association with grain zinc content explaining variance of more than 10% (Table 3). MLM analysis of the data during the year 2015 at Delhi, showed only one marker, Xipes 3016 to be significantly associated with grain zinc with more than 8.0% variation explained in the model, while Xpsmp 2261 was observed to be significantly associated with the mean over 2 years at Delhi. At Dharwad, during the year 2014, Xipes 0096 was observed to be associated with 12.7% variation, whereas, during the year 2015 at Dharwad, Xicmp 4006 revealed its significance with an R² of 10.4%, while Xipes 0096 and Xsinramp 6 were recorded as significant for Dharwad mean over 2 years.

Jodhpur, in the year 2014 recorded three SSRs (Xicmp 3004, Xipes 0224, and Xpsmp 2261) to be significantly associated with grain zinc content explaining more than 10% variation. In the year 2015 at Jodhpur, Xpsmp 2261 and Xipes 0180 were associated significantly, whereas the mean of Jodhpur over 2 years revealed three SSR markers (Xicmp 3004, Xpsmp 2213, and Xpsmp 2261) to be associated with grain zinc content. Mean zinc content of the years, 2014 and 2015 revealed Xipes 0096 and Xpsmp 2261 to be associated significantly with the trait explaining more than 10% variation each. When association were observed with the grand mean, over all locations and years, Xipes 0096 and Xpsmp2261 were found to be significantly associated with grain zinc content with R² more than 11.0% individually.

Table S7A shows the number of times a marker was observed as significantly associated with grain iron or zinc content via MLM analysis. Out of total 6 markers identified to be associated with grain iron content, Xpsmp 2261 recorded significance in highest number of times (Del-15, Jod-14, Jod-15, Del-M, Jod-M, Y14-M, Y15-M, and GM). Further, higher number of associations were observed with Xipes 0096 which recorded as significant in five datasets (Del-M, Jod-14, Y14-M, Y15-M, and GM) followed by Xipes 0180 which recorded significant associations in four datasets (Del-14, Del-M, Y14-M, and GM). Genic marker, Xsinramp 6 identified to be associated with iron content in two datasets (Del-14 and Del-M) while Xicmp 3092 and Xpsmp 2209 were significant at Dharwad location during the year 2014 and 2015, respectively.

For grain zinc content (Table S7B), among 10 markers identified as significantly associated with the trait, Xpsmp 2261 showed association in eight datasets (Del-14, Jod-14, Jod-15, Del-M, Jod-M, Y14-M, Y15-M, and GM). Next, Xipes 0096 was observed as significant for five times (DW-14, DW-M, Y14-M, Y15-M, and GM). Genic marker, Xsinramp 6 was associated with the trait in three datasets (DW-14, DW-15 and DW-M) while, seven SSR markers (Xipes 0180, Xpsmp 2086, Xipes 0224, Xpsmp 2213, Xicmp 3004, Xicmp 3016, and Xicmp 4006) showed significance only at one location.

For grain iron content alone, six markers were identified to be statistically significant at Bonferroni correction (p < 0.0037) by MLM analysis in all the combinations of environments studied (Table S8, Figure 5). Two SSR markers were from LG7 and one each marker belonging to LG3 and LG5 in the published pearl millet consensus map of Rajaram et al. (2013) were associated with iron content. Remaining two markers (Xpsmp 2209 and Xsinramp 6) were unmapped to any linkage group of the published consensus map of pearl millet.

For grain zinc content (Table S8, Figure 5), total 10 markers were observed to be associated in all the environments taken together. The maximum number of significant markers (4) were from unmapped markers (from Rajaram et al., 2013) followed by LG6, where Xipes 0224 and Xpsmp 2213 were observed to be significant. One each marker from LG3, LG4, LG5, and LG7 were observed to associate with grain zinc content in MLM.

In total, 12 markers (4 genomic SSRs, 7 EST-SSRs, and 1 genic marker) accounting for16 MTAs were observed to be significantly associated with either grain iron or zinc content at significant threshold of Bonferroni correction (p < 0.0037) using MLM analysis of data from different environments and combination of their means (Table S8, Figure 5). The maximum contribution was from LG6 and LG7, which included two SSRs each. Both of the markers from LG6 were associated with zinc content only, whereas one SSR, Xipes 0096 of LG7 was highly associated with both iron and zinc content, while another SSR, Xicmp 3092 was associated with iron content only. One SSR each from LG3, and LG5 were associated with both the traits. One genic marker, Xsinramp 6 also showed significance of association with both iron and zinc content. Remaining four unmapped markers in the published consensus map (Rajaram et al., 2013) were either associated with grain iron or zinc content.

Twelve significant markers resulted from MLM analysis were used to study their associations at sub-population level. For grain iron content, 13, 26, and 23 MTAs were recorded in sub-population A, B and C respectively (Table S9A). Similarly, 19, 24, and 18 MTAs were recorded for grain zinc content in sub-population A, B and C respectively (Table S9B). Maximum number of associations were observed in sub-population B for both traits, where most of promising lines like PPMI 1102, PPMFeZMP 199, PPMI 708 etc. were included.

Xpsmp 2261 at Del-15 and Xsinramp 6 at Del-M showed significant association with grain iron content in all the three sub-populations (comparing Table S9A with Table 3). Significance in at least two sub-populations were observed by Xicmp 3092 at DW-14; Xipes 0180 in Del-14 and GM; Xsinramp 6 in Del-14 and Xpsmp 2261 in Del-M, Jod-M, Y14-M, Y15-M, and GM datasets. Few associations viz., Xipes 0096 at Del-M, Jod-14, Y14-M, Y15-M, and GM; Xipes 0180 at Del-M and Y14; Xpsmp 2209 at DW-15; Xpsmp 2261 at Jod-14 and Jod-15 were restricted only to one sub-population, but resulted in significant associations using MLM analysis of the whole population. Few MTAs were not observed while analyzing the whole population but were recorded in two out of three sub-populations for some datasets.

Similarly for grain zinc content (comparing Table S9B with Table 3), Xpsmp 2261 in Y14-M, Y15-M, and GM; Xsinramp 6 in DW-M; Xsinramp 6 in DW-14 recorded associations in two sub-populations out of three, whereas Xpsmp 2261 in Del-14, Del-15, Jod-14, and Jod-M; Xipes 0096 in DW-14, Y14-M, Y15-M, and GM; Xicmp 3016 in Del-15; Xipes 0096 in DW-M; Xipes 0180 in Jod-15; Xicmp 4006 in DW-15; Xicmp 3004 and Xipes 0224 in Jod-14 datasets recorded significance only in one sub-population. MTAs, Xpsmp 2086 at DW-14; Xsinramp 6 in DW-15; Xpsmp 2213 in Jod-M did not show significance at sub-population level, though they are associated in the whole population. Xpsmp 2261 in DW-14 and DW-M; Xipes 0180 in Del-14 and Xsinramp 6 in Del-15 appeared in significant association with grain zinc content in two sub-populations, though they are not associated at whole population level.

Alleles having positive effect were considered as favorable for both grain iron and zinc content. Details of these alleles along with top three genotypes carrying them are given in Tables S10A,B. The favorable allele is expressed in terms of amplicon size of the SSR marker. A total of six alleles were detected for grain iron content, and five out of 10 alleles detected for grain zinc content had phenotypic effect of more than four. Highest phenotypic effect of alleles for grain iron and zinc content was observed by Xpsmp 2261-180. Among the top three performing genotypes, PPMI 1102 had highest number (7) of favorable alleles (Tables S11A,B), followed by PPMI 1104 (5), PPMFeZMP 199 (4), and PPMI 708 (4). PPMI 1102, PPMFeZMP 199, PPMI 708 and PPMI 683 had alleles which expressed across environments, whereas PPMI 1104 had Xicmp 3092-220 allele specific for Fe content, and Xicmp 4006-280 allele specific for Zn content at Dharwad.

Considering the average phenotypic effect of an allele approximately equal to four or greater, nine better crosses (Table 4) were suggested to pyramid maximum number of alleles in a single genotype. To account for maximum pyramiding of alleles even having minor effect, one single cross (PPMI 1102 × PPMI 1104) can capture as many as 10 alleles favorable for both grain iron and zinc content.

Out of three SSR markers selected for amplifying seven genotypes, Xipes 0096 was dropped for further analysis, because of its poor read quality. The amplicons of Xpsmp 2261 and Xipes 0810 for seven genotypes were compared with reference genome Tift23D₂B1−P1-P5. Among two markers studied, Xpsmp 2261 amplicon landed in intergenic region on pseudomolecule 5 and was having >50% GC content, while the other marker, Xipes 0810 was observed to be overlapping with gene on pseudomolecule 3. This flanking gene was annotated to be aspartic proteinase (Asp1) gene. On comparison with the reference sequence, two putative SNPs (Figure 6) and a 15 bp InDel were observed.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.