3.1. Linkage Map
Linkage maps are important pre-requisite for QTL mapping and for marker-assisted breeding programs. RIL is the most favored mapping population, but the accuracy of the mapping resolution depends on the population size. So far, in pearl millet, linkage maps have been predominantly constructed using F2
populations and, up to certain extent, RILs. The number of genotypes in individual populations of pearl millet was less than 190 individuals, which is lower than the theoretically required size to achieve precise mapping [37
]. The RIL population, namely, 81B × ICMP 451, presently accommodating 321 loci (258 DArTs and 63 SSRs) was developed from only 140 RILs [15
]. However, the here presented mapping population with 317 RILs is the largest mapping population in pearl millet.
During the polymorphism survey, a set of 235 DArT clones (3.3%) was found polymorphic between both parents. The number of polymorphic DArT clones was comparatively lower than that found by Supriya et al. [15
]. In the case of SSRs, 112 (33%) SSRs out of 342 were polymorphic and comprised 18% of Xipes
, 13% of Xpsmp
, and 2% of Xicmp
markers. To construct the framework linkage map, a small set of 33 polymorphic SSRs was used for genotyping the RILs. The SSR polymorphism level was lower than that determined by Senthilvel et al. [12
], which was 74%. The reason for a low polymorphism in the present study may be the closer relatedness of the mapping population parents in the present study compared to that of Senthilvel et al. [12
As far as the positioning of the SSR markers on the genetic map, all SSRs mapped on the same linkage group (LG) as in previous pearl millet maps reported by various researchers [12
]. A set of 75% of DArT loci was polymorphic, which is similar to the result of Kumar et al. [14
] but is lower than the percentage found by Supriya et al. [15
], who reported 80% polymorphism. Out of 235 polymorphic markers, 177 DArTs and 19 SSRs were assigned to seven previously established pearl millet linkage groups (base map), (Figure 1
a), in which LG4 had two sub-groups (LG4A and LG4B). Thirty-four loci (28 DArT and 6 SSRs) not linked to any of the existing LGs and were grouped in 14 small groups (LG A to LG N), (Figure 1
b), with markers ranging from two to five per group. Similar to the current study, Kumar et al. [14
] also reported small LGs in the pearl millet linkage map. Moreover, attempts for linking any such small segments to any major group drastically increased the length of that major group, with huge inter-marker distances. Hence, these 14 segments were kept separate and treated as small LGs. The 34 unmapped markers either had a too high degree of segregation distortion or were positioned too distantly in relation to the next markers assigned to the linkage groups and were presumably on the distal ends of the LGs. The un-mapping of polymorphic markers in pearl millet was also reported earlier [14
In the current study, the distribution of DArT throughout the chromosome was uniform. The map of seven LGs consisting of 196 loci was 964 cM in length, as shown in Table 1
. The average length of the LGs in this base map was 121 cM, and the average number of loci per LG was 24.5. LG1 was the longest, having 38 loci with a distance of 218 cM. The shortest LG in pearl millet on the basis of the previous studies of Gulia [39
] and Yadav et al. [38
] with four (30.2 cM) and nine markers (15.4 cM), respectively, was LG3; however, in the present study, LG6 was found to be the shortest LG (47 cM), holding 13 markers only. This indicates that the marker number was not sufficient in the previous investigations [14
] to cover the full length of LG3. The length of LG2 was one-third of that previously reported in the DArT map of pearl millet.
The present map had a length greater than the one reported by Yadav et al. [38
] and Gulia [39
], and smaller than in the other maps [14
]. The map covers 964 cM, which is comparable to the 815 cM of the map recently developed by Sehgal et al. [40
], and 84% of the intervals were less than 10 cM, which is comparable with the intervals reported by Supriya et al. [15
The average inter-marker lengths on the base map ranged from 3.6 cM (LG6) to 9.2 cM (LG5). Among the small segments, LG H was the longest (34 cM), while LG I was the shortest (1.9 cM). The average distance between markers in the base map was a little higher in the DArT linkage map of Supriya et al. [15
]. DArT markers in large-sized population constitute a significant improvement in marker density and, possibly, inter-marker distance. A positive correlation was found between the number of mapped markers and the length of linkage groups. For instance, LG1 spans a distance of 219 cM with 33 markers, and LG4B, with 5 loci, spans a distance of 26 cM. Similar observations were recorded in an earlier DArT-based map of pearl millet study [15
]. The markers that were unlinked to any of the major LGs will be merged into larger LGs with the availability of an additional set of polymorphic markers in the near future.
The improved inter-marker distance of the current map, as compared to previously reported pearl millet maps [14
], developed on a variable size population and a different number of markers, indicates that this map is one of the densest genetic maps for pearl millet.
3.2. Segregation Distortion and Inter-Marker Gaps
RILs usually show segregation distortion, because, during the RIL development process, many recessive lethal genes become homozygous, and their expression caused failure to contribute seeds to subsequent generations, consequently resulting in a skewed population [14
]. It was found that 60% of loci showed segregation distortion in the present linkage map. Distorted markers showing p
> 0.001 were also used for mapping in the framework linkage groups. Segregation distortion was found in both maternal and parental loci, although it was much higher in ‘ICMS’, the female parent of our mapping population. In previous studies, segregation distortion in favor of the female parent alleles was observed [15
]. Out of 196 markers on the base map, the segregations of 118 DArTs and 14 SSRs were distorted, and, of these, 60 were skewed towards AIMP, and these loci were distributed across all seven LGs. In pearl millet, pollen abortion is more common than abnormalities in the female gametes, leading to a relatively greater loss of male parent alleles and the ensuing skewness towards the female parent. The distorted markers mapped on LG1 and LG2 were skewed towards the alleles of the male parent (AIMP), while most of the distorted markers were skewed towards the alleles of the female parent (ICMS) in the remaining LGs. Most of the loci on linked fragments were skewed towards ICMS.
Distortion from expected Mendelian segregation has been observed previously in barley [43
], rice [45
], maize [47
], wheat [49
], and pearl millet [15
]. It has been suggested that the protogynous nature of pearl millet also contributes to segregation distortion [4
]. Residual heterozygosity and inbreeding depression during inbred development may also contribute, as previously demonstrated by Cloutier et al. [52
] and Livingstone et al. [53
]. At the same time, the residual heterozygosity existing in some RIL may be beneficial, since deleterious genetic combinations in the form of lethality or reduced fitness can be overcome. In contrast to previous studies, the segregation of almost all loci of LG3 was distorted, which may explain part of the increase in LG3 map length.
The characteristic feature of pearl millet linkage maps is the presence of large gaps between centromere and telomere. LG4 was split into two pieces, suggesting additional polymorphic markers are needed to fill-up the gap between these two pieces. Senthilvel et al. [12
] also reported a similar big gap in LG4. The previously constructed framework map for the cross ICMB 841-P3 × 863B-P2 had large gaps in LG2 and LG7 [12
]. This new population also showed a large gap (>25 cM) in LG2, LG5, and LG7, for which the most probable reason could be the extreme localization of recombination at the ends of LG2 and LG7 [38
]. According to various researchers, large gaps in the distal regions reflect regions of high recombination, rather than a lack of markers in these regions [12
It is, however, possible, on the other hand, that these linkage groups are still incomplete, and genomic resources can be extended to develop new markers that are located on the distal regions of the linkage groups. Moreover, areas of low marker density and gaps may correspond to regions of similar ancestry or identity by descent in the germplasm included in the initial diversity representation for the development of the DArT markers [56
]. In the case of potato, a gap was observed, and the authors postulated that this could be due either to recombination hot spots or to fixation (homozygosity) of the potato genome [57
]. The number of large gaps has decreased in the present study compared to previous linkage maps, although there is still the possibility to map more markers to saturate these gaps.
3.6. Quantitative Trait Loci for Grain Iron (Fe) and Zinc (Zn) Content and Epistasis
The base map accommodated eight putative QTLs out of 11 detected for Fe content of self-pollinated grain in the summer 2010 dataset. R2
values (phenotypic expression) for individual QTLs ranged from 9 to 31.9%, while the R2
for their final simultaneous fit was 74.6%. The range of R2
was higher than that found by Kumar et al. (2016). A major putative QTL with R2
of 31.9% and a LOD value of 25.36 was detected on LG1. It had an additive effect of 9.7 ppm, and the favorable allele was the from high-Fe parent AIMP 92901-deriv-08 (Table 4
). Compared to the present study, Kumar et al. [14
] reported a major QTL for Fe on LG3. This indicated the role of the genome content in the parental genotypes. The result of one major QTL on LG1 and numerous putative QTLs with small effect is indicative of the probable complexity of inheritance of this trait, with effects distributed across the whole genome contributing to the control of uptake, accumulation, and content of this mineral micronutrient. The existence of multiple QTLs for grain iron content was also identified in different crops, like rice [64
] and Medicago
]. On the basis of the single-environment data analysis, three putative QTL × QTL interactions (QQI) were observed for Fe.
Eight putative QTLs for selfed grain Zn content were mapped, with their final simultaneous fit providing an R2
of 65.4%. All favorable alleles were from AIMP 92901-deriv-08, except for two QTLs on LG4B at position eight and 22 cM. The R2
values for the individual putative QTLs for this trait ranged from 9.4 to 30.4% (Table 4
). Similarly, their additive effects ranged from 0.6 to 6.7. Using single-environment data, one putative digenic interaction was observed for Zn.
Pleiotropism and linkage in quantitative traits are responsible for the correlations among QTLs and, consequently, the detection of co-localized QTLs. Similar to Kumar et al. [14
], QTLs for Fe and Zn were co-mapped (Figure 3
) on LG1 (major QTL) and LG7 (minor QTL), and favorable alleles for these QTLs were contributed by the male parent. The co-mapping of QTLs for minerals content has been earlier observed in cereals and millet in many studies [14
]. Such outcomes indicate that some QTLs control the expression of multiple traits. The co-localization of QTLs also suggests the presence of QTLs with major effects for positively correlated traits [66
]. From the crop improvement point of view, co-localized QTLs are of importance for the simultaneous improvement of Fe and Zn content in the grain of pearl millet.