Characterization of entire germplasm for economically important traits is a daunting task for breeders. For the past two decades, germplasm characterization at ICRISAT has led to the establishment of core and minicore collections for pearl millet (Bhattacharjee et al., 2007; Upadhyaya et al., 2009, 2011a), finger millet (Upadhyaya et al., 2006), and foxtail millet (Upadhyaya et al., 2008, 2011b) while only core collections was established for other small millets (Upadhyaya et al., 2014) (Figure 1).

Trait-specific germplasm characterization is a prerequisite to identify genotypes contrasting for desirable traits. All India Coordinated Small Millets Improvement Project implemented by 1986 has focused mainly on varietal development for high yield and disease resistance with due negligence on nutritional quality (Shivran, 2016). Only by the start of the 21st century, millet germplasm received scientific attention for nutrition traits. Screening of pearl millet, foxtail millet, and finger millet accessions for grain nutrients revealed sufficient genetic variability. Multi-location on farm trials identified nutritionally superior lines with farmer preferred traits such as earliness to flowering and grain yield, adapted to local environments (Upadhyaya et al., 2011c; Muthamilarasan and Prasad, 2015). This process accelerates the pace of breeding in millets by studying the inheritance pattern and genotype–environment interaction for grain nutrients. Hybridization between diverse genotypes will generate mapping populations for DNA marker-assisted tagging of genomic regions linked to grain nutrients. However, molecular characterization of other minor millets for nutritional traits is limited and needs special attention. A thorough evaluation of small millets core collections for nutritional traits is the need of the hour to bring genetically diverse parents into mainstream breeding for generating recombinant inbred lines suitable for biofortification. The following section details the utilization of millet germplasm in biofortification in millets under the major headings: macronutrients, micronutrients, and antinutrients.

Millets are the primary source of carbohydrates in tropics and semi-arid tropics of India and sub-Saharan Africa (Shivran, 2016). Grain starch typically comprises of two polymers, amylose (15–30%) and amylopectin (70–85%). Based on the amylose content, millet accessions are classified into two major phenotypes, waxy and non-waxy. Waxy grains containing 0% amylose and nearly 100% amylopectin are glutinous in nature, easily digestible and therefore recommended as food for infants under 6 years of age (Dreher et al., 1984; Englyst et al., 1992). Waxy mutants in staple crops have evolved during the domestication of landraces by human selection (Olsen et al., 2006; Fan et al., 2008). They have been identified in cereals and millets including rice (Oryza sativa; Isshiki et al., 1998), barley (Hordeum vulgare; Domon et al., 2002), sorghum (Sorghum bicolor; McIntyre et al., 2008), maize (Zea mays), foxtail millet (S. italica), proso millet (P. miliaceum), and barnyard millet (Echinochloa sp.) (Fukunaga et al., 2002a; Kawase et al., 2005; Kim et al., 2009). Amylose synthesis in millets is controlled by a single dominant waxy allele (Wx), while the recessive loss-of-function allele (wx) leads to the waxy phenotype with near 0% amylose content (Nakayama et al., 1998; Fukunaga et al., 2002a). In polyploid crops, mutations in different alleles of Wx loci produce low amylose, non-waxy and waxy phenotypes. Low amylose lines contain <3% amylose content due to the residual activity of non-mutant alleles (Fukunaga et al., 2002b; Graybosch and Baltensperger, 2009). Precise identification of mutations in low amylose and waxy mutants have led to the development of waxy starch foods.

The waxy gene product named as granule-bound starch synthase 1 (GBSS 1) is the key enzyme catalyzing the formation of amylose (Sano, 1984). Mutations in GBSS 1 result from insertions/deletions (InDels), transposable elements, and single base pair mutations (Isshiki et al., 1998; Domon et al., 2002; Van et al., 2008). In millets, Wx gene was found to contain 14 exons and 13 introns (Figure 2). Foxtail millet with waxy grains are cultivated in Japan, Taiwan, Philippines, and Indonesia (Takei and Sakamoto, 1989). Molecular analysis of foxtail millet waxy landraces by polymerase chain reaction (PCR) covering 4,006 bp of Wx gene using exon–intron primer sets (ex1/ex2 or ex2int1/ex4r) identified the insertion of transposable elements [transcriptionally silent information (TSI)-2 and TSI-7] in intron 1 or exon 3. All the waxy phenotypes harboring transposable elements in Wx loci differed in their geographical origin of domestication (Van et al., 2008). Next-generation sequencing technology has accelerated the re-sequencing of various crops to identify genome-wide nucleotide variations (Huang et al., 2009). High throughput analysis of genes and transcripts using genotyping by sequencing, transcriptome, and epigenetic analysis has been well documented as a powerful tool for identifying the genetic basis of nutritional traits in millets (Muthamilarasan and Prasad, 2015). The release of draft genome sequences of two foxtail millet cultivars, Yugu1 and Zhang gu has advanced the research for genetic improvement in foxtail millet (Bennetzen et al., 2012; Zhang et al., 2012). A foxtail millet waxy landrace, Shi-Li-Xiang (SLX) was re-sequenced using Solexa sequencing technology and Genome Analyzer II (GA II) to explore the nucleotide variations covering agronomic traits-related genes. Alignment with the reference genomes identified SNPs (single nucleotide polymorphisms), InDels, and SVs (structural variants). Novel markers obtained by re-sequencing helped in genome mapping of starch synthase encoding GBSS 1 peptide. Sequencing of GBSS 1 gene found the insertion of transposable elements confirming its waxy nature (Bai et al., 2013).

Waxy proso millet varieties are in existence since 1885 (Hixon and Brimhall, 1968). Graybosch and Baltensperger (2009) identified six waxy accessions of proso millet from the germplasm collection of United States Department of Agriculture-Agricultural Research Service (USDA-ARS) North Central Regional Plant Introduction (PI) Station, Ames, IA, USA. Based on the crosses between two (PI 436625 and PI 436626) waxy accessions and several wild type lines, two recessive alleles designated as wx-1 and wx-2 were found to be responsible for waxy phenotype. Grain starch amylose concentration was found to be reduced by 7.2-folds in waxy lines. However, a low level of amylose (3.5%) was present in some waxy types. This is most likely due to the low level of GBSS 1 activity produced by one of the alleles. The molecular basis of waxy endosperm in proso millet was further investigated using 38 landraces. Of these, 14 waxy phenotypes from China and Korea had little or no GBSS 1 activity. Sequencing of GBSS 1 in waxy lines revealed two different forms designated as “L type” and “S type” (Figure 2). These two forms differ in their exon sequence coding for mature GBSS 1 peptides resulting from three polymorphisms. The “S type” contains 15-bp deletion leading to the loss of five amino acids from glucosyltransferase domain 1. The “L type” contains a frame shift mutation caused by the insertion of an adenine residue and by substitution of a cysteine to tyrosine amino acid (Hunt et al., 2010). Based on these mutations, Araki et al. (2012) developed a PCR-based marker system (length polymorphism and derived cleaved amplified polymorphic sequences (dCAPS)) to identify waxy landraces amongst 83 accessions from Japan and 15 accessions from other countries. Sequencing of wx gene confirmed the ability of dCAPS markers in locating the single-bp mutations in exon 7 and 9 of waxy landraces.

Japanese barnyard millet is an allohexaploid and mutations in three alleles are considered to confer waxy nature. The probability of spontaneous mutations in all three alleles to obtain waxy cultivars is difficult. Wheat, also an allohexaploid was found to carry spontaneous mutations at one or two waxy loci (Nakamura et al., 1993a,b). Mutagenesis by chemical treatment on a low amylose wheat cultivar with non-functional Wx-A1 and Wx-B1 genes produced a waxy mutant (Yasui et al., 1997). With wheat as the model system, Hoshino et al. (2010) attempted to mutagenize barnyard millet landrace (Noge-Hie) into waxy phenotypes by treatment with Co⁶⁰ gamma ray irradiation. One of the mutant plants showed significant reduction in amylose content. PCR analysis of waxy genes from this low amylose line using primers designed from the consensus sequences of foxtail millet, pearl millet, sorghum, and wheat detected the absence of a gene sequence specific for waxy phenotypes. Contrary to waxy wheat with more than one non-functional Wx allele, loss-of-function mutation in one allele was found to produce waxy cultivars in barnyard millet (Hoshino et al., 2010).

Traditional breeding in millets for waxy trait is a labor intensive and time consuming process. It took nearly 15 years to transform waxy trait into non-waxy elite foxtail millet cultivar Yugu1 through cross breeding (Quan et al., 2010). Marker-assisted selection accelerates the process of conventional breeding. But the information on molecular markers linked to waxy traits in millets lags far behind cereals. Henceforth, genotyping of diverse millet germplasm by high throughput re-sequencing will facilitate the development of new molecular markers to map the waxy trait. Molecular analysis of waxy gene in foxtail millet, proso millet, and barnyard millet has identified mutations in one or more alleles. With recent advancements in genome editing, application of programmable site-specific nucleases is a straightforward approach to induce genetic mutations in non-waxy elite cultivars for transforming them into waxy phenotypes. Thus genomics approaches will speed up the genetic improvement in millets in a cost effective manner to produce biofortified varieties.

High quality proteins are essential for physical and mental well-being of humans, especially children (Heine et al., 1995; Tomé and Bos, 2007). Cereal proteins deficient in essential amino acids like methionine, lysine, and tryptophan lead to malnutrition in developing countries (Nirgude et al., 2014). Cereal proteins contain 1.5–2% lysine and 0.25–0.5% tryptophan while estimated average requirement is 5% and 1.1% for lysine and tryptophan (Young et al., 1998). Finger millet on the other hand is high in essential amino acids than cereals (Mbithi-Mwikya et al., 2000). It is therefore a suitable model system to elucidate the genetic control of protein quality. High lysine and tryptophan in finger millet is attributed to the transcriptional regulation of amino acid catabolism genes by Opaque2 (o2), a basic leucine zipper (bZIP) transcription factor. o2 modifiers (Opm) downregulate lysine ketoglutarate reductase dehydrogenase (Kemper et al., 1999) and upregulate aspartate kinase (Brennecke et al., 1996) resulting in free lysine and tryptophan in endosperm. Genetic differences in Opm alleles of finger millet germplasm remained largely uncharacterized until 2012. In the preliminary study, finger millet genotypes were evaluated for diversity in seed protein content using PCR-based markers. Random amplified polymorphic DNA (RAPD) and simple sequence repeat (SSR) profiles revealed few unique bands discriminating high and low protein genotypes. This study laid the foundation to select superior genotypes for use in traditional breeding to produce high quality proteins (Kumar et al., 2012).

Molecular characterization of Opm alleles using SSRs and SNPs can effectively identify quantitative trait loci (QTLs) influencing amino acid content (Goron and Raizada, 2015). Genic SSRs are powerful tools to link the genetic maps of related species and alleles of genic SSRs are associated with structural variations in the gene that affect transcription and translation of proteins (Kashi and Soller, 1999; Andersen and Lübberstedt, 2003). Nirgude et al. (2014) reported higher level of polymorphism in Opm genes using genic SSRs than RAPD, genomic SSRs and Cytochrome P₄₅₀ markers as reported by Panwar et al. (2010b). Utilizing the functional potential of comparative genomics, high tryptophan finger millet genotypes were identified from global collection using genic SSRs for Opm genes derived from expressed sequence tag (EST) sequences of rice, maize, and sorghum. Association mapping of SSR loci found two QTLs for tryptophan and one QTL for protein content. Interestingly, a 220-bp allele of SSR locus OM5 marker designed from the 27-kDa γ-zein gene of Opm was present mostly in the high tryptophan-containing genotypes (Babu et al., 2014). This marker could be employed in marker-assisted breeding for introgression of Opm allele into high yielding cultivars. Fine mapping of Opm genes linked to QTLs could lead to genetic enhancement of seed protein quality in cereals and small millets. Recently, 16 prolamin encoding genes called setarins have been identified in foxtail millet using computational approaches (Muthamilarasan and Prasad, 2015). Sequence alignment of setarin genes with other millets and cereals revealed least homology indicating their uniqueness in improved protein quality. Quantitative RT-PCR analysis of setarin genes confirmed their role in seed protein accumulation by overexpression in developing spikes (Muthamilarasan and Prasad, 2015). Preliminary clues of such crop-specific genes for high protein could be useful in engineering other millets and cereals for protein enrichment.

Iron (Fe) deficiency in India report 79% of pre-school children and 56% of women to be anemic (Krishnaswamy, 2009). Fe supplementation program in India since 1970 failed to address the issue of iron deficiency (Anand et al., 2014). Recognizing biofortification as a feasible alternative for Fe delivery, HarvestPlus has developed high Fe pearl millet by conventional breeding (HarvestPlus, 2009). The first step in breeding crops for better nutrition is to evaluate the genetic diversity of available germplasm for target nutritional trait. ICRISAT, a member of HarvestPlus undertook the process of screening pearl millet germplasm for sources of high Fe density. Early reports revealed positive correlation of Fe and Zn grain content with no significant correlation on grain yield and seed size (Velu et al., 2007, 2011; Gupta et al., 2009; Govindaraj et al., 2013). This suggested increase in Zn grain content as an associated trait while breeding for high Fe pearl millet.

Prospects of breeding pearl millet for high Fe content began in the year 2004 with screening of germplasm accessions, seed parents, open pollinated varieties (OPV), improved populations, and population progenies. Progenies derived from AIMP 92901, a high-yielding OPV, exhibited large intra-population variability with highest levels of Fe (about 76 mg/kg) and Zn (about 65 mg/kg) than their parental control (Velu et al., 2007). All the improved progenies were derived from iniadi germplasm. Iniadi landraces from African subcontinent have farmer preferred agronomic traits like early maturing, large seed size, compact panicles, disease resistance, and drought tolerance (Andrews and Anand Kumar, 1996; Rai et al., 2008). Hence, iniadi germplasm was exploited for further hybridization experiments to breed for high Fe content. Velu et al. (2008) identified ICTP 8203, a commercial OPV (Rai et al., 1990) released in India by 1998 and still under cultivation to possess highest Fe and Zn content. By the year 2012, high Fe biofortified variety “ICTP 8203 Fe 10-2” developed by progeny testing was commercially released in India. ICTP 8203 Fe 10-2 had 9% more Fe (71 mg/kg) and 11% more grain yield than the parental control (Rai et al., 2013).

Taking advantage of cross-pollination in pearl millet, breeders are in continuous search for nutritionally elite varieties suited for local environments. Rai et al. (2014) screened seed parent progenies and restorer parent progenies for Fe and Zn variability using X-ray fluorescence spectroscopy. The mean Fe density of these progenies increased by 5–66% than the control cultivars. As Iniadi germplasm is a promising source for high grain Fe and Zn densities, Rai et al. (2015) and Upadhyaya et al. (2016b) evaluated germplasm accessions and landraces using the inductively coupled plasma atomic emission spectroscopy. Significant variability was observed for Fe (51–121 mg/kg) and Zn (46–87 mg/kg) with positive correlation. Novel sources for developing nutrient-rich pearl millet include IP 17521 of Gnali (106.9 mg/kg), IP 11523 of Idiyouwe (106.5 mg/kg) and IP 17518 of Gnali (79.1 mg/kg), IP 11535 of Iniadi (78.4 mg/kg) for iron (Fe) content and zinc (Zn) content, respectively. These germplasm accessions with greater nutrient density than the commercial cultivar ICTP 8203 are valuable sources for introgressing high Fe and Zn into elite breeding lines. Continuation of breeding program with planned crosses will help in identifying nutrient-dense parental lines suited for rapidly changing climatic conditions.

High Fe biofortified pearl millet provides twofolds higher iron than modern wheat varieties. This led to increase in iron absorption by 5–10% in around 35 million people consuming biofortified pearl millet (Cercamondi et al., 2013; Kodkany et al., 2013). Feeding trial by Haas et al. (2013) revealed that consumption of 232 g iron biofortified pearl millet flour/day resolved 65% more iron deficiency in Indian school children. Similarly, a randomized efficacy trial of iron-biofortified pearl millet was conducted for 6 months on secondary school children in Maharashtra to assess the improvement in Fe status. The study population included children with high risk of iron deficiency, low-iron diets, and regular consumption of pearl millet. Iron intake of children consuming biofortified pearl millet far exceeded the EAR within 4 months. There was a phenomenal increase in their serum ferritin and total body iron levels. The dose of iron delivered was comparable to that achieved by supplementation trials (Finklestein et al., 2015). Promising results from bioefficacy studies is the clear evidence for the use of biofortified millets as a sustainable intervention in high risk populations to overcome the iron deficiency.

Zinc (Zn) deficiency affects 50% of the world population resulting in diarrhea, impairment of physical growth, and suppressed immune function (Hotz and Brown, 2004; Gibson, 2006; Prasad, 2007; Qaim et al., 2007; Gibson et al., 2008). Breeding approach to improve the Zn grain content is discussed in the previous section. Genetic enhancement of grain Zn content is possible by modulating the metal transporters that facilitate their uptake, translocation, and storage. Members of Zn-regulated transporters and Iron (Fe) regulated transporter-like protein (ZIP) family contribute to Zn homeostasis by either uptake or remobilization in intracellular compartments (Guerinot, 2000; Mäser et al., 2001; Colangelo and Guerinot, 2006; Figure 3). ZIP transporters increase Zn uptake in several higher plants including Arabidopsis thaliana (Plaza et al., 2007), rice (O. sativa; Ishimaru et al., 2005; Bashir et al., 2012), barley (H. vulgare; Pedas et al., 2008), soybean (Glycine max; Moreau et al., 2002), and tomato (Solanum lycopersicum; Eckhardt et al., 2001). ZIP transporters are differentially regulated in various tissues under Zn deprivation and abundance in soils (Ramesh et al., 2003). Cereals and millets with high Zn seeds can be engineered by seed-specific expression of ZIP transporters. Initial success in transgenic development for seed Zn accumulation was recorded in rice (Ramesh et al., 2004). Recently, high zinc accumulating finger millet transgenic plants were produced by overexpression of OsZIP1 driven by constitutive (35S) and endosperm-specific promoters (Bx17). Seeds of T₁ transgenic plants showed 10–15 mg/kg higher Zn than wild type and the difference further increased to 20 mg/kg in T₂ generation. Interestingly, higher Mn (5–10 mg/kg higher than wild type) accumulation was also recorded in the seeds of transgenic plants (Ramegowda et al., 2013). The in planta evidence of Zn accumulation in seeds by upregulation of ZIP transporters is the key to develop high Zn biofortified millets. As millets exhibit high synteny with cereals, expression of heterologous Zn transporters need further investigation to enhance grain Zn content.

Calcium (Ca) deficiency resulting from high intake of cereals leads to osteoporosis in women (Ross et al., 2011). Plants absorb calcium ions from the soil solution and translocate to different organs via xylem transport predominantly controlled by transpiration pull (Grusak, 2002). Non-exposed structures like developing seeds exhibit negligible rate of transpiration and are low in calcium as they acquire minerals through phloem. In reproductive tissues, ion transporter proteins direct calcium transport (White and Broadley, 2003; White, 2005; Panwar et al., 2010a). Henceforth, elucidation of role of calcium transporters in plants favors the development of Ca biofortified cereals. Finger millet containing about 5–30 times higher Ca than wheat and rice serves as a model system to understand seed calcium accumulation. Singh et al. (2014) undertook transcriptomics approach to characterize calcium sensor gene family from the developing spikes of finger millet using Illumina paired-end sequencing methods. This study included characterization, identification, classification, phylogeny, and pathway analysis of calcium sensor genes of two genotypes, GP-1(low calcium) and GP-45 (high calcium). In total, 82 calcium sensor proteins identified in the transcriptome of finger millet spikes were grouped into 25-calmodulin (CaM) and calmodulin-like proteins (CaML), 9-CDPK-related protein kinases (CRK), 9-calcineurin B-like protein (CBL), 23-CBL interacting protein kinases (CIPK), and 14-Ca²⁺-dependent and CaM-independent protein kinases (CDPK) genes. Comparative phylogenetic analysis of calcium sensor gene family in finger millet identified 12 calcium sensor genes diverse from the rice orthologs. In addition, abundance of calcium sensor gene expression in GP-45 compared to GP-1 proved the polygenic nature of calcium accumulation. Quantitative real-time PCR analysis revealed higher expression of EcCIPK2, EcCIPK9, and EcCIPK11 genes in GP-45 in response to external calcium supplied to the rhizosphere resulting in increased seed calcium content. In a similar study, Mirza et al. (2014) identified higher expression of Ca²⁺/H⁺ antiporter (CAX1), two pore channel (TPC1), CaM-stimulated type IIB Ca²⁺ ATPase and two CaM-dependent protein kinase (CaMK1 and CaMK2) genes in GP-45 resulting in greater uptake, translocation, and accumulation of calcium (Figure 3). Higher expression of CAX1 genes and CaM isoform in developing spikes strongly influence positive seed calcium content. Multigene engineering for co-expression of calcium sensor genes in cereal crops and small millets will play a significant role in the production of transgenic biofortified varieties.

Vitamin A deficiency is a serious threat to well-being of children and pregnant women in developing countries (World Health Organization [WHO], 2011). Biofortification of staple crops such as sweet potato, cassava, and maize for high provitamin A by conventional breeding has achieved great success (Bouis et al., 2011). Millet collections on the other hand lack sufficient genetic variation for beta carotene content (Buerkert et al., 2001; Upadhyaya et al., 2011a). Hence traditional breeding for high provitamin A in millets is not feasible. In such cases, transgenic approach favors metabolic engineering of vitamin biosynthesis in plants. Golden rice is a notable example for genetic modification of provitamin A in grains (Paine et al., 2005). Regulatory issues concerning the genetically modified (GM) crops are being revisited with scientific evidences to facilitate their commercial production. Therefore, research on GM millets with enriched vitamins is gaining momentum.