Discovery and validation of candidate genes for grain iron and zinc metabolism in pearl millet [Pennisetum glaucum (L.) R. Br.]

Pearl millet is an important crop for alleviating micronutrient malnutrition through genomics-assisted breeding for grain Fe (GFeC) and Zn (GZnC) content. In this study, we identified candidate genes related to iron (Fe) and zinc (Zn) metabolism through gene expression analysis and correlated it with known QTL regions for GFeC/GZnC. From a total of 114 Fe and Zn metabolism-related genes that were selected from the related crop species, we studied 29 genes. Different developmental stages exhibited tissue and stage-specific expressions for Fe and Zn metabolism genes in parents contrasting for GFeC and GZnC. Results revealed that PglZIP, PglNRAMP and PglFER gene families were candidates for GFeC and GZnC. Ferritin-like gene, PglFER1 may be the potential candidate gene for GFeC. Promoter analysis revealed Fe and Zn deficiency, hormone, metal-responsive, and salt-regulated elements. Genomic regions underlying GFeC and GZnC were validated by annotating major QTL regions for grain Fe and Zn. Interestingly, PglZIP and PglNRAMP gene families were found common with a previously reported linkage group 7 major QTL region for GFeC and GZnC. The study provides insights into the foundation for functional dissection of different Fe and Zn metabolism genes homologs and their subsequent use in pearl millet molecular breeding programs globally.


Results
In silico identification of Fe and Zn metabolism genes. A total of 114 Fe and Zn-responsive gene sequences from different plant species including Setaria italica (11), Zea mays (11), Oryza sativa (39), Sorghum bicolor (3) and Arabidopsis thaliana (49) were downloaded from NCBI. BLAST was performed against pearl millet sequences which resulted in the identification of 29 putative Fe and Zn metabolism genes. The reliability of these porters was checked by confirming the presence of Fe and Zn conserved domains. These genes were subdivided into five sub-families based on their conserved domains. The YSL was found to be a large family with 12 genes (with the exception of PglYSL2 which was not considered due to a smaller length), followed by ZIP superfamily with 9 genes, NRAMP family with 6 genes and 1 each gene in NAS and ferritin-like (FER) families. For convenience, they are named as PglYSL for Yellow stripe-like proteins (PglYSL1-12), ZIP (PglZIP1-9), Nramp (PglNRAMP1-6), Ferritin like (PglFER-1) and NAS (PglNAS-1) (Supplementary Table S1). chromosomal location and structure of fe and Zn metabolism genes. The 29 genes identified for Fe and Zn transporters in the present study were distributed on all 7 chromosomes of pearl millet and their homologs were found in Setaria italica, Sorghum bicolor, Oryza sativa and Zea mays which are shown in CIRCOS 31 Fig. 1. Of the 29 genes, PglZIP1 was found on chromosome 1; 5 genes on chromosome 2; 13 on 3, 2 on 4, 3 on 5, 2 on 6 and 3 on chromosome 7. Results indicated that chromosome 3 is the hot spot for these genes as it is homing maximum genes (Fig. 2). The gene structures analysis revealed that three genes are intron less while remaining displayed many introns. The exons ranged between 1 (Pgl_GLEAN_10002987) and 13 (Pgl_GLEAN_10000860) (Supplementary Table S1).

Scientific RepoRtS
| (2020) 10:16562 | https://doi.org/10.1038/s41598-020-73241-7 www.nature.com/scientificreports/ Sequence analysis and conserved motif analysis of fe and Zn metabolism genes. The number of amino acids in the proteins ranged between 65 (PglYSL2, not considered as full-length gene) and 883 (Pgl-NAS1); and the pI ranged from 5.32 (PglFER1) to 10.02 (PgYSL2). Out of 29 proteins, 10 were found acidic and the remaining basic. Molecular weights varied between 7146.53 (PglYSL2) and 96,605.68 (PglNAS1) daltons. Except for PglFER1 and PglNAS1, all Fe and Zn transporter proteins were found hydrophobic. They showed the highest aliphatic index that varied from 132.59 (PglNRAMP5) to 82.52 (PglFER1). Transmembrane helices ranged from 1 to 15, but 2 proteins (PglFER1 and PglNAS1) exhibited no helices. Subcellular localization revealed that of 29, 21 were in the plastid, 4 in the vacuole, 2 in the chloroplast, 1 each in the cytosol and extracellular region. The phosphorylation sites of these proteins (Supplementary Table S2) indicated that they are phosphorylated at serine residues, followed by threonine. But, the YSL subgroups were found phosphorylated at tyrosine residues. The PKC, CKII, PKA, UNSP, and CDC2 were common kinases for Fe and Zn protein phosphorylations. The identified Fe and Zn transporter proteins belonging to various subgroups did not display inter-subgroup similarity, hence each family was submitted to MEME separately for conserved motif analysis.  www.nature.com/scientificreports/ The MEME analysis revealed ten conserved motifs, which showed various compositions and distribution patterns among the family members. All the YSL group proteins exhibited ten conserved motifs except YSL2, which showed 2 motifs. Motifs 6 and 10 were conserved at N terminus, while 5 and 8 at C terminus. Out of 29 identified genes, 10 showed conserved motifs among the YSL members. Motifs 6 and 3 are the most conserved at N terminus, while motifs 7 and 8 are commonly present at C terminus. In the ZIP family, motif 3 was found as the most conserved at N terminus, followed by motif 5, while 3 and 4 represent NES motifs. In the Nramp family, 10 were identified as the conserved motifs. While motif 3 and 9 were the most conserved at N terminus, motif 6 and 7 conserved at C terminus. Among all, motif 1 and 3 appeared as the utmost conserved (Fig. 3).
phylogenetic analysis. A phylogenetic tree was constructed for 29 identified Fe and Zn metabolism genes, belonging to 5 groups in pearl millet. The analysis revealed four clades, of which five subfamilies clustered and showed group-specific distribution based on the conserved domains (Fig. 4a). While FER1 and NAS1 were clustered into one group, YSL2 fell into the ZIP family, ZIP3 and ZIP6 into the NRAMP subgroup. A total of 9 paralogs, 4 tandem duplications (PglYSL1/PglYSL4, PglYSL5/PglYSL6, and PglYSL10/PglYSL11) on chromosome 3, one PglNRAMP5/PglNRAMP6 on chromosome 7, and 5 segmental duplication events were noticed. The phylogenetic tree was constructed to identify the evolutionary relationships of 5 gene families of Fe and Zn metabolism genes in pearl millet with other crop plants ( Fig. 4b-f). The trees showed species-specific, group-specific and class-specific clades. Pearl millet showed 4 orthologous events, of which one with Oryza (Os11G0138400/Pgl_GLEAN_10003023), while remaining with Setaria (Si024505/Pgl_GLEAN_10026644, Si010411/Pgl_GLEAN_10012759 and Si03616/Pgl_GLEAN_1005377). All four genes belonged to the ZIP family (Fig. 4d).
promoter analysis. Promoter analysis revealed that the majority of them contain Zn deficiency-responsive (ZDRE), iron-responsive (IRO2) and iron deficiency elements (IDE), which are responsible for iron and zinc deficiencies. Interestingly, two IRO2 elements were found in PglFER1 gene which is a potential candidate gene for GFeC. In PglZIP2 and PglZIP4, single ZDRE element was found, whereas for PglZIP2 two, PglZIP4 one, PglNRAMP5 two, and PglZIP9 three IDEs were found. They also contained dehydration-responsive elements (DRE) for drought, low temperature-responsive elements (LTE), salt-regulated elements (SRE, GT1GM) and abscisic acid-responsive elements (ABRE). They exhibited the highest number of ABRE along with Cu-responsive elements. Among all, genes 3, 7, 11, 22, and 27 contained large amounts of stress-responsive elements (Supplementary Table S3).
Grain fe and Zn content. The grain Fe and Zn content in bi-parental RIL mapping population parents AIMP 92901-S1-183-2-2-B-08 (hereafter referred to as AIMP 92901) and ICMS 8511-S1-17-2-     www.nature.com/scientificreports/ compared to vegetative stages. In ICMS also, the up-regulation of genes was observed in roots and leaves at the vegetative stage. The PglFER1, PglZIP2, PglZIP8, PglNRAMP2, PglYSL1 were found to be the candidate genes with high expressions in seeds showed in CIRCOS 31 (Fig. 5b). They showed moderate expressions in the roots at the vegetative stage, indicating their Fe and Zn restoration nature in seeds. In ICMS, the majority of them were expressed at the vegetative stage compared to the reproductive stage, while in AIMP they were expressed in roots at the reproductive stage similar to MRC, the check.
Correlation of differentially expressed genes with previously reported QTLs. The earlier identified LG1 and LG7 high grain QTL regions for GFeC and GZnC 7 were scanned for presence of PglZIP, Pgl-NRAMP, PglYSL and PglFER family genes that were differentially expressed between the two parents. Upon sequence comparison between these genes and the QTL regions defined by the set of linked markers, we could www.nature.com/scientificreports/ not locate any common set of genes for LG1 QTL interval. However, interestingly, for LG7 QTL interval two gene families, PglZIP and PglNRAMP were found to be common. miR159 associated with Fe. A host of plant miRNAs were screened in addition to checking the miRNAs from miRBASE. miR159 is the only miRNA that appeared to be associated with Fe, but, no miRNAs could be found for Zn. Surprisingly, careful dissemination and search from PMRD yielded only one miRNA in pearl millet. Before this, we have considered all the miRNAs in Arabidopsis and made a Smith-Waterman/Blast search to find any potential miRNAs, but in vain except miR159.
protein-protein interaction analysis predictions. The predicted protein-protein interaction network revealed interactions of Fe and Zn accumulation associated proteins with several other transporter proteins of pearl millet (found ontology with Sorghum and Seteria). Fe and Zn transporter proteins have been noticed as centers for interactions and displayed interactions with various other proteins associated with the metabolism of fructose, galactose, and also with ATP binding protein, hexokinase and protein kinase (Supplementary Table S5; Fig. 6).

Discussion
The deficiencies and improper consumption of micronutrients like minerals and vitamins lead to multiple disorders and diseases in humans and are collectively stated as hidden hunger or malnutrition. The connecting link between health and food has been well documented as human beings require several organic and mineral nutrients for their proper growth and development. Hyper and hypo levels of iron and zinc in the soil cause many physiological and metabolism-related complications in plants. But plants have developed a tightly regulated cellular iron and zinc homeostasis, including the uptake, distribution within different tissues and their utilization 32 .
Since both iron and zinc are the vital nutrients, their deficiencies can lead to a variety of problems like retarded growth, skeletal abnormalities, learning capacity and wellbeing of humans 33 . Their deficiencies cause nutritional disorders in human beings, especially in developing world 34 . Edible parts of many cereal cultivars have inherently low iron and zinc content. Millets are nutritionally rich and can serve potentially as crops of nutritional security in the developing and underdeveloped countries. Therefore, biofortification of cereals or increasing their bioavailability in the edible parts like endosperm can cost-effectively solve the problem of malnutrition and save millions of lives across the globe 35 . Among the Fe and Zn superfamily, genes such as YSL, ZIP, NRAMP, FER, NAS have been identified in many plants like Arabidopsis thaliana 36 , Setaria italica 29 and Oryza sativa 28 . These genes were responsible for Fe and Zn translocation from soil to the grain filling sight. Genome-wide studies of different families of Fe and Zn have been characterized individually in cereals, but there were no reports on the identification and gene expressions of Fe and Zn in pearl millet, though it has high micronutrients 37 . In www.nature.com/scientificreports/ the present study, 12 PglYSL , 9 PglZIP, 6 PglNRAMP and a single gene for PglFER and PglNAS were identified and characterized from P. glaucum. However, form PglYSL family, one gene PglYSL2 was found to be of much shorter length (just 65 amino acids). This may be because of differential divergence of this gene in pearl millet, or because of lack of capture of complete sequence during the genome assembly process. Hence, PglYSL2 was not considered in further studies. Palmer et al. 38 identified 520 genes that belong to 40 different transporter classes in the Panicum virgatum (C 4 switchgrass) genome. There are five sub-families for Fe and Zn transport based on their conserved domains. Attempts were made earlier for in silico identification and expression analysis of either YSL 39 , or ZIP 40 or NAS gene family 41 or NRAMP members 42 in a wide array of plants. The YS1 or YSL is a membrane-bound Fe (III)phytosiderophores (PS) transporter 43 and is distantly related to the Oligopeptide Transporter (OPT) family proteins. Though non-grass species do not synthesize PS, many YSL genes were detected in monocots, dicots, and lower groups of plants. Meena et al. 44 identified 8 and 18 YSL family members in Arabidopsis thaliana and Oryza sativa, respectively. Similarly, in the present study, 12 genes belonging to the YSL family were detected suggesting that they are highly conserved across plant species. Species-specific variations in the gene numbers were also known to exist for many gene families in cereals 45 . The ZIP family members have been characterized in higher plants and are involved in metal uptake, transport, detoxification and storage of Fe and Zn in plant cells 46 . In Oryza sativa, 17 ZIP genes 47 , 15 in Arabidopsis thaliana 10 , 12 in Poncirus trifoliata 48 , and 9 in Zea mays 46 were reported. Alagarasan et al. 29 on the other hand, identified 36 ZIP family members in Setaria italica. In contrast, in the present study, 8 ZIP family members (out of a total of 29 iron and zinc transporter genes) were noticed in the genome of P. glaucum. This number is close to the published data in maize that belongs to the same family 46 . Members of the NRAMP family genes are also widespread from bacteria to plants to humans and transport Fe and Mn (manganese) across cellular membranes 49 . In the present study, 6 NRAMP members were identified in pearl millet like that of Arabidopsis 10 , while Qin et al. 42 reported 13 in soybean. In Arabidopsis, 4 FER genes were detected 50 , and 3 in Lupinus luteus 51 . Gross et al. 52 identified 2 FER genes in maize and rice, respectively, but only one is detected in the present study. A systematic analysis for NAS members could identify 3 and 4 NAS members in Oryza sativa and Arabidopsis thaliana, respectively 53 , while in the current study merely one NAS member was identified in pearl millet. But in maize and barley, 9 NAS genes were detected 41 . This suggests that NAS members in pearl millet are encoded by a few genes instead of a large gene family members. Zhou et al. 41 pointed out that large NAS gene members in maize and barley may be due to gene duplication events that evolved during the evolution of new species.
YSL members are acidic with a large number of glutamic and aspartic acid residues. It was observed that these proteins are highly hydrophobic (27 out of 29) and are membrane-bound transporters. Except for two, the rest of the Fe and Zn porters have transmembrane helices in pearl millet. Such variability can occur between transmembrane domains 7 and 8 and also at the N-terminal portion of the YSLs. Chu et al. 39 reported that ysl1/ ysl2 double mutants result in strong interveinal chlorosis and also reproductive defects implying that these two YSL members are needed for loading metals into the reproductive tissues. Menna et al. 44 found that the intron/ exon structure and domains of YSL members from rice and Arabidopsis are conserved during evolution. They also noticed less variation in sequence lengths of amino acid residues of rice (554-728) and Arabidopsis (664-724) YSL members. Though they are transport proteins, plasma membrane localization of these proteins could not be detected among the 29. But, others 44 could detect plasma membrane-bound porters/proteins as the probable subcellular localization. This difference could be due to the diverse bioinformatics tool(s) that we use for this purpose. Guerinot 9 pointed out that a metal-binding domain is seen between transmembrane 3 and 4. A phylogenetic tree was constructed to find out the evolutionary relationships of these porters. The Arabidopsis YSL family members share high sequence similarity. Based on sequence similarity, the YSL family can be divided into three conserved groups 39 , while we noticed only two groups (Fig. 3). It was noticed that amino acid sequences of PglZIPs were related to ZIPs identified in other plant species. This infers that PglZIPs share a common evolutionary ancestor along with other cereals. Two subfamilies appear for the members of Nramp that are common both in dicots and monocots as noticed in the present study and also in soybean 41 . Phylogenetic analysis revealed that NASs from the members of Poaceae were divided into two classes. Further, more members existed in class I in maize and barley than in rice 41 . Schnable et al. 54 suggested that nearly 25% of the genes in maize possess closely related paralogs due to genome duplication events. Many NAS genes in some of the monocots indicate that there may be functional redundancy as also noticed by Mizuno et al. 55 . Nevertheless, the evidence is lacking at present to exclude or include the possibility of functional redundancy and if they are pseudogenes or not. Strozycki et al. 51 working with Lupinus suggested that eudicot ferritins are both structurally and functionally diverse. They found out that 14-bp-long iron-dependent regulatory sequence (IDRS) is the characteristic feature for all higher plant ferritin promoters. They also found that a highly conserved IDRS can be extended (extIDRS) up to 22 bp. Such an analysis of ferritin gene promoters for specific or conservative elements and their validation only can resolve such a complexity. In the present study, the analysis of promoter regions was carried out for all the 29 genes. The presence of Fe and Zn deficiency elements, hormone-responsive, metal-responsive and salt-regulated elements indicate that Fe and Zn transporters are associated with diverse stresses. However, the promoters need to be validated further in the lab. The grain Fe and Zn content were higher in AIMP compared to ICMS indicating that AIMP should be utilized for future breeding programs. In pearl millet, QTLs for high GFeC and GZnC revealed a co-localization of QTLs on chromosome 3. Major QTLs were noticed earlier in other populations also, and LG3 and LG1 served as co-localizations for high grain Fe and Zn 56 . Current work thus forms a perfect computational framework for in silico characterization of five gene families that can be utilized in marker-assisted selection (MAS) of important cereal crops for enriching grain Fe and Zn.
The present work reveals that major differential gene expressions were recorded in PglYSL2, PglZIP2, PglZIP8, PglNRAMP2, PglFER1 in the two contrasting parents. Such Fe and Zn transporter gene expressions in diverse tissues under various developmental stages provide evidence in the acquisition of iron and zinc, and tight regulation  44 reported a preferential pattern of YSL gene expressions in rice and Arabidopsis which suggest spatiotemporal regulation of YSL genes during plant development. Interestingly, both AtYSL1, 3 and 6 were highly expressed in senescent leaves indicating that they are implicated in the remobilization of metals 44 . It is known that some ZIP genes in Arabidopsis and rice play crucial roles in transporting Zn or Fe. ZIP genes exhibit different expression profiles like tissue specificity and response to the availability of Fe and Zn in the environment 57 . The gene expression patterns of PglZIPs indicate diverse functions carried out by ZIPs. While expression of MtZIP1 was reported in Zn-deficient roots and leaves 58 , MtZIP2 was stimulated in roots and stems by Zn deficiency 59 . Yang et al. 60 observed that OsZIP7a was induced mostly in roots where Fe-deficiency is common, but OsZIP8 was activated in Zn-deficient shoots as well as roots. In Arabidopsis, bZIP19 and bZIP23 regulated the adaptation to Zn deficiency by increasing the transcription of ZIPs 61 . In maize, ZmZIP4, 5, 7 and 8 were upregulated in shoots and roots after excess Fe was applied 46 . They pointed out that ZmZIP4, 5, 7 and 8 may be associated with excessive Fe detoxification and storage. Also, ZmZIP4 and ZmZIP5 were stimulated during embryo development and they were repressed in later stages 46 . Such an expression suggests that these two ZIPs are essential for the growth of plumule and radicle in maize. Qin et al. 42 studied the soybean NRAMP gene expressions and suggested these genes function in many tissues and developmental stages. GmNRAMP genes were found to be differentially regulated upon N, P, K, Fe and S deficiencies as well as toxicities of Fe, Cu, Cd, and Mn. These results infer that NRAMPs are functionally associated with many nutrient stress-responsive pathways. Zhou et al. 41 found that class I NAS genes were induced under Fe deficiency, but suppressed under excess Fe conditions in Zea mays. In contrast, they noticed that the expression pattern of class II NAS genes was a reversal of class I genes, thus complementing each other. While ZmNAS1; 1/1; 2 was mainly expressed in cortex and stele of root tissues under sufficient Fe conditions, but under deficit conditions, its expression was noticed also in the epidermis, and shoot apices. The expression patterns suggest that the two classes may be regulated at the transcriptional level and respond to the demand for iron uptake and transport 40 . Fobis-Loisy et al. 62 noticed in maize differential accumulation of FM1 and FM2 (two sub-classes) mRNAs in response to iron, ABA and water deficit conditions which could be due to differential transcription of ZmFer1 and ZmFer2. Bournier et al. 63 found out that a molecular link exists between the control of iron and phosphate homeostasis in Arabidopsis. Further, genome-wide analysis of gene expression profiling revealed that an intact COP9 signalosome is essential for correct expression of Fe homeostasis genes in Arabidopsis 64 . Such a possibility cannot be ruled out in pearl millet. Our in silco analysis has further validated key genes that are differentially expressed like ZIP, and NRAMP family members which have been found significantly on LG 7 QTL by annotation study and found to be a candidate gene. Besides, miRBASE, other plant miRNAs were screened in the present study for identifying miRNAs associated with the regulation of Fe, Zn transporters 65 . We found that pgl-miR159 is known to be the only matured miRNA sequence that was identified from the Plant miRNA database (PMRD) 66 . With the plant miRNAs predominantly targeting transcription factors, we aimed to identify functional targeted genes responsible for various physiological processes using psRNAtarget 67 . For the lone miRNA, we found that 1-amino-cyclopropane-1-carboxylate synthase 8 (ACS8) is the probable target. While this gene was found in Arabidopsis, it is not yet annotated in pearl millet. We believe that ACS transcripts are responsible for Fe deficiency, which is invariably associated with increased abundance of ACS transcripts in plant leaves 68 . Our study revealed that among the 29 proteins, 6 proteins (PglFER1, PglZIP2, PglZIP3, PglZIP4, PglZIP5, and PglZIP9) formed tight networking with other proteins (Fig. 6). The protein Pgl_GLEAN_10005377 (PglZIP2) displayed the highest interactions. (Supplementary Table S5). PglFER and PglZIP might play a key role in mineral transportation and accumulation in seeds 69 . conclusions Pearl millet is a climate-resilient nutricereal with substantial genetic variability for grain Fe and Zn content. Our results from the gene expression studies, corroborated by QTL interval annotation revealed that the gene families PglYSL, PglZIP, PglNRAMP, PglNAS, and PglFER play an important role in Fe and Zn homeostasis in pearl millet. Ferritin-like gene PglFER1 along with other key candidate genes should be further investigated in diverse pearl millet germplasm. This study provides insights into the foundation for functional dissection of different Fe, Zn metabolism genes homologs. The genomic resources developed in this study may be used for the development of breeder-friendly marker systems for improvement of grain Fe and Zn content in the pearl millet breeding programs globally. www.nature.com/scientificreports/ ware was used for obtaining the Fe and Zn gene structures-exons, introns, and untranslated sequence regions (UTRs) based on the alignments of their coding sequences. MEME software 72 was employed to analyze the new sequence patterns, their distributions, significance 73 and identify the motifs by setting different default parameters like the number of motifs from 1-10, with a motif width of 5-50, and the number of motif sites from 5-10. in silico protein analysis. The molecular weight (MW), isoelectric point (pI), and grand average of hydropathy (GRAVY) of Fe and Zn metabolism genes were identified using ProtParam of Expasy tools 74 (https :// web.expas y.org/protp aram), while the phosphorylation sites were predicted by employing NetPhosK3 software of Expasy tools 75 . The putative transmembrane helices within genes were identified using TMHMM software 76 . For finding out the subcellular localization of transporters, the WOLFPSORT program (https ://wolfp sort.org/) was used 77 . in silico prediction of cis-regulatory elements and phylogenetic analysis. Genomic sequences measuring 2000 bp of upstream of the start codon of Fe and Zn metabolism genes were extracted for the identification of cis-acting elements. The elements were analyzed by employing PLANTCARE software 78 80 . The ground samples were digested in closed tubes, and Fe and Zn in the digests were analyzed using Inductively Coupled Plasma Optical Emission Spectrometry (ICP-OES). RNA was isolated at the vegetative and reproductive stages. The flag leaf and root were harvested first at the vegetative stage once and again the same was collected at the mature embryo (reproductive) stage and frozen in liquid nitrogen and stored at − 80 °C until further use. The values are represented as means ± standard errors of the replicated data.

Materials and methods
RNA isolation, cDNA synthesis and qRT-PCR analysis. Total RNA was extracted from different tissues of pearl millet using the Invitrogen RNA isolation kit according to the manufacturer's instructions. A total of 2.5 μl RNA with 5 μg concentration was taken for conversion to cDNA using a first-strand cDNA synthesis kit (Thermo Scientific) for qRT-PCR analysis of pearl millet Fe and Zn transporters. The template was diluted with nuclease-free water (1:10). Gene expression analysis was performed for 29 gene-specific primers, along with three reference genes (Supplementary Table S6), namely elongation factor 1-alpha (EF-1α), eukaryotic initiation factor 4A (EIF4A) and an acyl carrier protein (ACP) 81 for gene normalization in 96-well optical PCR plates. The SYBR Green PCR master mix (Applied Biosystems, CA, USA) was used for gene expression analysis according to the manufacturer's instructions. The ABI 7500 real-time PCR system (Applied Biosystems, USA) was used with the following thermal cycling conditions; the cycle 1 at 95 °C for 10 min, followed by 40 cycles at 95 °C for 15 s and 60 °C for 1 min (variable based on primers gradient temperatures) (Supplementary Table S6). The amplicon dissociation curves were recorded with a fluorescent lamp after the 40 th cycle by heating from 58 to 95 °C within 20 min. Three biological replicates were taken in addition to two technical replicates each time.
The relative expression values of 29 genes in five tissues as mentioned earlier at two developmental stages (vegetative and reproductive) of 3 contrasting genotypes (low and high Fe and Zn content) were calculated using qbase + software 82 .
Correlation of differentially expressed genes with previously reported QTLs. We correlated differentially expressed genes in the plant materials (ICMS 8511 and AIMP 92901) which also happen to be parents of a bi-parental RIL mapping population. This RIL population was earlier used to identity LG1 and LG7 high grain QTL regions for GFeC and GZnC 7 . Using in silico approaches, the LG1 and LG7 QTL interval was scanned for presence of common genes that were differentially expressed and were also present in the QTL region.
In silico identification of miRNA(s). For the in-silico identification of miRNA associated with Fe and Zn we have used 29 putative gene sequences as input to find perfect or complementary mi RNA binding to its target sequences. The pearl millet miRNA targets were identified through pairwise homolog search by subjecting mature miRNA sequences as query against manually curated reference mRNA sequences (RefSeq, NCBI) and assembled EST sequences of pearl millet using Plant small RNA Target 67 (psRNATarget). Analysis server (https ://plant grn.noble .org/psRNA Targe t/) of Dai and Zhao 83  www.nature.com/scientificreports/ and Plant miRNA database 66 (PMRD): https ://bioin forma tics.cau.edu.cn/PMRD/. To further confirm this, we have also done a UVA Fasta/Smith-Watermann algorithm search using the Pearl millet and Arabidopsis whole genomic sequences taking the matured sequences as query from miRNA databases. The brief methodology representing in pictorial form in supplementary Fig. 1. protein-protein interaction. The protein-protein interaction (PPI) was prepared using 29 identified genes. As the pearl millet PPI network details are not available in the STRING database, initially all the orthologs of P. glaucum in the S. italica were listed out. For each gene (29), corresponding and interacting proteins were listed by searching in the STRING database and redundant proteins were discarded. Network analysis and functional enrichment in the network were performed by Cytoscape 85 .

Data availability
All the data presented in the manuscript are publicly available. The gene expression data generated during this study are available from the corresponding author(s) on request. All authors have read and accepted the MS. www.nature.com/scientificreports/