Among all major enzymes involved in starch biosynthesis in plants, ADP-glucose pyrophosphorylase (AGPase) is the first key regulatory and rate limiting enzyme (Smidansky et al., 2002; Tuncel and Okita, 2013). Using glucose-1-phosphate (G-1-P) and ATP as the sole precursor molecules, AGPase synthesizes ADP-glucose (ADP-G) moieties, which are the basic building blocks of starch (Soliman et al., 2014). This enzyme controls the amount of starch in cereal grain endosperm as well as in other plant organs (for reviews see James et al., 2003; Saripalli and Gupta, 2015). Various abiotic stresses including heat and drought inhibit the activity of AGPase leading to significant losses in crop yield (Boehlein et al., 2008; for a review see Saripalli and Gupta, 2015). Therefore, attempts have also been made in maize to engineer the AGPase to obtain its thermostable variants (Hannah et al., 2012; Boehlein et al., 2015). However, the precise nature of the metabolic and developmental changes that result due to genetic engineering of AGPase remains to be elucidated (Smidansky et al., 2002, 2003, 2007; Hannah et al., 2012).

AGPase occurs in both, cytosol and plastids. In cereals, a major part of ADP-glucose (ADP-G) is synthesized mainly in the cytosol of the developing endosperm cells, although some ADP-G is also synthesized in the plastids (Denyer et al., 1996). However, in dicotyledonous plants, AGPase-mediated ADP-G synthesis primarily occurs in the plastids (Beckles et al., 2001a,b). The plastidial AGPase also encodes transit peptides for transport of the newly-synthesized AGPase subunits into the plastids. Both plastidial and cytosolic AGPases function as heterotetramers (α2β2), each composed of two similar small subunits, SS (α2) and two similar large subunits, LS (β2) (Preiss et al., 1990). The SS and the LS differ only slightly in their molecular weights, and are encoded by different nuclear genes (for a review see Georgelis et al., 2007). The SS has both catalytic and regulatory functions, whereas the LS have largely a regulatory function (Kavakli et al., 2001). The relative roles of the two subunits suggest that the LS in the seed endosperm plays an important role in determining grain yield in cereals (Lee et al., 2007; Li et al., 2011; Hannah et al., 2012; Kang et al., 2013), while SS in the leaf contributes to plant biomass (Li et al., 2011; Schlossar et al., 2014). Two different types of AGPase SS genes (Type 1 and Type 2) have been reported in most of the cereals. Type 1 genes give rise to two transcripts; one is cytosolic endosperm and second is leaf plastidial. Similarly, Type 2 genes encode endosperm plastidial protein (Johnson et al., 2003; Rösti and Denyer, 2007; Kang et al., 2010). In case of rice and maize, AGPase LS is encoded by four different genes; one of these four genes is mainly expressed in embryo and remaining three genes are either expressed in embryo or leaf or both (Bae et al., 1990; Lee et al., 2007). However, in wheat (Johnson et al., 2003; Kang et al., 2010) and barley (Burton et al., 2002; Johnson et al., 2003), LS is known to be encoded by two genes which are mainly expressed in endosperm and embryo.

The genes encoding the LS were cloned from maize and hexaploid wheat (Shaw and Hannah, 1992; Thorneycroft et al., 2003; Rose et al., 2016); however, genes for the SS were cloned only from maize and potato (Nakata et al., 1994; Hannah et al., 2001). Recently, the primary and secondary structures of the AGPase subunits have been analyzed in silico in some selected monocots (maize, wheat, rice and barley) and dicots (potato and Arabidopsis) (Rani et al., 2013), where major emphasis was laid on the study of physico-chemical properties of amino acids and their importance in conferring stability to proteins. Using the crystal structure of the potato SS (Jin et al., 2005), in silico 3D structure of SS in several monocots and dicots as well as that of heterotetramer in potato and rice and heterodimer in wheat have also been reported (Tuncel et al., 2008; Danishuddin et al., 2011; Dawar et al., 2013; Sarma et al., 2014).

The monomer units of AGPase LS and SS, each consists of an N-terminal catalytic domain (ADP_Glucose_PP) and a C-terminal β-helix domain (Lbh_G1P_AT_C). The catalytic domain mainly consists of a largely parallel but mixed seven-stranded β sheet covered by α helices, which represent a fold reminiscent of the dinucleotide binding Rossmann fold. N-terminal catalytic domain makes strong hydrophobic interactions with the C-terminal β-helix domain via an α helix (Jin et al., 2005; Rani et al., 2013).

Although considerable information about AGPase and its genes is available, but detailed information about the diversity and variation in structure, function and evolution of the underlying genes among different plant species is missing. Therefore, the present study was carried out using well characterized maize AGPase genes as a reference to identify and characterize its “true” orthologs from 10 plant species including both cultivated and model species along with two progenitors of wheat (Triticum urartu and Aegilops tauschii). Results for the following aspects in 11 species (including maize) are presented in this communication: (i) identification of the “true” orthologs of maize LS and SS genes in 10 species; (ii) synteny and collineraity analyses across the 11 species; (iii) evolution of gene structure for both subunits among the monocots and the dicots; (iv) identification of motifs and domains of AGPase protein that are critical for proper functioning of the enzyme and the subtle differences between and within AGPases of monocots and dicots; (v) promoter and expression analysis, revealing novel features that might have bearing on the spatial and temporal expression of AGPase genes. Possible use of this knowledge for crop improvement is also discussed.

The present study demonstrated a high level of conservation of AGPase and its subunits (LS and SS) not only at the protein level, but also at the gene level in both monocots and dicots. At the protein level, this conservation was evident through estimates of similarity, identity, coverage of a sequence, the presence of two similar conserved domains (ADP_Glucose_PP and Lbh_G1P_AT_C) and 3D structure analysis across all the 11 species examined (8 monocots and 3 dicots). The data at the DNA level (gene sequence) suggested a higher level of conservation in exonic regions relative to that in introns, which is understandable, because introns are not involved in encoding the protein.

The level of conservation was higher for SS than for LS both at the protein and gene levels. This is in agreement with the results of an earlier study (Smith-White and Preiss, 1992). At the DNA level, value of Ka/Ks for SS was 2.61-fold that of LS, although the values of Ka/Ks for LS was higher in two earlier studies; Ka/Ks value in LS was 3.18-fold that in SS in one study (Corbi et al., 2012), and 2.7-fold in the other study (Georgelis et al., 2007, 2008), which may be attributed to the study of a larger set of species by the latter two studies than the present study. The higher similarity among SS than in LS may be due to its higher functional importance in the enzyme complex or due to a higher purifying selection exercised on SS during the evolution as suggested by Georgelis et al. (2007). Many more exons and introns, and relatively higher variability in their number in LS can be attributed to many more duplications in LS genes than in SS genes belonging to a number of eukaryotic plant systems (Georgelis et al., 2008), suggesting relatively higher conservation of SS than LS. Our analysis on intron phase gives further support to this conclusion. The prevalence of phase 0 in several genes was earlier attributed to primitive and conserved nature of the corresponding genes, since it allowed conservation of codons in the reading frame (Long and Deutsch, 1999; Figures 1, 2).

A comparison of intron length variation revealed that introns were generally longer in SS (34 to 3532 bp) than in LS (66 to 605 bp except third last intron of maize). Available evidence also suggested that during the course of evolution, division of introns in LS genes have given rise to smaller introns in monocots. For instance, splitting of sixth intron (537 bp) of LS on 1AL in T. urartu, the progenitor contributing A sub-genome to wheat, gave rise to a smaller sixth intron (102 bp) and two extra exons and introns (7 and 8) in wheat ortholog on chromosome arm 1AL. This variation in intron lengths may also be due to one or more of the following reasons: (i) presence of transposable elements, (ii) the changes in frequency and size of deletion events, and (iii) presence of regulatory elements and RNA genes (Zhu et al., 2009).

The low GC content in the introns than in the exons of genes for both subunits (LS and SS) was conspicuous, although it differed for the two subunits. Birdsell (2002) suggested that introns, intergenic regions and pseudogenes tend to have lower GC contents than ORFs. Differences in GC content allow discrimination between exons and introns and allow marking of exons for the splicing machinery (Amit et al., 2012). A positive correlation between the size of introns/exons (except introns of LS) with the GC content, observed in the present study, was in agreement with similar correlation reported in case of plant genomes of rice and Arabidopsis and the genomes of fly, zebrafish and worms (Zhu et al., 2009).

The sequences of genes for LS and SS of monocots and dicots differed, and seem to be more conserved in monocots. For instance, the number of exons and introns in monocots largely remained constant during evolution except in case of LS of T. urartu (15 exons in LS). However, variation seems to have occurred in dicots (12–14 exons in LS, 8–9 exons in SS), which was accompanied with loss of introns, leading to fusion of exons. It appears, that after the divergence of monocots and dicots from a common ancestor 140–150 MYA (Chaw et al., 2004), there was intron loss in dicots (Roy and Gilbert, 2006), which receives support from the fact that the exons were longer in dicots (Figures 1, 2). Dicot-monocot divergence is also apparent in our dendrogram, and in an earlier study (Rani et al., 2013). This is understandable since the domestication syndromes differ in monocots and dicots due to difference in the storage organs for starch. Higher variations in dicots also suggest a higher level of conservation in monocots than in dicots.

The lack of shared synteny (gene content) and collinearity (gene order) in most species except rice and sorghum suggest that the homoeologous chromosomes of different species carrying AGPase genes have undergone extensive reshuffling during evolution. The retention of some degree of synteny in rice and sorghum is understandable, since relative to other species, these are more closely related to maize (Salse et al., 2004). However, complete absence of shared synteny in other monocot species including T. aestivum (wheat), T. urartu, Ae. tauschii, barley and Brachypodium was a bit surprising, because at the whole genome level, extensive conservation of synteny has been reported among all grasses (Pfeifer et al., 2012).

Synteny and microsynteny information from the available literature (Ahn et al., 1993; Sorrells et al., 2003; Bennetzen and Chen, 2008) also suggest that although synteny at the whole genome level is conserved among grasses, but there are genomic regions, where synteny and microsynteny is lacking (Ahn et al., 1993; Tarchini et al., 2000; Vision, 2005; Bennetzen and Chen, 2008). This loss of shared synteny is attributed to gain/loss of genes and other structural changes within chromosome segments.

The superimposition of the heterotetrameric structures (predicted through MD simulation analyses) of each species over the known potato homotetramer (Jin et al., 2005) showed that the predicted heterotetramer structures are accurate and confirms identity of the true orthologs identified in the study. The rice heterotetrameric structure obtained during the present study appears to be better than the corresponding structure reported earlier (Dawar et al., 2013). The N-terminal region of the proteins for both subunits was hypervariable in all species examined. The N-terminal regions of SS in all dicots and in few monocots (T. urartu, wheat ortholog on 7BS, rice and sorghum) also contain a “QTCL” motif, which was earlier reported in potato and Arabidopsis (Linebarger et al., 2005; Hädrich et al., 2012). The absence of this motif in the remaining monocots suggested that this is present in plastidial AGPases only. It appears that after the divergence of monocots and dicots, this motif was retained by the plastidial copies of AGPases and was lost by the cytosolic copies since the motif is available in common ancestor (Amborella trichopoda) of monocots and dicots. Further study involving more dicot and monocot species may resolve whether or not this motif occurs more frequently in plastidial AGPase than in cytosolic AGPase. The important role of this motif in providing thermo-tolerance to AGPase was demonstrated in maize through the observation that QTCL motif was present in heat stable AGPase and not in heat labile AGPase (Hannah et al., 2001). The role of QTCL in providing heat stability was further substantiated through substitution of the “QTCL” motif for the native “STYL” motif through site directed mutagenesis, which provided heat stability (Linebarger et al., 2005).

It is known that the cysteine residue in the “QTCL” motif forms intermolecular SS disulphide bridge between the two small subunits (SS-SS dimerization), which facilitates the heterotetrameric structure of AGPase leading to its improved thermo-stability, albeit with reduced activity (Linebarger et al., 2005). The role of cysteine residue in SS-SS dimerization and in AGPase turnover was also reported in Arabidopsis (Hädrich et al., 2012).

In the present study, ligand-binding sites including the “LGGG” motif were identified in the ADP_Glucose_PP domain of the LS subunit in all of the species except T. urartu, the three wheat homoeologues, Ae. tauschii and chickpea. In all the three homoeologues of wheat and Ae. tauschii, the “LGGG” motif was present in the N-terminal region. The ATP, ADP, Mg⁺⁺, 3-phosphoglyceric acid (3-PGA), fructose-6-phosphate (F-6-P) and glucose-6-phosphate (G-6-P) ligands are known to bind to the above sites and control the AGPase activity through allosteric and redox modifications (Boehlein et al., 2010). The occurrence of these identified ligand binding aa residues corresponds the similar lignad binding aa residues of potato homotetramer (1YP4), which interacts with ATP (Jin et al., 2005). Further, from among the several aa residues involved in the ligand binding sites, arginine has been implicated in controlling the allosteric regulation of maize AGPase (Boehlein et al., 2010). In an earlier study of SS of several monocot and dicot species, five times more ligand binding sites were reported in the ADP_Glucose_PP domain than in the LbH_G1P_AT_C domain (Sarma et al., 2014). Since ADP_Glucose_PP domain is mainly involved in catalyzing the first key rate-limiting step of starch biosynthesis in plants, the larger number of ligand binding sites in this particular domain may be contributing to its catalytic activity and may be involved in differential regulation of its enzymatic activity. However, the exact role of these residues may be elucidated through wet-lab experiments.

In promoter region, several cis-regulatory elements responsible for response to light, abiotic stress and endosperm expression were identified. Light responsive elements may have an important role, since starch is synthesized in the chloroplast of leaves during the day and degraded during the night following circadian rhythm. Thus, a tight regulation of pathways of starch synthesis and degradation in response to light signals is required. The response regulators in the LS and the SS 5′ regulatory regions may be responsible for the high level of expression of these genes in developing endosperm (Burton et al., 2002; Ohdan et al., 2005). Similarly in wheat LS, first 400 bp at the 5′ region along with the first intron of 88 bps was found to be important for its expression in the endosperm (Thorneycroft et al., 2003). The functional significance of the several other identified response elements in the present study need to be validated using wet lab approaches.

The higher expression of both the LS and the SS in developing endosperm during anthesis and at grain filling stage and low expression in the vegetative tissue (roots and leaves) revealed its role during the starch biosynthesis in developing grain. Similar results through RT-PCR analyses were reported in wheat endosperms-embryo, roots and leaves (Burton et al., 2002) and in the developing rice endosperm (Ohdan et al., 2005). However, in Arabidopsis, a higher expression of the AGPase genes was shown in leaves at the bolting stage, suggesting leaves to be the main organs for the carbohydrate accumulation (Streb and Zeeman, 2012), and not the grain endosperms as in the case of cereals.

The available data also suggested temporal expression of the wheat homoeologues and also down-regulation of both subunits of AGPase under conditions of abiotic stress (e.g., heat and drought) in the spikes and developing endosperm. However, similar data for expression of AGPase subunits in endosperm of cereals under abiotic stress was not available for any of the species studied. An earlier study involving RT-PCR analyses in wheat cv. Butte86 provided evidence of an early and sharp decline in the expression of AGPase and other starch synthesizing enzymes at a higher day temperature (37°C) (Hurkman et al., 2003). We are currently carrying out a detailed study of the expression of LS and SS in the developing wheat endosperm using a set of heat tolerant and sensitive genotypes under varying levels of heat stress. In the long run, this will allow us to develop gene-based functional markers for use in breeding programs for development of heat tolerant wheat cultivars.

The present study identified the “true” orthologs of AGPase genes along with its structural and functional evolution among monocots and dicots. Overall, the comparative study revealed the structural conservation of LS and SS of AGPase, although some variation was observed in the length, number and phases of introns; N-terminal regions were relatively more variable. Among the two domains (ADP_Glucose_PP and LbH_G1P_AT_C) in each of the LS and SS, the domain ADP_Glucose_PP (with many more ligand binding sites carrying a conserved “LGGG” motif) perhaps plays an important role in the regulation of the AGPase activity. Some specific features of the LS and SS also suggested a species-specific evolution among monocots and dicots. Promoter and expression analyses revealed endosperm to be the main site of expression among cereals and leaves in Arabidopsis. Expression analysis also suggested that the expression of the genes for AGPases is regulated in a temporal manner during abiotic stresses.

RB and GS performed the data analyses; MD simulation analysis was performed by SG and PV. AM assisted RB and GS in conducting the data analyses and preparing the first draft. KG, AM, HB, and PG conceived, supervised, edited, and finalized the manuscript.

This work was supported by United States Agency for International Development, Feed the Future Innovation Lab-Climate Resilient Wheat (Grant number AID-OAA-A-13-00008).