Glu-370 in the Large Subunit Influences the Substrate Binding, Allosteric, and Heat Stability Properties of Potato ADP-glucose Pyrophosphorylase

ADP-glucose pyrophosphorylase (AGPase) is a key allosteric enzyme in plant starch biosynthesis. Plant AGPase is a heterotetrameric enzyme that consists of large (LS) and small subunits (SS), which are encoded by two different genes. In this study, we showed that the conversion of Glu to Gly at position 370 in the LS of AGPase alters the heterotetrameric stability along with the binding properties of substrate and effectors of the enzyme. Kinetic analyses revealed that the affinity of the LSE370GSSWT AGPase for glucose-1-phosphate is 3-fold less than for wild type (WT) AGPase. Additionally, the LSE370GSSWT AGPase requires 3-fold more 3-phosphogyceric acid to be activated. Finally, the LSE370GSSWT AGPase is less heat stable compared with the WT AGPase. Computational analysis of the mutant Gly-370 in the 3D modeled LS AGPase showed that this residue changes charge distribution of the surface and thus affect stability of the LS AGPase and overall heat stability of the heterotetrameric AGPase. In summary, our results show that LSE370 intricately modulate the heat stability and enzymatic activity of the AGPase. Abbreviations AGPase ADP-glucose pyrophosphorylase BSA bovine serum albumin DTT dithiothreitol G1P glucose-1-phosphate IPTG isopropyl-P-D-thiogalactopyranoside LS large subunit 3PGA 3-phosphoglyceric acid Pi inorganic phosphate SS small subunit TBS Tris-buffered saline WT wild type


Introduction
Starch is a staple in the diet of much of the world's population and is also widely used in different industries as a raw material [1]. The first committed step of starch biosynthesis in plants is catalyzed by ADP-glucose pyrophosphorylase (AGPase, EC 2.7.7.27).
AGPase, an allosteric enzyme, catalyzes the rate limiting reversible reaction and controls the carbon-flux in the α-glucan pathway by converting Glucose-1-phosphate (G1P) and ATP to ADP-glucose and pyrophosphate using Mg 2+ as the cofactor [1--3]. Regulation of almost all AGPases depends on the 3-phosphoglyceric acid to inorganic phosphate ratio (3PGA/P i ). While 3PGA functions as the main stimulator, P i inhibits the activity of enzyme [4][5][6]. Plant AGPases consist of pairs of small (SS, or α) and large (LS, or β) subunits, encoded by two distinct genes, thereby constituting a heterotetrameric structure (α 2 β 2 ) [7]. In potato tuber AGPase, the sequence identity between the LS and the SS is 53% suggesting a common ancestral gene [8,9]. The molecular mass of tetrameric AGPases ranges from 200 to 240 kDa depending on the tissue and plant species.
Specifically, molecular masses of LS and SS in potato tuber AGPase are 51 kDa and 50 kDa, respectively [10].
When expressed in bacteria, the SS of potato tuber AGPase can form a catalytically active homotetrameric enzyme with defective allosteric properties [11,12]. The LS was incapable of self-assembly and therefore was assumed to be essential for modulating the allosteric properties of the active heterotetrameric enzyme [13]. Evidence indicates that the LS may bind to the G1P and ATP substrates [14--16]. The binding of the LS to its substrates may allow the LS to interact cooperatively with the SS for binding substrates and effectors, which, in turn, influences net catalysis [14,15,17,18].
Manipulation of the AGPase activity and its heat stability gets much attention in increasing the starch yield of plants because it catalyzes the first committed step as well as the rate limiting step of starch biosynthesis [19--22]. Modulating the heat stability is especially important for cereal AGPases because elevated temperature is a major environmental factor that greatly reduces grain yield and seed weight [23,24]. Among AGPases, the potato tuber AGPase is much more heat stable than many others [10,23]. It has been shown that the lack of a QTCL motif, which is found on the SS of potato tuber AGPase, is one of the reasons for the heat labile profile of the maize AGPase [25].
Thermal stability of the enzyme has been shown to not only depend on the robustness of the LS-SS interactions but also on the binding of the effector molecules to the allosteric sites [23,26,27].
The structure of homotetrameric potato SS [28] and A.tumefaciens [29] AGPases were solved by X-ray crystallography. Neither the LS nor the heterotetrameric AGPase (α 2 β 2 ) structure have been solved yet. It is, therefore, homology-modeling approach was taken to understand structure-functions of heterotetrameric plant AGPases [30--34]. The modeling studies enable us to identify critical amino acid residues at the interface of the AGPase that are important for the LS-SS interaction and allosteric regulations [30--34].
Furthermore, based on the described heterotetrameric AGPase assembly model, LS-SS dimers are formed by side-by-side interactions and then LS-SS dimers interact in a head to tail orientation to form heterotetrameric AGPases [34]. Although there are several studies showed the importance of heterotetrameric assemblies in enzyme functions [13, 35--37], amino acids that contribute structural stability of heterotetrameric AGPase is not well defined. In this study, we demonstrated that an amino acid (LS E370 ), close to the LS interface, is important for maintaining the heat stability of the heterotetrameric structure of the AGPase. Computational analysis on the 3D model of LS AGPase revealed that LS E370G mutation disrupts hydrogen bond interaction network and electrostatic interactions in that region. In addition, biochemical characterization showed that LS E370G SS WT AGPase was less heat stable and showed altered affinities towards G1P and 3PGA. Our data demonstrate Glu-370 is critical for the structural stability of the LS subunit itself and the overall heterotetrameric AGPase, and influences the allosteric, substrate binding and heat stability properties of the AGPase.

Site-directed mutagenesis
Glutamate was replaced into glycine at the position of 370 in the potato LS AGPase cDNA using PCR with appropriate primers (Table S1) as previously described [38]. The PCR products were digested with DpnI to remove template plasmid DNA and transformed into E. coli DH5α. The presences of the mutations were confirmed by DNA sequencing (Macrogen Inc, Netherlands).

Protein expression
The wild type and LS E370G AGPases were expressed in E.coli AC70R1-504 (glgC -) cells as previously described [38]. E.coli AC70R1-504 cells were transformed with the plasmids pML7 and pML10 containing potato LS and SS cDNAs, respectively. Cells were grown in LB medium containing 50 µg/ml spectinomycin and 50 µg/ml kanamycin.
Once absorbance reached 1.0-1.2 at OD 600 , the cells were induced with 0.2 mM IPTG and 10 mg/L nalidixic acid at room temperature for 20 h.

Native PAGE analysis
The heterotetrameric assembly properties of the LS E370G mutant AGPase with wild type SS were investigated using 3-13% gradient native polyacrylamide gels followed by Western blot using potato anti-LS and anti-SS antibodies as previously described [27].

Heat stability
Cells, expressing mutant and wild type LSs in pML7 plasmids and wild type SS in pML10 plasmid, were harvested in buffer containing 50 mM HEPES at a pH of 7.5, 5 mM phosphate buffer (K 2 HPO 4 /KH 2 PO 4 ) 5 mM MgCl 2 and 10% glycerol. Aliquots of cell-free extracts of the wild type and mutant AGPases were placed in a water bath at 65°C for 5 and 10 min and then cooled on ice. All preparations were subsequently analyzed with native page followed by Western blot as described above using 0.5 µg, 1 µg and 2 µg of total protein.
AGPase activities in the crude cell extracts were also measured using forward direction assay, as described in the kinetic characterization part, in the presence of saturated substrates (5 mM) and the 3PGA concentration (10mM). Blank samples were complete reaction mixtures without enzyme. To get measurable activity a total of 10 µg of protein were used in the assay. Duplicate samples were left on ice and their activity was taken to be 100%.

His-tagged protein expression and purification
The recombinant polyhistidine-tagged wild type and mutant AGPases were expressed and purified as described herein. Plasmids pSH228 and pSH275 containing the AGPase SS and LS coding sequences [39], respectively, were sequentially transformed into E.coli EA345 cells lacking endogenous AGPase activity due to the null mutation in its structural gene glgC [18]. Three colonies were inoculated into 25 ml of Luria Broth (LB) medium containing 0.4% glucose, 100 µg/ml ampicillin, and 50 µg/ml kanamycin. The starter culture was grown overnight at 37°C and transferred to 1 L of LB medium. When the OD 600 of the culture was reached 1.0, the cells were induced with 0.1 mM IPTG, and the protein expression was induced for 18 h at room temperature. The cells were harvested and disrupted by sonication in 25 ml of binding buffer (25 mM HEPES-NaOH, pH8.0, 5% glycerol) supplemented with 0.5 mg/L lysozyme, EDTA-free protease inhibitor cocktail (Sigma), and 1 mM PMSF [40]. The soluble fraction was separated by centrifugation and passed through Macroprep DEAE resin (BIORAD). After extensive washing in binding buffer, AGPase was eluted with binding buffer containing 0.3 M NaCl and then loaded onto an immobilized metal affinity column (TALON metal affinity resin, Clontech). The column was washed with binding buffer containing 0.3 M NaCl and 5 mM imidazole. The protein was eluted with a 5-100 mM imidazole gradient in binding buffer containing 0.3 M NaCl. Fractions containing AGPase activity were combined and concentrated using 30 kDa cut-off Amicon ultra centrifugal filters (Millipore). The concentrated protein was divided into small aliquots and stored at -80°C until use.

Kinetic characterization
A non-radioactive endpoint assay was used to determine the amount of PP i produced by monitoring the decrease in the NADH concentration. The standard reaction mixture contained 50 mM HEPES-NaOH at a pH7.4, 15 mM MgCl 2 , 5.0 mM ATP, 5.0 mM G1P, and 3 mM DTT in a total volume of 100 µl. The assays were initiated by the addition of the enzyme, and the reactions were performed in a water bath at 37°C followed by termination by boiling for 2 min. AGPase activities were linear with respect to time and the amount of enzyme. The reactions were developed by adding 150 µl of coupling reagent to each tube, and the A 340 was determined using a 96-well microplate reader. The coupling reagent was prepared as described in [41]. with minor modifications. All The K m and A 0.5 values were determined in reaction mixtures in which one substrate or effector was added in varying amounts, and the other reaction components were saturated (5 mM). Kinetic constants were calculated by non-linear regression analysis using Prism software (Graph Pad). I 0.5 values for P i were determined in the presence of 1.0 and 2.5 mM 3PGA by adding increasing amounts of P i . Students' t-test was used.

Reducing and non-reducing conditions for AGPase
Cell-free extracts were prepared as described previously [26]. For reducing conditions, samples were heated at 95 o C for 5 min with Laemmli sample buffer containing 4 mM DTT. For non-reducing conditions, Laemmli sample buffer without any reducing agent was used. The samples were exposed to 10% SDS-polyacrylamide gels followed by Western blotting.

Minimizing the Energy of AGPase structure
3D structures of large subunit and heterotetrameric assembly of AGPase modeled by our group previously [34] were used for computational analysis. Using the NAMD (v. 2.6) [42] and VMD (v. 1.9.1) program packages structure of hetero tetrameric AGPase was solvated in a rectangular box with TIP3P water molecules having a minimum of 10Å distance from the closest atom of the protein to the boundary, and then counter ions were added to neutralize the system. Initially only the side chains were minimized for 10000 steps. Subsequently all atoms were minimized for 10000 steps without pressure control.
Then the system was heated up to 310K by increasing the temperature 10K and 10ps simulation was performed at each step.

Protein Stability Calculations
FoldX uses statistical energy terms, structural descriptors and an empirical potential obtained from addition of weighted physical energy terms (e.g. van der Waals interactions, solvation, hydrogen bonding, and electrostatics) to predict protein energetics and to provide quantitative estimation of intermolecular interactions promoting the protein stability depending on the availability of 3D structures [43]. Furthermore FoldX can calculate the interaction energy between subunits of a protein complex via the structure based energy function. Previously it has been shown that FoldX is able to produce accurate predictions about the stability change upon point mutations [43] and interaction of globular proteins [44][45][46]. After minimizing the structures by following the procedure mentioned above, we used 'Repair Object' module of FoldX which rearranges the side chain positions but not backbone of the structure, to obtain better optimized and minimized structure. In all of the calculations structures minimized by NAMD and then 'repaired' by FoldX were used. To validate the stability data obtained from FoldX, we used I-Mutant 3.0 (structure mode) (http://gpcr2.biocomp.unibo.it/cgi/predictors/I- [47] and POPMUSIC server (https://soft.dezyme.com/home).
To predict the effect of mutations on the Tm value of AGPase tetramer we used HOTMUSIC server. HOTMUSIC is very recently developed novel tool to predict thermal stability of proteins by predicting the change in Tm [48]. This tool uses standard and temperature dependent statistical potentials in combination with the neural network.
Totally 1600 mutations with experimentally measured ΔTm data were used to obtain the parameters of the model [48].

The large subunit E370G mutation impairs glycogen production in a bacterial complementation assay
Modeling studies followed by mutagenesis of the potato AGPase identified a hotspot mutant AGPase for the interaction of head to tail LS-SS dimers, where an Arg residue at position 88 was replaced by Ala in the LS (LS R88A ) [30,34]. In our previous study we employed an error prone-PCR random mutagenesis approach using the LS A88 AGPase cDNA as template to select second-site LS suppressor that could reverse the iodine staining deficiency of the E.coli glgCcontaining wild type SS AGPase cDNA [27].
Stained colonies were picked up and plasmids were isolated. To identify mutations each mutant were subjected to the sequencing as described in [27]. To see the effect of the second site revertants independently in AGPase function, plasmids containing mutant LS cDNAs were then purified and subjected to site-directed mutagenesis by PCR to convert the primary mutation Ala88 back to the corresponding wild type arginine residue.
The LS cDNAs containing only the secondary mutations were then co-expressed with wild type SS in E.coli glgCand their activity were assessed by iodine staining. This study yielded the following two group of mutants: one group of the LS mutants was able to complement bacterial AGPase expressed with wild type SS cDNA in E.coli glgC -, and the other group of the LS mutants was not able to complement bacterial AGPase when the cells were exposed to the iodine vapor in E.coli glgC -. The first group of the LS AGPase mutants was characterized by biochemical methods and found to have enhanced heterotetrameric assemblies and comparable kinetic properties to wild type AGPase [27].
In this study, five revertants from the second group (RM4, RM5, RM25, RM27 and RM29) of the LSs were studied. We wished to identify the role of the interface amino acids of the LS in AGPase function we therefore mapped the position of each mutation on the modeled heterotetrameric AGPase. The results indicated that there are five mutations (LS A113T , LS F123L , LS F324L , LS A113T , and LS E370E ) on the interface or close the interface of the LS AGPase. We then introduced these mutations on wild type LS cDNA by site directed mutagenesis. The LS cDNAs containing only these mutations were then co-expressed with the WT SS in E. coli glgC -, and their activity was assessed by iodine staining. The results indicated that the cells containing LS A113T , LS F123L , LS F324L , LS A113T , and WT SS stained as dark as cells that contained WT cDNAs of the LS and SS (data not shown) whereas the cells with LS E370G displayed no iodine staining phenotype compared with cells containing the WT LS and SS AGPase (Fig. S2). The LS E370G mutation is shown on the 3D homology model of potato AGPase [34], where the residue appears not to be a part of the interface residues but rather is very close to the subunit interface residues (Fig. S1). We then decided to further characterize the function of LS-Glu370 in AGPase function.

Glutamate at position 370 in the LS modulates stability of heterotetrameric AGPase
We determined the heterotetrameric assembly properties of the LS E370G AGPase with the wild type SS AGPase using native-PAGE and Western blot analysis. Before native-PAGE analysis, the integrity and expression levels of both subunits of LS E370G SS WT and LS WT SS WT AGPases were assessed by Western blot analyses. The LS and the SS proteins of the LS E370G SS WT and wild type AGPases in cell free extract were detected around their predicted molecular mass (50 kDa), and their expression levels were comparable with the wild type AGPase (data not shown). Then, variable amounts (0.5, 1, and 2 µg of total protein) of the mutant and wild type AGPase samples were subjected to the 3-13% gradient native-PAGE followed by Western blot using anti-LS potato AGPase and anti-SS AGPase antibodies (Fig. 1A). The first uppermost bands correspond to approximately  (Fig. 2A). The presence of the Gly instead of the Glu at the position of 370 in LS drastically changes surface charge distribution towards positive around that region (Fig.   2B). Then, we decided to replace Glu-370 with Arg in the LS AGPase to further shift surface towards more positive and investigate the effect of such mutation on the heterotetrameric stability of the AGPase. Native-PAGE analysis revealed that LS E370R SS WT AGPase has enhanced LS and SS dimer formation rather than heterotetrameric assembly (Fig. 3A). The quantity of the gel showed that the second band, indicating the presence of dimers, consists of 40%, whereas the amount of the dimer in wild type AGPase was approximately 20% (Fig. 3B). All these experiments let us hypothesize that a change in the surface charge of LS might affect the stability of the AGPase.
Computational analysis was performed to understand how the LS E370G and LS E370R mutations affect stability of the LS AGPase. Over the last decade various programs were developed to predict the protein stability upon point mutations by applying different approaches, namely EGAD [49], I-Mutant [50], CC/PBSA [51], FoldX [52], Rosetta [53], SDM [53], POPMUSIC [54] and others. FoldX produces quantitative estimation of protein stability upon mutation [43] and calculates the interaction energies between the subunits of a protein complex accurately [44][45][46]. We therefore used FoldX to calculate  (Table S2). These results imply that hydrogen-bond network and electrostatic interactions around the 370 th residue are disturbed in both mutants causeing less stable LS AGPase. To further validate these results, we used I-Mutant 3.0 (structure mode) [47] and POPMUSIC server (Table S3) where they consistently show that these mutations have destabilizing effect on the LS. All these computational results explain how 370 th residue contributes the stability of the LS AGPase.
Then we would like to calculate the interaction energies between subunits of AGPase in wildtype and mutants to investigate the effect of mutations on the interaction of subunits.
Our results (Table 1) show that LS E370G mutation does not cause overall interaction energy change between the subunits of AGPase (Table S4). On other hand LS E370R mutation slightly decreases the interaction energy and indicates that mutation favors the formation of the heterotetrameric AGPase (Table S5).
According to the heterotetrameric AGPase model [30, 32--34], such interactions are only possible between LS and SS in a head to tail orientation (Fig. 2). On the other hand, mutant LS R370 , may interact with SS E124 because of the opposite charged amino acids on these subunits and this may elevate their affinities to each other and result more LS-SS dimer formation. Our computational analysis also indicated that this LS R370 mutation favors head to tail interaction between LS R370 and the SS (Table S5). This might explain why we observed more dimer formation between LS R370 and the SS ( Fig. 2A).

Heterotetramers of LS E370G SS WT AGPase are heat labile
We used a tag-less expression system described in [13] to investigate the effect of temperature on the enzyme stability and activity. Cell free extracts of the wild type and mutant AGPases were incubated at 65°C for 5 and 10 minutes. Then, samples were centrifuged and subjected to the 3-13% native-PAGE followed by Western blot. Fig. 4 clearly showed that the heterotetrameric form of the wild type AGPase was preserved while the heterotetrameric form of the LS E370G SS WT AGPase totally disappeared at the end of the heat treatment. We also tested the heat stability of LS E370R SS WT AGPase under the same conditions. The heterotetrameric assembly of the mutant AGPase disappeared at the end of heat treatment, which was similar to the LS E370G SS WT AGPase (Fig. S3). Then, the samples were used to measure the remaining activities at saturated concentrations of substrates (5mM) and 3PGA (20mM) with forward direction assay. There were no detectable remaining activities of LS E370G SS WT and LS E370R SS WT (data not shown).
To explore the possibility that the LS E370G SS WT AGPase has altered redox status mediated by Cys-12 it was subjected to non-reducing SDS-PAGE followed by Western blot. The results indicated that both WT and LS E370G SS WT AGPases had all LS subunits as monomers and had comparable disulfide bond between the SS (Fig. S4).
Finally we investigated the heat stability of our mutant AGPases by using the recently developed HOTMUSIC tool which predicts melting temperature changes as a result of point mutations [6]. HOTMUSIC analysis indicated that both mutations cause protein to adopt lower Tm which means decrease in the heat stability (Table 1).

Kinetic and allosteric characterization
Polyhistidine tagged heterotetrameric AGPases were purified using DEAE weak anion exchange chromatography followed by TALON metal affinity chromatography with a final purity of 60-70%. The purified recombinant wild type and mutant AGPases were assayed in the forward reaction direction to determine the effects of mutations on the enzyme activity. The kinetic and allosteric parameters of the wild type and mutant AGPases are shown in the Table 2. Kinetic properties of wild type AGPase were comparable with previously published kinetic values of the recombinant heterotetrameric AGPase [13].
Analysis of the substrate binding properties of the AGPases showed that there were no significant differences in the ATP K m values between the heterotetrameric wild type and LS E370G SS WT AGPases (Fig. S5A). In contrast, the affinity of the enzymes towards the other substrate G1P showed more variation. The wild type AGPase has a K m for G1P of 0.044 mM. However, the G1P affinity of LS E370G SS WT AGPase was approximately 3-fold lower than of the wild type, with a K m of 0.13 mM (Fig. S5B).
Allosteric regulatory properties varied considerably between the wild type and mutant AGPases. The A 0.5 values for 3PGA with wild type and LS E370G SS WT AGPases are 0.28 and 0.83 mM, respectively (Fig. S6A) (Fig. S6B,C).
Our results indicated that the structural change due to the E370G mutation in the LS affects the kinetic and allosteric properties of the heterotetrameric AGPase. Previous studies have shown the importance of the LS region close to the subunit interface residues for the allosteric regulation of AGPase [27,33,41,55,56]. These data are in agreement with the previous studies indicating that both the LS and SS are equally important in enzyme catalysis and regulation [55--57].
To determine the degree of the conservation of the Glu residue at position 370 in the LS, multiple alignments from the mature LS sequences of various plants were performed using CLUSTAL (Fig. S7). Conservation analysis revealed that the glutamate residue at position 370 in the LS is not highly conserved among different plant species. However, the neighboring amino acids are conserved, and valine and isoleucine residues are mostly observed at this position. In fact the maize counterpart of this residue was previously identified with a phylogenetic analysis as a candidate residue (positively selected type II site), which might have an important effect for the function of the protein [33,57].
Maize LS V416I AGPase leads to increased heat stability without affecting the kinetic constants, whereas maize LS V416E AGPase significantly impairs 3PGA binding but does not change the heat stability. Those together with our results, which reveal the contribution of Glu370 for the heat stability, substrate and effector binding properties, specify the importance of this residue for the structure-function relationship of AGPase.
Additionally this study showed that Glu370 affects oligomeric status of the AGPase by enhancing the formation of the LS-SS dimer. In 3D homology modeled potato heterotetrameric AGPase we observed that replacement of Glu into Gly drastically shifted surface charge distribution towards positive in the LS and this region is close proximity with SS E124 with distance 5Å (Fig. 2A). It is most likely that SS E124 strongly interacts with this positively charged region and dimer formation between the LS E370R and WT SS yet to be shown by crystal structure study. Further kinetic characterization showed that the LS E370G SS WT AGPase has reduced affinity in both substrate (G1P) and activator (3PGA). Therefore, the expression of the LS E370G AGPase with SS WT AGPase in bacterial system fails complement glgC in E.coli (Fig S2).
In conclusion, our results show that LS E370 intricately modulate the heterotetrameric assembly, heat stability and enzymatic activity of the AGPase.   Table 2: Kinetic and allosteric parameters of the wild type and LS E370G SS WT AGPase. Kinetic and regulatory properties were determined in the ADP-glucose synthesis direction. All values are in mM. K m and A 0.5 for substrates and 3PGA, respectively, correspond to the concentration of these molecules required for the enzyme activity to attain 50% of maximal activity. I 0.5 is the amount of P i required to inhibit the enzyme activity by 50% of maximal activity in the presence of either 1 mM or 2.5 mM 3PGA. Depicted are the mean values with standard error of at least two independent experiments. showing the ratio of monomer, dimer and tetramer to total band intensity, respectively.

Figure Legends
Average values with standard error for >4 independent experiments were compared with Student's t test with regard to the wild type. ***P <0.0001; *P<0.05.