Phylogenetic analysis of ADP-glucose pyrophosphorylase subunits reveals a role of subunit interfaces in the allosteric properties of the enzyme

ADP-glucose pyrophosphorylase (AGPase) catalyzes a rate-limiting step in glycogen and starch synthesis in bacteria and plants respectively. Plant AGPase consists of two large and two small subunits that were derived by gene duplication. AGPase large subunits have functionally diverged, leading to different kinetic and allosteric properties. Amino acid changes that could account for these differences were identified previously by evolutionary analysis. In this study, these large subunit residues were mapped onto a modeled structure of the maize endosperm enzyme. Surprisingly, of 29 amino acids identified via evolutionary considerations, 17 were located at subunit interfaces. Fourteen of the 29 amino acids were mutagenized in the maize endosperm large subunit (SH2) and resulting variants were expressed in Escherichia coli ( E. coli ) with the maize endosperm small subunit (BT2). Comparisons of the amount of glycogen produced in E. coli , and the kinetic and allosteric properties of the variants to wildtype SH2/BT2 indicate that 11 variants differ from wildtype in enzyme properties or in vivo glycogen level. More interestingly, 6 of 9 residues located at subunit interfaces exhibit altered allosteric properties. These results indicate that the interfaces between the large and the small subunit are important for the allosteric properties of AGPase and changes at these interfaces contribute to AGPase functional specialization. Our results also demonstrate that evolutionary analysis can greatly facilitate enzyme structure-function analyses. and type-II sites during the large subunit evolution et al., 2008). We then replaced the SH2 residues with amino acids of a group different from the SH2 family. Several amino acid sites important for the kinetic and allosteric properties and heat stability of AGPase were identified. Our results indicate that the subunit interfaces between the large and the small subunit are important for the allosteric properties of AGPase. They also indicate that amino acid changes at subunit interfaces have been important for AGPase specialization in terms of allosteric properties.


Introduction
subunit, the identity of the amino acid sites in the large subunit that account for these kinetic and allosteric differences was not pursued. Georgelis et al. (2008) presented supporting evidence for AGPase large subunit specialization by identifying positively selected amino acid sites in the phylogenetic branches following gene duplication events. We also identified amino acid residues that were conserved in one large subunit group but not conserved in another large subunit group (type-I functional divergence) (Gu 1999) and amino acid residues that are conserved within large subunit groups but variable among large subunit groups (type-II functional divergence) (Gu 2006). Positively selected, type-I and type-II sites could have contributed to specialization of the different large subunit groups. Indeed, positively selected and type-II sites in several proteins have been proven via site-directed mutagenesis (Bishop, 2005;Norrgard et al., 2006;Cavatorta et al., 2008;Courville et al., 2008) to be important for protein function and functional specialization. Additionally, several positively selected, type-I and type-II amino acid sites in the large AGPase subunit identified in our previous evolutionary analysis ( Georgelis et al. 2008) have been implicated in the kinetic and allosteric properties and heat stability of AGPase. The role of these sites was demonstrated by site-directed mutagenesis experiments of large subunits from Arabidopsis, maize endosperm and potato tuber (Ballicora et al., 1998;Kavakli et al., 2001a;Ballicora et al., 2005;Jin et al., 2005;Linebarger et al., 2005;Ventriglia et al., 2008). These analyses indicate that the rest of the amino acid sites identified as positive, type-I and type-II sites in our previous evolutionary analysis (Georgelis et al., 2008) represent promising candidate targets for mutagenesis.
To identify large subunit amino acids that are possibly important in controlling enzyme properties and that may have contributed to large subunit specialization, we conducted site-directed mutagenesis of the maize endosperm large subunit encoded by Shrunken-2 (Sh2). We specifically identified amino acids of SH2 that correspond to amino acid sites that were detected as positive, type-I and type-II sites during the large subunit evolution (Georgelis et al., 2008). We then replaced the SH2 residues with amino acids of a group different from the SH2 family. Several amino acid sites important for the kinetic and allosteric properties and heat stability of AGPase were identified. Our results indicate that the subunit interfaces between the large and the small subunit are important for the allosteric properties of AGPase. They also indicate that amino acid changes at subunit interfaces have been important for AGPase specialization in terms of allosteric properties.
These experiments also support the idea that the majority of positively selected sites as detected by codon substitution models (Nielsen and Yang, 1998;Yang et al., 2000) and type-II sites are not false positives. Site-directed mutagenesis of such sites can greatly facilitate enzyme structure/function analyses.

Previous phylogenetic analysis and structural mapping of type-II and positively selected sites
AGPase large subunits can be placed into five groups depending on sequence similarity and tissues of expression (Supplemental Figure 1) (Georgelis et al., 2007;2008). Group 4 includes only two sequences whose role has not been studied. Accordingly, we had restricted further evolutionary analysis to the remaining four groups. We identified 21 type-II sites (Supplemental Table 1 Georgelis et al., 2008). These amino acid sites potentially contributed to the functional divergence among AGPase large subunits. Type-II sites 96 and 106 were shown to play an important role in enzyme catalysis (Ballicora et al., 2005), while site 506 has been implicated in the allosteric properties of the potato tuber AGPase (Ballicora et al., 1998).
We also detected 18 amino acid sites upon which potential positive selection may have taken place in the tree branches following the gene duplications that led to the different large subunit groups (Supplemental Table 1, Figure 1) (Georgelis et al., 2008). These sites could also be important in large subunit specialization since functional diversification among different large subunits could have been beneficial for the fitness of the plant. Positively selected sites 104, 230, 441 and 445 are implicated in the allosteric properties of AGPase (Kavakli et al., 2001a;Ballicora et al., 2005;Jin et al., 2005). Finally, we identified 91 type-I sites. These sites are apparently important for AGPase function in one group but not in another group and they could contribute towards subfunctionalization or specialization or both among large subunit groups. The usefulness, however, of type-I sites in detecting functional divergence has been disputed (Philippe et al., 2003) since type-I divergence between orthologous and paralogous groups is indistinguishable in some instances (Gribaldo et al., 2003). It would be expected that paralogues should exhibit more functional divergence compared to that revealed by orthologues. Therefore, there may be a considerable number of false-positive sites of type-I divergence that do not necessarily represent functional divergence.
The fact that several type-II and positively selected sites have already been shown to be important for the kinetic and allosteric properties of AGPase strongly suggests that the remaining type-II and positively selected sites may also be important for enzyme function. To gain insight into the potential role of these sites, we placed them on the modeled structure of the SH2 protein. The type-I sites were excluded from this initial analysis since a potential inclusion of a high number of false-positives could confound the results.
Although the only plant crystal structure available is a potato tuber small subunit homotetramer (Jin et al., 2005), the high degree of identity (40-45%) and similarity (55-65%) between the small and the large subunits strongly suggests that the structure of the physiologically relevant AGPase heterotetramer will be very similar or identical to the resolved homotetramer structure. Superimposition of SH2 and BT2 on the solved potato structure agrees with this conjecture (Figure 2A, 2B). Additionally, the potato tuber AGPase heterotetramer was modeled after the homotetrameric structure and a molecular dynamics study was conducted to determine the most thermodynamically favorable interactions between the large and the small subunit ( Figure 1) (Tuncel et al., 2008).
Superimposition of the potato tuber large subunit on SH2 indicates that the two structures are virtually identical (Figure 3). This enables us to use the potato tuber large subunit modeled structure to determine the areas of SH2 that interact with BT2. According to the modeled potato tuber heterotetramer (Tuncel et al., 2008), a SH2 molecule makes direct contacts with one molecule of maize endosperm small subunit (BT2) through its Cterminal domain (tail-to-tail interaction) and to the second BT2 protein through its Nterminal catalytic domain (head-to-head interaction) as shown in Figure 2A and 2B. We observed that 17 out of 29 amino acid sites (type-II and positively selected) were at or near the subunit interfaces ( Figure 4A, 4B). The areas of SH2 that participate in subunit interactions do not constitute more than 30% of the SH2 monomer structure. Hence, almost 60% of the residues that were selected through evolutionary analysis were located in less than 30% of the SH2 monomer. This preferential localization of the residues at subunit interfaces points to the possibility that subunit interfaces are important for the functional specialization of the AGPase large subunit.

Site-directed mutagenesis
To determine the role, if any, of type-II and positively selected amino acid sites, and particularly the ones found at the subunit interfaces, in AGPase function and gain insight into their potential roles in large subunit specialization, we performed site-directed mutagenesis on 12 sites in SH2. We mutagenized 7 SH2 sites (4 type-II, 1 positively selected, and 2 both type -II and positively selected) located at the subunit interfaces and 5 sites (3 type-II, 2 positively selected) located in the rest of the SH2 monomer. In all cases, the residue of SH2, which belongs to group 3b (Supplemental Figure 1), was changed to a residue found in other groups. To gain more information about the subunit interfaces, we scanned type-I sites for the ones that are located in the subunit interfaces.
We selected type-I site 149 as a target because SH2-containing group 3b, contains a His while other groups contain the physicochemically different Ala or Ser. His was changed to a Ser. We also selected type-I site 361, which is also located in subunit interfaces. This site is invariant in Group 2 but variable in Group 3b. Group 3b can be subdivided in two subgroups, one that contains only endosperm specific large subunits including SH2 and one that includes mostly embryo large subunits. SH2 along with the other members of the former subgroup contain a Thr at site 361 while the latter subgroup contains a Cys. The Thr of SH2 was changed to a Cys, which has different physicochemical properties.

Glycogen production
Wildtype SH2 and the 14 SH2 variants created by site-directed mutagenesis were expressed with wildtype BT2 in E. coli strain AC70R1-504 (Materials and Methods) and resulting glycogen of each genotype was quantified. The majority (10/14) of the SH2 variants resulted in altered amounts of glycogen ( Figure 5). This strongly suggests that the majority of the mutations introduced in SH2 were not neutral, at least when expressed in E. coli, despite the fact that the substituted amino acid residues are present in other large subunit groups.
Expression of the Sh2 mutants without the presence of the BT2 protein in E.coli resulted in no glycogen production (data not shown) indicating that potential SH2 homotetramers are inactive. It is also known that wildtype SH2 and BH2 homotetramers do not produce any glycogen in E.coli (Georgelis and Hannah 2008). Hence, the changes in glycogen production of the Sh2 mutants are most likely due to altered properties of the SH2/BT2 heterotetramer.

Characterization of kinetic and allosteric properties of SH2 variants
Glycogen levels suggested that some of the mutants alter AGPase function at the protein/enzyme level. Therefore, the SH2 variants and wildtype SH2 were expressed in (Materials and Methods). The affinity of the SH2/BT2 complexes for the allosteric activator 3-PGA (K a ) was determined in the forward direction (G-1-P +ATP ADPglucose +PPi). Interestingly, 7 out of the 14 SH2 variants had a higher Ka compared to wildtype SH2/BT2 (Table 1). The overwhelming majority of them (6/7) had an amino acid change in a site at the subunit interfaces. Two changes were in the head-to-head interaction areas (H149S, S163F) while four (T361C, D368S, P372A, C382F) were in the tail-to-tail interaction areas ( Figure 4). One change (Q213H) was in the N-terminal catalytic domain far from the subunit interfaces (Supplemental Figure 2). The affinity for the deactivator Pi (K i ) was also determined in the presence of 15mM 3-PGA by use of Dixon plots. Higher K a s in the variants described above were accompanied by a lower K i (Table 1). It has been proposed that 3-PGA and Pi are competing for binding to AGPase and they may even bind to the same site (Boehlein et al., 2008). Therefore, the lower affinity for 3-PGA may maximize the efficiency of Pi inhibition.
The K m s for G-1-P and ATP were determined for all variants at 15mM 3-PGA. Except for a 4-fold lower affinity of BT2/S163F for ATP all the other variants exhibited indistinguishable and wildtype K m values. Similarly, most K cat values were similar to wildtype BT2/SH2 except for BT2/C424V (~150%), BT2/V227R (~40%), and BT2/D368S (~60%). This indicates that the allosteric changes in the variants affected the affinity for effectors to a much greater extent than the effect on the mechanism of activation.
The purified complex of SH2/BT2 is present in three forms, a heterotetramer, a heterodimer and monomers of SH2 and BT2 (Boehlein et al., 2005). The SH2 and BT2 monomers have negligible activity at the levels of 3-PGA, G-1-P and ATP used in the present study (Burger et al., 2003). Additionally, Greene and Hannah (1998a) showed that the SH2/BT2 heterodimer is inactive. Therefore, all extant evidence strongly suggests that the AGPase activity in this study comes from the SH2/BT2 heterotetramer.

Heat stability
The structures of the large and the small subunit are almost identical. It has been shown that the loop connecting the C-terminal described above (from aa 362 to 399), the heat stability of the resulting variants was determined. The variants and wildtype BT2/SH2 were heated for various amounts of time at 39 o C and remaining activity was determined by assaying in the forward direction using 20mM 3-PGA and saturating amounts of substrates. With the exception of BT2/S163F that showed a 3-fold increase in heat stability all the other variants were similar to wildtype BT2/SH2 ( Figure 6). These results indicate that the majority of the mutagenized sites at the subunit interfaces have a specific role only on the allosteric properties of AGPase.

Correlation of kinetic and heat stability data with glycogen production
In general, the amount of glycogen produced by the variants in E. coli was consistent with the kinetic data. Six of seven allosteric variants produced lowered amounts of glycogen compared to wildtype. In the case of the exceptional BT2/S163F, the K a was increased and hence decreased glycogen production might have been expected. This was not observed. However, the higher heat stability of BT2/S163F may counteract the increased K a . As a result BT2/S163F produces wildtype amounts of glycogen.
BT2/M172T, BT2/C114A, and BT2/E438Q had wildltype kinetic properties and heat stability. Not surprisingly they produced wildtype amounts of glycogen. BT2/V227R and BT2/C424V had lower and higher K cat and glycogen production compared to wildtype respectively. BT2/V502T and BT2/A508S exhibited wildtype kinetic properties and heat stability. However, glycogen production was markedly reduced in these mutants. Perhaps these variants have reduced solubility and/or increased susceptibility to proteases in E. coli or perhaps transcription or translation is reduced in these mutants. These possibilities would predict reduced amounts of SH2 and/or BT2 protein in E. coli extracts. To investigate these possibilities, a Western blot analysis was conducted on total and soluble protein extracts from E. coli expressing wildtype BT2/SH2, BT2/V502T and BT2/A508S. The amount of SH2 and BT2 in both total and soluble protein extracts is indistinguishable from wildtype in these two variants (Figure 7). Therefore, the possible explanations discussed above for the reduced glycogen produced by BT2/V502T and BT2/A508S should be excluded. The underlying reason for reduced glycogen production in these variants remains unresolved.
Interestingly, none of the SH2 variants gave a null phenotype in E.coli. The exact ratio of 3-PGA/Pi in our E.coli system is not known. Some tentative amounts for 3-PGA and Pi are 0.5-0.75mM and 5-10mM respectively, depending on the type of cells and the growth conditions (Moses and Sharp, 1972;Ugurbil et al, 1978;Ishii et al., 2007). Since the ratio of 3-PGA/Pi is low, it maybe expected that our AGPase variants have very low to almost no activity in E.coli. However, maize endosperm AGPase is known to have some low activity even in the absense of 3-PGA (5-10% compared to the presence of  (Greene et al., 1996a;1996b;Greene and Hannah, 1998b;Laughlin et al., 1998;Kavakli et al., 2001a;2001b;Ballicora et al., 2007;Georgelis and Hannah, 2008;Hwang et al., 2008). Additionally, random mutagenesis of these variants has led to the identification of intragenic suppressors of initial mutants and resulted in the identification of additional sites that are important for allosteric properties of AGPase (Greene et al., 1998;Kim et al., 2007). Site-directed mutagenesis has also greatly facilitated structure-function analysis of AGPase. The resolved structure of the potato tuber small subunit homotetramer (Jin et al., 2005) along with structure modeling has been used to identify candidate sites for mutagenesis (Bejar et al., 2006;Hwang et al., 2006;2007). Additionally, evolutionary comparison of AGPase with other pyrophosphorylases has identified conserved amino acid sites that have undergone site-directed mutagenesis (Ballicora et al., 1998;Fu et al., 1998;Frueauf et al., 2001;2003;Ballicora et al., 2005).
Herein, we have identified positively selected sites in the large subunit of AGPase and amino acid sites that are conserved within large subunit groups but variable between groups. We argue that these sites may have been important in functional diversification of the large subunits and subsequently AGPases in plants. After  sites on the modeled structure of SH2, we observed that the majority of them was localized at tail-to-tail or head-to-head subunit interfaces. This strongly suggests that the subunit interfaces have been important in AGPase isoform specialization. To gain insight into the potential role of the amino acid sites, especially the ones located at subunit interfaces, we mutagenized 12 of these sites in SH2. We also mutagenized 2 type-I sites that were located at subunit interfaces for reasons discussed above. The original amino acid in SH2 was replaced with an amino acid found in other large subunit groups. Surprisingly, 6 of 9 changes in SH2 that are located at subunit interfaces had altered allosteric properties. Indeed, these variants showed a 2.4 to 10-fold increase in 3-PGA K a and similar decreases in the Pi K i . One substitution, S163F, also resulted in 4-fold increase in the ATP K m while change D368S resulted in a slight decrease in K cat (~60% wildtype). Overall, K cat s were not appreciably altered in the allosteric variants. Therefore, most of the changes in these variants were quite specific for K a and K i . This indicates that the changes introduced by mutagenesis primarily affected the affinity for the allosteric effectors rather than the mechanism of activation after effector binding. These results clearly indicate that that subunit interfaces (both head-to-head and tail-to-tail) are Since 9 out of 15 SH2 variants were at or near subunit interfaces, we asked whether these alterations affected AGPase heat stability. Alteration in heat stability has been noted with at least one mutant (Greene and Hannah, 1998b)  helix of the small subunit participates in tail-to-tail interactions with the large subunit and it is important for AGPase heat stability (Boehlein et al., 2009). However, only one variant, BT2/S163F showed altered heat stability. Either the specific changes were not sufficient to change heat stability or the specific sites are not important for heat stability in BT2/SH2. However, the present study pinpoints specific amino acids and areas in the large subunit that account for specialization in allosteric properties. This study also indicates that the specialization has taken place mainly in terms of the affinity for 3-PGA and Pi. Indeed, the mechanism of activation seems unaffected since the K cat s in the presence of activators or inhibitors of the vast majority of these mutants are very similar to wildtype AGPase.
Our mutagenesis also resulted in a change in site 424 that increased the specific activity of AGPase. This mutation may lead to agronomic gain when expressed in plants.
The allosteric sites detected can also be targets of mutagenesis to obtain variants with low 3-PGA K a and high Pi K i that would also be of agronomic interest. Variants with low K a and high K i can also be achieved by isolation of intragenic suppressors of our allosteric variants through random mutagenesis.
Additionally, the results show that evolutionary analysis can substantially benefit structure-function studies. The present study is among the few examples where a large collection of positively selected and type-II sites initially detected by phylogenetic analysis were verified biochemically. The fact that the majority of the changes in these sites are not neutral should be encouraging to biochemists to use more evolutionary analysis to analyze enzyme structure/function analyses. In this study, evolutionary analysis led to the selection of several amino acid sites that are important for enzyme function. These sites would not have been selected based solely on the structure of AGPase. The majority of the type-II and positively selected sites alter amounts of glycogen synthesized and/or enzymatic properties. Type-I sites can also be useful in detecting sites important for function since changes in type-I sites 149 and 361 altered enzyme properties. However, a large-scale study should be conducted on type-I sites to determine the false-positive rate.
Overall, the study enhanced our understanding of the evolution and structure/function relationships in AGPase, set the stage for protein engineering that may lead to increased starch yield in crops and provided support for use of evolutionary analysis to understand protein function. SH2 (Accession #: P55241), and potato tuber large subunit (Accession #: CAA43490) protein sequence alignment were obtained using the MEGA software (Kumar et al., 2004) with BLOSUM matrix followed by manual inspection. The large subunit amino acid numbers used throughout the report correspond to SH2.

Structure modeling
BT2, SH2, and the potato tuber large subunit monomer structures were modeled after the potato small subunit in the recently published 3D structure of the potato tuber homotetrameric AGPase (RCSB Protein Data Bank #: 1YP2c). SWISS MODEL was used for performing homology modeling (Schwede et al., 2003;Arnold et al., 2006). The potato tuber large subunit and SH2 were modeled from amino acid # 34 and # 94 to the end respectively due to poor alignment of the N termini. WHATCHECK (Vriend, 1990) and VERIFY3D (Luthy et al., 1992) were used to structurally evaluate the models. The corresponding WHATCHECK values (z-values for Ramachandran plot, backbone conformation, chi-1/chi-2 angle correlation, bond lengths, and bond angles) were within acceptable range. The high quality of the models was verified by the positive values assigned by VERIFY3D throughout all the structures. Visualization and superimposition of models and structures was done with Chimera (Pettersen et al., 2004).

Site-directed mutagenesis
The PCR reactions for site-directed mutagenesis were done with high fidelity Vent polymerase (New England Biolabs) by using pMONcSh2 as a template.

Glycogen quantitation
Glycogen quantitation was performed by phenol reaction (Hanson and Phillips, 1981) as described by Georgelis and Hannah (2008).

Enzyme expression and purification
The SH2 wildtype and variants were expressed along with wildtype BT2 in bacterial cells AC70R1-504 cells (Iglesias et al., 1993). Briefly, AC70R1-504 cells containing wildtype Bt2 were transformed with plasmids containing each of the various variant Sh2 genes.
Transformation mixes were diluted and grown at 37°C in Luria-Bertani medium containing 75 µg/mL spectinomycin and 50 µg/mL kanamycin. Glycogen amounts were determined in cells grown in the presence 2%w/v glucose until OD600 was around 2.0.
For enzyme purification, cells were induced for 4 hours with 0.2 mM isopropyl-β-Dthiogalactoside and 0.02 mg/mL nalidixic acid after cells had reached an OD600 of 0.6.
The resulting enzymes were purified as described by Georgelis and Hannah (2008 brief, cell paste was harvested by centrifuging at 5000xg for 10 min and the cells were lysed by using French press. AGPase was purified by protamine sulfate and ammonium sulfate fractionation, followed by anion-exchange and affinity (hydroxyapatite) chromatography. AGPase was concentrated and desalted and exchanged into 50 mM HEPES, pH 7.4, 5 mM MgCl 2 , 0.5 mM EDTA before use. Bovine serum albumin (0.5 mg/mL) was added to confer stability to AGPase. Purified AGPase was stored at -80°C.

Enzyme kinetics
The forward direction of the reaction was used (G-1-P + ATP → ADP-glucose + PPi) for estimating K cat , K m for ATP and G-1-P, and affinities for 3-PGA (Ka) and Pi (Ki).
Specifically, 0.04-0.12 μ g of purified enzyme was assayed, at 37 °C for 10 min, in the presence of 50 mM HEPES pH 7.4, 15 mM MgCl 2 , 2.5 mM ATP, and 2.0 mM G-1-P and varying amounts of 3-PGA to determine Ka. Ki was determined in the presence of 15mM 3-PGA. Kms for G-1-P and ATP were estimating by varying the amount of G-1-P and ATP respectively in the presence of 15mM 3-PGA. The reaction was terminated by boiling for 2 min and PPi was coupled to a reduction in NADH concentration using a

Heat stability
Heat stability of the SH2 wildtype and variants expressed with wildtype BT2 was determined as described by Georgelis and Hannah (2008). The enzyme was heated at 39 o C.

Western detection of SH2 and BT2
A Western blot detection of both BT2 and SH2 in BT2/SH2, BT2/V502T and BT2/A508S variants was performed as described by Georgelis and Hannah (2008). A polyclonal antibody against SH2 (1:2000 (v/v)) was used in addition to a polyclonal antibody against BT2 to detect both SH2 and BT2.