This Article Appeared in a Journal Published by Elsevier. the Attached

Home , Mutase

This article appeared in a journal published by Elsevier. The attached copy is furnished to the author for internal non-commercial research and education use, including for instruction at the authors institution and sharing with colleagues. Other uses, including reproduction and distribution, or selling or licensing copies, or posting to personal, institutional or third party websites are prohibited. In most cases authors are permitted to post their version of the article (e.g. in Word or Tex form) to their personal website or institutional repository. Authors requiring further information regarding Elsevier’s archiving and manuscript policies are encouraged to visit: http://www.elsevier.com/copyright Author's personal copy

http://dx.doi.org/10.1016/j.jmb.2012.08.013 J. Mol. Biol. (2012) 423, 831–846

Contents lists available at www.sciencedirect.com Journal of Molecular Biology journal homepage: http://ees.elsevier.com.jmb

Conservation of Functionally Important Global Motions in an Enzyme Superfamily across Varying Quaternary Structures

Emily K. Luebbering 1, Jacob Mick 1, Ranjan K. Singh 2, John J. Tanner 1, 2, Ritcha Mehra-Chaudhary 3 and Lesa J. Beamer 1, 2⁎

1Department of Biochemistry, University of Missouri, Columbia, MO 65211, USA 2Department of Chemistry, University of Missouri, Columbia, MO 65211, USA 3Structural Biology Core, University of Missouri, Columbia, MO 65211, USA

Received 3 May 2012; The α‐D‐phosphohexomutase superfamily comprises enzymes involved in received in revised form carbohydrate metabolism that are found in all kingdoms of life. Recent 16 August 2012; biophysical studies have shown for the first time that several of these accepted 17 August 2012 enzymes exist as dimers in solution, prompting an examination of the Available online oligomeric state of all proteins of known structure in the superfamily (11 27 August 2012 different proteins; 31 crystal structures) via computational and experimental analyses. We find that these proteins range in quaternary structure from Edited by A. Panchenko monomers to tetramers, with 6 of the 11 known structures being likely oligomers. The oligomeric state of these proteins not only is associated in Keywords: some cases with enzyme subgroup (i.e., substrate specificity) but also phosphohexomutase; appears to depend on domain of life, with the two archaeal proteins existing normal mode analysis; as higher‐order oligomers. Within the oligomers, three distinct interfaces oligomeric interface; are observed, one of which is found in both archaeal and bacterial proteins. global fluctuations; Normal mode analysis shows that the topological arrangement of the conformational flexibility oligomers permits domain 4 of each protomer to move independently as required for catalysis. Our analysis suggests that the advantages associated with protein flexibility in this enzyme family are of sufficient importance to be maintained during the evolution of multiple independent oligomers. This study is one of the first showing that global motions may be conserved not only within protein families but also across members of a superfamily with varying oligomeric structures. © 2012 Elsevier Ltd. All rights reserved.

*Corresponding author. Department of Biochemistry, University of Missouri, Columbia, MO 65211, USA. E-mail address: [email protected]. Abbreviations used: BaPNGM, Bacillus anthracis PNGM; CaPAGM, Candida albicans PAGM; CSS, complexation signiﬁcance score; DLS, dynamic light scattering; FtPNGM, Francisella tularensis PNGM; NMA, normal mode analysis; OcPGM, Oryctolagus cuniculus PGM; PAGM, N-acetylglucosamine phosphate mutase; PaPMM, Pseudomonas aeruginosa phosphomannomutase/phosphoglucomutase; PDB, Protein Data Bank; PGM, phosphoglucomutase; PMM/PGM, phosphomannomutase/phosphoglucomutase; PNGM, phosphoglucosamine mutase; PhPMM, Pyrococcus horikoshii phosphomannomutase/phosphoglucomutase; SAXS, small‐angle X-ray scattering; St-mutase, Sulfolobus tokodaii phosphohexomutase; StPGM, Salmonella typhimurium PGM; TtPGM, Thermus thermophilus PGM.

832 Conservation of Global Motions in Enzymes

Introduction Herein, we analyze the oligomeric state of all known structures of enzymes from this superfamily The α‐D‐phosphohexomutase enzyme superfam- currently found in the Protein Data Bank (PDB), ily is ubiquitous in organisms from all kingdoms of representing a total of 11 proteins and 31 crystal life. Enzymes in this superfamily catalyze the structures. Crystal packing analyses and biophysical reversible conversion of phosphosugar substrates, characterization show that more than half of these from the 1-phospho to 6-phospho form. Four proteins are likely oligomers, with the two archaeal subgroups of the superfamily have been well enzymes adopting the largest tetrameric assembly. characterized: phosphoglucomutase (PGM), phos- Two distinct dimeric arrangements that are specific phomannomutase/phosphoglucomutase (PMM/ to different subgroups of the superfamily are PGM), phosphoglucosamine mutase (PNGM), and observed, suggesting independent evolutionary N-acetylglucosamine phosphate mutase (PAGM).1,2 origins. The fluctuation dynamics of the various Although all proteins in the superfamily catalyze oligomers were characterized using normal mode the same reaction, they have differing preferences analysis (NMA). For all assemblies, the most mobile for the sugar moiety of the substrate, as implied by region of the protein is domain 4, consistent with the names of the various subgroups. Due to their different conformers of this domain observed in roles in numerous biosynthetic and metabolic various crystal structures. The conservation of low‐ pathways, including those involved in virulence frequency global motions characterized in this study of human pathogens, many of these enzymes are of is consistent with the known mechanistic impor- 3–6 interest as potential drug targets and may also tance of conformational change of the α‐D‐phospho- – have utility in metabolic engineering.7 10 hexomutases. This study supports recent evidence The reaction mechanism of the α‐D‐phosphohex- for the evolutionary conservation of protein vibra- omutases involves two successive phosphoryl tional dynamics in homologous proteins27 but – transfers.11 15 Initially, the enzyme donates a extends this concept to include a superfamily with phosphoryl group from a conserved active‐site varying quaternary structures. Thus, evolutionary phosphoserine residue to substrate, forming a pressure to maintain conformational flexibility bisphosphorylated sugar intermediate. The inter- across varying molecular shapes may be a factor mediate then reorients in the active site and must affecting the evolution of oligomers in other protein rebind in the opposite orientation so that the serine families. can accept the alternate phosphoryl group from the intermediate, forming product and regenerating active, phosphorylated enzyme. Structural studies Results have shown that conformational change of the enzyme is required at several points in the multi- step reaction, including upon binding of substrate, Overview of the superfamily and structure of the to permit reorientation of the intermediate, and for protomers – the release of product.16 19 Over the last two decades, crystal structures Enzymes in the α‐D‐phosphohexomutase super- have been determined for at least one protein in family are ubiquitous in all organisms including 1 each subgroup of the α‐D‐phosphohexomutase bacteria, archaea, and eukaryotes. They participate – superfamily.20 24 These studies have shown that in a variety of key biosynthetic and metabolic theenzymesshareaconservedfour-domain pathways, determined by the specificity for the architecture, with a large, centrally located active‐ sugar moiety of their substrate. The PGM proteins site cleft. Another feature commonly observed in have high specificity for glucose; the PMM/PGMs the crystal structures is conformational variability can utilize either glucose or mannose; the PNGMs of the C-terminus, which moves via a hinge-type prefer glucosamine; and the PAGMs utilize N- rotation relative to the rest of the protein, and is acetylglucosamine phosphate. Additional specific- correlated with ligand binding.16,18,20,21,25,26 Until ities/activities have been reported for some proteins recently, the quaternary structures of these pro- in the superfamily, including phosphopentomutase teins had been largely unexamined, perhaps and glucose 1,6-bisphosphate synthase activity.28 because the best characterized enzymes were The currently available crystal structures for known to be monomers. However, biochemical enzymes in the α‐D‐phosphohexomutase super- characterization of Salmonella typhimurium PGM family include 11 proteins from 10 different (StPGM) and Bacillus anthracis PNGM (BaPNGM) organisms, and a total of 31 crystal structures demonstrated that these two proteins exist as (Table 1). These structures reflect the widespread dimers in solution.20,21,25 The discovery of oligo- phylogenetic distribution of these proteins, with mers within α‐D‐phosphohexomutase superfamily six from bacteria, two from archaea, and three prompted the current examination of all known from eukaryotes. Sequence comparisons of these structures of these enzymes. proteins show profound diversity, consistent with Author's personal copy

Conservation of Global Motions in Enzymes 833

Table 1. Summary of α‐D‐phosphohexomutase of known structure

Subgroup PISA (EC no.) Organism Abbreviation Lineage UniProtKB PDB ID assembly PGM (5.4.2.2) S. typhimurium StPGM Bacteria Q8ZQW9 2FUV, 3NA5, 3OLP Dimer T. thermophilus TtPGM Bacteria Q5SLE0 2Z0F Dimer O. cuniculus OcPGM Eukaryote P00949 3PMG, 1VKL, 1JDY, 1LXT, Monomer 1C4G, 1C47 Paramecium tetraurelia PtPGM Eukaryote P47244 1KFl, 1KFQ Monomer PMM/PGM P. aeruginosa PaPMM Bacteria P26276 1K35, 1K2Y, 1P5D, 1P5G, 1PCJ, Monomer (5.4.2.8) 1PCM, 3BKQ, 3C04, 2H4L, 2H5A, 2FKF, 2FKM P. horikoshii PhPMM Archaea O58651 1WQA Tetramer PNGM (5.4.2.10) B. anthracis BaPNGM Bacteria Q81VN7 3PDK Dimer F. tularensis FtPNGM Bacteria Q5NII8 3I3W Dimer PAGM (5.4.2.3) C. albicans CaPAGM Eukaryote Q9P4V2 2DKA, 2DKC, 2DKD Monomer ambiguousa S. tokodaii St-mutase Archaea Q976E4 2F7L Octamer T. thermophilus — Bacteria Q5SKJ3 1TUO Monomer PDB IDs in boldface correspond to structures shown in Fig. 1. a Ambiguous indicates proteins that could not be clearly placed into one of the four well-characterized subgroups of the superfamily based on sequence relationships (see the text).

an ancient evolutionary origin of the superfamily. well characterized biochemically, although a few – Amino acid identities between the enzymes in reports have appeared recently.36 38 Table 1 range from a high of 55% to b10% All known proteins in the superfamily share (Supplementary Fig. S1a). remarkably similar tertiary structures, especially Despite their overall sequence diversity, proteins considering their sequence divergence. A superpo- in the superfamily can be unambiguously identified sition of one structure from each of the superfamily basedonhighconservationofkeyactive‐site subgroups shows that these proteins have the same residues. The sequence-conserved regions include tertiary architecture, comprising four domains the phosphoserine residue that participates in arranged in an overall heart shape (Fig. 1a). The phosphoryl transfer and a metal-binding loop that active site is located in the large central cleft, found typically coordinates a Mg2+ required for enzyme at the confluence of the four domains, and is activity. Several other conserved regions are in- typically positively charged.20,21,23 As reflected in volved in binding either the hydroxyl or phosphate their sequence conservation, the arrangement of key groups of the phosphosugar ligands. These con- active‐site loops and catalytic residues is highly served regions are evident on a multiple sequence similar in all of the known structures (Fig. 1b). alignment of 10 proteins from Table 1 (Supplemen- Consistent with its known role in enzyme mecha- tary Fig. S1b). nism, a common theme of conformational variability, Most of the structurally characterized proteins can particularly for the C-terminal domain of the proteins, be easily categorized into one of the four known has been revealed by structural studies. Differing subgroups of the superfamily (Table 1), based on conformers have been seen in comparisons of ligand- previous biochemical characterization or clear se- bound versus apoenzyme structures for PaPMM and quence relationship with functionally characterized Candida albicans PAGM (CaPAGM).22 They have also homologs. Both Oryctolagus cuniculus PGM been observed in comparisons of different polypep- (OcPGM) and Pseudomonas aeruginosa PMM/PGM tide chains in the asymmetric unit of the crystal (PaPMM) have been the subject of extensive structures for BaPNGM and Francisella tularensis – biochemical and kinetic studies.11,12,15,17,19,29 34 The PNGM (FtPNGM)25 and between different crystal PAGM and PNGM proteins of known structure forms of StPGM and PtPGM.21,26 Typically, the have clear sequence homology to functionally different conformers are produced via an interdo- characterized homologs.35 However, based on main rotation of domain 4, with a magnitude of ~10– sequence comparisons, we find two proteins in 15°. However, in the PNGM proteins, the magnitude Table 1 difficult to place in a specific subgroup of the of the rotation is much larger (30–55°),20,25 while in superfamily and therefore designate them as “am- PAGM, the conformers appear to differ mostly via biguous”. (This designation refers only to substrate ordering of active‐site loops in domain 4.22 specificity, not whether they are members of the superfamily, which is not in question.) Notably, The assemblies these two proteins are from extremophiles, including one of the archaeal sequences. In general, To characterize potential oligomeric assemblies for superfamily members from archaea have not been the α‐D‐phosphohexomutases of known structure, Author's personal copy

834 Conservation of Global Motions in Enzymes

Fig. 2. The experimentally confirmed oligomeric assemblies found to date in the α‐D‐phosphohexomutase superfamily. Structures that exhibit the respective assemblies are listed by PDB ID. Both schematics (heart shapes) Fig. 1. (a) A ribbon diagram showing an overlay of one and ribbon diagrams are shown for each assembly. protomer from four subgroups of the α‐D‐phosphohex- Domain 4 is indicated by “4” on each schematic to omutase superfamily, highlighting their similar domain indicate relative orientation of the protomers within the architecture. The four domains of the proteins are shown assembly. Blue arrows and labels refer to interfaces in different colors: domain 1 in green, domain 2 in red, described in Table 2. (a) Dimer A; (b) dimer B; and (c) domain 3 in yellow, and domain 4 in blue. The proteins tetramer. Note that interface B is found in both dimer B illustrated are CaPAGM (2DKD), PaPMM (1P5D), and the tetramer. FtPNGM (3I3W), and StPGM (3NA5). (b) A close-up view of the active site of the four proteins in (a). Shown in a stick model are side chains of the catalytic phosphoserine residue and three aspartates that coordinate the metal ion (spheres in shades of pink) to highlight the structural summary of their physiochemical features is given in similarity of these key active‐site residues in all subgroups Table 2. Below, we describe each of the assemblies, of the superfamily. features of the oligomeric interfaces, and experimental evidence supporting the predicted quaternary structures. we used the program PISA39 to analyze crystal packing for each of the 11 proteins in Table 1. PISA Monomer uses thermodynamic calculations to predict the stability of all molecular interfaces observed in the The monomers in the α‐D‐phosphohexomutase crystal lattice with a reported accuracy of 80–90%.40 superfamily include bacterial PMM/PGM, as well In total, 6 of the 11 proteins of known structure as the eukaryotic PGM and PAGM proteins. Several appear to be oligomers (Table 1), with predicted of the predicted monomers have been experimen- quaternary assemblies ranging from monomeric to tally verified, including PaPMM,30,31 CaPAGM,22 octameric. However, as the octamer could not be and OcPGM.41 The enzyme from the Gram-negative confirmed experimentally (see below), it is not eubacterium Thermus thermophilus (found in ambig- considered in detail herein. As shown in Fig. 2, uous subgroup in Table 1; PDB 1TUO) is also three different oligomers are found, including two predicted to be a monomer. PtPGM (also known as distinct dimers and a tetramer. Within these oligo- parafusin) has been suggested to be a dimer based mers, three unique interfaces can be identified and a onWesternblotsbuthasnotbeenconfirmed Author's personal copy

Conservation of Global Motions in Enzymes 835

Table 2. Physiochemical characteristics of the interfaces in the α‐D‐phosphohexomutase oligomers

No. of Protein No. of SAS/% H-bonds/ion % Polar/ Gap volume Similar Interface (interface no.) Chain CSS residues total (Å2) pairs charged % NP (Å3), index (Å) interface Q-score A 3NA5 (1) A, B 1.0 33 1095 (5.3%) 15/5 34.4/28.1 37.5 6723, 3.1 2Z0F 0.67 2Z0F (3) A, B 1.0 21 731 (3.8%) 9/5 30.0/25.0 45.0 6990, 4.8 3NA5 0.67 B 3PDK (1) A, B 1.0 36 1198 (6.9%) 16/7 28.1/15.6 56.3 5052, 2.1 2F7L 0.61 3I3W 0.60 1WQA 0.53 2F7L (1) A, B 1.0 34 1310 (7.1%) 10/6 28.6/25 46.4 7857, 4.0 1WQA 0.66 3I3W 0.56 3PDK 0.61 3I3W (1) A, B 1.0 31 1125 (5.9%) 4/0 34.5/10.3 55.2 5542, 2.4 2F7L 0.55 1WQA 0.54 3PDK 0.60 1WQA (1) A, B 1.0 34 1290 (6.9%) 12/8 25.0/28.1 46.9 3834, 1.5 2F7L 0.66 3I3W 0.54 3PDK 0.53 C 1WQA (3) D, A 1.0 23 899 (4.8%) 12/4 27.3/22.7 50.0 8228, 4.5 2F7L 0.66 2F7L (3) A, C 1.0 28 987 (5.3%) 21/5 28.6/25.0 46.4 46,332, 23.5 1WQA 0.66 Interface number as assigned by PISA; CSS, complexation significance score; SAS, solvent‐accessible surface area; % total, contribution of interface to total surface area of protein; NP, nonpolar. The % polar/nonpolar/charged residues, gap volume, and index were obtained from ProtorP; all other quantities were calculated by PISA. The Q-score quantifies the overall similarity between two interfaces for the structural pair listed under “Protein” and “Similar Interface”. otherwise.26 Additional studies on other represen- the highest possible, for both StPGM and TtPGM. tatives of the superfamily will be needed to This score, as calculated by PISA, estimates the determine whether the monomeric form is common significance of each interface in the assembly of the to specific subgroups (e.g., PAGM) or lineages (e.g., respective oligomer. The area of interface A repre- eukaryotic PGM). sents ~5% of the total surface area of both proteins, and the interface consists of roughly equal amounts Dimer A of polar and apolar regions. The Q-score, which quantifies the similarity between interfaces in One of the two dimeric assemblies in the different proteins, is 0.67 when comparing StPGM superfamily (hereafter called dimer A; Fig. 2a) and TtPGM, indicating high similarity (see also Fig. was first observed for a bacterial PGM protein from S2a) (Q-scores in the range of 0.2–0.4 indicate S. typhimurium (PDB ID: 3NA5).21 Dimer A has C2 moderate structural similarity39). symmetry, with the two protomers arranged “up- ” side down relative to each other, such that the two Dimer B active sites are found on opposite sides of the dimer. Dimer A has an overall “S” shape, and the residues The second, distinct dimer in the superfamily was involved in the dimeric interface come mostly from first noted in the structure of BaPNGM (PDB ID: domain 1 of each chain. This dimer has been 3PDK) and confirmed in solution by DLS and experimentally confirmed in solution for StPGM SAXS20,25 (Fig. 2b). As in dimer A, dimer B also via dynamic light scattering (DLS), analytical has C2 symmetry, and the interacting residues ultracentrifugation, and small‐angle X-ray scatter- reside mostly in the first domain of the protein. ing (SAXS) studies.21 PISA calculations predict that Dimer B has an overall “3” shape, with the two the same dimer is found for the PGM protein from protomers arranged in a side-by-side fashion, such T. thermophilus (TtPGM; PDB ID: 2Z0F), which has that both active sites open to the same side of the 55% sequence identity with StPGM (Fig. S1a). A dimer. Even a cursory comparison shows that structural superposition of dimer A from StPGM dimers A and B are structurally unrelated, and and TtPGM is shown in Fig. S2a, highlighting their thus likely arose independently in separate branches high degree of structural similarity. of the superfamily during evolution. Within dimer A, a protomer–protomer interface is Structural comparisons show that dimer B is found identified, termed interface A. This interface is in the other representative of the PNGM subgroup found only in the structures that comprise dimer from F. tularensis (PDB ID: 3I3W); a detailed A. A detailed description of the residues and structural comparison of FtPNGM and BaPNGM interactions of this interface in StPGM can be dimer interfaces is found in Ref. 25. To confirm the found in Ref. 21. As can be seen in Table 2, interface dimeric arrangement of the FtPNGM in solution, we A has a complexation significance score (CSS) of 1.0, analyzed purified protein (see Materials and Author's personal copy

836 Conservation of Global Motions in Enzymes

Table 3. DLS measurements of α‐D‐phosphohexomutases arranged upside down relative to each other (i.e., back-to-back pair of 3's). Hence, within the tetramer, Estimated Calculated molecular mass molecular mass % all four active sites are accessible to solution, with Protein Rh (Å) (kDa) (kDa) Difference two opening on one side and two on the other. The FtPNGM 44 106.9 101.8 4.8 contacts between protomers in the tetrameric PhPMM 57 193.9 208.2 6.9 assembly involve residues from domain 1 (the St-mutase 58 204.9 209.1 2.0 dimer B interface between protomers) and from Conditions for measurements are described in Materials and domain 2 (the interface between the two dimers, Methods. Estimated molecular mass is obtained from Rh hereafter interface C). measurements by DLS. Calculated molecular mass is based on As this was the first suggestion of a higher‐order amino acid sequence of the protomer (including affinity tag) oligomer in the superfamily, we sought experimen- multiplied by the expected oligomeric assembly. Protein names tal confirmation. The gene for PhPMM was com- use the abbreviation from Table 1. mercially synthesized, and the protein was expressed and purified (see Materials and Methods). DLS measurements of purified protein Methods) by DLS (Table 3). This shows that the show a hydrodynamic radius of 57 Å, correspond- protein has a hydrodynamic radius of 44 Å, corre- ing to an estimated molecular mass of 194kDa sponding to an estimated molecular mass of 107kDa. (Table 3). This is within 7% of the calculated This is within 5% of the calculated molecular mass molecular mass for a tetramer based on amino for a dimer of FtPNGM based on its amino acid acid sequence. Hence, the DLS results experimen- sequence. tally confirm the prediction by PISA for a tetrameric Dimer B is also found within the tetrameric assembly of this protein. assemblies of Pyrococcus horikoshii phosphomanno- To further investigate the PhPMM assembly in mutase (PhPMM) and the protein of ambiguous solution, we determined its oligomeric state and subgroup from Sulfolobus tokodaii (St-mutase; PDB quaternary structure using SAXS. The composite ID: 2F7L), which are discussed further in the scattering profile from the lowest protein concen- following section. A structural superposition of tration sample is shown in Fig. 3a. The Guinier plot dimer B from all four of these proteins, which have (Fig. 3a, inset) exhibits good linearity (R2 =0.9899) overall sequence identities ranging from 27% to and suggests Rg of 47Å. The Rg values estimated 45%, is shown in Fig. S2b. This figure highlights the from the two higher protein concentrations samples common dimeric interface (interface B) shared by all are 47 Å (4.8mg/mL) and 46Å (7.2 mg/mL). The four proteins. In Table 2, it can be seen that interface linearity of the Guinier plots and the lack of a strong B is consistently the largest of the three oligomeric dependence of Rg on protein concentration suggest interfaces, comprising N30 residues and involving that the protein sample is free of substantial ~6% or more of the total surface area of the aggregation and interparticle interference effects. protomer. It has a CSS of 1.0 for all proteins and Calculations of the pair distribution function yield – involves both polar and apolar interactions. Q- an Rg of 47 48Å and a maximum particle dimension scores comparing interface B for the 12 possible of 158 Å. For reference, the Rg values of the two structural pairs of these proteins range from 0.53 to dimeric assemblies that are predicted by PISA to be 0.66, all indicating highly significant structural stable in solution are only 32Å and 36Å. Thus, the similarity. Unlike dimer A, the mechanism of Rg analysis clearly suggests that the oligomeric state oligomerization for dimer B does not involve an of PhPMM is larger than a dimer. The Rg calculated obvious sequence motif or structural insertion. from the tetramer shown in Fig. 2c is 40Å, which is Contacts in the dimer interface are mediated by more consistent with the SAXS data, albeit signifi- α β residues in helices that are part of the core / fold cantly smaller than the SAXS estimate. The Rg from of domain 1 of these proteins.20 the crystallographic tetramer is likely an underesti- mate of the true Rg because it does not account for The tetramer the histidine tag and flexibility of domain 4. Theoretical scattering profiles were calculated from Calculations by PISA for the phosphohexomutase dimer and tetramer models of PhPMM in order to from the archaeal organism P. horikoshii (PhPMM; gain additional insight into the solution structure PDB ID: 1WQA) predicted that this protein would (Fig. 3a). The profiles calculated from the dimer exist in a tetrameric arrangement (Fig. 2c). This models deviate substantially from the experimental protein has been biochemically characterized and one. In contrast, the profile calculated from the belongs to the PMM/PGM subgroup.36 It is found tetramer shows substantially better agreement with in a hyperthermophilic organism that lives in the experimental profile. Finally, the envelope hydrothermal ocean vents at an optimal tempera- determined from a shape reconstruction calculation ture of 98°C. The predicted tetramer of PhPMM has matches the size and shape of the tetramer. It is D2 symmetry and consists of two copies of dimer B, concluded that PhPMM forms a tetramer in solution Author's personal copy

Conservation of Global Motions in Enzymes 837

Fig. 3. SAXS analysis of PhPMM. (a) Comparison of the experimental SAXS proﬁle with those calculated from dimeric and tetrameric assemblies that are predicted to be stable by PISA. The inset shows a Guinier plot covering the region ≤ ≤ 0.53 qRg 1.24. (b) Superposition of the crystallographic tetramer with the shape reconstruction (con- toured at 1.0σ). Author's personal copy

838 Conservation of Global Motions in Enzymes and the four-body assembly identified in the crystal coincided well with functionally relevant conforma- lattice represents the solution structure of this tional change. To gain further insights, we con- protein. ducted NMA of representative members of the As noted earlier, PISA predicted that one protein superfamily using the elastic network model as in Table 1 would exist as an octamer in solution. implemented by elNémo (Materials and Methods).42 This is St-mutase from S. tokodaii, which is the A monomer and one of each oligomer were second archaeal enzyme of known structure in the examined using the following structures: monomer, superfamily. (S. tokodaii is a thermoacidophilic 1K2Y; dimer A, 3NA5; dimer B, 3I3W; and tetramer, organism that grows in sulfur-rich hot acid 1WQA. In particular, we were interested in charac- terrestrial spring at an optimum temperature of terizing the global motions of these proteins in the 80°C.) To characterize the oligomeric state of this context of their respective assemblies and assessing enzyme in solution, we commercially synthesized the potential impact of oligomerization on the the gene for St-mutase, and the protein was functionally important conformational change of expressed and purified (see Materials and domain 4. Unless noted otherwise, the global Methods). DLS measurements of purified protein fluctuations described below are those from the show a hydrodynamic radius of 58 Å, correspond- first (lowest frequency) normal mode, which are ing to an estimated molecular mass of 205kDa highly correlated with biological relevance in many – (Table 3). Rather than an octamer, this molecular systems.43 45 mass corresponds to a tetramer, similar to PhPMM. Although we were unable to experimentally Monomer confirm the predicted oligomer, it is possible that it could exist for some fraction of molecules in NMA for a monomeric representative of the solution or under different experimental condi- superfamily was performed for PaPMM (PDB ID: tions. However, we note that while PISA predicts 1K2Y). Figure 4a shows its structure colored St-mutase to be a stable octamer, the tetramer– according to the relative amplitudes of motion; tetramer interface has a CSS of only 0.37, signifi- vectors indicate the direction of motion for residues cantly lower than that for the experimentally with the largest movements. It is apparent on this verified interfaces. figure that the largest amplitude fluctuations (in red) The interface between dimers within the tetramer occur within domain 4 of the protein, with some (interface C) is found only in the PhPMM and St- smaller motions also occurring in domains 1 and 2. mutase structures and involves residues from Moreover, the direction of the domain 4 fluctuation domain 2 of the proteins. A superposition highlight- serves to open/close the active site (see animation in α ing this interface is shown in Fig. S2c; Q-scores File S1), accounting for a decrease in the overall C indicate high interface similarity for the two pro- RMSD between the two structures from 3.4 to 1.6Å2 teins. Interface C has a calculated CSS of 1.0 for both (Fig. S3). This global fluctuation is consistent with a proteins and involves ~5% of the total surface area previously characterized hinge-type movement of (Table 2). Similar to interfaces A and B, interface C domain 4 (~10° rotation) for this protein that occurs involves a mixture of polar and apolar interactions. upon ligand binding.16,18,30 Although both of these proteins are from thermo- A correlation matrix showing residue fluctuations philes, we find no distinct differences in the within PaPMM in more detail is found in Fig. S4a. physiochemical nature of their oligomeric interfaces This matrix allows for extraction of correlated atomic compared to those from the mesophiles in Table 2. motions from the NMA: blue coloring indicates For example, interface B of PhPMM and St-mutase residues that move in similar directions (positively has a similar size (as % total surface) and numbers of correlated), while red indicates groups of residues hydrogen bonds/ion pairs, relative to FtPNGM and that move in different directions (negatively corre- BaPNGM. Residues in interface C are generally lated). For reference, a schematic highlighting the conserved among proteins closely related to four-domain structure of the protein is also shown PhPMM and St-mutase (data not shown), but there (right of figure). A few sizable residue groupings of are no obvious sequence motifs or structural correlated motion (e.g., large patches of blue or red) features that clearly distinguish this interface from can be seen; however, in general, the fluctuations for other (non-tetrameric) proteins in the superfamily. the monomer tend to correspond to small residue clusters, in both sequential (along the diagonal) and Conformational variability and low‐frequency nonsequential regions of the protein chain. In Fig. S5 motions (left column), we show quantitatively through a correlation of eigenvectors that the first normal Given the various quaternary structures found for mode fluctuations for the PaPMM monomer are the α‐D‐phosphohexomutases, we were interested in highly similar to those of the isolated protomers from assessing their global motions, as earlier studies on dimer A, dimer B, and the tetramer, as would be two of these enzymes had shown these fluctuations expected due to their similar molecular architecture. Author's personal copy

Conservation of Global Motions in Enzymes 839

Fig. 4. Ribbon diagrams of the α‐D‐phosphohexomutases colored according to global motions from the NMA calculations. Structures are shown for the monomer and each distinct oligomeric assembly. (a) Monomer of PaPMM; (b) dimer A of StPGM (3NA5); (c) dimer B of FtPNGM (3I3W); (d) tetramer of PhPMM (1WQA). Color of semitransparent surface represents amplitude of vibrations for the ﬁrst normal mode (blue, least; red, greatest). Black arrows show overall direction of movement for regions of high ﬂuctuation; length of arrow indicates relative magnitude.

Dimer A showing the correlated motions between residues in all possible pairs of chains (A–A, B–B, A–B, and B– Previous NMA of the protomer of StPGM (PDB A). In dimer A, domains 1 and 2 of the protein, and ID: 3NA5) showed that fluctuations corresponding parts of domain 3, tend to move as a large positively to the lowest‐frequency normal mode recapitulate correlated region (large blocks of blue). Within each the crystallographically observed and catalytically polypeptide chain (upper left and lower right relevant conformational change of domain 4 in this panels), the motion of domains 1–3 is anticorrelated protein.21 Here, we examine the global motions of with that of domain 4 (see red region inside yellow dimer A of StPGM, which, due to its different rectangle). The same is true when comparing the molecular shape, are distinct from those of the motion of domains 1–3 in chain A with domain 4 of protomer. As seen in Fig. 4b, residues in domain 4 of chain B. However, when comparing the movement both polypeptide chains exhibit the largest ampli- of domain 4 between chains (e.g., A versus B, upper tude fluctuations in this dimer. For the lowest‐ right‐hand panel of Fig. S4b), positive correlation is frequency normal mode, the movement of this observed (see blue region inside green square), domain occurs in a plane perpendicular to the 2- consistent with the similar directions of motion fold axis of the dimer and has the effect of opening/ highlighted by the vectors in Fig. 4b. For compar- closing the active‐site cleft in both protomers (see ison, the correlation map of the protomer of StPGM animation in File S2). The fluctuations of domain 4 is also shown (see single panel at the bottom of Fig. of both protomers are seen to occur in the same S4b), which differs in having many small regions of direction (note generally similar direction of vectors correlated movement, rather than the domain‐sized on two chains). However, due to the “S” shape of the clusters observed in the dimer. dimer, these fluctuations lead to alternate opening/ closing of the two active sites—as one site closes, the “ other opens and vice versa. Such dynamically Dimer B coupled” motions have been previously observed in several other systems25,46,47 and may have potential NMA for dimer B was conducted for FtPNGM functional consequences (see Discussion). (PDB ID: 3I3W). The results of these calculations are The correlated motions within dimer A can be very similar to those previously described for the ascertained in more detail on its correlation matrix related BaPNGM protein,25 which shares the dimer (Fig. S4b). For the dimer, this matrix has four panels, B assembly. However, we present a brief synopsis Author's personal copy

840 Conservation of Global Motions in Enzymes here for the sake of completeness. As is found for the assembly, and it can be seen that domain 4 tends dimer A, domain 4 in dimer B is the most affected by to move independently and with a hinge-like the lowest‐frequency normal mode fluctuations motion, particularly in the third mode. Fluctuations (Fig. 4c). The fluctuations result in an overall from the low‐frequency normal modes were twisting motion, perpendicular to the long axis of assessed for their ability to generate the functionally the dimer (see animation in File S3). However, in significant conformational change of domain 4, by contrast to dimer A, domain 4 of each protomer in superimposing one protomer from the tetramer dimer B moves in the opposite direction to its (which adopts an open conformer in the PhPMM counterpart in the other chain (note vectors in structure), with a closed conformer from the 1P5D Fig. 4c). In this dimer of FtPNGM and as previously structure of PaPMM (an enzyme–ligand complex noted for BaPNGM,25 dynamic coupling is also with glucose 1-phosphate). A combination of mo- observed. However, in the case of dimer B, the tions from the first and third modes largely re- opening/closing of the two active sites is concerted, produces the closed conformer observed upon rather than alternating as seen for dimer A. This ligand binding in this related protein (Fig. S6). contrasting result is due to the very different spatial Hence, it seems that for this large assembly, several arrangement of the protomers in the dimer A and of the lowest‐frequency normal modes may be dimer B assemblies, despite the similar tertiary important for the functionally relevant motion of structures of their protomers. this domain. This result is consistent with previous The anticorrelated motion of domain 4 in the two analyses showing that a small number of low- polypeptide chains of dimer B is evident on its frequency modes typically suffice for qualitative correlation matrix (Fig. S4c). Similar to that of analysis of protein dynamics.48 dimer A, this matrix also shows large, domain- As seen in Fig. S4d, the correlation matrix of the sized groupings of residues with correlated mo- tetramer has 16 panels, corresponding to the two tions. Domains 1–3 of each chain within the dimer copies of dimer B (chains A/B and chains C/D) tend to form a rather large, single unit with within this assembly. In contrast to dimer B, the positively correlated motion, which is negatively domain-sized groups of correlated motion in the correlated to domain 4 within the same chain (see lowest‐frequency normal mode of the tetramer are yellow rectangle), while positively correlated to not quite as pronounced, although domain 4 still domain 4 of the other chain. When comparing the tends to fluctuate as a single group, as to some motions of domain 4 in chain A versus chain B, extent do domains 2 and 3. Relative to the protomer, negative correlation is observed (see green square), the tetramer still shows enhanced blocks of corre- consistent with the different directions of motion lated motion (Fig. S4d). For the second and third indicated in Fig. 4c. Comparing the correlation map normal modes of the tetramer, different patterns of of the FtPNGM dimer with that of its protomer correlated motion are seen (data not shown), but (Fig. S4c) shows that the protomer generally lacks both feature fluctuation of domain 4 as a single unit. the domain-sized groupings of positively correlat- Because of the multiple subunits of the tetramer and ed motions. the multiple normal modes involved in the functionally important motion, it is difficult to assess the Tetramer relative motions of domain 4 in each chain with respect to the others; hence, we cannot make a NMA of the PhPMM tetramer shows that domain conclusion regarding the possibility of dynamic 4, found on the four outer corners of the oligomer, is coupling within the tetramer. the location of the highest‐amplitude, low‐frequency fluctuations in this assembly (Fig. 4d). However, as can be seen qualitatively in the animation in File Discussion S4, the motion of domain 4 from this normal mode does not quite capture the hinge-like opening/ closing of the active site, as observed for the Features and evolution of the monomer. Moreover, as seen in Fig. S5a, the α‐D‐phosphohexomutase oligomers different molecular shape of the tetramer contri- butes to a weaker, although still significant, corre- Until recently, most proteins in this enzyme lation of eigenvectors with the monomer or its superfamily were believed to be monomers. Howev- protomer. The contribution of other low‐frequency er, as shown herein and in several other recent normal modes to the functional motion of domain 4 studies,20,21,25 both experimental and computational in the tetramer was therefore considered. analyses suggest that many of these proteins are Fluctuations from the second and third normal oligomeric. In addition to the two previously charac- modes of the tetramer were calculated, and their terized dimeric forms, we show here that a higher‐ animations are included in File S4. For these modes, order assembly, a tetramer, is also found in this domain 4 fluctuations are also the most significant in enzyme superfamily. Of the three experimentally Author's personal copy

Conservation of Global Motions in Enzymes 841 verified assemblies, dimer A seems to be localized to from archaea, and also in the bacterial PNGM the most limited set of proteins of the superfamily. proteins, it appears to be an ancient oligomeric The two dimer A proteins (StPGM and TtPGM) both interface, potentially predating the divergence of belong to the PGM subgroup, are both found in bacteria and archaea. Gram-negative bacteria, and share relatively high sequence identity. To date, interface A has only been Advantages of oligomeric found in the dimer A assembly. Dimer B, on the other phosphohexomutases? hand, appears to be widespread in the superfamily, occurringinboththePNGMandPMM/PGM As in other systems, the presence of multiple subgroups and in both archaeal and bacterial oligomeric forms within the α‐D‐phosphohexomu- organisms. Moreover, dimer B is found in both tase superfamily suggests evolutionary benefits. In dimeric and tetrameric assemblies, and only moder- general, protein oligomers have been associated ate sequence identity is found between the four with advantages such as cooperativity, increased proteins. To date, the tetramer has only been coding efficiency, and additional opportunities for observed in structures of archaeal members of the regulation.51 Oligomerization is also a convenient superfamily. Its subgroup distribution is not clear, way to build large proteins, which are associated since only one of the two proteins (PhPMM) has with increased stability against denaturants and a been biochemically characterized.36 The two tetra- reduction in surface area relative to smaller meric proteins share moderate sequence identity proteins.51 While it is not yet known which, if any, (45%) but come from organisms in different king- of these may apply to the α‐D‐phosphohexomutases, doms of the archaea: P. horikoshii belongs to the several possibilities are discussed below. Euryarchaeota, while S. tokodaii is from Crenarch- Based on the currently available structures in the aeota. Overall, even though a relatively small α‐D‐phosphohexomutase superfamily, it appears number of these enzymes have been structurally that higher‐order oligomers may be more common characterized to date, it is clear that oligomers are in the proteins from archaeal organisms. This raises spread throughout the different subgroups and the issue of whether higher‐order oligomers may be phylogenetic branches of the superfamily. generally correlated with thermostability. Such a Despite their varying assemblies, a common correlation has been noted in several other protein 52–55 feature of all of the observed oligomers is that the families. It appears that the α‐D‐phosphohex- protomers are arranged with their active sites omutases could add to this list. However, in other – distant from the oligomeric interfaces and exposed studies,56 58 oligomers were not noted to be more to solution. This distinguishes the α‐D‐phosphohex- common in extremophiles, as might be expected if omutases from other protein families that undergo they contribute to thermal stability. (As a cautionary large conformational changes, such as transferases, note, these studies utilized the biological assembly which appear to utilize oligomerization as a way to listed in the header of the PDB file, without any shield ligand from solvent.49 Moreover, the in- additional analyses for verifying the stated assem- teractions between protomers involve almost exclu- bly, and therefore may have underestimated the sively residues in domain 1 or 2 of the enzymes, occurrence of oligomers.) Therefore, while a number thereby leaving domain 4 topologically unencum- of protein families appear to exist as higher‐order bered and free to vary its conformation. This oligomers in extremophiles, it is currently unclear observation emphasizes the importance of domain whether this is a common feature of thermophilic 4 conformational variability in enzyme function in proteins. the α‐D‐phosphohexomutases. Another possible, although currently speculative, Recent surveys have shown that most proteins in advantage of the oligomeric versions of the α‐D‐ the PDB are oligomeric, with dimers being the most phosphohexomutases is the opportunity for intro- commonly observed assembly.40 The two distinct ducing cooperative behavior in enzyme mechanism. dimers observed in the α‐D‐phosphohexomutase Cooperativity has potential benefits such as alloste- superfamily, found in 4 of the 11 known structures, ric regulation.51 The low‐frequency correlated mo- are consistent with this observation. It has also been tions for the two dimers, which result in dynamically shown that most homologous proteins (N30% coupled opening/closing of domain 4 in the poly- sequence identity) have conserved quaternary struc- peptide chains, may be relevant to this. These ture (~70% of the time). In cases where the coupled movements are manifested differently on quaternary structure differs, a pathway linking the active sites, depending on the assembly. In dimer observed assemblies and evolution has been A, the two active sites open in an oscillating fashion noted,50 whereby the largest interface is maintained (one open, the other closed), while in dimer B, the 25 during evolution. In the α‐D‐phosphohexomutase two active sites open/close in concert. Both superfamily, this relationship holds true for dimer B versions of dynamic coupling could be envisioned and the tetramer, both of which share interface B. As to produce cooperative effects on enzyme activity this interface is found in the two tetrameric proteins (i.e., binding of substrate to one active site may Author's personal copy

842 Conservation of Global Motions in Enzymes induce the opening or closing of the other). flexibility may select proteins for which function However, this intriguing possibility has not yet depends on physically robust normal modes. been addressed experimentally for any proteins in Maguid et al. note that conservation of normal the superfamily. modes could occur either because of their importance to function or because these modes are more Evolutionary conservation of global motions robust with respect to mutational perturbations.27 For the α‐D‐phosphohexomutases, either scenario Previous examination of global motions via NMA could potentially apply, as the low‐frequency for the protomer of StPGM and protomer and dimer vibrational modes are functionally important as of BaPNGM showed that they reproduced varying well as conserved in superfamily members with conformers observed in crystal structures of the very diverse sequence relationships. It is tempting proteins.21,25 Here, we show that this is also the case to speculate that a synergistic combination of a foramonomericmemberofthesuperfamily, normal mode that is both physically robust and of PaPMM. Based on these comparisons with structur- clear functional importance may underlie the al data, it is now well established that the functional success of the α‐D‐phosphohexomutases on an motions of proteins within this superfamily are evolutionary time scale. largely collective in nature. These findings are An interesting question that emerges from this similar to those seen in many protein systems study is whether the varying oligomers in the where biologically relevant conformational changes superfamily have evolved simply to allow low‐ of proteins are found to correspond to the lowest‐ frequency vibrations of domain 4, or whether frequency vibrational mode(s) of the molecule (for oligomeric assemblies have been selected to enhance reviews, see Refs. 43–45). As vibrational normal the conformational fluctuations of this domain. Due modes are imposed by protein topology,59 and the to their larger size, collective motions within oligo- structures of the known protomers in the superfam- mers are naturally enhanced, as can be seen here and ily are quite similar, it is not surprising that high‐ in several other cases where large, domain-sized amplitude fluctuations of domain 4 are commonly groupings of concerted motions within oligomers are observed. Earlier studies have established that seen relative to those of the protomers.25,65,66 Should proteins with similar architecture display similar these enhanced collective motions confer functional large-scale dynamic behavior60 and, moreover, that advantages, such as increasing the amplitudes of evolutionary subspace (e.g., structural differences potential fluctuations or by affecting the time scale of between homologous proteins) overlaps significant- dynamic changes, they may have contributed to the ly with subspace from calculated low‐frequency emergence of multiple oligomeric assemblies within normal modes.61 this enzyme superfamily. It is therefore interesting to In the present study, we look beyond the global note that dimer B, the most widespread of the known fluctuations of the protomer and find that low‐ oligomers, displays the largest conformational differ- frequency vibrational modes produce functionally ence between domain 4 of its protomers in its crystal relevant conformational changes across the known structures.20,25 Similar functional advantages can be oligomeric assemblies in the α‐D‐phosphohexomu- easily envisioned to apply to other systems as well. tase superfamily. Despite their different molecular Recent studies of GroEL, for example, have shown shapes and interfaces, low‐frequency global motions that a model containing all subunits is necessary to of the monomer, dimer A, dimer B, and tetramer all capture its allosteric transition.67 result in large amplitude fluctuations of domain 4. In summary, we find that the α‐D‐phosphohex- Thus, just as the topological arrangement of omutase enzyme superfamily comprises proteins of oligomers in this superfamily has preserved the quaternary structure ranging from monomers to conformational freedom of domain 4, the low‐ tetramers. All of the known oligomeric assemblies frequency vibrational modes of each oligomer are arranged such that domain 4 can adopt varying contribute to the conformational variability of this conformers, as required for ligand binding and domain. It therefore seems likely that low‐frequency catalysis. Low‐frequency vibrations are found to fluctuations resulting in domain 4 motions are a contribute to the functionally important motion of conserved feature of the entire α‐D‐phosphohexo- domain 4, despite varying quaternary structures of mutase superfamily, regardless of the oligomeric the proteins. Although the functional relevance of state of individual proteins, and consistent with the vibrational normal modes has been demonstrated in known importance of this domain rotation in the many protein families and superfamilies, the current catalytic mechanism of these proteins. study is one of only a few examples showing Based on recent studies showing that the lowest, possible evolutionary conservation of global mo- most collective normal modes of proteins tend to tions in homologous proteins of varying oligomeric be conserved within protein families and state.63,65 This intriguing result may lead to new – superfamilies,27,61 65 it has been suggested that insights regarding the function and evolution of the evolutionary optimization of conformational protein oligomers. Author's personal copy

Conservation of Global Motions in Enzymes 843

Materials and Methods Dynamic light scattering Protein samples at 1–2mg/mL were prepared and put Computational analysis of quaternary structure and through a 0.22‐μm polyvinylidene fluoride filter if oligomeric interfaces necessary. Data were collected on a Protein Solutions DynaPro 99 instrument at a wavelength of 8363Å for at The PISA web server† was used to calculate the likely least 100s (5s each for 20 or more acquisitions) at 25°C. oligomeric assembly and interface characteristics for each Polydispersity of samples ranged from 0% to 25% (for of the proteins in Table 1. The ProtorP protein–protein PhPMM and St-mutase, respectively). The moderate interaction analysis server68 was used to characterize polydisperity of the St-mutase protein appears due to additional features of the interfaces‡. All calculations were some higher‐molecular‐weight species coexisting in solu- performed using coordinates for the various assemblies as tion, but these could not be clearly identified. generated by PISA, unless otherwise noted. Structural figures were made with PyMOL.69 SAXS analyses

Protein expression and purification SAXS experiments were performed at beamline 12.3.1 of the Advanced Light Source via the mail-in program.71 Genes for the PhPMM and St-mutase proteins were Prior to analysis, samples of PhPMM were run over a synthesized commercially (GenScript) and inserted into Superdex 200 size‐exclusion column, and protein from the the pET14b vector (Novagen) with an N-terminal hex- major peak corresponding to the presumed tetramer was ahistidine tag. The construct for FtPNGM in the pMCSG7 collected and dialyzed as described above. Scattering vector was a generous gift from the Center for Structural intensities (I) were measured at three protein concentra- Genomics of Infectious Diseases. For expression, Escher- tions (2.4, 4.8, and 7.2mg/mL) using exposure times of 0.5, ichia coli BL21(DE3) cells were transformed with the 1.0, 3.0, and 6.0s. The scattering curves collected from the corresponding plasmid and grown at 37°C to an OD600 protein sample were corrected for background scattering (optical density at 600nm) of 0.8 in LB medium with using intensity data collected from the dialysis buffer. A 0.1mg/L ampicillin. Expression of target protein was composite scattering curve for each protein concentration induced by the addition of IPTG to a final concentration of was generated with PRIMUS72 by scaling and merging the 0.4mM, grown at 37°C for 4h, and harvested by background-corrected high‐q region from the 3.0‐s expo- centrifugation. Cell pellets were stored at −80°C. sure with the low‐q region from the 0.5‐s exposure. For protein purification, cell pellets were resuspended GNOM73 was used to calculate the pair distribution in Buffer A (20mM Na phosphate, pH7.8, and 300mM function [P(r)] in order to estimate the radius of gyration 73 NaCl) on ice with 2mM MgSO4 and CaCl2, 14.4mM (Rg) and the maximum particle dimension (Dmax). Shape BME, 0.5mM TLCK and PMSF, and 10μg/mL DNase. reconstruction calculations were performed with the Lysis was performed with a French press, followed by sastbx.shapeup module of the Small Angle Scattering centrifugation for 30min at 25°C. In the case of PhPMM ToolBox (SASTBX) via the SASTBX server.74 The shape and FtPNGM, the lysate was stirred with protamine reconstruction calculations employed the PISA database sulfate (5mg/g cell pellet) for 30min at 25°C to remove of shapes, and we note that the 1WQA structure was not in nucleic acids. For St-mutase, the lysate was instead the basis set used for shape reconstruction. MOLEMAN75 incubated at 65°C for 30min. Precipitation was removed was used to calculate Rg from atomic coordinates. The by centrifugation; the lysate passed sequentially through FoXS server76 was used to calculate SAXS curves from 0.45‐ and 0.22‐μm filters and incubated with pre- atomic coordinates. equilibrated (His Select) Ni2+ affinity beads for 30min. The resin was transferred into a column and allowed to Normal mode analyses settle, and the column was washed with Buffer A containing 5mM and then 10mM imidazole, pH7.8. Proteins were eluted with buffer A containing 125– NMAs were performed using the elastic network model 42 250mM imidazole, pH7.8. The proteins were dialyzed as implemented in the elNémo server§. For these into 50mM Mops, pH7.4, and 1mM MgCl2. For PhPMM analyses, coordinates for the various assemblies were and St-mutase, 100mM NaCl was included in the final taken from PISA, with the exception of the 3I3W structure. dialysis buffer. Protein concentration was determined by For this crystal structure, which is highly asymmetric with Bradford assay.70 For the St-mutase construct, the N- regard to its conformer of domain 4, a dimer with two fi terminal His6 af nity tag was removed following open monomers was generated by superimposing mono- TM – purification using the Thrombin CleanCleave Kit mer A (the open conformer) on residues 1 370 of 77 (Sigma-Adrich), to ensure that it did not interfere with monomer B (the closed conformer) using Coot. Corre- 78 the formation of a potential octamer (see the text). lation matrices were calculated with ProDy using the vector output of elNémo for mode 7 (lowest‐frequency normal mode). Vectors in Fig. 4 were made using modevector.py script from the PyMOLWiki website.

† http://www.ebi.ac.uk/msd-srv/prot_int/pistart. html ‡ http://www.bioinformatics.sussex.ac.uk/protorp § http://igs-server.cnrs-mrs.fr/elnemo/ Author's personal copy

844 Conservation of Global Motions in Enzymes

Acknowledgements 6. Wu, S., de Lencastre, H., Sali, A. & Tomasz, A. (1996). A phosphoglucomutase-like gene essential for the optimal expression of methicillin resistance in Staph- This work was supported by a grant from the ylococcus aureus: molecular cloning and DNA sequenc- National Science Foundation (MCB-0918389) to ing. Microb. Drug Resist. 2, 277–286. L.J.B and an MU-Howard Hughes Medical Institute 7. Kim, B. G., Sung, S. H. & Ahn, J. H. (2012). Biological 3 C undergraduate research fellowship to E.K.L. We synthesis of quercetin 3-O-N-acetylglucosamine con- thank Yingying Lee, Charles Jones, and Alex jugate using engineered Escherichia coli expressing Hopkins for assistance with protein expression and UGT78D2. Appl. Microbiol. Biotechnol. 93, 2447–2453. puriﬁcation. The Center for Structural Genomics of 8. Nic Lochlainn, L. & Caffrey, P. (2009). Phosphoman- Infectious Diseases has been funded in whole or in nose isomerase and phosphomannomutase gene part with Federal funds from the National Institute disruptions in Streptomyces nodosus: impact on ampho- of Allergy and Infectious Diseases, National In- tericin biosynthesis and implications for glycosylation engineering. Metab. Eng. 11,40–47. stitutes of Health (http://www3.niaid.nih.gov/ 9. Rodriguez-Diaz, J., Rubio-Del-Campo, A. & Yebra, research/resources/sg/), Department of Health M. J. (2012). Metabolic engineering of Lactobacillus and Human Services, under Contract No. casei for production of UDP-N-acetylglucosamine. HHSN272200700058C. We thank Kevin Dyer for Biotechnol. Bioeng. 109, 1704–1712. collecting the SAXS data through the SIBYLS mail-in 10. Sanfelix-Haywood, N., Coll-Marques, J. M. & program. X-ray scattering and diffraction technolo- Yebra, M. J. (2011). Role of alpha-phosphogluco- gies and their applications to the determination of mutase and phosphoglucose isomerase activities at macromolecular shapes and conformations at the the branching point between sugar catabolism and anabolism in Lactobacillus casei. J. Appl. Microbiol. SIBYLS beamline at the Advanced Light Source, 111 – Lawrence Berkeley National Laboratory, are sup- , 433 442. 11. Naught, L. E. & Tipton, P. A. (2001). Kinetic ported in part by the DOE program Integrated mechanism and pH dependence of the kinetic Diffraction Analysis Technologies under Contract parameters of Pseudomonas aeruginosa phosphoman- Number DE-AC02-05CH11231 with the U.S. nomutase/phosphoglucomutase. Arch. Biochem. Bio- Department of Energy. phys. 396, 111–118. 12. Naught, L. E. & Tipton, P. A. (2005). Formation and reorientation of glucose 1,6-bisphosphate in the PMM/PGM reaction: transient-state kinetic studies. Supplementary Data Biochemistry, 44, 6831–6836. 13. Ray, W. J., Jr, Burgner, J. W., 2nd, Deng, H. & Supplementary data to this article can be found Callender, R. (1993). Internal chemical bonding in solutions of simple phosphates and vanadates. Bio- online at http://dx.doi.org/10.1016/j.jmb.2012.08.013 chemistry, 32, 12977–12983. 14. Ray, W. J., Jr, Burgner, J. W., 2nd & Post, C. B. (1990). Characterization of vanadate-based transition-state- References analogue complexes of phosphoglucomutase by spectral and NMR techniques. Biochemistry, 29, 1. Shackelford, G. S., Regni, C. A. & Beamer, L. J. 2770–2778. (2004). Evolutionary trace analysis of the alpha‐D‐ 15. Ray, W. J., Jr, Post, C. B. & Puvathingal, J. M. (1989). phosphohexomutase superfamily. Protein Sci. 13, Comparison of rate constants for (PO3-) transfer by 2130–2138. the Mg(II), Cd(II), and Li(I) forms of phosphogluco- 2. Whitehouse, D. B., Tomkins, J., Lovegrove, J. U., mutase. Biochemistry, 28, 559–569. Hopkinson, D. A. & McMillan, W. O. (1998). A 16. Regni, C., Naught, L. E., Tipton, P. A. & Beamer, L. J. phylogenetic approach to the identification of phos- (2004). Structural basis of diverse substrate recogni- phoglucomutase genes. Mol. Biol. Evol. 15, 456–462. tion by the enzyme PMM/PGM from P. aeruginosa. 3. Barreteau, H., Kovac, A., Boniface, A., Sova, M., Structure, 12,55–63. Gobec, S. & Blanot, D. (2008). Cytoplasmic steps of 17. Regni, C., Schramm, A. M. & Beamer, L. J. (2006). The peptidoglycan biosynthesis. FEMS Microbiol. Rev. 32, reaction of phosphohexomutase from Pseudomonas 168–207. aeruginosa: structural insights into a simple processive 4. Felek, S., Muszynski, A., Carlson, R. W., Tsang, T. M., enzyme. J. Biol. Chem. 281, 15564–15571. Hinnebusch, B. J. & Krukonis, E. S. (2010). Phospho- 18. Regni, C., Shackelford, G. S. & Beamer, L. J. (2006). glucomutase of Yersinia pestis is required for autoag- Complexes of the enzyme phosphomannomutase/- gregation and polymyxin B resistance. Infect. Immun. phosphoglucomutase with a slow substrate and an 78, 1163–1175. inhibitor. Acta Crystallogr., Sect. F: Struct. Biol. Cryst. 5. Jolly, L., Wu, S., van Heijenoort, J., de Lencastre, H., Commun. 62, 722–726. Mengin-Lecreulx, D. & Tomasz, A. (1997). The 19. Schramm, A. M., Mehra-Chaudhary, R., Furdui, C. M. femR315 gene from Staphylococcus aureus, the inter- & Beamer, L. J. (2008). Backbone flexibility, confor- ruption of which results in reduced methicillin mational change, and catalysis in a phosphohexomu- resistance, encodes a phosphoglucosamine mutase. tase from Pseudomonas aeruginosa. Biochemistry, 47, J. Bacteriol. 179, 5321–5325. 9154–9162. Author's personal copy

Conservation of Global Motions in Enzymes 845

20. Mehra-Chaudhary, R., Mick, J. & Beamer, L. J. (2011). 34. Ray, W. J., Jr, Mildvan, A. S. & Long, J. W. (1973). Crystal structure of Bacillus anthracis phosphogluco- Arrangement of the phosphate- and metal-binding samine mutase, an enzyme in the peptidoglycan subsites of phosphoglucomutase. Intersubsite rela- biosynthetic pathway. J. Bacteriol. 193, 4081–4087. tionships by means of inhibition patterns. Biochemis- 21. Mehra-Chaudhary, R., Mick, J., Tanner, J. J., Henzl, try, 12, 3724–3732. M. T. & Beamer, L. J. (2011). Crystal structure of a 35. Pang, H., Koda, Y., Soejima, M. & Kimura, H. (2002). bacterial phosphoglucomutase, an enzyme involved Identification of human phosphoglucomutase 3 in the virulence of multiple human pathogens. (PGM3) as N-acetylglucosamine-phosphate mutase Proteins, 79, 1215–1229. (AGM1). Ann. Hum. Genet. 66, 139–144. 22. Nishitani, Y., Maruyama, D., Nonaka, T., Kita, A., 36. Akutsu, J., Zhang, Z., Tsujimura, M., Sasaki, M., Fukami, T. A., Mio, T. et al. (2006). Crystal structures Yohda, M. & Kawarabayasi, Y. (2005). Characteriza- of N-acetylglucosamine-phosphate mutase, a member tion of a thermostable enzyme with phosphomanno- of the alpha‐D‐phosphohexomutase superfamily, and mutase/phosphoglucomutase activities from the its substrate and product complexes. J. Biol. Chem. 281, hyperthermophilic archaeon Pyrococcus horikoshii 19740–19747. OT3. J. Biochem. (Tokyo), 138, 159–166. 23. Regni, C., Tipton, P. A. & Beamer, L. J. (2002). Crystal 37. Namboori, S. C. & Graham, D. E. (2008). Acetamido structure of PMM/PGM: an enzyme in the biosyn- sugar biosynthesis in the Euryarchaea. J. Bacteriol. 190, thetic pathway of P. aeruginosa virulence factors. 2987–2996. Structure, 10, 269–279. 38. Rashid, N., Kanai, T., Atomi, H. & Imanaka, T. (2004). 24. Dai, J. B., Liu, Y., Ray, W. J., Jr & Konno, M. (1992). The Among multiple phosphomannomutase gene ortho- crystal structure of muscle phosphoglucomutase logues, only one gene encodes a protein with refined at 2.7-angstrom resolution. J. Biol. Chem. 267, phosphoglucomutase and phosphomannomutase ac- 6322–6337. tivities in Thermococcus kodakaraensis. J. Bacteriol. 186, 25. Mehra-Chaudhary, R., Mick, J., Tanner, J. J. & Beamer, 6070–6076. L. J. (2011). Quaternary structure, conformational 39. Krissinel, E. & Henrick, K. (2007). Inference of variability and global motions of phosphoglucosa- macromolecular assemblies from crystalline state. mine mutase. FEBS J. 278, 3298–3307. J. Mol. Biol. 372, 774–797. 26. Muller, S., Diederichs, K., Breed, J., Kissmehl, R., 40. Venkatakrishnan, A. J., Levy, E. D. & Teichmann, S. A. Hauser, K., Plattner, H. & Welte, W. (2002). Crystal (2010). Homomeric protein complexes: evolution and structure analysis of the exocytosis-sensitive phos- assembly. Biochem. Soc. Trans. 38, 879–882. phoprotein, pp 63/parafusin (phosphoglucomutase), 41. Joshi, J. G. & Lane, R. (1978). Rabbit muscle from Paramecium reveals significant conformational phosphoglucomutase is a monomer. Biochem. Biophys. variability. J. Mol. Biol. 315, 141–153. Res. Commun. 85, 729–736. 27. Maguid, S., Fernandez-Alberti, S. & Echave, J. (2008). 42. Suhre, K. & Sanejouand, Y. H. (2004). ElNemo: a Evolutionary conservation of protein vibrational normal mode web server for protein movement dynamics. Gene, 422,7–13. analysis and the generation of templates for molecular 28. Maliekal, P., Sokolova, T., Vertommen, D., Veiga-da- replacement. Nucleic Acids Res. 32, W610–W614. Cunha, M. & Van Schaftingen, E. (2007). Molecular 43. Bahar, I. & Rader, A. J. (2005). Coarse-grained normal identification of mammalian phosphopentomutase mode analysis in structural biology. Curr. Opin. Struct. and glucose-1,6-bisphosphate synthase, two members Biol. 15, 586–592. of the alpha‐D‐phosphohexomutase family. J. Biol. 44. Ma, J. (2005). Usefulness and limitations of normal Chem. 282, 31844–31851. mode analysis in modeling dynamics of biomolecular 29. Naught, L. E., Regni, C., Beamer, L. J. & Tipton, P. A. complexes. Structure, 13, 373–380. (2003). Roles of active site residues in P. aeruginosa 45. Tama, F. & Brooks, C. L. (2006). Symmetry, form, and phosphomannomutase/phosphoglucomutase. Bio- shape: guiding principles for robustness in macromo- chemistry, 42, 9946–9951. lecular machines. Annu. Rev. Biophys. Biomol. Struct. 30. Sarma, A. V., Anbanandam, A., Kelm, A., Mehra- 35, 115–133. Chaudhary, R., Wei, Y., Qin, P. et al. (2012). Solution 46. Jana, B., Adkar, B. V., Biswas, R. & Bagchi, B. (2011). NMR of a 463-residue phosphohexomutase: domain 4 Dynamic coupling between the LID and NMP domain mobility, substates, and phosphoryl transfer defect. motions in the catalytic conversion of ATP and AMP to Biochemistry, 51, 807–819. ADP by adenylate kinase. J. Chem. Phys. 134, 035101. 31. Schramm, A. M., Karr, D., Mehra-Chaudhary, R., Van 47. Merlino, A., Vitagliano, L., Ceruso, M. A. & Mazzarella, Doren, S. R., Furdui, C. M. & Beamer, L. J. (2010). L. (2004). Dynamic properties of the N-terminal Breaking the covalent connection: chain connectivity swapped dimer of ribonuclease A. Biophys. J. 86, and the catalytic reaction of PMM/PGM. Protein Sci. 2383–2391. 19, 1235–1242. 48. Krebs, W. G., Alexandrov, V., Wilson, C. A., Echols, 32. Ray, W. J., Jr & Long, J. W. (1976). The thermodynamic N., Yu, H. & Gerstein, M. (2002). Normal mode and structural differences among the catalytically analysis of macromolecular motions in a database active complexes of phosphoglucomutase: metal ion framework: developing mode concentration as a effects. Biochemistry, 15, 4018–4025. useful classifying statistic. Proteins, 48, 682–695. 33. Ray, W. J., Jr & Long, J. W. (1976). Thermodynamics 49. Koike, R., Amemiya, T., Ota, M. & Kidera, A. (2008). and mechanism of the PO3 transfer process in the Protein structural change upon ligand binding corre- phosphoglucomutase reaction. Biochemistry, 15, lates with enzymatic reaction mechanism. J. Mol. Biol. 3993–4006. 379, 397–401. Author's personal copy

846 Conservation of Global Motions in Enzymes

50. Levy, E. D., Boeri Erba, E., Robinson, C. V. & 65. Maguid, S., Fernandez-Alberti, S., Ferrelli, L. & Teichmann, S. A. (2008). Assembly reflects evolution Echave, J. (2005). Exploring the common dynamics of protein complexes. Nature, 453, 1262–1265. of homologous proteins. Application to the globin 51. Goodsell, D. S. & Olson, A. J. (2000). Structural family. Biophys. J. 89,3–13. symmetry and protein function. Annu. Rev. Biophys. 66. Keskin, O., Durell, S. R., Bahar, I., Jernigan, R. L. & Biomol. Struct. 29, 105–153. Covell, D. G. (2002). Relating molecular flexibility to 52. Grabarse, W., Vaupel, M., Vorholt, J. A., Shima, S., function: a case study of tubulin. Biophys. J. 83, Thauer, R. K., Wittershagen, A. et al. (1999). The crystal 663–680. structure of methenyltetrahydromethanopterin cyclo- 67. Zheng, W., Brooks, B. R. & Thirumalai, D. (2007). hydrolase from the hyperthermophilic archaeon Allosteric transitions in the chaperonin GroEL are Methanopyrus kandleri. Structure, 7, 1257–1268. captured by a dominant normal mode that is most 53. Hess, D., Kruger, K., Knappik, A., Palm, P. & Hensel, R. robust to sequence variations. Biophys. J. 93,2289–2299. (1995). Dimeric 3-phosphoglycerate kinases from hy- 68. Reynolds, C., Damerell, D. & Jones, S. (2009). ProtorP: perthermophilic Archaea. Cloning, sequencing and a protein–protein interaction analysis server. Bioinfor- expression of the 3-phosphoglycerate kinase gene of matics, 25, 413–414. Pyrococcus woesei in Escherichia coli and characterization 69. DeLano, W. L. (2002). The PyMOL Molecular Graphics of the protein. Structural and functional comparison System. DeLano Scientific, San Carlos, CA. with the 3-phosphoglycerate kinase of Methanothermus 70. Bradford, M. M. (1976). A rapid and sensitive method fervidus. Eur. J. Biochem./FEBS, 233,227–237. for the quantitation of microgram quantities of protein 54. Villeret, V., Clantin, B., Tricot, C., Legrain, C., utilizing the principle of protein-dye binding. Anal. Roovers, M., Stalon, V. et al. (1998). The crystal Biochem. 72, 248–254. structure of Pyrococcus furiosus ornithine carbamoyl- 71. Hura, G. L., Menon, A. L., Hammel, M., Rambo, R. P., transferase reveals a key role for oligomerization in Poole, F. L., 2nd, Tsutakawa, S. E. et al. (2009). Robust, enzyme stability at extremely high temperatures. Proc. high-throughput solution structural analyses by small Natl Acad. Sci. USA, 95, 2801–2806. angle X-ray scattering (SAXS). Nat. Methods, 6, 55. Walden, H., Bell, G. S., Russell, R. J., Siebers, B., 606–612. Hensel, R. & Taylor, G. L. (2001). Tiny TIM: a small, 72. Konarev, P. V., Volkov, V. V., Sokolova, A. V., Koch, tetrameric, hyperthermostable triosephosphate isom- M. H. J. & Svergun, D. I. (2003). PRIMUS: a Windows erase. J. Mol. Biol. 306, 745–757. PC-system for small angle scattering data analysis. 56. Kumar, S., Tsai, C. J. & Nussinov, R. (2000). Factors J. Appl. Crystallogr. 36, 1277–1282. enhancing protein thermostability. Protein Eng. 13, 73. Svergun, D. I. (1992). Determination of the regulari- 179–191. zation parameter in indirect-transform methods using 57. Baldasseroni, F. & Pascarella, S. (2009). Subunit perceptual criteria. J. Appl. Crystallogr. 25, 495–503. interfaces of oligomeric hyperthermophilic enzymes 74. Liu, H., Hexemer, A. & Zwart, P. H. (2012). The Small display enhanced compactness. Int. J. Biol. Macromol. Angle Scattering ToolBox (SASTBX): an open-source 44, 353–360. software for biomolecular small-angle scattering. 58. Maugini, E., Tronelli, D., Bossa, F. & Pascarella, S. J. Appl. Crystallogr. 45, 587–593. (2009). Structural adaptation of the subunit interface 75. Kleywegt, G. J. (1997). Validation of protein models from of oligomeric thermophilic and hyperthermophilic Calpha coordinates alone. J. Mol. Biol. 273,371–376. enzymes. Comput. Biol. Chem. 33, 137–148. 76. Schneidman-Duhovny, D., Hammel, M. & Sali, A. 59. Lu, M. & Ma, J. (2005). The role of shape in determining (2010). FoXS: a web server for rapid computation and molecular motions. Biophys. J. 89, 2395–2401. fitting of SAXS profiles. Nucleic Acids Res. 38(Suppl.), 60. Keskin, O., Jernigan, R. L. & Bahar, I. (2000). Proteins W540–W544. with similar architecture exhibit similar large-scale 77. Emsley, P. & Cowtan, K. (2004). Coot: model-building dynamic behavior. Biophys. J. 78, 2093–2106. tools for molecular graphics. Acta Crystallogr., Sect. D: 61. Leo-Macias, A., Lopez-Romero, P., Lupyan, D., Biol. Crystallogr. 60, 2126–2132. Zerbino, D. & Ortiz, A. R. (2005). An analysis of 78. Bakan, A., Meireles, L. M. & Bahar, I. (2011). ProDy: core deformations in protein superfamilies. Biophys. J. protein dynamics inferred from theory and experi- 88, 1291–1299. ments. Bioinformatics, 27, 1575–1577. 62. Echave, J. & Fernandez, F. M. (2010). A perturbative 79. Pei, J., Kim, B. H. & Grishin, N. V. (2008). view of protein structural variation. Proteins, 78, 173–180. PROMALS3D: a tool for multiple protein sequence 63. Laberge, M. & Yonetani, T. (2007). Common dynamics and structure alignments. Nucleic Acids Res. 36, of globin family proteins. IUBMB Life, 59, 528–534. 2295–2300. 64. Leo-Macias, A., Lopez-Romero, P., Lupyan, D., 80. Waterhouse, A. M., Procter, J. B., Martin, D. M., Zerbino, D. & Ortiz, A. R. (2005). Core deformations Clamp, M. & Barton, G. J. (2009). Jalview Version 2—a in protein families: a physical perspective. Biophys. multiple sequence alignment editor and analysis Chem. 115, 125–128. workbench. Bioinformatics, 25, 1189–1191.