Induced structural disorder as a molecular mechanism for enzyme dysfunction in phosphoglucomutase 1 deficiency

Overview of the wild-type structure

The high-resolution structure of human WT PGM1 is reported here for the first time. Previous models of human PGM1 relied on the known structure of the rabbit enzyme, which has been refined to a nominal resolution of 2.4 Å (PDB code: 3PMG) [12]. Our 1.85 Å structure (Table 1 and Materials and Methods) confirms many of the features of this model, but provides additional detail on inter-residue interactions, as well as solvent and sulfate binding sites. We first describe the overall structure of the enzyme, and then the context of the glycines affected by the disease-related mutations G121R and G291R. Although the two copies of the polypeptide chain found in the asymmetric unit are chemically identical, chain A is more highly ordered in the crystals, and is used as the reference in structural descriptions below, unless otherwise noted.

Human PGM1 is a monomeric protein that shares the four-domain architecture common to all enzymes in the α-D-phosphohexomutase superfamily [9]. Its four domains are of roughly equal size, and arranged in an overall heart-shape (Fig. 2a). As noted in the structure of rabbit PGM [13], domains 1–3 share a common mixed α/β core, with a central 4-stranded β-sheet flanked on each side by an α-helix. The active site of PGM1 is located in a large central cleft, formed at the confluence of its four structural domains. This cleft has a molecular surface area of 1670 Ų as calculated by CASTp [14], and involves 64 residues. It is quite hydrophilic, with nearly 70% of residues being charged or polar (including 11 arginines and lysines), as would be expected for binding of the phosphosugar substrates.

Proposed functional roles of four critical loops within the active site of PGM1 have been previously described (for detailed review see [15]). These loops are from domains 1–4 (D1-D4), respectively, and contain: i) phosphoserine 117 that participates in phosphoryl transfer; ii) the metal-binding loop including the three coordinating aspartates (residues 290, 292 and 294); iii) a sugar-binding loop; and iv) the phosphate-binding site that interacts with the phosphate group of the substrates. Fig. 2b shows a close-up view of these regions in the active site of PGM1; their context in the structure is described below.

Context of glycines 121 and 291 in the WT structure

Gly121 is located immediately following loop (i), just after the highly conserved TASHNP motif. In the WT structure, Gly121 is in a partially buried region between two turns, and has ϕ/ψ angles of ~130/175°, which are uncommon except for glycine. Gly121 makes no direct contacts to other residues, but is located at the interface between D1 and D2 of the protein (Fig. 2b), adjacent to a short helix (residues 257–259) and extended region (residues 260–269) in D2.

Gly291 is found within loop (ii), the metal binding loop of PGM1. This loop is deeply buried within the active site cleft, and adjacent to residues from both D1 and D3 (Fig. 2b). Gly291 is the intervening residue between Asp290 and Asp292, both of which coordinate the metal ion. Its ϕ/ψ angles are ~80/10°, values preferred for glycine. Gly291 makes several interactions with other residues, including a hydrogen bond between its backbone amide proton and the side chain of Asp288, and two hydrogen bonds from its backbone carbonyl to the amide of Gly391 and a water-mediated interaction to the backbone amide of Leu393.

Structure of the G121R missense variant

The crystal structure of the G121R missense variant was determined to 2.35 Å (Table 1; Materials and Methods). Fig. 3a,b show an overview of the structure of the mutant, superimposed with that of WT enzyme. While the two structures are quite similar with an overall root-mean-square-deviation (rmsd) of 0.51 Å for 540 C_α pairs), the mutation produces both local and longer-range structural changes in the enzyme. In the vicinity of residue 121, the introduced arginine displaces the side chain of a nearby residue, Phe257, which packs in the interface between D1 and D2 in the WT structure (Fig. 3c). In its place, Arg121 makes adventitious hydrogen bond interactions with the backbone carbonyls of Leu254 and Asp62 (Fig. 3d). These fortuitous interactions appear to compensate for the introduction of the polar, positively charged side chain of the arginine, which would otherwise be unfavorable in this buried environment. Despite the proximity of the G121R mutant to active site loops (i) and (ii), few changes to this region are observed, at least in chain A in the crystal structure. In chain B, which is less well ordered, neither the side chain of Arg121 nor loop (i) is clearly observed.

In addition to this local structural rearrangement, longer-range changes also occur, particularly in the adjacent region of D2. Electron density for the displaced side chain of Phe257 cannot be located in the maps, and subsequent to this residue, density for the entire polypeptide chain disappears and does not clearly reappear until approximately residue 265, presumably due to dynamic disorder in the crystal. A comparison of this area in the G121R and WT enzyme structures is shown in Fig. 3c. Although these residues have limited regular secondary structure in the WT enzyme, a number of well-defined inter-residue interactions involving this region are evident, all of which are lost in the structure of the mutant. The induced structural disorder appears to originate from the steric rearrangement of Phe257 forced by the arginine substitution, and subsequent accommodation of its side chain in the D1-D2 interface. Despite the extended region of disorder in the middle of the sequence, the polypeptide chain eventually returns to its proper course, adopting an essentially WT fold following residue 270. In total, the propagated effects of the G121R substitution extend more than 17 Å from the site of the mutation, with induced disorder of seven amino acids (Table 2).

Structure of the G291R missense variant

The crystal structure of the G291R missense variant was determined to 2.75 Å (Table 1; Materials and Methods). Despite crystals of similar size, diffraction data collected from these crystals were noticeably lower in resolution compared to WT (1.85 Å) and the G121R mutant (2.5 Å). The structure of the G291R missense variant superimposed with that of WT enzyme is shown in Fig. 4a,b. As is the case for G121R, the overall structure of G291R is quite similar to that of WT (C_α rmsd of 0.83 Å for 525 residues). Nevertheless, the single arginine substitution causes widespread structural changes, including induced disorder in three loops near the active site, as well as conformational rearrangements and differences in residue packing (Fig. 4c), as detailed below.

In the G291R structure, loop (ii) shows changes in the conformation of both its polypeptide backbone and side chains. The corresponding areas of the WT and G291R structures are displayed in Fig. 4d,e. The backbone angles of the polypeptide chain of G291R have altered to accommodate the introduced arginine side chain and the less flexible backbone of this residue relative to glycine, moving the loop to a more solvent-exposed position in the active site cleft. Accompanying the backbone changes are significant differences in the orientations of the three metal-coordinating aspartates, with concomitant destruction of the metal-binding site. Aspartates 290 and 292, in particular, reorient dramatically: Asp290 becomes fully solvent exposed, while Asp292 makes a novel hydrogen bond interaction with Thr93 (Fig. 4e). The introduced arginine side chain is apparent in the electron density maps of chain A, where it points directly toward the center of the positively charged active site cleft. This unfavorable electrostatic environment may be mitigated by interactions with a sulfate ion and the high pH of the crystals (see Materials and Methods), which would tend to deprotonate Arg291 and/or other nearby positive residues.

Another apparent outcome of the conformational rearrangement of loop (ii) is induced disorder in the phosphoserine loop (i) of D1. As can be seen by comparing Fig. 4d and e, the native conformation of this loop would be in steric conflict with the rearranged position of loop (ii) in the mutant. Consequently, residues 117–125 are disordered in the electron density maps of the G291R mutant. In an apparent series of propagated changes, several other nearby loops are also disordered, including another in D1 (residues 64–65) and a larger region in D2 (residues 256–267). In total, 23 residues in two domains are affected by induced disorder in the structure of G291R mutant (Table 2), including some that are nearly 20 Å away from the site of the mutation. Other structural effects include changes in the packing of residues, such as Trp359 (Fig. 4c), which is one of several residues near loop (ii) that appear to have adjusted in response to its rearranged conformation. Thus, while the overall structure of the enzyme remains largely intact, changes due to this single residue substitution propagate across the protein, affecting multiple residues and domains.

Correlation between structural and biochemical data

The crystal structures of the G121R and G291R missense variants provide a valuable opportunity for understanding biochemical data on these disease-related mutants [11]. Previous work on G121R showed that this protein is well expressed in E. coli, soluble, and its apparent molecular weight in solution is similar to WT enzyme. Likewise, its near-UV circular dichroism spectrum and T_m (only 1° lower) are quite similar to those of WT, although its interaction with the fluorescent dye 1-anilinonaphthalene-8-sulfonic acid (ANS) shows a moderate increase. Its catalytic activity is nevertheless significantly compromised, with k_cat / K_m only 0.4% of WT (Table 2), although Ser117 in the mutant protein can still be phosphorylated by the activator glucose 1,6-bisphosphate. The structural studies of this variant, showing various perturbations near the active site cleft, are consistent with its catalytic impairment, which is significant but not complete. At least in chain A of the crystal structure, the position of Ser117 and the metal-binding loop are quite similar to WT, implying the mutant should be competent for catalysis. Disorder of loop (i) in chain B of the crystals, however, suggests that the mutant protein may exist as an inactive conformer more frequently than the WT enzyme. Together with other effects, including many small changes in/near the active site environment due to the disordered regions, it is not difficult to rationalize an overall decrease in the catalytic efficiency of G121R.

In the case of G291R, the structural disruption is more extensive, and this is reflected in the extreme catalytic defect of this mutant, which is the most impaired of all the PGM1 missense variants characterized to date (activity below detection in in vitro kinetic assays; Table 2) [11]. The inactivity of G291R is consistent with structural effects of the mutation on key loops within the active site, perturbing both metal binding and phosphoryl transfer, two essential features of the reaction. Defects in these processes are also reflected by the inability of the G291R variant to be phosphorylated under the conditions tested. While not directly linked to catalysis, the disordered loop containing residues 64–65 disrupts a conserved salt-bridge between Asp62 and Arg64, which in turn makes multiple interactions with loop (i) in the WT structure. Absence of these anchoring interactions with the critical phosphoserine loop would be a further impediment to phosphoryl transfer by this mutant protein.

The structural effects of the G291R missense variant are also interesting to consider in light of other biochemical data on this protein [11]. The mutant is generally well expressed in E. coli and soluble, although not quite at WT levels. In solution, the purified protein is well behaved with a similar apparent molecular weight to WT enzyme, while its T_m is a bit lower (~3°). Notably, however, its binding to ANS is significantly increased, and the G291R mutant is also more susceptible to limited proteolysis. These results make sense based on the observed changes in the crystal structure, particularly the regions of induced disorder, which could be easily proteolyzed. Similarly, the perturbation of the 3-way domain interface in which this residue is involved, and concomitant changes in residue packing (Fig. 4c), can be easily envisioned to allow enhanced binding of ANS, which typically occurs in apolar crevices in protein structures [19]. In the future, it may be informative to conduct studies such as ANS-binding and limited proteolysis in the presence and absence of ligand, to determine whether ligand binding can mitigate the disordered regions of structure in mutants such as G121R and G291R.

Missense variants of PGM1 cluster in domain-domain interfaces

In light of the multi-domain effects of the G121R and G291R mutants, we analyzed other disease-related missense variants of PGM1 relative to the interdomain interfaces of the protein. The conserved four-domain architecture of PGM1 and the location of the active site cleft at the juncture of these structural domains (Fig. 2A) imply that mutations in these regions could be disproportionately disruptive to enzyme function. Moreover, in addition to the 13 originally characterized [11], several new missense mutants of PGM1 associated with disease have recently been identified [6]. These include G230E, T337M, P336R, R422W, R503Q, and R515L (Fig. 5). It can be seen that these new variants expand the structural regions associated with enzyme dysfunction: while many are near the active site cleft, others, such as G230E and R422W, are more distant from known regions of functional importance.

To assess the potential correlation between domain interfaces and the pathogenicity of missense variants, residues in the structure of WT PGM1 involved in each possible domain-domain interface were identified using LigPlot+ [20] and are detailed in Supplementary Fig. S1. Of the six potential interfaces, four have contacting residues in the crystal structure: D1-D2, D1-D3, D2-D3, and D3-D4. Each interface includes at least one known disease-related mutation, although most are in either the D1-D2 or D3-D4 interface (see following paragraph). Notably, only one of the known missense variants resides in two distinct domain interfaces: G291R, which contacts residues in D1, D2, and D3 (see also Fig. 4c). Although not affected by known mutations, 13 other residues in PGM1 are also involved in 3-way domain interactions (Supplementary Fig. S1) and might also be expected to be especially sensitive to structural perturbation by mutation.

As many of the known missense mutations localize to the D1-D2 and D3-D4 interfaces, it is convenient to consider domains 2 and 3 (residues 192–421) as a single structural unit (D2/3), and the interface between this unit and the rest of the enzyme as a single, extended surface (blue in Fig. 5). This large interface contains 101 residues out of 562 in the protein (18%), including much of the active site cleft, but also extends to include regions between D1-D2 and D3-D4 that are relatively far from the catalytic center. Using this convention, it can be seen that the interface with the D2/3 unit encompasses 11 of the 18 residues affected by missense variants of PGM1, while four others are immediately adjacent (a combined total of 83%). Thus it appears that mutations in this interface account for a striking proportion of the currently recognized disease-related variants of PGM1. It remains to be seen whether this distribution will change as new missense mutants are identified.

In addition to G121R and G291R, these interdomain mutations include many previously categorized as deleterious to catalysis (D62H, T115A, G121R, G230E, D263G/Y, G291R, T337M, and R503Q) [6,11]. However, several also appear to cause folding problems (P336R, E377K, E388K, and R422W) [6,11]. This suggests that a gradation of effects can occur due to substitutions at this interdomain interface, with the functional outcome dependent on factors such as the physicochemical change of the mutation and the structural interactions of the WT residue. It should be noted that several missense variants of PGM1 fall well outside of the D2/3 interface, suggesting the presence of additional “residue clusters” (e.g., residues 38 and 41, Fig. 5) with distinct functional roles, perhaps in protein folding. Mutants affecting residue 38 and 41 also tend to have milder clinical phenotypes than those affecting the interface residues [6], although the heterozygous backgrounds and limited number of patients prohibits a definitive analysis.

Protein expression, purification, and crystallization

Human WT PGM1 and the G121R and G291R missense variants were expressed recombinantly in E. coli and purified to homogeneity via an N-terminal histidine tag, as previously described [11]. Prior to crystallization, the histidine tag was removed by cleavage with tobacco etch virus (TEV) protease as follows. Proteins were dialyzed into Buffer 1 (0.3 M NaCl and 50 mM Tris, pH 8.0), then mixed at 10–20 fold ratio (w/w) with TEV protease (Buffer 1 with 0.5 mM DTT and 0.5 mM EDTA) and glycerol (10% v/v final). The mixture was incubated at room temperature for 3–4 hours, kept overnight at 4° C, and extensively dialyzed into Buffer 1 before loading onto a pre-equilibrated Ni-NTA column. Cleaved protein eluted in the flow through; TEV and any residual His-tagged proteins were retained on the column, and eluted with Buffer 1 plus 300 mM imidazole, pH 8.0. Successful cleavage was confirmed by electrospray ionization mass spectrometry; all proteins were the expected molecular weight with no indication of proteolysis. The purified, cleaved proteins were dialyzed into a solution of 50 mM MOPS, pH 7.4, with 1 mM MgCl₂, and concentrated to ~10 mg / mL or higher. If not used immediately, samples were flash-frozen in liquid nitrogen and stored at −80° C. Prior to crystallization, the protein was diluted as desired and 0.22 μm filtered.

Initial crystallization screens were set up with Crystal Screen kits 1 and 2 (Hampton Research) and Wizard screen kits 1 and 2 (Emerald BioSystems Inc.) at protein concentrations ranging from 10–18 mg / mL using the hanging drop vapor diffusion method at 20° C. Drops containing 2 μl protein solution and 2 μl crystallization buffer were sealed over a 0.5 mL reservoir. The WT protein crystallized successfully in a range of conditions from ammonium or lithium sulfate (1.35 – 1.85 M) with either 0.1 M buffer of MES, pH 6.0, or Tris HCl, pH 7.5 – 8.5. Crystals grew in approximately one week. The G121R mutant crystallized in ~1.5 M ammonium sulfate with lithium sulfate (0.12–0.18 M) and 0.1 M CAPS, pH 10.5. The G291R mutant crystallized in ~1.0 M sodium citrate with 0.1 M CHES buffer, pH 9.7.

All crystals were cryoprotected using a solution of well buffer supplemented with 30% glycerol (v/v), mounted on Hampton loops, and flash cooled in liquid nitrogen. WT and mutant proteins crystallized in space group P4₁2₁2 with ~60% solvent (V_M = 3.04 ų/Da). The asymmetric unit contains two copies of the polypeptide chain.

X-ray Diffraction Data Collection and Refinement

Diffraction data were collected at a wavelength of 1.00003 Å from single crystals on beamline 4.2.2 of the Advanced Light Source using a Taurus-1 CMOS detector in shutterless mode. The data were processed using XDS [40] and AIMLESS [41] via CCP4i [42]. Data processing statistics are listed in Table 1. Values of CC_1/2 > 0.30 [43] and R_pim [44] were considered when determining the high resolution cutoff, due to the large number of images (1800–3600 per data set) and high redundancy obtained with the shutterless data collection. In addition, all data sets were anisotropic, especially those of G121R and G291R, which contributes to poor R_merge statistics at high resolution for these two data sets, although CC_1/2 is acceptable in the chosen range.

Crystallographic refinement calculations for the WT enzyme were initiated using coordinates derived from the 2.4 Å resolution structure of rabbit PGM (PDB code: 3PMG), which crystallizes isomorphously with human PGM1 (97% sequence identity). Refinement was performed with REMAC 5.0 [45] and PHENIX [46]; progress was monitored by following R _free with 5% of each data set was set aside for cross validation. The B-factor model consisted of an isotropic B-factor for each atom; TLS refinement was used as automated in PHENIX. COOT [47] was used for model building. Additional details of the refinement are on Table 1. The structures of the G121R and G291R missense variants were determined using the model of WT enzyme, starting with molecular replacement using MOLREP [48] or rigid-body followed by restrained refinement. The R_free data sets for both mutants were constrained to match those of the WT data. The structures were validated using MolProbity [49]. Refinement statistics are listed in Table 1. Structural figures were prepared with PYMOL [50]. Coordinates and structure factor amplitudes have been deposited in the PDB under the accession numbers listed in Table 1.