Crystal structure of phosphoglucose isomerase from pig muscle and its complex with 5-phosphoarabinonate


  • Christopher Davies,

    Corresponding author
    1. Department of Biochemistry and Molecular Biology, Medical University of South Carolina, Charleston, South Carolina
    • Department of Biochemistry and Molecular Biology, Medical University of South Carolina, 173 Ashley Avenue, Charleston, SC 29425
    Search for more papers by this author
  • Hilary Muirhead

    1. Department of Biochemistry, School of Medical Sciences, University of Bristol, Bristol, United Kingdom
    Search for more papers by this author


Phosphoglucose isomerase (PGI) (EC is a key enzyme in glycolysis and gluconeogenesis that catalyzes the interconversion of glucose 6-phosphate (G6P) and fructose 6-phosphate (F6P) by transfer of a carbon-bound hydrogen between C1 and C2. The mechanism is believed to be a proton transfer via a cis-enediol intermediate. Crystallographic structures of the enzyme from rabbit 1 (Davies and Muirhead, unpublished) and human2 suggest that Glu-357 is the base responsible for this transfer. Other aspects of the reaction mechanism, however, remain sketchy, including the identity of the group responsible for catalyzing ring opening.

Interest in PGI has increased in recent years after a number of discoveries linking it to various cytokine factors [e.g., autocrine motility factor (AMF)].3 However, several inconsistencies remain; the most notable is how the established dimeric structure of the enzyme is related to its cytokine form, which appears to be monomeric.

Here we present the structure of PGI from pig muscle solved at 2.5 Å resolution together with a 3.5 Å structure of the enzyme bound to the transition-state mimic, 5-phosphoarabinonate (PAB). A comparison of these structures reveals conformational changes in the enzyme that occur during catalysis.

Materials and Methods.

PGI from pig muscle crystallizes from ammonium sulfate in space group P43212 with one monomer in the asymmetric unit.4 The native data used in this study were collected at the SRS, Daresbury, using film methods. Data were collected at room temperature from a total of seven crystals and were processed with MOSFLM (CCP4). These data are 75.7% complete to 2.18 Å (Table I). Because most of the absent data are at the high-resolution end, only data to 2.5 Å were used for refinement. The structure of pig PGI was solved originally at 3.5 Å resolution,5 but refinement at higher resolution was impeded by a lack of quality phases. With a high-resolution structure of the rabbit enzyme in hand (Davies and Muirhead, unpublished), a new pig structure was generated by applying a transformation to one monomer of the dimeric rabbit structure. The sequence was then changed to correspond to that of pig, and the structure was refined at 2.5 Å resolution, initially using XPLOR but later with REFMAC. The final model is composed of 554 residues, 141 water molecules, and 1 sulfate molecule (PDB code 1GZD). Residues 555–557 are not visible in the electron density map and are excluded from the model. The crystallographic R and free R factors are 17.5% and 24.0%, respectively, with good geometry (Table I). 88.8% of the residues lie in the most favored region of the Ramachandran plot with no outliers.

Table I. X-ray diffraction data and model refinement statistics
Resolution range (Å)31–2.1917.7–3.5
Rmerge (%)7.34.5
Completeness (%)75.792.8
Resolution range for refinement (Å)15.0–2.515.0–3.5
% reflections used for Free R set10.05.0
No. of protein atoms44224422
No. of hetero atoms515
No. of waters1410
R factor (R work + R free)(%)17.519.3
R free (%)24.125.0
Overall Biso2)17.242.1
R.M.S deviation for bond length (Å)0.0180.019
R.M.S deviation for angles (Å)1.811.98

Data were collected previously from crystals of pig PGI that had been crystallized in the presence of PAB6 extending to 3.5 Å resolution. To model the inhibitor-bound enzyme, the pig structure (without the water and the sulfate molecules) was refined using XPLOR and REFMAC. Manual adjustments were made to the model by using the O program. The final model, which contains residues 1–555 and one molecule of PAB, has a crystallographic R factor of 19.3% (Rfree = 24.9%) with good geometry (PDB code 1GZV)(Table I). 88.1% of the residues lie in the most favored region of the Ramachandran plot with no outliers.

Results and Discussion.

Structural studies of PGI from pig muscle first began 30 years ago, culminating in a polyalanine structure at 3.5 Å resolution.5 When the sequence for PGI became available, refinement of this structure at higher resolution stalled, largely because of the lack of quality phases. The problem was resolved by determining a whole new structure, that of rabbit PGI, by heavy atom methods (Davies and Muirhead, unpublished) and using this to correct model errors in the pig structure. Because the original 3.5 Å structure was solved in advance of any sequence information for the enzyme,5 this model was incomplete and contained several connectivity errors. These were corrected in the current structure by reversing the directions of five α-helices and one β-strand: A9, A14, A15, A16, A17, and B6 (1977 nomenclature). In doing so, the unusual left-handed connection between B4 and B5 no longer exists. Several new elements of secondary structure were identified, including an additional β-strand and α-helix in the small domain, which accounts for most of the 43 residues missing in the 3.5 Å model.

Pig PGI crystallizes with one monomer in the asymmetric unit, whereas the native form of the enzyme is dimeric. Applying a symmetry operation can generate the dimer, and this is prerequisite for viewing either active-site region, each of which is comprised of chains from both monomers. This arrangement agrees with biochemical data showing that the dimer is required for enzymatic activity7 but is at odds with data suggesting that AMF exhibits PGI enzyme activity and yet appears to exist as a monomer.3

As expected, the structure of porcine PGI is highly similar to those of PGIs from other mammalian sources. It is most highly similar to that of human PGI2: the root mean square in all main-chain atoms between these structures is 0.4 Å. A molecule of sulfate from the crystallization solution is observed in the active site of pig PGI, which appears to mimic the phosphate moiety of the glucose 6-phosphate or fructose 6-phosphate substrates. A similar sulfate was also observed in the structure of human PGI,2 and when this structure was compared with that of native rabbit PGI, some structural differences were noted in the small domain that were attributed to conformational changes induced by sulfate binding. It was suggested that similar changes would occur when the enzyme recognizes the phosphate moiety of the substrate. The fact that pig PGI also contains bound sulfate and also adopts the same conformation in this region supports this hypothesis.

The new structure for pig PGI was used to reexamine 3.5 Å data collected from crystals soaked in 5-phosphoarabinonate (PAB).6 This is a competitive inhibitor of PGI (Ki = 2.7 × 10−7) that mimics the cis-enediolate intermediate. At that time, difference Fouriers established the location of the active site, but without a complete model, little could be inferred about the identity of active-site residues. In the new structure, the enzyme-inhibitor contacts can be seen clearly including the putative base catalyst, Glu-357, which contacts one of the oxygens of the C1 carboxylate group. Comparison of the native and PAB-bound structures shows that residues 512–520 (inclusive), which comprise the N-terminal half of helix α23, shift significantly toward the active site in the inhibitor-bound structure (Fig. 1). A similar change has also been observed in the crystal structures of rabbit PGI,8 but this is the first time a comparison has been made with a native structure of mammalian PGI. Of particular note are the large shifts of Glu-515 and Lys-518. The new position for Glu-515 equates to a 7 Å movement of its epsilon oxygens toward the active site. Similarly, the zeta nitrogen of Lys-518 has moved 3.1 Å to make contact with both O4 and O5 of the inhibitor. Lys-518 likely plays a significant role in the catalytic mechanism, either by protonating O5 during ring opening or by transferring a proton between the C1 and C2 hydroxyls.1

Figure 1.

The movement of helix α23 on binding the competitive inhibitor, 5-phosphoarabinonate (PAB). Shown is a backbone superimposition of the native (dark gray) and PAB-bound (light gray) PGI structures in the active-site region. Important active-site residues are shown as solid bonds for the native structure and dashed bonds for the PAB-bound structure. The PAB inhibitor is shown in ball-and-stick form. For clarity, most loops in the active site have been removed.


Taken together with previous investigations, these data suggest there are two conformational changes associated with the catalytic activity of PGI. Recognition of the phosphate group of the substrate induces two loops in the small domain to close down onto the active site and, near the C-terminus, α23 also moves toward the active site to bring Lys-518 into contact with the substrate.