Crystal structure of the Streptococcus pneumoniae phosphomevalonate kinase, a member of the GHMP kinase superfamily


  • Michael J. Romanowski,

    1. Laboratories of Molecular Biophysics, The Rockefeller University, New York, New York
    Search for more papers by this author
  • Jeffrey B. Bonanno,

    1. Laboratories of Molecular Biophysics, The Rockefeller University, New York, New York
    2. Howard Hughes Medical Institute, The Rockefeller University, New York, New York
    Search for more papers by this author
  • Stephen K. Burley

    Corresponding author
    1. Laboratories of Molecular Biophysics, The Rockefeller University, New York, New York
    2. Howard Hughes Medical Institute, The Rockefeller University, New York, New York
    Current affiliation:
    1. Structural GenomiX, Inc., 10505 Roselle St., San Diego, CA 92121
    • The Rockefeller University, 1230 York Ave., New York, NY 10021
    Search for more papers by this author


S. pneumoniae phosphomevalonate kinase (PMK, PMVK, MvaK2; EC; 335 residues; predicted molecular mass = 37026.5 Da) is a cytoplasmic enzyme in the mevalonate-dependent isoprenoid biosynthesis pathway.1, 2 The enzyme catalyzes transfer of the γ-phosphoryl group of ATP to the phosphate oxygen of (R)-5-phosphomevalonate to form (R)-5-pyrophosphomevalonate.3 PMK is a member of the GHMP kinase superfamily [named for galactokinase (GK), homoserine kinase (HSK), mevalonate kinase (MK), and PMK], which includes at least 19 30%-identical clusters of proteins with various enzymatic activities. Previously determined structures of two GHMP kinases, Methanococcus jannaschii HSK (PDB ID 1FWK) and Saccharomyces cerevisiae mevalonate-5-diphosphate decarboxylase (MDD; PDB ID 1FI4), produced low-accuracy homology models for a number of PMKs.4 Phylogenetic analyses of nearly 200 superfamily members suggested that an experimental structure of at least one PMK would be required to produce more accurate models for members of this GHMP kinase subfamily. We cloned and expressed two PMKs: one from Staphylococcus aureus and one from S. pneumoniae. S. aureus PMK proved to be largely insoluble under standard purification conditions, whereas the S. pneumoniae homolog emerged as the better structure determination target.

PMK, like HSK and MDD, is an α/β protein (dimensions 59 Å × 53 Å × 48 Å) composed of two domains. The linear arrangement of the secondary structural elements is β1, 310, β2, β3, αA, β4, αB, αC, αD, αE, β5, αF, αG, β6, β7, αH, αI, αJ, αK, β8, β9, αL, and β10 (Fig. 1). The N-terminal domain contains a six-stranded β-sheet of mixed polarity (antiparallel β6-β5-β2-β1-β4 followed by β3 parallel to β4) arching over an α-helical region (αC, αE, αA, αD, αG, and αF) (Fig. 2). β1 (11 residues) and β2 (17 residues containing a β-corner) further subdivide the six-stranded β-sheet into two four-stranded β-sheets: one antiparallel (β6-β5→N-terminal portion of β2→C-terminal portion of β1) and the other of mixed polarity (antiparallel N-terminal portion of β1→C-terminal portion of β2→β4 followed by β3 parallel to β4). The C-terminal domain is composed of an antiparallel four-stranded β-sheet (β8-β9-β7-β10) flanked by αJ, αK, and αL. Visual inspection of the surface of PMK revealed a deep cleft between two β-sheets (β6-β5-β2-β1 and β8-β9-β7-β10). This feature almost certainly represents the enzyme active site because it contains three conserved GHMP kinase motifs5 (I, II, and III).

Figure 1.

Alignment of S. pneumoniae PMK homologs and related proteins identified during an iterative PSI-BLAST14 search (20 iterations; January 2002). The secondary structural elements are shown above the aligned sequences. Gray spheres represent disordered residues. Color-coding denotes sequence conservation among the proteins (white-to-green ramp, 30–100% identity). PMKSTRPN (GI: 15457898), S. pneumoniae (strain R6) PMK; PMKSTRPY (GI:9937402), Streptococcus pyogenes PMK (Z-score = 2e-77; 50% identity); PMKENTFC (GI:9937388), Enterococcus faecalis PMK (Z-score = 2e-81; 29% identity); PMKLISIN (GI:16799091), putative Listeria innocua PMK (Z-score = 3e-68; 28% identity); PMKSTAHA (GI: 9937374), Staphylococcus haemolyticus PMK (Z-score = 8e-85; 29% identity); PMKSTAUR (GI: 9937366), S. aureus PMK (Z-score = 2e-84; 26% identity); MvKMETJA (GI: 15669275), M. jannaschii MK (Z-score = 1e-57; 21% identity), MvKENTFC (GI: 9937386), E. faecalis MK (Z-score = 6e-66; 18% identity).

Figure 2.

RIBBONS15 drawing of PMK with labeled N- and C-termini and secondary structural elements color-coded for α-helices (blue) and β-strands (red). Motifs I, II, and III lining the putative active site cleft are shown in green.

Previously, the New York Structural Genomics Research Consortium (NYSGRC) reported homology modeling of various PMKs.4 With the X-ray structure of S. pneumoniae PMK, we can evaluate the accuracy of these modeling attempts (PMK versus an HSK-derived model gave a root-mean-square deviation (RMSD) = 2.8 Å; PMK versus an MDD-derived model gave an RMSD = 3.6 Å). It is not surprising that the PMK models proved to be of relatively low accuracy because they were both calculated with distantly related templates (sequence identity between PMK and HSK is 16%, between PMK and MDD 8%). In both cases, modeling was compromised by errors in the alignment of the modeled sequence with the template and by the presence of insertions and deletions.

Automated homology modeling with MODPIPE6 using PMK as a template yielded 250 protein models (model score > 0.7, model length > 170 residues) reflecting conservation of the GHMP kinase fold. High- or medium-accuracy models in the 50–100% and 30–50% sequence identity ranges were obtained for prokaryotic PMKs. Lower-accuracy models in the 20–30% sequence identity range represent other PMKs and some proteins annotated as MKs. Models with < 20% sequence identity include fungal PMKs, various MKs, HSKs, GKs, and MDDs, plus proteins of unknown function.

The X-ray structure of PMK provided a useful test of the GHMP kinase homology modeling. In addition, modeling with the PMK template vindicated earlier predictions concerning target selection for structural genomics. Further studies of PMK will be required to better understand the enzyme's mechanism of action, which should be facilitated by the X-ray structure.


The full-length gene encoding PMK was amplified by polymerase-chain reaction using the forward primer containing an NdeI restriction site (GTCGTCATATGATGATTGCTGTTAAAACTTGCGGAAAAC), a reverse primer containing a BamHI restriction site (GTCGTGGATCCTCACGATTTGTCGTCATGTCCTATC), and S. pneumoniae strain R6 DNA as template using standard protocols.7 The amplified insert was cloned into the corresponding sites of a modified pGEX-6P-1 plasmid. The resulting Se-Met-labeled fusion protein was expressed in E. coli BL21 and purified on glutathione and Sepharose Q resins following established procedures.8 Protein for crystallization (24 mg/mL) was dialyzed extensively against 20 mM HEPES pH 7.0, 100 mM KCl, and 3 mM DTT. Gel filtration indicated that PMK is monomeric in solution (data not shown). MALDI-MS confirmed the identity of the purified recombinant protein (measured mass = 37747.3 ± 30 Da, predicted mass = 37703.3 Da).

Diffraction-quality Se-Met PMK crystals (hexagonal rods, Table I) were obtained by hanging-drop vapor diffusion at 20°C against a reservoir containing 0.1 M HEPES pH 7.5, 0.8 M NaH2PO4, 0.8 M KH2PO4, and 30 mM unbuffered ATP. Crystals were cryoprotected by soaking in the mother liquor supplemented with 25% (v/v) glycerol for 2–3 min and immersion in liquid propane (see Table I for space group and unit cell dimensions).

Table I. Crystallographic Data and Refinement Statistics
Crystal characteristics and data collection statistics
Space group: P63; 6 molecules per asymmetric unit
Cell constants: a = b = 231.6 A˚, c = 88.6 A˚
X-ray source: NSLS X25 Beamline
 λ1 (Se-Met peak)
Wavelength (A˚)0.97920
Resolution (A˚)20.0–2.40
Number of observations3616027
Number of reflections103387
Completeness (%)98.3
(2.4–2.49 A˚ shell)92.4
Mean I/σ(I)24.1
(2.4–2.49 A˚ shell)2.5
R-merge on Ia0.082
(2.4–2.49 A˚ shell)0.409
Cutoff criteriaI<−3σ(I)
Figure of meritb0.22 (20.0–2.40 A˚ resolution) for 98297 reflections
Model and refinement statistics
 Data used in structure refinementλ1 (Se-Met peak)
 Resolution range20–2.42 A˚
 Number of reflections202394 (192349 in working set; 18573 in test set)
 Completeness95.0% (85.9% in working set; 9.2% in test set)
 Cutoff criterion|F| > 0.0
 Number of residues1932c
 Number of water molecules779
 Rcrystd0.215R.m.s. deviations
 Rfree0.257Bond lengths (A˚)0.007
  Bond angles (°)1.32
  Luzzati error (A˚)0.30
  • a

    Rmerge = ∑hkli|I(hkl)i − 〈I(hkl)〉|/∑hkliI(hkl)i〉.

  • b

    Figure of merit calculated by using MLPHARE.

  • c

    Number of residues includes 12 amino acids present in a cloning artifact (six His-Met).

  • d

    Rcryst = ∑hkl|F0(hkl) − Fc(hkl)|/∑hkl|F0(hkl)|, where F0 and Fc are observed and calculated structure factors, respectively.

  • e

    Computed with PROCHECK.16

  • f

    Computed with MODPIPE.6 Models are publicly available from MODBASE ( via advanced search with the keyword NYSGRC_1K47.

Ramachandran plot statisticse
 Residues in most favored regions1583 (91.3%)
 Residues in additional allowed regions135 (7.8%)
 Residues in generously allowed regions10 (0.6%)
 Residues in disallowed regions6 (0.3%)
 Overall G-factor0.3
MODPIPE statisticsf
 Total number of models250
  (model score > 0.7, model length > 170 residues)
 Models with >50% sequence identity to template1
 Models with 30–50% sequence identity to template7
 Models with 20–30% sequence identity to template21
 Models with <20% sequence identity to template221

Diffraction data were collected under standard cryogenic conditions on Beamline X25 (National Synchrotron Light Source, Brookhaven National Laboratory), processed and scaled by using Denzo/Scalepack.9. The structure of the Se-Met protein was determined with data from a single-wavelength anomalous diffraction experiment.10 Selenium positions (41/42) (7/monomer) were located with SnB11 and anomalous difference Fourier syntheses and refined by using MLPHARE12 (figure of merit of 0.22 at 2.4 Å resolution). Density modification, combined with sixfold noncrystallographic averaging, yielded an experimental electron density map suitable for model building with O.13 Residues 51–59, 330–335, and three residues from an N-terminal cloning artifact for each monomer were not visible in the electron density map and were omitted from refinement. The final model, consisting of 1932 of 2010 residues and 779 water molecules, was refined at 2.42 Å resolution to an R factor of 21.5% with an Rfree value of 25.7% (Table I).


We thank Dr. K. Rajashankar for help with data collection. This work was supported by NIGMS grant P50-GM62529 (S.K.B.) and NIH grant GM20276 (M.J.R.). S.K.B. is an Investigator in the Howard Hughes Medical Institute. S. pneumoniae PMK represents NYSGRC target T27.