Structural genomics of Pyrococcus furiosus: X-ray crystallography reveals 3D domain swapping in rubrerythrin



Rubrerythrin was first isolated from the anaerobic bacterium Desulfovibrio vulgaris [Protein Data Bank (PDB) entry: 1RYT].1, 2 It is a homodimeric protein and each subunit contains two different types of metal sites. One is mononuclear and located in the C-terminal domain. It is rubredoxin-like and consists of 1 metal ion bound to the side-chains of 4 cysteinyl residues. The second site, in the N-terminal domain, is binuclear. It is hemerythrin-like and consists of 2 metal ions bound by the side-chains of 5 glutamyl and 1 histidinyl residue. The nature of the metal ions in the 2 sites of D. vulgaris rubrerythrin is not clear, since both 3Fe3, 4 and 1Zn/2Fe forms5 of the protein have been isolated. The mononuclear site appears to be occupied only by Fe in both forms.5 A variety of catalytic activities have also been proposed for D. vulgaris rubrerythrin, including pyrophosphatase,3 ferroxidase,6 and nicotinamide adenine dinucleotide hydride (NADH) peroxidase.7–9 Its physiological function has also not been established, although a role in oxidative stress has been implicated for the D. vulgaris protein, as well as for those from several other anaerobic and microaerophilic bacteria.10

The nature and function of rubrerythrin in anaerobic archaea was recently clarified by the purification and biochemical characterization of the protein from the hyperthermophile, Pyrococcus furiosus.11 It was shown that the native protein purified from P. furiosus contains only iron, although the recombinant protein, obtained from Escherichia coli, contained one zinc atom/subunit even when the organism was grown on a minimal medium supplemented with iron. In addition, native (but not the zinc-containing recombinant forms of) P. furiosus rubrerythrin had peroxidase activity and reduced hydrogen peroxide to water using P. furiosus rubredoxin as the electron donor. An NADH-dependent peroxidase system was reconstituted using rubrerythrin, rubredoxin, and NADH rubredoxin oxidoreductase (from P. furiosus). This system is proposed to be part of a pathway for the detoxification of reactive oxygen species involving the novel enzyme superoxide reductase.12

A structural study of rubrerythrin from P. furiosus was initiated as part of the structural genomics intiative that is under way with this hyperthermophilic archaeon.13 The sequence of P. furiosus rubrerythrin shows 32% sequence identity with that of the D. vulgaris protein.14 A direct alignment is shown in Figure 1 and reveals a 13-residue deletion in the P. furiosus sequence but very similar secondary structural motifs,16 suggesting virtually identical structures. However, we show here that the structure of P. furiosus rubrerythrin, determined to a resolution of 2.35 Å, reveals a previously unobserved swapping of its structural domains.

Figure 1.

Sequence alignment with CLUSTALW15 and secondary structure for PDB entries 1NNQ (P. furiosus, red/top) and 1RYT (D. vulgaris, blue/bottom).


Expression and purification: The gene encoding P. furiosus rubrerythrin (PF1283) was cloned and expressed in E. coli and the recombinant protein was purified according to the high-throughput protocols established for P. furiosus protein production at the Southeast Collaboratory for Structural Genomics (SECSG).13 The resultant protein contained an N-terminal His6 tag. Protein identity and purity were assessed using mass spectroscopy and polyacrylamide gel electrophoresis (PAGE).

Crystallization: Crystallization experiments were performed using the modified microbatch under oil method.17 For these experiments, 0.5 μL of crystallization reagent were mixed with 0.5 μL of protein (20 mg/mL) solution, 10 mM NaCl, 2 mM dithiothreitol (DTT), and 20 mM Tris, pH 8. The wells were sealed with a 70:30 paraffin:silicone oil mixture to retard dehydration. Initial crystallization conditions were identified using the Wizard-1 screening kit (Emerald Biostructures, EBS-WIZ-1) and optimized. Diffraction-quality crystals were obtained from a 1.2 μL drop containing 0.5 μL of the above protein solution, 0.5 μL precipitant solution, 20% w/v polyethylene glycol (PEG-1000), 200 mM calcium acetate, 100 mM imidazole, pH 8, and 0.2 μL of additive solution, and 25 mM K2PtCl4 (Hampton Research, HR2-422).

Data Collection: A single crystal was harvested with a cryoloop (Hampton Research, HR4-747) and briefly immersed in a 1 μL drop containing a 1:4 mixture of glycerol and the above-described precipitant solution prior to incubation (10 min) under 100 PSI xenon (Rigaku/MSC CryoXeSiter), before being flash cooled in liquid Freon, retrieved, and stored in liquid nitrogen.

Data that lead to the structure solution were collected at cryogenic temperatures on beamline 22ID (SER-CAT), Advanced Photon Source, Argonne National Laboratory using a MAR 165 charge-coupled device (CCD) detector and 0.97 Å X-rays. The detector was positioned 170 mm from the crystal at a 2θ angle of 0°, providing data to the 2.35 Å resolution, the limit of the detector. Two data sets consisting of 200 images each (0.5° ω steps) were collected with ϕ offsets of 0° and 180°, respectively. The exposure time was 2 s. Data were indexed, integrated, and scaled using the HKL2000 software suite;18 see Table I.

Table I. Data Collection Parameters and Statistics
Wavelength (Å)0.97
Resolution (outer shell, Å)50-2.35 (2.43–2.35)
〈I〉/〈σ(I)〉36.7 (6.3)
Completeness (%)99.9 (100.0)
Rsym = Σ(ABS(I-〈I〉))/Σ(I)7.1 (36.9)
Unique HKLs data (Rfree)18575 (1004)
Number of atoms in model2744
Rfactor (Rfree, %)21.1 (25.3)
RMSD bond lengths (Å)/angles (°)0.013/1.23

The positions of 2 anomalous scatterers and initial phases based on these sites were obtained with SOLVE V2.0219 using the single-wavelength anomalous scattering option. Phase refinement and initial tracing of the peptide chains were accomplished using RESOLVE.20 The model was adjusted manually using XFIT,21 refined using REFMAC5,22 and validated using MOLPROBITY23 and PROCHECK.24 The refined model is available from the PDB,25 entry 1NNQ.

Results and Discussion.

A crystal of recombinant rubrerythrin was incubated under high pressure (100 psi) xenon, flash-cooled in liquid Freon, and shipped to SER-CAT for data collection. A set of high-resolution (2.35 Å) data was collected on the xenon-incubated crystal using the SER-CAT undulator beamline and 0.97 Å X-rays for refinement as described. The SER-CAT data set was also submitted to the SECSG Sca2structure high-throughput structure determination pipeline. The pipeline spawned 180 SOLVE/RESOLVE jobs, on the SECSG Linux cluster. Parameters surveyed include space group (P41212, P42212, and P43212), high-resolution data cutoffs for both the initial phasing (4.0–2.4 Å in 0.4-Å increments) and phase refinement, and extension (also screened from 4.0 to 2.4 Å in 0.4-Å increments, but always at the same or higher resolution cutoff than used in initial phasing), the number of anomalous scatterers within the asymmetric unit (2 and 4), and the number of molecules in the asymmetric unit (2 or 3). All anomalous scatterers were assumed to be xenon for the calculations. The total phasing and fitting process (180 parallel jobs) took approximately 3.5 h on a 128 processor IBM Linux cluster with the initial fitted map being produced in under 4.5 h from mounting the crystal on the goniometer.

The SCA2STRUCTURE results gave two sites from the SOLVE analysis and a fitted sequence (80% complete) from automated RESOLVE fitting. The inspection of the RESOLVE maps and the model produced quickly revealed that the anomalous scatterers identified by SOLVE were in fact metal ions located in the protein's iron-binding sites. Whether the unsuccessful attempts at derivatization with K2PtCl4 or xenon were in fact relevant to the successful structure determination has not been revisited. The experimental details for these procedures are listed here for completeness only.

Overall structure and domain swapping: The P. furiosus crystals contain 1 rubrerythrin homodimer in the asymmetric unit. The rubrerythrin monomer (cartoon of homodimer shown in Fig. 2) is mostly helical, consisting of 2 helical domains, each containing a 2-helix bundle (HA-HB and HC-HD; see Table II), separated by a short 4-residue linker. A 7-residue linker connects helix HD with a small C-terminal domain, comprising residues 138 through 171, that contains a short, 3-strand β-sheet.

Figure 2.

Cartoon representation of the P. furiosus homodimer generated with PyMOL.26 Metal cations are shown as gray spheres.

Table II. Comparison of the P. furiosus and D. vulgaris 4-Helix Bundlesa
HelixPyrococcus furiosusDesulfovibrio vulgaris
  • a

    PDB entry 1RYT contains coordinates for a single peptide chain.


During examination of initial electron density maps, discrepancies between the initial P. furiosus trace and the topology of D. vulgaris structure (1RYT) became apparent. While the topologies of both structures are conserved to residue 62 (the end of helix HB), the structures are distinct as to how the peptide chain connects to the next helix (HC). This is due to the 13-residue deletion observed in the P. furiosus sequence (Fig. 1) that corresponds to the loop linking HB and HC in the D. vulgaris structure. Because of this deletion, the relationship of the 2 helical domains in the monomer is markedly different in the P. furiosus structure compared with D. vulgaris structure and affects the overall dimer structure (Fig. 3). The D. vulgaris dimer can be described as a simple 2-fold (crystallographic) related dimer (HD↓-HC↑-HB↑-HA↓ · HA′↑-HB′↓-HC′↓-HD′↑). In the P. furiosus case, domain swapping occurs between the 2-rubrerythrin monomers, creating a very similar (HD′↓-HC′↑-HB&uaarr;-HA↓ · HA′↑-HB′↓-HC↓-HD↑) dimer. Because of this, the relationship of the two 4-helix bundles comprising the rubrerythrin dimer remains essentially unchanged [root-mean-square deviation (RMSD) Cα 1.5 Å] in the P. furiosus and D. vulgaris structures (Fig. 4).

Figure 3.

3D domain swapping27 shown in topology diagrams of PDB entries 1NNQ (P. furiosus, top) and 1RYT (D. vulgaris, bottom). Colors orange, purple, green, and gray were applied to distinguish the monomers. The linker segments relevant to domain swapping are colored red (1NNQ) and blue (1RYT). The diagram was generated with the program TOPDRAW.28

Figure 4.

Stereographic representation of the superimposition of helix bundles in PDB entries 1NNQ (P. furiosus) and 1RYT (D. vulgaris). The monomers are colored in orange/purple for 1NNQ and green/gray for 1RYT, respectively. This image was created using PyMOL.26

During the preparation of this manuscript, three-dimensional (3D) domain swapping in the rubrerythrin-like protein sulerythrin from the aerobic archeon Sulfolobus tokodaii was reported.29 The report included a prediction that “rubredoxin-like proteins from Pyrococcus species” would not exhibit domain swapping. The authors of the sulerythrin structure emphasize the “as isolated” in vivo character of their model. In this context, it is noteworthy that the P. furiosus structure discussed here is the result of recombinant expression. Whether 3D domain swapping in this case is an artifact27 of E. coli expression will be conclusively addressed with the structure determination of native P. furiosus rubrerythrin. However, a nonswapped assembly conserving the typical helix bundle is unlikely due to the very short loop connecting helices HB and HC.

Metal-binding sites: Each peptide chain accommodates 3 metal ions in 2 metal-binding centers. The coiled C-termini of each subunit accommodate a single metal ion in a rubredoxin-like CX2CXnCPXC center. As a consequence of domain swapping27 in the structure of the P. furiosus protein, the glutamyl and histidinyl side-chains that form the binuclear metal center are contributed by residues from both rubrerythrin monomers.

For refinement, all metal ions found in the metal-binding sites were assigned as divalent zinc, since inductively coupled plasma spectroscopy detected significant amounts of zinc in the crystallization sample (data not shown11). The presence of zinc is also supported by Bijvoet difference Fourier analysis using data collected with 0.97 Å radiation. The peaks corresponding to the 3 metal sites were stronger for data collected using 0.97 Å X-rays than for data collected with 1.54 Å X-rays (data not shown). This would be expected for zinc-containing crystals, since zinc (Δf″Zn = 2.4) should show anomalous signal from the λ = 0.97 Å data and not from the λ = 1.54 Å data (Δf″Zn = 0.7). In the case of iron, one would expect a stronger signal from the λ = 1.54 Å data (Δf″Fe = 3.2) than from the λ = 0.97 Å data (Δf″Fe = 1.5), which runs counter to what was observed. The resolution of the structure prevents identification based on binding geometry.


Data were collected at Southeast Regional Collaborative Access Team (SER-CAT) 22-ID beamline at the Advanced Photon Source, Argonne National Laboratory. Supporting institutions may be found at