Structural studies on metal-containing enzymes: T4 endonuclease VII and D. gigas formate dehydrogenase

Many biological processes require metal ions, and many of these metal-ion functions involve metalloproteins. The metal ions in metalloproteins are often critical to the protein's function, structure, or stability. This thesis focuses on two of these proteins, bacteriophage T4 endonuclease VII (EndoVII) and D. gigas fonnate dehydrogenase, which are studied by X-ray crystallography. The structure of EndoVII reveals how a magnesium or calcium ion is used to cleave several kinds of irregular but flexible DNA, while a zinc ion maintains the structural integrity of this DNase.The formate dehydrogenase contains a tungsten ion and a seleno-cysteine at the active site, that catalyses the oxidation of formate to carbon dioxide. The two released electrons are transferred through four [4Fe-4S] clusters before they can be handed over to another protein. Two of the [4Fe-4S] and the selenium have been overlooked by other techniques, but could be localised and identified by crystallography.Chapter 1 gives a general introduction on metals in biological systems, X-ray crystallography and also describes the biological background of both proteins.Chapter 2 presents the structure of the four-way DNA-junction resolving enzyme T4 endonuclease VII, and that of the inactive N62D mutant. The betterexpressed mutant was solved first, using seleno-methionine, mercury and gold derivatives. These mercury and gold derivatives bind to the sulphurs that also ligand the zinc. The wild-type was solved with help of a single mercury derivative since molecular replacement with the mutant structure failed.On its own, the EndoVII monomer would not represent a stable fold, as it exposes many hydrophobic residues to the solvent. But two monomers intertwine to form a dimer without this problem. In this dimer, the monomers are aligned head-to-tail; the N-terminus of one monomer interacts with the C-terminus of the other monomer and vice versa . The major dimerization element, unique to EndoVII, is the "four-helix-cross" domain, which consists of helix-2 and helix-3 from each monomer. It contains an extended hydrophobic core.Another feature is the "beta-finger", residues 38-56. Its stability depends critically on the zinc. This zinc ion is tetrahedrally co-ordinated to four cysteines, linking helix- I through residues C23 and C26 firmly to the N-terminal part of helix-2 (C58, C61). Indeed, interfering mutations inactivate the protein. Finally, the calcium ion, which marks the active site, is liganded to aspartate-40 and asparagine-62. Mutation studies show that these amino acids are essential for activity: The N62D mutant is completely inactive.The EndoVII structure has been docked to a "stacked-X" four-way DNA junction, one of its many substrates. This model is not refined, since both the DNA and the protein are known to be flexible and might undergo conformational changes. However, its overall features confirm experimental data: 1) The EndoVII dimer binds to the minor groove side of the four-way junction; 11) Basic residues on helix-2 can interact with phosphates on the exchanging strands and those on the C-terminal domain can interact with phosphates in the continuous strands, consistent with observed foot-printing patterns; 111) The C-terminus binds up to nine base pairs away from the junction, confirming the minimal length of two arms of the substrate; IV). The active sites do not cleave both the scissile phosphates simultaneously.Surprisingly, the N62D mutant shows a major rearrangement in the "four-helixcross" domain, when compared to the wild-type: helices-2 are translated by half a turn each, in opposite direction and the opening of the "bays", between each helix-2 and betafinger, is wider. These differences might be attributed to the point-mutation, which introduces an extra charge in the active site, to differences in crystallisation conditions, to the different pH employed, to crystal contacts or perhaps they are simply a sign of the intrinsic flexibility of EndoVII.This dilemma is partly solved in chapter 3, which presents the crystal structure of wild-type EndoVII in a different space group, which contains less solvent. It crystallised in the same drop, so that differences observed between the two wild-type structures cannot be attributed to the mutation, pH or salt concentrations. Since the helical-cross region of this second structure is very similar to that of the mutant, rearrangements in this region must be seen as a consequence of intrinsic flexibility of EndoVII. The widening of the "bays", however, might still be a consequence of the mutation, different pH, absence of Ca 2+ or crystal packing. An investigation of the flexibility of EndoVII with TLS- refinement, i.e. anisotropic refinement of rigid bodies, provides only limited insight. However, it confirms that rotations along the axes of the helices 2 and 4 and along the beta-finger are a main source of flexibility and also that the C-terminus, helix-4, 5 and 6, behave as a rigid body.The high-resolution structure of the N62D mutant brings more clarity towards the reaction mechanism of the nuclease. This model contains important water molecules and reveals the position and orientation of 14 sulphate ions, which may indicate favoured phosphate (DNA) binding sites. Supported by new mutation data (Birkenbihl, unpublished), these sulphate and water positions, combined with the Ca 2+ positions in the wild-type structures, suggest a reaction-mechanism similar to those proposed for some other magnesium dependent nucleases.2 Asparagine-62, glutamate-65 and aspartate-40 are important to position Mg 2+ or Ca 2+ next to the scissile phosphate of the DNA substrate. Histidine-41 activates a water molecule, which in turn executes a nucleophilic attack on the phosphor atom. Histidine-43 stabilises this phosphate directly through a hydrogen bond. Unfortunately it is still unclear why the N62D shows no DNase activity at all; an aspartate would also be able to ligand/position a divalent cation. The extra charge that this mutation introduces in the active site might distort the geometry of the active site, and repel the DNA. A more attractive, albeit more speculative hypothesis, assumes that the amino group of asparagine-62 donates a hydrogen-bond to the phosphate, which would also stabilise the transition state.At present, there are no known proteins with significant sequence homology to EndoVII, though nucleases with structural similarities do exist. One group consists of magnesium-dependent nucleases, which have a similar geometry of liganding sidechains around the magnesium (or calcium) ion in the active site; e.g. the E. coli proteins RuvC (Ariyosi et al., 1994) and RNase H (Katayanagi et al., 1990). However, these nucleases have no resembling fold. Most likely, this just shows that magnesium-dependent nucleases need a certain geometry to function.A more interesting group shares a folding motif similar to the beta finger and helix-2: Serratia Nuclease (Miller et al., 1994), Ppol (Flick et al., 1998) and perhaps even Colicin E9 (Kleanthous et al., 1999). Asparagine-62 and histidine-41 are conserved between Serratia nuclease, Ppol and T4 endonuclease VII. Ppol has also been crystallised in complex with DNA. If one superimposes this with the EndoVII structure, it turns out that the Ca 2+ in EndoVII is buried deeper within the protein, but small rotations (10-20 degrees) along helix 2 and the beta-finger suffice to superimpose them. These two nucleases act on different substrates, and maybe the larger DNA junctions of EndoVII need a wider and deeper binding groove than the double stranded DNA of Ppol. However, it could also be the source of EndoVII's specificity; flexible DNA might impose this conformational change of EndoVII upon binding, readying the enzyme for cleavage, while the magnesium or calcium ion might be too far away if EndoVII approaches more rigid DNA. A structure of EndoVII in complex with DNA would solve these questions.Chapter 4 presents the major part of the determination of the 3D structure of the tungsten-containing formate dehydrogenasc (W-FDH) from Desulfovibrio gigas, one of the first tungsten-containing enzymes isolated from a mesophile. The large subunit (92 kDa) is structurally related to several tungsten- and molybdenumcontaining enzymes and X-ray structures have been determined for two of them. One of these, the periplasmic nitrate reductase (Dias et al, 1999), could be used to obtain a molecular replacement solution. But the quality of phasing was not sufficient to generate a clear, interpretable electron density map. Furthermore, the amino acid sequence of W-FDH has not yet been determined, what makes model building complicated. Multiple wavelength diffraction (MAD) measurements were undertaken at the absorption edges of W and Fe to define unambiguously the number, positions and identity of these anomalous scatterers and to improve the X-ray phases. The MAD-analysis revealed one W-atom with a Se-cys ligand and one [4Fe-4S] cluster bound to the large subunit, and three [4Fe-4S] clusters in the small subunit. The four [4Fe-4S] clusters are ca. 10 Å apart, creating a feasible electron transfer pathway, which connects the exterior of the protein to the W/Se site in the large subunit. Two of the four iron-sulphur clusters had not been predicted before by spectroscopic techniques (Almendra et al., 1999). A reinvestigation of the spectroscopic data was performed, but gave the same results as before. If these data were correct, this means that the [4Fe-4S] clusters are instable, and that only protein with fully occupied clusters crystallises.The formate dehydrogenase H (FDH-H) from E. coli catalyses the same reaction as W-FDH, but uses a molybdenum instead of tungsten. Both are liganded to two molybdopterin-cofactors and to a seleno-cysteine, so the question remains why W-FDH prefers tungsten to the more common molybdenum. The full structure will allow a comparison of the two enzymes in atomic detail, and perhaps, it will shed some light on this phenomenon.X-ray crystallography has been used to characterise the nature of metal-centres in proteins, their coordination geometry and even their identity. Sometimes, the way metal ions are bound to the protein already clarifies its role in the protein. In other cases it has to be supplemented with other studies before the role can be fully understood. Either way, crystallography provides a powerful tool for the study of metalloproteins.ReferencesAlmendra, M.J., Brondino, C.D., Gavel, 0., Pereira, A.S., Tavares, P., Bursakov, S., Duarte, R., Caldeira, J., Moura, J.J.G., Moura, 1. (1999) Biochemistry, 38 , 16366-16372Ariyosi, M., Vassylyev, D., Iwasaki, H., Shinagawa, H. and Morikawa, K. (1994) Cell, 78 , 1063-1072.Flick, K.E., Jurica, M.S., Monnart Jr, R.J. and Stoddard, B.L. (1998), Nature, 394 , 96-101.Katayanagi, M., Miyagawa, M., Matsushima, M., Ishikawa, M., Kanaya, S., Ikehara, M., Matsuzaki, M. and Morikawa, K. (1990) Nature, 347 , 306-309.Kleanthous, C., Kuhlmann, U.C., Pornmer, A.J., Ferguson, N., Radford, S.E., Moore, G.R., James, R. and Hemmings, A.M. (1999) Nature Struct. Biol., 6 , 243-252.Miller, M.D., Tanner, J., Alpaugh, M., Benedik, M.J. and Krause, K.L. (1994) Nature Struct Biol., 1 , 461-468.