A comparison of the structures of horse oxy- and human deoxyhaemoglobin

Medical Research Council, Laboratory of Molecular Biology, Cambridge, G.-B.

Abstract

Mammalian haemoglobins undergo a reversible transition between two alternative structures. One of these is taken up when a ligand occupies the sixth co-ordination position at the iron atoms; the other when this position is empty. X-ray studies have shown that crystals of the liganded form, which includes oxy- and carbonmonoxy-haemoglobin and all the derivatives of methaemoglobin, are isomorphous. A three-dimensional Fourier synthesis of oxy- or methaemoglobin of horse showed its two pairs of polypeptide chains to be arranged tetrahedrally, forming a spheroid. The conformation of each chain closely resembled that of myoglobin (Kendrew el al. 1960). The haem groups were embedded in separate pockets in the surface of the spheroid (Perutz et al. 1960; Cullis, Muirhead, Perutz, Rossmann et North, 1962).

The ligand-free structure is crystallographically different from the liganded one, and on present evidence appears to be unique for deoxyhaemoglobin. The absence of a ligand in the sixth co-ordination position has not been observed in deoxyhaemoglobin itself, but can safely be inferred from its absence in deoxymyoglobin (Nobbs, Watson et Kendrew, 1966). First results of the present three- dimensional X-ray study of human deoxyhaemoglobin showed its β-chains to be arranged differently from those in horse oxyhaemoglobin, and indicated that the distance between the two iron atoms in the β-chains increased by over 6 Å on deoxygenation (Muirhead et Perutz, 1963). However, the imperfections of the first Fourier synthesis made it impossible to analyse the difference between the two structures accurately.

The results have now been clarified by the inclusion of an additional isomorphous heavy atom derivative, by an improved method of refinement of the heavy atom parameters and phase angles, and by averaging the electron density about the molecular dyad axis. The structures of oxy- and deoxyhaemoglobin have been compared in detail by shifting and turning the electron density distributions representing the individual chains of oxyhaemoglobin until they fitted those of the deoxy form best. This was done by a computer program which superimposed the two electron density distributions and minimized the differences between them as a function of their relative rotations and translations by a method of least squares. The comparison shows that both the α-and β-chains move on deoxygenation, though the dyad symmetry of the molecule is preserved.

Mathematically, any movement of a body can be most simply represented by a rotation about, combined with a shift along, a suitably placed screw axis. In order to describe the change from the oxy- to the deoxy conformation in this way, we chose a Cartesian co-ordinate system with its origin at the centre of mass of the four haem groups in the oxyhaemoglobin molecule. We then found the position and orientation of the separate screw axes belonging to the a and β subunits, as well as the rotations and shifts of each of the subunits relative to its screw axis. The results show the shifts along the screw axis to be negligible (< 0.2 Å), so that the movements of the subunits can be described simply as rotations. The a-chains rotate by 9.4° and the β-chains by 7.4°. However, since the β-rotation axes lie near the edge of the β-chains while the a-rotation axes lie inside the α-chains, the β-chains actually move more than the a-chains.

To find out whether these movements are caused by detectable conformational changes within the subunits themselves, the electron density distributions representing the individual subunits of deoxyhaemoglobins were superimposed on those of oxyhaemoglobin in the position of optimum fit and then compared section by section. No significant shift of the haem group relative to the globin chain is visible in either of the subunits, and the globin chains themselves look very similar indeed. There appear to be small changes in the FG-corners of the β-chains, and in the GH-corners of the α-chains. These changes are suggestive, since they lie in regions of contact between unlike chains, but at the present limited resolution one cannot be sure that they are geniune.

Nothing that is known about the structure gives us any indication of the manner by which these structural changes might affect the oxygen affinity of the iron atoms, nor indeed how the changes might be triggered by the reactions with ligands. For the solution of this problem X-ray analysis at higher resolution of both the deoxy and oxy forms will be required.