130 J. C. KENDREW [9 



obvious unique site is the single haem group itself, but we found that although 

 some of the usual haem group reagents (such as isocyanides and nitroso- 

 compounds) could be synthesized with heavy atoms in their molecules, it 

 was very difficult attaching them stoicheiometrically to the haem group, 

 partly because of the low solubility of reagents containing heavy atoms such 

 as mercury and partly on account of the very high affinity of myoglobin 

 for oxygen. Unless the most stringently anaerobic conditions are main- 

 tained the complexes of myoglobin with, for example, isocyanides rapidly 

 dissociate: and we have not so far been able to carry through the whole 

 process of preparation, crystallization and X-ray photography without an 

 unacceptable amount of decomposition. 



The approach which finally proved successful, and which was mainly 

 developed by Drs Bodo and Dintzis, was to crystallize myoglobin in the 

 presence of small amounts of suitable ions containing heavy metals (gener- 

 ally one or two moles of ion per mole of protein) ; the ions were selected 

 in the light of such general information as is available about the chemical 

 affinities of different protein side-chains. Crystals thus prepared were tested 

 directly by means of X-rays, and not by chemical methods. If their diffrac- 

 tion pattern proved to be identical with that of a normal crystal it was 

 evident that no combination had taken place. But if there were appreciable 

 changes in the intensities of the reflexions, it could be concluded that attach- 

 ment had taken place. The next stage would be to see whether the com- 

 bination was at a single site, and to locate that site in the unit cell; for, as 

 already indicated, indiscriminate combination with a large number of sites 

 on the protein surface — with all the lysine amino groups, for example — 

 would be useless for the present purpose. To test this, Fourier methods are 

 used; just as we can prepare a difference vector projection of the heavy- 

 atom/heavy-atom vectors from the changes in diffracted intensity, so by 

 using the changes in amplitude as terms in a Fourier synthesis (once the 

 signs of the reflexions have been established) we can in effect subtract out 

 the protein and locate the positions of the heavy atoms responsible for the 

 changes in the diffraction pattern. An example of the resulting 'difference- 

 Fourier projection' is given in Fig. 3 ; this shows the effect of crystallizing 

 myoglobin with /?-chloro-mercuribenzene sulphonate (PCMBS), and indi- 

 cates that a single mercury atom has been attached to each of the two myo- 

 globin molecules in the unit cell (the other atoms in PCMBS are too light 

 to show up in the difference-Fourier projection). Of course, PCMBS is a 

 reagent for sulphydryl groups, and there was no a priori reason to expect 

 that it would combine with myoglobin at all: indeed we discovered by 

 accident that it does so, and we still have no certain knowledge of the chem- 

 istry of the process. The same is true of most of the other reagents we have 

 investigated. These include the ions AUCI4-, Ag+ (in silver nitrate), \{g\^^-, 

 Hg(NH3)^+ (made by dissolving mercuric oxide in concentrated ammonium 

 sulphate), and AUI4-. The results of these experiments are summarized in 



