A DETAILED X-ray study of collagen fibres obtained from different sources (namely, shark ray, rat tail tendon and kangaroo tail tendon) and a re-examination of the published wide-angle patterns indicate that the unit cell of collagen is hexagonal with a = 12–16 A. and c = 9.5–9 A., the actual values depending on the moisture content. The essential difference of this indexing from those reported earlier 1- 3 is that the 2.86- A. meridional arc is here interpreted not as a true meridional reflexion, but as arising from the superposition of two close non-meridional reflexions. A calculation of the angular spread of the arc, using the tilt of the c-axis deduced from the spread of the equatorial reflexions, confirms this interpretation. Table 1 shows the good agreement between the calculated and observed spacings of an air-dried and a wet specimen from kangaroo tail tendon.
   A structure has been obtained (Fig. 1) which fits the above unit cell and which appears to be in good agreement with infra-red, X-ray and chemical data for collagen. It consists of nine amino-acid residues per unit cell, which corresponds to the observed density. These are linked together to form cylindrical rods, which occur in a hexagonal array. All the residues have the trans configuration, and the latest values of Corey and Pauling4 for the dimensions of the amide group were used for the calculations. The residues are arranged in the form of three helical chains, each of pitch 9.5 A. (= c) and containing three residues per turn, with the symmetry 31 . The three helixes are also arranged with a 31 symmetry about the c-axis, and they are held together by means of hydrogen bonds to form the cylindrical rods. Two of the three NH groups in each turn of a chain are linked by hydrogen bonds to an oxygen of each of the other two chains, the NH . . . O distance being 2.80 A. The third NH group points outward from the cylinder, and the nitrogen atom forms part of a proline ring. Of the three a-carbon atoms per turn, a hydrogen attached to one of them (R, Fig. 1) could be replaced by a general R group to form an amino-acid residue such as arginine or lysine; another (P) takes part in forming the proline ring, while the third (G) is in such a position that there is no space for either of its hydrogens to be replaced by Table I.
 

any other group, so that it could only form part of a glycine residue. These features of the structure could explain the observed range of amino-acid composition for collagen. The NH- and CO-bonds are almost exactly perpendicular to the fibre axis, the angle made with the c-axis being about 85° in both cases. This agrees with the observed large infra-red dichroism5 . Structure- factor calculations show fairly good agreement with observation. The individual helixes are unstable, and the stability of the cylindrical rods arises from the hydrogen bonds between the helixes. If these are broken, for example, by heating, the structure would crumble down, which would explain the thermal contraction of collagen. The structure could, however, re-form on cooling, as the chain of amino-acid residues in a single helix need not be ruptured in this process. It may be mentioned that this structure is essentially different from the three-chain structure of Pauling and Corey2. The value c = 9.5 A. is not critical, and a variation of about 15 per cent is permissible without affecting the main features of the structure. A detailed paper, containing also a discussion of other features of this structure and a critical comparison with previously proposed structures, will be published elsewhere.
 

1. Astbury, W. T., J. Int. Soc. Leather Chem., 24, 69 (1940).
2. Pauling, L., and Corey, R. B., Proc. U. S. Nat. Acad. Sci., 37, 272 (1951).
3. Bear, R. S., “Advances in Protein Chemistry”, 7, 69 (1953).
4. Corey, R. B., and Pauling, L., Proc. Roy. Soc., B, 141, 10 (1953).
5. Randall, J. T., Fraser, R. D. B., and North, A. C. T. Proc. Roy. Soc., B, 141, 62 (1953).
   
    

Fig. 1. Projection of the proposed structure along the fibre axis. The numbers denote the heights of the atoms above the plane of projection. Only a part of each of the three individual chains is shown. The thin lines indicate the N . . . O directions in the hydrogen bonds

symmetrically disposed with respect to one another. Their configuration is most easily described by saying that they are displaced along the fibre axis by + 28.6 A. and – 28.6 A. with respect to the first chain. They are also related to the latter by a rotation of ± 108° about the axis of the cylinder and a translation of ± 2.86 A. parallel to the axis. For the mathematical discussion of the structure, each chain may be described as consisting of a minor helix (following the nomenclature of Crick3) having ten residues in three turns, with a projection of 2.91 A. per residue along its axis, wound to form a major helix of radius 2.5 A., there being thirty residues per turn of the major helix. The major helix is wound in a direction opposite to that of the minor helix, the latter being a left-handed screw if all the residues are of the L-type. Thus, while the thirty residues make nine turns in the co-ordinate system of the minor helix, they make ten turns in the fixed co-ordinate system of the major helix; this ensures that every third residue is similarly situated with respect to the fibre axis.
   Because of the above relationships, there are only two types of inter-chain hydrogen bonds, marked A and B in Fig. 1, which are repeated by screw symmetry. In both bonds, the N . . . O distance is close to 2.9 A. and the angle between NH and N . . . O directions is about 30°. This angle would be less if the NH-bond were not coplanar with the rest of the amino-acid residue.
   As in the earlier structure, two of every three NH groups are hydrogen bonded to an O of a CO group, but unlike it, the two oxygens to which they are linked are different. Thus, of every three carbonyl oxygens, only one does not take part in internal hydrogen bonding in the triple-chain cylindrical rods. This

Fig. 2. Radial intensity distribution along the equator calculated from the structure. The abscissae are propor- tional to the reciprocal of the spacing. The positions corresponding to the 100, 110 and 200 reflexions of the hexagonal unit cell with a = 12 A. are also marked

oxygen points outward from the rod and plays an important part in the cross-linking of the rods by side- chains, as shown below. As before, every third residue must be a glycine residue, and similarly a maximum of one-third of the residues may be proline or hydroxyproline.
   The essential difference introduced in the structure, so far as its X-ray pattern is concerned, is that the 2.86 A. reflexion is meridional, and is the only meridional reflexion apart from its higher orders. Further, the exact repetitive unit along the fibres axis is ten residues (28.6 A.), which agrees well with the observed layer- line spacings4. The intensity distribution among the spots on the layer lines has been found to agree with the structure qualitatively. The distribution of intensity along the equator has been worked out in detail and is shown in Fig. 2. This is in good agreement with observations (Ramachandran, G. N., and Ambady, G. K., unpublished results), particularly with regard to the large peak at 4.7 A. and the minimum at 7.4 A. The absence of the expected spot at 7 A. is thus explained. In wet collagen, the water would be expected to occur between the cylindrical rods and, under these conditions, the minimum shifts towards smaller angles and continues to occur in the region of the unobserved reflexion of 7–8.5 A., even though the lattice expands. Further, the three-dimensional cylindrical Patterson diagram for collagen published recently by Yakel and Schatz5 is in excellent agreement with the structure. The peak at about 1 A. in the level z = 0 is due to the projected component of bonds joining neighbouring atoms perpendicular to the fibre axis; that at 2.7 A. is due to the component of the almost horizontal hydrogen bond interactions; the strong one at 4.5 A. is due to the distances between the three approximately parallel amino-acid residues at the same level.
  The recent infra-red data6 of Sutherland et al. are also in agreement with the present structure. The expected dichroism of the non-coiled 3–10 helix agrees with the observed data7, and the further coiling of this helix produces only small changes and the results are not materially different.
  The distribution of the proline and hydroxyproline rings can be calculated exactly, and they are found to be such that the oxygen of hydroxyproline is nearly as far away from the axis of the cylinder as is possible. Thus internal hydrogen bonding of the OH in hydroxyproline is ruled out. Rather, these groups take part in linking the cylindrical rods with one another; the hydroxyproline OH group in one rod is nearly at the same level as the exposed non-hydrogen-bonded carbonyl oxygen of the neighbouring rod, so that they could readily be linked by a hydrogen bond. This is in agreement with the results of Gustavson8 on the nature of the hydroxyproline OH . . . O bond. Assuming these bonds are 2.7 A. long, the distance between the centres of neighbouring rods works out to be 11.6 A., which is close to 12 A. (10.4/0.866) found for well- dried collagen9. The theoretical value is not precise, for it would change if the bond angles in the amino-acid residue and the a-carbon atoms are altered by even a small amount.
   If the coiled coils described above are simply put together in a hexagonal array, then only some of the hydroxyproline OH groups are bonded to carbonyl oxygens of neighbouring rods. However, it is found that a very slight change in the pitch of the major helix is sufficient to make the number of hydroxyproline� carbonyl hydrogen bonds a maximum. Instead of a rotation angle (say f) of 36° per three residues about the axis of the major helix, the angle must be changed to 35°. This immediately introduces a 61 symmetry for the major helix, which now repeats after 216 residues (618 A.). The spacing 618 A. is highly suggestive of the 640 A. spacing found in the small-angle X-ray pattern and in electron micrographs of collagen. The condition for stabilization of the structure by forming the maximum number of hydroxyproline cross-links thus automatically leads to the long spacing of collagen. Further, on working out the number of hydroxyproline residues which occur in such a position that the hydrogen bonds are fairly short, the number is found to be twelve for every thirty-six of the P type residues. The percentage of hydroxyproline residues occurring in collagen is variable with the source of the material8; but if we assume that the hydroxyproline residues which actually occur are all hydrogen-bonded, then the maximum proportion of hydroxyproline residues should be 11 per cent. It is noteworthy that the maximum observed (namely, in bovine collagen) is about 10 per cent. Further, the correlation of the shrinkage temperature with hydroxyproline content reported by Gustavson8 is also to be expected if hydroxyproline is mainly responsible for the cross- linking of the cylindrical rods. When collagen shrinks, the individual chains in the triple chain are not likely to be separated, but the triple chain itself would take up a highly folded configuration. The X-ray pattern of gelatin, in fact, supports this. The detailed evidence will be discussed elsewhere.
   Having fixed the configuration of the rods with reference to hydroxyproline cross-linkages, a consideration of the length of the R side-chain of the third residue needed to form hydrogen-bonded linkages between one rod and another shows that it is variable from 5.5 to 7.5 A. In view of this, there must be a periodic distribution of the R-type residues. The details are not easy to work out, but an interesting result emerges, namely, that the distribution of R-groups along the fibre axis will not have a period of 618 A., but one-sixth of this. It is indeed noteworthy that electron micrographs of collagen stained with phosphotungstic acid exhibit10 such a six-fold subdivision of the gross period of 640 A.
   The agreement of 618 A. with 640 A. observed for dry collagen is not close enough, nor is the six-fold subdivision exact11. However, it must be mentioned that it is not at all certain that the unit cell of collagen is hexagonal. It is probably monoclinic even in the dry state with slightly differing a and b (taking c as monoclinic axis), as is found to be the case with wet collagen12. The subdivision of 618 A. would then not be exactly into six equal parts, and further the angle f may be different from 35° and the macro-period may also differ from 618 A. The variation of the long spacing with moisture9 is to be attributed to the changes in the angle of the monoclinic cell.
  Further details of this work are reserved for a separate communication.

June 3

1. Ramachandran, G. N., and Kartha, G., Nature, 174, 269 (1954).
2. Cowan, P., North, A. C. T., and Randall, J. T., “Nature and Structure of Collagen”, 241 (Butterworth, 1953).
3. Crick, F. H. C., Acta Cryst., 6, 689 (1952).
4. Ramachandran, G. N., and Ambady, G. K., Curr Sci., 23, 349 (1954).
5. Yakel, H. L., and Schatz, P. N., Acta Cryst., 8, 22 (1955).
6. Sutherland, G. B. B. M., Tanner, K. N., and Wood, D. L., J. Chem. Phys., 22, 1621 (1954).
7. Ramachandran, G. N., J. Chem. Phys., 23, 600 (1955).
8. Gustavson, K. H., Nature, 175, 70 (1955).
9. Rougvie, M. A., and Bear, R. S., J. Amer. Leather Chem. Assoc., 48, 735 (1953).
10. Nutting, G. C., and Borasky, R., J. Amer. Leather Chem. Assoc., 43, 96 (1948).
11. Kaesberg, P., and Shurman, M. M., Biochim. Biophys. Acta, 11, 1 (1953).
12. North, A. C. T., Cowan, P. M., and Randall, J. T., Nature, 174, 1142 (1954).