THE PURIFICATION OF ANTIBODY 243 



are relatively rigid structures, by reason of the large number of interatomic forces 

 which, acting together, stabilize the geometric form of the molecule, they are 

 also capable of some distortion, which may or may not be reversible. In much 

 of the earlier work on protein molecules, the assumption was made that protein 

 molecules behaved as spheres, i.e. were isodimensional. It is now evident that 

 some molecules, including the globulins, are markedly anisodimensional, behaving 

 as though they were rod-shaped, or at any rate larger in one dimension than in 

 any other. The term " globular," appUed to the non-fibrous proteins, does not 

 imply that they necessarily behave as spherical or nearly spherical units. For 

 example, in the serum reactions, the shape of the molecule, and the degree to 

 which it may be changed, will affect the range of permissible speculation about 

 the structure of the antigen-antibody complexes. 



The molecular weight of jjroteins may be determined from their sedimentation 

 rate in the intense gravitational fields of high-speed (" ultra "-) centrifuges. Those 

 of non-fibrous proteins range from 17,500 for lactalbumin, to over 6 millions for 

 mollusc hsemocyanius. The molecular weight of the normal serum globulins 

 and the majority of antibody globulins in the horse and rabbit are of the order 

 of 150,000 (see Svedberg 1937, Heidelberger and Pedersen 1937). Neurath (1939) 

 estimated that the physicochemical properties of rabbit antibody globulin were 

 consistent with those of an ellipsoid, 27 m[.i in length, and a maximum width of 4 m//. 

 Svedberg has pointed out that the molecular weights of the non-fibrous proteins 

 cluster round certain values which are whole-number multiples of the lowest 

 weight recorded, 17,500, and suggests that they are built up from units having 

 a molecular weight of approximately 35,000. This relationship of the molecular 

 weights of proteins, whether animal or vegetable in origin, presumably reflects 

 a similarity of the enzyme systems that take part in protein-synthesis. On this 

 view, the rabbit antibody globulin molecule would correspond to four Svedberg 

 " units." As we shall see, this possibility has been raised in connection with 

 the valency of antibody by Hooker and Boyd, who argue that if the four units 

 each carried an antibody receptor, and were arranged linearly, the whole molecule 

 would be multivalent, and markedly anisodimensional (p. 251). 

 The Purification of Antibody. 



Antibodies are precipitated from serum by agents that bring down the serum 

 globulin. Precipitation of antisera with sodium or ammonium sulphate, with 

 alcohol at various temperatures, or by dialysis, brings down the greater part of 

 the contained antibody with the globulin fractions. (See Banzhaf and Gibson 

 1907, Gibson and ColUns 1907, Ledingham 1907, Mellanby 1908, Hartley 1914, 

 Felton 1926, 1928, 1932, Maitland and Burbury 1927, Barr, Glenny and Pope 

 1931, Barr and Glenny 1931, Laidlaw and Dunkin 1931, Barr 1932, and many 

 others.) 



A considerable degree of purification of antibody solutions is possible where 

 the antibody is confined to one particular fraction. By fractionation Felton 

 (1926, 1932) purified the antibody to the Type I pneumococcus, and Laidlaw 

 and Dunkin (1931) the antibody to distemper virus. 



Specific precipitation of serum globulins by antisera prepared in an animal 

 of another species also brings down the antibodies (Landsteiner and Prasek 1911, 

 Eisler 1920, Smith and Marrack 1930). Moreover, the inactivation of antibodies 

 by heat corresponds in general with the denaturation of proteins by heat (see Streng 

 1909, Madsen and Streng 1910, Marrack 1938). Many of the earlier studies of 



