296 F. M. BURNET 



of MEL and NWS) is a new and untried combination that is only going to 

 "work" adequately when many minor mutations and perhaps recombinations 

 are sorted out by survival in an appropriate environment. 



The second would be to look at the process by which neuropathogenicity 

 developed as one demanding a series of mutations at several loci, so that the 

 difference between G and g might be expressed as G 1 G 2 G 3 G* G 5 and g 1 g 2 g 3 g* 

 g 5 . If in the replicating pool of the doubly infected cell genomes carrying these 

 alleles underwent a series of matings in the special sense used by Levinthal 

 (1954) for phage, and were repeatedly subject to selection for survival, both 

 in the intracellular pool and in the intercellular phase as the virus particle, 

 then we might expect the emergence of some such complex as g l G 2 g 3 G i G 5 in 

 the neuro-MEL form best fitted to survive. In essence, the second suggestion 

 is only an attempt to visualize one way in which the requirements of the 

 first might be provided in conventional genetic terms. 



From the more general point of view, we can also consider the reasons why, 

 when a virulent and an avirulent strain are crossed, the commonest finding is 

 an asymmetrical redistribution. If we have two serotypes P and Q, with 

 virulence expressible as 4 and 0, respectively, then the progeny from 

 recombination will show most Q serotypes avirulent, with only a small 

 proportion representing low to moderate, grades 1 and 2, virulence. 



The P serotypes corresponding to the virulent parent will, on the other 

 hand, show a larger proportion of P serotypes, with diminished virulence 

 running right down to zero (Burnet and Lind, 1954c). 



It is of special interest that in the system which does not show redistribution 

 of virulence (MEL/WSE), mixed infection gives a high proportion of recom- 

 binants which can be isolated without recourse to selective processes. With a 

 very similar system, CAM/WSE, in which only a small yield of recombinants 

 can be obtained, there was evidence that redistribution of mouse lung 

 virulence could occur (Burnet and Lind, 1955). This was expressed as: 



ACDF--CE X adf-ce 



AcDFcE, AcD-fce, ad-FCE, adf-CE 



In other words, the redistribution of mouse and chick pathogenicity on the 

 recombinants characterized in vitro as A c -c or a-C could take almost any form. 

 The only limitation was that the virulence of the derivative was always 

 lower than that of the parent. 



All these findings fit into the general pattern that virulence for the mouse 

 lung is dependent on a delicate harmony within the genotype. While it is in 

 process of being established by passage or when recombination is occurring, 

 deviations are possible which will sharply reduce the virulence. At any 

 recombination all the progeny of doubly infected cells will probably contain 

 some mixture of genetic elements, and even those which superficially retain 



