GENETIC INTERACTIONS BETWEEN ANIMAL VIRUSES 295 



majority are serologically homozygous. If most viable particles showing 

 pheuotypic mixtures are heterozygotes, as Hirst and Gotlieb's experiments 

 suggest, some means must be available by which infection with a heterozygote 

 gives a homozygous yield. An obvious suggestion from analogy with what is 

 known from work with bacterial viruses is that the initiation of replication by 

 one virus genome inhibits replication by any other type of genome which 

 may also be present. The existence of recombination in itself makes this 

 impossible or highly unlikely. The simplest interpretation is probably to make 

 the assumption that the chance of any genome initiating multiplication is 

 quite low and that the virus particle contains n genomes to ensure the likeli- 

 hood of one being effective. In line with this is Donald and Isaacs' (1954) 

 finding that in the average fluid only 1 of 10 visible particles initiates infection. 

 This provides a simple explanation for the difference between American and 

 Australian results. In our system, slightly less than 1/n genomes succeed in 

 initiating infection, so that virtually all yields are homozygous, in Hirst and 

 Gotlieb's system about 2/n succeed and heterozygotes are frequent. Other 

 possible explanations are more conveniently deferred till the relation between 

 genetic changes and virus multiplication is discussed. 



b. Virulence. Virulence in a virus is necessarily a phenomenon of multiple 

 causation. If we consider a virus of maximal virulence for the standard host 

 as the normal form, then any genetic change which reduces the effectiveness 

 of its functioning will diminish virulence. To be recognized as virulent an 

 influenza virus particle must be able (1) to attach itself to and enter the host 

 cell, and (2) to bring viral components not concerned with (1) into necessary 

 relationships with synthetic mechanisms in the cell. When a virus is trans- 

 ferred to a new host, many modifications will be necessary in both respects 

 before it can be brought to maximal virulence; all detailed studies indicate 

 that any such development of a new virulence is the end result of many small 

 mutational changes, probably involving nearly all genetic determinants. 



With this background we can consider the influence of genetic interaction 

 on virulence in typical systems, concentrating particularly on the interaction 

 between the neuropathogenic NWS and MEL. Here the usual result is 

 recombination to give ABDF-ce (g) (neuro-MEL), but in the allantoic cavity 

 the recombinant has not in our experience been able to kill 4-6 week mice 

 inoculated intracerebrally (Fraser and Burnet, 1952). To obtain mouse 

 virulent forms the interaction must take place in the mouse brain or within 

 the tissues of the chick embryo. There is also very considerable variability in 

 the manifestation of neuropathogenicity among replicate limit dilution 

 fluids. There are two possible approaches. One is to avoid any attempt to 

 provide a detailed genetic interpretation by recognizing that the new complex 

 ABDF-ceg (the symbols being italicized to indicate that we are dealing with 

 the genetic determinants that were responsible for the phenotypic characters 



