THE MUTABLE UNIT OF HEREDITY 49 



dealint:; with macromolecuk's which may sometimes be present only once 

 per cell, has no precedent. 



The Poisson formulation ought then to predict the numbers of muta- 

 tions occurring among different clones of cells if they take place at 

 random. It is important here to distinguish numbers of mutations from 

 numbers of mutants which we saw were distributed among different 

 clones according to another formulation, that of Lea and Coulson. 



It is possible to make a rather direct estimate of the distribution of 

 numbers of mutations among different clones by using a solid culture 

 medium instead of liquid for growing the cells, and thereby counting 

 mutant clones instead of mutant cells. When bacteria which cannot use 

 lactose are grown on an amino acid medium, mutants appear which can 

 use that sugar. On the surface of agar containing amino acids and 

 lactose, the nonutilizing parents form a colony; the mutants arising 

 within it, by also using the lactose present, overgrow to form papillae 

 on the colony surface. If an indicator dye is incorporated in the 

 medium, the lactose-utilizing papillae turn red because they form certain 

 specific waste products. On the other hand, the colony background is 

 almost white, allowing a ready enumeration of the papillae. Here each 

 colony is a separate culture and each papilla represents one mutation, 

 irrespective of the number of mutant bacteria it contains. In this way 

 the mutations themselves may be counted, even though a constant frac- 

 tion may not express them as papillae. If almost all the parental lactose- 

 nonutilizing bacteria have the same chance of mutating per division, then 

 a Poisson distribution of the numbers of papillae (but not, of course, of 

 the number of mutant cells composing them) should be observed. The 

 numbers of colonies with no mutations, or with one, two, or more, are 

 shown in Figure 2.5, along with the Poisson distribution expected for 

 the average number of mutations observed. The close similarity of the 

 two distributions indicates that the mutations were randomly distributed 

 among the bacteria making up the various colonies. 



Still one more component of statistical randomness exists; it involves 

 independence in the mutation of one gene from mutation in another. 

 This criterion seems to be met by mutation in plants and animals, but 

 it can be shown most clearly to govern mutation in bacteria. Double 

 auxotrophic mutants can be obtained which have changes in two differ- 

 ent genes, each one causing a requirement for a different substance 

 needed for growth. For example, mutant bacteria requiring both the 

 amino acids histidine and methionine have been obtained in two steps. 

 First the strain became histidineless (histidine-requiring, or his ) by one 

 mutation; then a second mutation occurred which gave rise to a histi- 

 dineless-methionineless type (his' met'). This double auxotroph back- 



