348 GENETICS: S. BENZER Proc. N. A. S. 



plaques appearing on K are picked and retested, they fall into three categories: 

 (1) a type which, like the original mutant, produces very few plaques on K and r- 

 type plaques on B; (2) a type which produces plaques (often smaller than wild 

 type) on K with good efficiency but r-type plaques on B; and (3) a type indis- 

 tinguishable from the original wild. These three types are understood to be due 

 to the following: (1) "leaking" effects, i.e., ability of the mutant to grow slightly 

 on K, so that there is a chance for a few visible plaques to form; (2) a mutation 

 which partially undoes the effect of the rll mutation, so that multiplication in K 

 is possible, but the full wild phenotype is not achieved; and (3) apparent reverse 

 mutation, which may or may not be genuine, to the original wild type. 



The proportion of each type occurring in a stock is characteristic and reproducible 

 for a particular rll mutant but differs enormously from one rll mutant to another. 

 There is no evident correlation in the rates of occurrence of the three types. 



Reversion Rates of rll Mutants. — Reversion of r mutants to a form indistinguish- 

 able from wild type was demonstrated by Hershey,^ who made use of the selective 

 advantage of wild type on B to enrich its proportion in serial transfers. Given the 

 inability of rll mutants to produce plaques on K, such reversions are easily de- 

 tected, even in very small proportion. An index to the frequency of reversion of a 

 particular rll mutant can be obtained by preparing a lysate from a small inoculum 

 (about 100 particles, say, so that there is very little chance of introducing a wild- 

 type particle present in the stock). If S is used as the host, both rll mutant and 

 any reversions which arise can multiply with little selection, as shown by control 

 mixtures. The average fraction of wild-type particles present in several lysates is 

 an index which can be shown to be roughly proportional to the probability of re- 

 version per duplication of the rll mutant. Under the conditions of measurement 

 the index is of the order of 10-20 times the probability of reversion per duplication. 

 The plaques appearing on K must be tested by picking and replating on B. This 

 eliminates the "spurious" plaques produced by partial reversions and by leaky 

 mutants, which show up as r type on B. As may be seen in Table 2, the reversion 

 indices for rll mutants vary over a very wide range. One mutant has been found 

 which reverts 10 times more frequently than r51, so that the reversion rates cover 

 a known range of over 10^-fold. 



It has not been proved that these apparent reversions constitute a genuine re- 

 turn to the original wild type. However, the possibility of suppressor mutations 

 distant from the site of the rll mutation has been ruled out by backcrosses to the 

 original wild type. Krieg'' found very few, if any, r-type recombinants in back- 

 crosses of several reversions, localizing the reverse changes to within a few tenths 

 of a per cent linkage distance from the original rll mutations. One case of "par- 

 tial reversion" has also been tested by backcrossing, and failure to observe rll- 

 type recombinants localized the "partial reverse mutation" to within the rll region. 



Mapping of the rll Region. — A cross between two rll mutants is made by infect- 

 ing a culture of B with equal multiplicities (three per bacterium) of each type. 

 The yield after lysis contains the two parental types and, if the parents are geneti- 

 cally distinct, two recombinant types, the double mutant and wild type. In the 

 average yield from many cells, the recombinant types occur in equal numbers.'' 

 In all cases thus far tested, double rll mutants, like single mutants, do not produce 

 plaques on K. On the assumption that this is generally true, the proportion of 



213 



