Rebuttal: Growth under Selection Stimulates Lac⁺ Reversion (Roth and Andersson)

Like Rosenberg and Hastings (5), Roth and Andersson (6) set up a false trichotomy among directed mutation, hypermutation, and amplification. They then argue that neither directed mutation nor hypermutation can be correct, and, therefore, their amplification model must be. But amplification is not the only alternative. In my review (3) I present another model. In addition, the amplification model simply does not fit the data.

Roth and Andersson maintain that during lactose selection, Lac⁺ revertants of FC40 arise at a rate of 1 per 10⁸ replications of the lac allele. They then have to explain how these replications fail to be detected. In their latest attempt, Roth and Andersson postulate that (i) the proportion of cells with lac duplications is 1/100, (ii) the proportion of these that amplify lac and produce a microcolony is 1/100, (iii) the number of cells in a microcolony is 10⁴, and (iv) and the number of lac copies per amplifying cell is 100. To explain how respreading failed to detect these microcolonies, Roth and Andersson state that only 1 of 10⁴ cells with lac duplications produces a Lac⁺ colony. But this does not explain the results, because microcolonies would have been detected. During my respreading experiment (2), an average of one Lac⁺ colony per plate appeared each day; according to Roth and Andersson, this requires 100 microcolonies. If the steady-state number of microcolonies was, say, 4 (100 microcolonies/24 h), then after spreading, 400 new microcolonies (4 × 10⁴ cells/100) would arise. Since each of the 40 plates that were respread was subsequently examined at a magnification of ×30, had these microcolonies existed they would have been detected.

Another way in which Roth and Andersson explain how growth and amplification could evade detection is illustrated in their Fig. 2. They imagine that the only evidence that the Lac⁻ population is stable is from sampling cells between Lac⁺ colonies. This is incorrect. We have incubated Lac⁻ FC40 cells (without scavenger cells) in liquid lactose medium for 3 days without detecting any increase in turbidity or in cell number (1, 2). In similar experiments, no increase in the amount of lac DNA was detected (4). Not only do these results put severe limits on the amplification model, they also demonstrate that Lac⁻ FC40 cells do not grow on lactose, contrary to what Roth and Anderson assert (6).

John Cairns and I abandoned the amplification model because of the evidence summarized here, and also because Lac⁺ mutations arise at a constant rate during lactose selection. The only biologically feasible way that the rate could be constant is if amplification and deamplification of lac were always in equilibrium (2). If true, this would preclude the very growth advantage due to amplification that is central to the Roth-Andersson model.