细菌从病毒敏感到病毒抗性的突变

When a bacterial culture is attacked by a bacteriophage, almost all the cells die, but a few resistant cells survive and found resistant colonies. Two explanations were on the table. By the acquired-immunity view, contact with the virus itself induces a small, fixed fraction of cells to become resistant. By the mutation view, rare resistant mutants already exist in the culture, having arisen by chance during earlier growth, independently of the virus.

Luria and Delbrück realised the two views make different statistical predictions, and that the difference shows up not in the average number of survivors but in how that number fluctuates from one culture to another. If resistance is induced at the moment of exposure, each culture is an independent series of rare events and the survivor counts should follow a Poisson distribution, with variance about equal to the mean. If resistance is inherited from a chance mutation during growth, a mutation that happened early is passed to a large clone of descendants, so an occasional culture carries a huge 'jackpot' of resistant cells and the counts fluctuate far more widely than Poisson allows.

They grew many small parallel cultures of Escherichia coli from tiny inocula, let each grow undisturbed, then plated each entire culture on a lawn of bacteriophage and counted the resistant colonies. As a control, they sampled a single large culture many times over. The single culture, sampled repeatedly, gave counts that varied only by sampling (Poisson) error; the independent parallel cultures gave counts that fluctuated enormously, including rare jackpots.

The wide fluctuation between independent cultures was incompatible with acquired immunity and matched the mutation hypothesis. From the relationship between the mutation rate and the distribution of survivors — in particular the fraction of cultures with no resistant cells at all — the authors could even estimate the rate at which the resistance mutation occurs per cell division. Resistance, they concluded, arises by spontaneous mutation before the virus is ever applied; the virus only selects the mutants already present.

The full paper develops the probability theory of the mutant distribution, tabulates the parallel-culture and single-culture data, and derives the mutation rate; it runs to about twenty pages and is available in full at the source below.