A spotlight on bacterial mutations for 75 years

In the debate about how bacterial mutations arise, an experiment in 1943 showed that they can occur spontaneously and independently of a selection pressure. This study also popularized the use of maths-driven analysis of biological data.
Manoshi S. Datta is in the Faculty of Biology, Technion–Israel Institute of Technology, Haifa 3200003, Israel.

Search for this author in:

Roy Kishony is in the Faculty of Biology, Technion–Israel Institute of Technology, Haifa 3200003, Israel.

Search for this author in:

Do bacteria acquire mutations randomly, or do mutations arise adaptively as a direct response to environmental pressures? This question has wide implications in areas ranging from evolution to the treatment of bacterial infections. In 1943, writing in Genetics, Luria and Delbrück1 revealed, by a combination of experimental analysis and profound mathematical insight, that bacteria evolve through random mutations that arise independently of an environmental stress, and that occur even before bacteria encounter such selective conditions. Their study was a milestone in a debate about the nature and causes of bacterial evolution that is still ongoing. Moreover, this work has inspired the fields of microbial evolution and quantitative biology.

Luria and Delbrück worked at a time when scientists disagreed on the fundamental nature of bacterial evolution2, despite tremendous advances in molecular biology and microbiology. For plants and animals, there was a general consensus that, consistent with Charles Darwin’s theory of evolution, natural selection acted on mutations that arose randomly, regardless of their benefit to the organism. However, the unusual nature of bacterial genetics — such as the absence of sexual reproduction — sparked a vigorous debate about whether the principles that drive animal evolution also apply to bacteria (see The main alternative hypothesis was Lamarckian evolution, named after the French biologist Jean-Baptiste Lamarck. In this model, the specific mutations that provide an advantage to an organism are acquired directly in response to the organism’s environment3.

For present-day microbiologists, this debate might seem strangely contrived — after all, if other organisms evolve in a manner consistent with the Darwinian principles of randomly occurring organismal variation that selection can act on, why should bacteria be an exception? Yet, it’s worth having sympathy for our scientific predecessors. Even though we now accept that bacteria evolve through Darwinian mechanisms, ‘quasi-Lamarckian’ processes of bacterial evolution are still being discovered and debated46.

Luria and Delbrück themselves encountered some difficulties when they entered the debate about how bacterial evolution occurs. To establish an approach to study mutations in bacteria, they allowed individual Escherichia coli cells to grow into large populations in individual test tubes, and added the cells from each of these tubes to Petri dishes containing agar coated with viruses known to kill the bacteria. Luria and Delbrück monitored the number of visible bacterial colonies on each of the plates. Each of these virus-resistant colonies arises from a cell and its descendants that had a mutation enabling the cells to survive the viral attack. Yet, for a simple experiment, their results were initially confusing: the number of colonies was highly variable between the different plates, a result that the authors initially attributed to an experimental error (see But in a moment of clarity, Luria realized7 that the high variability in the number of bacterial colonies might be an important clue, not an error.

Let’s consider the experimental variance in the number of virus-resistant colonies per Petri dish expected under the process of either adaptive or random mutation. If mutations arise by an adaptive process, each bacterial cell would have a chance of acquiring a resistance mutation only on encountering the virus. Assuming each cell’s chance of becoming resistant is small, the prediction would be that the number of virus-resistant colonies per Petri dish would vary according to a Poisson distribution (a standard probability distribution for random events, in which the standard deviation of the data equals the square root of the mean).

But, if evolution is driven by random mutations, mutations that confer viral resistance would arise during the growth of the bacterial population before viral exposure. In this case, the experimental variance in the number of virus-resistant bacterial colonies between different Petri dishes would be much higher than in the adaptive-mutation scenario, because the number of virus-resistant bacteria in a given test tube would depend on the random timing of when mutations occurred. A single virus-resistance mutation that occurred early in the growth of the bacterial population would result in a large number of virus-resistant bacterial descendants of the original mutated cell, whereas mutations that arose much later during the growth of the bacterial culture, just before viral encounter, would produce many fewer virus-resistant bacteria.

On the basis of this insight, Luria and Delbrück generated a statistical distribution (the Luria–Delbrück distribution) to describe the prevalence of virus-resistant bacterial mutants that would be expected if mutations arose randomly before the bacterial population came under selective pressure from the virus. Compared with a Poisson distribution expected for adaptive mutations, this Luria–Delbrück distribution has a long ‘tail’ at the end of the distribution pattern. In the context of the authors’ experiments, this tail would correspond to Petri dishes that have a high number of bacterial colonies, corresponding to early mutational events that lead to a large number of mutant descendants.

The 1943 paper reported the results of the authors’ experiments, termed fluctuation tests, that took this mathematical approach to analyse the number of virus-resistant colonies in E. coli populations. The authors’ findings were consistent with mutations following a Luria–Delbrück distribution rather than a Poisson distribution, demonstrating that bacterial mutations arose randomly, and independently of an encounter with a virus.

Luria and Delbrück’s work shaped subsequent studies of biology and evolution in many ways. Luria himself was reported as saying that their fluctuation test removed bacteria from “the last stronghold of Lamarckism” (see The fluctuation test is still a standard procedure for accurately measuring mutation rates in diverse systems, from bacteria8 and yeast9 to cancer cells10. Their study also popularized the use of E. coli and the viruses that attack it as a simple experimental model system for biology11. Beyond its direct impact in laboratories, the experiment became a textbook example of how mathematical thinking combined with simple experimentation can lead to profound biological insights12. For their contributions to bacterial and viral genetics, Luria and Delbrück won the Nobel Prize in Physiology or Medicine in 1969 (which they shared with the biologist Alfred Hershey).

Their insight into mutational processes also has implications in settings such as the clinic. In analogy to the original experiment, imagine a population of patients who have the same type of bacterial infection and who are being treated with the same antibiotic (the antibiotic replaces the virus as the selection pressure here). According to the random-mutation model, even if all else is equal among the patients, the number of antibiotic-resistant bacterial mutants initially present will vary highly between the patients, which could lead to markedly variable treatment outcomes. Because such high inherent variability in treatment efficiency reflects resistance mutations arising in a population before treatment, using DNA sequencing or other types of analysis to identify the presence and number of antibiotic-resistant bacterial mutants before treatment could improve our ability to predict treatment outcome.

Did the Luria and Delbrück study really close the door on Lamarckism? As far as bacteria are concerned, the answer is much more complicated than the duo could probably ever have anticipated.

It is undeniable today that randomly occurring mutations and natural selection are central tenets of how bacterial evolution occurs5. However, scientists are uncovering and debating an increasing array of other evolutionary processes at work in bacteria, some of which are suspiciously Lamarckian in character46. For example, we now know that the genome-wide mutation rate, and even the mutation rates of specific genes, can be shaped by evolution and affected by the environment13–15. An even more striking example is bacterial adaptation through the CRISPR–Cas viral-defence system, in which bacteria can incorporate viral genetic material into their own genomes and use it, as an adaptive mechanism, to protect themselves and their descendants against current and subsequent viral attacks16,17. These quasi-Lamarkian mechanisms presumably evolved by random mutations and natural selection. They do not necessarily undermine the lessons learnt from Luria and Delbrück’s work, but rather, show the power of evolution to sculpt living organisms in endlessly interesting ways.

It is intriguing to imagine an alternative scientific history that might have occurred if Luria and Delbrück had stumbled upon one of these quasi-Lamarkian mechanisms. The CRISPR–Cas defence mechanism is mainly repressed in the E. coli that they studied, but it is active in other bacterial species, such as Streptococcus thermophilus. A fun challenge would be to repeat the Luria–Delbrück experiment under conditions that might favour the evolution of resistance by such adaptive mechanisms, for example by replacing E. coli with S. thermophilus. Would the distribution of the number of resistant mutants indicate random or adaptive mutations? What would Luria and Delbrück have concluded had they used a species that had the CRISPR–Cas system? The contingency of this historic choice underscores the fact that, like evolution, science perhaps also progresses both adaptively and randomly.

Nature 563, 633-644 (2018)


  1. 1.

    Luria, S. E. & Delbrück, M. Genetics 28, 491–511 (1943).

  2. 2.

    Zheng, Q. Chance 23 (2), 15–18 (2010).

  3. 3.

    Lenski, R. E. PLoS Genet. 13, e1006668 (2017)

  4. 4.

    Koonin, E. V. & Wolf, Y. I. Biol. Direct 4, 42 (2009).

  5. 5.

    Sniegowski, P. D. & Lenski, R. E. Annu. Rev. Ecol. Syst. 26, 553–578 (1995).

  6. 6.

    Rosenberg, S. M. Nature Rev. Genet. 2, 504–515 (2001).

  7. 7.

    Luria, S. E. A Slot Machine, a Broken Test Tube: An Autobiography (Harper & Row, 1984).

  8. 8.

    Bell, G. & MacLean, C. Trends Microbiol. 26, 471–483 (2018).

  9. 9.

    Farlow, A. et al. Genetics 201, 737–744 (2015).

  10. 10.

    Powers, S. & Pollack, R. E. Nature Rev. Cancer 16, 266–270 (2016).

  11. 11.

    Cairns, J., Stent, G. S. & Watson, J. D. Phage and the Origins of Molecular Biology (Cold Spring Harb. Press, 2007).

  12. 12.

    Murray, A. Genetics 202, 367–368 (2016).

  13. 13.

    Tenaillon, O., Denamur, E. & Matic, I. Trends Microbiol. 12, 264–270 (2004).

  14. 14.

    Moxon, E. R., Rainey, P. B., Nowak, M. A & Lenski, R. E. Curr. Biol. 4, 24–33 (1994).

  15. 15.

    Wielgoss, S. et al. Proc. Natl Acad. Sci. USA 110, 222–227 (2013).

  16. 16.

    Marraffini, L. A. Nature 526, 55–61 (2015).

  17. 17.

    Koonin, E. V. & Wolf, Y. I. Biol. Direct 11, 9 (2016).

Download references

Nature Briefing

An essential round-up of science news, opinion and analysis, delivered to your inbox every weekday.