How bacteria outrun antibiotics

Seed E. coli along the edges of a dish whose centre holds three thousand times the dose of trimethoprim that stops the ordinary strain from growing. Ten days later the bacteria are spreading across the centre. Nothing in them learned anything. Chance kept producing variants, and the drug stopped every cell except the ones that could grow in it.

What is being claimed

A population of bacteria is enormous, and each cell division copies the genome with the occasional error. A few of those errors happen to make a cell harder to kill. Put a drug in the way and it does not aim mutations at resistance; its main effect is to stop the cells that cannot grow. What is left is made of the lucky ones, and their descendants. Resistance is a property of the bacteria, never of the patient.

Why it is worth knowing

Resistance arrived almost as soon as the drugs did. Davies and Davies record that sulfonamide resistance was first reported in the late 1930s, “and the same mechanisms operate some 70 years later.” Fleming discovered penicillin in 1928; in 1940, several years before it was used as a treatment, two members of the penicillin team had already identified a bacterial enzyme that destroys it. So the genes are old. Many, the review notes, are “components of natural microbial populations,” and it asks “Which came first, the antibiotic or resistance?” What human use of antibiotics changed is how far and how fast those genes spread.

Watching it happen

Baym and colleagues built a dish 120 by 60 cm, which they call the MEGA-plate, and filled it with bands of agar. The drug-free outer bands were seeded with E. coli. Moving inwards, each band held ten times more drug than the last: 0, 3, 30, 300 and 3000 times the minimum inhibitory concentration (MIC, the lowest dose that stops the ordinary strain from growing) of trimethoprim, or 0, 20, 200, 2000 and 20000 times the MIC of ciprofloxacin. Bacteria swim towards food, so the colony spreads until it meets a band it cannot grow in, and there it stalls.

In their words, “As resistant mutants arise in the population, their descendants migrate into the next step.” When those descendants stall at the next wall, secondary mutations arise and the process repeats. Cells taken from the innermost band had MICs 10,000 times higher than their ancestors for trimethoprim, after 10 days, and 100,000 times higher for ciprofloxacin, after 12.

Two separate steps

Keep the two halves apart. The variation comes first: a copying error is not aimed at the drug, and the error that would help is not favoured over any other, so a mutant can arise far from any drug long before it is useful. Selection comes second, and selection is the part that matters here: at each band the drug stops every cell whose tolerance is too low. (At low doses antibiotics can also speed up gene exchange between cells, Davies and Davies note, but they do not steer which mutations appear.) The drug decides which mutants are left, not which mutants appear.

The steps between bands matter. When the plate went straight from no drug to the highest level, the bacteria “were unable to adapt directly.” A middle band made adaptation possible, unless the middle band was itself too high. Intermediate doses let partly resistant mutants build up large numbers, and a large population is what makes the next rare mutant likely.

The most resistant cell does not always win either. Many resistance mutations slowed growth, and later compensatory mutations restored it. Some of those compensated mutants were more resistant than the cells leading the front, but they arose behind it, in ground already taken, and stayed “trapped.” The authors conclude that what drives the population is the mutants “that are both sufficiently fit and arise sufficiently close to the advancing front.” They also found strains with a broken proofreading gene, dnaQ, carrying more than 60 mutations against fewer than 12 in the rest. Those mutator strains arose at least six times independently.

Where else resistance comes from

A new mutation is only one route. Davies and Davies tabulate how each drug class is defeated, and the entries fall into four families (our grouping): destroy the drug, pump it out, change its target, or rebuild the targeted machinery so the drug has nothing to hit. Genes can also move between cells. Transferable resistance was identified in Japan in the mid-1950s, with genes passed by conjugation “throughout an entire population” of pathogens, and the review judges such transfer in nature “probably several orders of magnitude” more frequent than in the lab.

What it does not show

The MEGA-plate is a laboratory device; its authors say it “is not intended to directly simulate natural or clinical settings.” And a resistant strain, on this plate or in one patient during one course of treatment, is not a new species. It is still E. coli, with some changed genes.

In short

Random copying errors supply variants before any drug meets them. The drug only stops the cells that cannot grow. Give the survivors a moderate dose to multiply in, and the next rare mutant arrives, then the next, until a 3000-fold dose is no barrier.