Antibiotic Resistance: Natural Selection in Real Time

A dose is a selection experiment

Every antibiotic course is, whether anyone intends it or not, a controlled evolution experiment. Each drug has a minimum inhibitory concentration (MIC) — the lowest concentration that stops a given strain from growing. Above the MIC, susceptible cells die faster than they divide and the population collapses; below it, they simply carry on. A resistant mutant has a higher MIC than its susceptible neighbours, so the same dose that clears the wild type may do almost nothing to it. Put a colony under that pressure and you are not just killing bacteria — you are running a filter that keeps only the cells able to survive the concentration you chose.

Four ways to dodge a drug

Resistance is not one trick — it is a toolbox, and different lineages reach for different tools depending on the drug class:

1. Enzymatic inactivation   β-lactamases hydrolyse the β-lactam ring before it acts
2. Target modification      mecA → altered PBP2a with 100x lower drug affinity (MRSA)
3. Efflux pumps             AcrAB-TolC expels the drug before it reaches its target
4. Target bypass            vanA rebuilds the cell-wall precursor the drug can no longer bind

A single gene acquisition can flip a cell from susceptible to resistant outright — mecA alone is what turns ordinary Staphylococcus aureus into MRSA, because the altered penicillin-binding protein it encodes barely notices β-lactam drugs at all.

The race between two growth curves

Escherichia coli mutates at roughly 10⁻⁹ to 10⁻¹⁰ errors per base pair per replication; across a four-million base-pair genome that is a few new mutations per thousand divisions. Most do nothing useful. But once antibiotic pressure is applied, the population splits into two competing exponentials:

Susceptible:  dN_S/dt = (g_S − k) · N_S     k > g_S under the drug → decline
Resistant:    dN_R/dt =  g_R · N_R          g_R < g_S (fitness cost) → but not killed

Resistant cells are rare without the drug (~10⁻⁸ of the population) because
their lower growth rate g_R is normally a losing strategy. Add the antibiotic
and the susceptible term flips negative — now the slower-growing resistant
line is the only one still climbing.

The 2016 Kishony-lab "MEGA-plate" experiment made this visible on a single agar tray: E. coli crossed a thousand-fold antibiotic gradient in about ten days, colonising each zone step by step as slightly more resistant lineages arose from the population that had just adapted to the zone before it — evolution filmed, essentially, in real time.

Resistance that spreads without mutating

Vertical inheritance — parent to daughter cell — is the slow route. Horizontal gene transfer (HGT) is the fast one, and it can cross species boundaries in hours. Conjugation passes an entire resistance plasmid, sometimes carrying five to ten different resistance genes at once, through a physical channel between two cells. Transformation lets naturally competent species such as Streptococcus pneumoniae pick up naked DNA released by dead bacteria nearby. Transduction lets a bacteriophage accidentally package host DNA and inject it into the next bacterium it infects. The NDM-1 carbapenemase gene, first found in a single patient in 2008, had appeared on multiple plasmid backbones in bacteria on six continents within two years — a resistance gene, not a particular bacterium, going global.

Why "just take the full course" actually matters

Under-dosing, or stopping treatment as soon as symptoms fade, is close to the worst-case scenario from a selection standpoint: drug concentration lingers near the MIC for longer, which is precisely the regime where a partially resistant subpopulation survives while the fully susceptible majority is not yet cleared. A full course keeps concentration comfortably above the MIC for long enough that the susceptible population collapses before resistant variants can expand into the space it leaves behind. Combination therapy pushes the same idea further: pairing two drugs with independent targets means the probability of a cell acquiring resistance to both simultaneously is roughly the product of the two individual probabilities — around 10⁻⁸ × 10⁻⁸ = 10⁻¹⁶ — which is the logic behind tuberculosis's four-drug regimen.

Frequently asked questions

Try it live

Adjust the dose, schedule, mutation rate and fitness cost in Antibiotic Resistance and watch susceptible and resistant subpopulations rise and fall under your chosen pressure. Everything runs client-side, in your browser.