Mutation-biased adaptation is consequential even in large bacterial populations.

Because mutation rates vary widely across genomes and environments, natural selection is typically presented with highly biased variation. Yet, the idea that mutational tendencies can influence adaptation is still controversial. While mutation-driven adaptation has been observed in diverse taxa, critics contend it reflects small populations or weak-effect mutations. Therefore, the importance and generality of this phenomenon remain unclear, largely due to a lack of empirical tests across broad population-size gradients and multiple fitness-relevant traits. Here, we address this gap using a system in which two Escherichia coli mutator lineages evolve antibiotic resistance via two mutationally favoured, yet genetically distinct, routes. Simulations and experiments show that the scaling of mutation-biased adaptation with population size is complex, highly dependent on biological details, and - most critically - on how closely mutation bias aligns with selection. Contrary to the common view, we find that mutation-biased adaptation may not wane in large populations, but instead intensify depending on the bias. Crucially, we demonstrate that distinct mutation biases produce markedly different collateral sensitivity profiles to multiple antibiotics, even at large population sizes. Our findings suggest that mutation-biased adaptation may be widespread, with far-reaching and unpredictable consequences both within and beyond the original selective context.