OK, everyone recognizes the problem that we face with drug-resistant bacteria. MRSA (methicillin-resistant Staphylococcus aureus) is the most well-known variety, and it's bad news. Penicillin was introduced in the 1940s, and methicillin was brought to market in 1959, largely because so many infections were becoming resistant to penicillin by then. The first methicillin-resistant strains were noted in the early 1960s, which would seem to make the story pretty clear.
But it isn't. This paper has gone back into the details, sequencing the genomes of many of the earliest available resistant isolates, freeze-dried since their collection. And they find that MRSA bacteria actually predate the introduction of methicillin. So does that mean that they aren't a product of antibiotic use? Not at all - it's just that the antibiotic that brought them on was penicillin, not methicillin. Here's how that seems to have worked:
Penicillin resistance is largely due to beta-lactamase expression in bacteria, the well-known enzyme that cleaves the beta-lactam ring common to all such antibiotics. (Dosing along with a beta-lactamase inhibitor, such as clavulanic acid, is a widely used strategy that's often still effective). Methicillin is not a substrate for the enzyme (it's the dimethoxyaryl amide in its structure that does it). Beta-lactamase is encoded by the blaZ gene, but it turns out that even in the 1940s, an alternative resistance mechanism was spreading, too, the mec gene cassette. In that you find mecA, which codes for a penicillin-binding protein (PBP2a), and that's the one that does the job against methicillin.
The sequencing work here supports some other phylogenetic analysis, suggesting that these genes transferred once into S. aureus at some point in the mid-1940s, almost certainly from some coagulase-negative staphylococci. Strains with the blaZ gene were outcompeting wild-type bacteria by then, under selection pressure from the widespread use of penicillin, and strains with both the mec and blaZ genes were starting to outcompete them in turn. So it appears to have been that same use of penicillin that prepared the way for methicillin resistance. It worked fine against the earlier resistant strains that had only beta-lactamase as a weapon, but by the time it was introduced, those weren't the whole problem.
To make things worse, streptomycin resistance seems to have come in at almost the same time, from the same branch of the family. The paper concludes that:
It therefore appears that the first MRSA clone emerged, and developed resistance to two of the earliest antibiotics-streptomycin and penicillin-almost immediately after the S. aureus population would have been first exposed to them.
At the time of its discovery, the incidence of MRSA in the general population is likely to have been very low. This is demonstrated by the fact that screening of over 5000 samples at Public Health England yielded only three methicillin-resistant isolates. Therefore, it is likely that when methicillin was introduced to circumvent penicillin resistance in S. aureus, it did not select for emergence of MRSA at that time, but instead provided the selective pressure, which drove the nosocomial spread of a pre-existing variant, at a time when infection control measures in UK hospitals were limited.
This adds to the usual feeling that we get, looking back at the earlier use of antibiotics and wanting to yell at everyone "Don't do that!" Too late now. The whole story is a festival of unintended consequences, which we're living with now.