I also think that this is an interesting question. Some work along these
lines has been done with plasmids and transposons and I believe some is
discussed in one of the chapters in the book "Mobile DNA" edited by Berg
et al.
What I remember reading somewhere is that if you introduce a plasmid
which codes for antibiotic resistance into a bacterium, then you can
expect that replicating that plasmid will be costly in metabolic terms
and that in the absence of selection, the plasmid will be lost.
However, if you grow the bacterium in a chemostat for a lengthy period
of time WITH selection, the plasmid seems to adapt itself to its new
host, and be maintained afterwards without selection.
Getting back to the original article, if the resistance was of the
plasmid-encoded type, even if it dwindled to a small frequency in the
population, it would rapidly increase as soon as the antibiotic was used
again.
MH
