Science, 25. Feb 2000, Vol 287, No. 5457, page 1409
Mice Are Not Furry Petri Dishes
James Bull and Bruce Levin
It is not as though microbiologists really believe that what is true in vitro is also true
in vivo. Like most other scientists, they realize that progress depends on developing
model systems that, although not faithful replicas of the in vivo environment, enable the
easy and repeatable study of biological phenomena. Microbial culture methods have
facilitated the isolation and selective propagation of microbes and control of their
genetics and metabolism. Unfortunately, it has become clear that interactions between
microbes and the complex environment of their hosts cannot be gleaned from in vitro
studies alone. The genes expressed by microbes inhabiting a mammalian host are different
from those expressed by microbes living in a petri dish (1). On page 1479 of this issue,
Björkman et al. (2) now show that the processes of mutation and selection (the basic
elements of evolution) in bacterial populations differ depending on whether the bacteria
grow in vivo or in vitro.
Mutations that enable bacteria and other microbes to grow in the presence of antimicrobial
agents commonly engender a cost that is manifest as a reduced growth rate (competitive
disadvantage) in environments where the drug is absent. The cost incurred by drug
resistance has been touted as a route to combatting the ever-increasing numbers of
drug-resistant pathogenic microbes. The rationale is that because resistance incurs a
cost, its incidence will wane if we administer antimicrobial drugs more prudently (3-5).
Unfortunately, bacteria and viruses adapt to the cost of drug resistance through secondary
mutations that compensate for the loss of fitness but usually do not reduce the level of
resistance (6-9). The Björkman et al. study provides compelling evidence that the process
of adaptation to the costs of antibiotic resistance in Salmonella are different depending
on whether the bacteria grow in mice or culture medium (broth).
A mutation in elongation factor G (which decreases protein synthesis and slows growth)
confers resistance to fusidic acid on Salmonella. The authors recovered 26 independent
compensatory mutations and two revertants when Salmonella with this drug-resistance
mutation were grown in culture broth. (Revertants are bacteria in which the
drug-resistance mutation has been lost and the original wildtype DNA sequence has been
restored.) In contrast, 11 compensatory mutations and 14 revertants were obtained when
bacteria with the drug-resistance mutation were grown in mice. The amino acid
substitutions in the 11 mouse-derived compensatory mutations differed from those in the
26 broth-derived compensatory mutations. In general, most of the compensatory mutations
in bacteria grown in mice provided only partial recovery of fitness, explaining the
preponderance of drug-sensitive revertants. These results are somewhat surprising because
it would be expected that the effects of compensatory mutations (which correct the defects
in protein synthesis that accompany fusidic acid resistance) would be independent of the
bacteria's environment.
The investigators also looked at streptomycin resistance conferred by a mutation in the
rpsL gene, which encodes ribosomal protein S12. Although broth and mice were not treated
with streptomycin, the adaptation to the costs of streptomycin resistance was solely
through compensatory mutations and not through reversion to a drug-sensitive phenotype.
Intriguingly, in broth bacteria all 14 compensatory changes were located in the rpsD and
rpsE genes (extragenic), and not in the rpsL gene. In contrast, in all 10 mice studied,
the compensatory mutations were located in rpsL (intragenic), within the same codon. The
original rpsL drug-resistance mutation was a substitution (AAA to AAC) at the 42nd codon;
two base changes converting AAC to AGA (which maintained drug resistance) compensated for
the effects of this substitution. Unlike the case for fusidic acid resistance compensatory
mutations, all of the streptomycin resistance compensatory mutations were accompanied by
relatively high bacterial fitness regardless of whether bacteria were grown in broth or in
mice. This led the authors to conclude that the differences between mice and broth
Salmonella in the evolution of streptomycin resistance compensatory mutations lay in the
mutation process itself, rather than in selection of mutants. An immediate implication of
this finding is that making predictions about the evolution of drug-resistant pathogens in
vivo requires that at least some experiments be performed in vivo. Despite the benefits of
in vitro experiments, we cannot yet abandon animal models.
Regarding the problems of drug resistance, the results of the Björkman study cannot be
interpreted in an optimistic light. In the case of streptomycin, at least, all of the
adaptations to the cost of resistance were through amelioration of the drug-resistant
mutations rather than by reversion to drug sensitivity. These findings also have
implications beyond drug resistance. They suggest that in vivo the mutants generated are,
quite different and the mutation rate is higher than in vitro. Do compensatory mutations
contribute to both acquired resistance in drug-treated hosts and the virulence of
infecting microbes? Evolution of bacterial population in an infected host may be
completely different from that taking place in a habitat outside of the host. For example,
the same gene may be favored in one habitat and selected against in the other. The
important findings of Björkman and co-workers raise a number of questions about why
mutation and selection, the fundamental elements of bacterial evolution, are different in
vivo and in vitro. Answering those questions should keep microbiologists deliciously
occupied for some time to come.