Evolution and ecology of antibiotic resistance

Antibiotic resistance is a growing public health concern. In ecological settings, primarily in the soil environment, naturally produced antibiotics have important ecological roles, mediating warfare and signaling interactions between microbial cells. While antibiotics appear in both clinical and ecological settings, the evolution of antibiotic-resistance is profoundly different in these two environments. In the clinic, use of antibiotics has been followed by rapid spread of resistance. In the natural environment, on the other hand, resistant and sensitive bacterial strains live in ecological balance–even a single grain of soil contains thousands of coexisting bacterial species with resistance and sensitivity to different antibiotics. The reasons for this difference are unknown. If we could understand the mechanisms that lead to ecological-balance in the natural setting, we might be able to identify new strategies for clinical treatment that would be less likely to generate rapid resistance.

One striking difference between the ecosystem of the soil and the clinical environment is that a very large combinatorial collection of antibiotics and other toxins is present in the soil, whereas in medical treatments one or a few drugs are used. Is it possible that the drug combinations in the soil are intrinsically better at keeping the evolution of resistance in check? We reported an observation that makes us think that this mechanism is at least possible in principle: “suppressive” antibiotic combinations, in which the effect of a drug combination is lower than the effect of one of the individual drugs alone, actually invert selection for resistance. Because the combination of drugs reduces growth less than does a single drug, mutants that acquire resistance to one drug actually grow slower than their non-resistant ancestors. We have identified the genetic mechanism underlying such suppressive drug interactions, showing that they result from non-optimal regulation of ribosomal genes under DNA stress (Bollenbach et al, Cell 2009). We further showed that such suppressive interactions makes resistant bacteria lose in competition with sensitive strains (Chait et al, Nature 2007), reduce the range of drug concentrations selective for resistance (Michel et al, PNAS 2008), and slow evolution through spontaneous mutations (Hegreness et al, PNAS 2008; see also recent review: Yeh et al, Nature Reviews Microbiology 2009).

Another interesting difference between antibiotics in the clinic and the natural environment is that in nature antibiotics have much more time to degrade chemically into a range of degradation compounds, which can profoundly affect selection for antibiotic resistance. We found that while tetracycline selects for tetracycline resistance, its soup of degradation products actually makes tetracycline resistant bacteria lose in competition with tetracycline sensitive strains. Since the degradation products are much longer lived, this selection against resistance vastly outweigh the initial selection in favor of resistance, leading to an overall net selection in favor of drug sensitivity (Palmer et al, Nature Chemical Biology, 2010).

We are now developing screens to directly identify novel chemical entities in the natural environment that can select against antibiotic resistance.

Together, these efforts provide a starting point for a deeper and more comprehensive investigation into the mechanisms that can keep antibiotic-resistance in check in the wild and in the clinic.

Genetic and pharmacological interactions

We are combining theoretical and experimental approaches to study epistasis networks – networks that describe how perturbations (mutations or drugs) in a given biological system are combined to aggravate or alleviate the phenotypic consequences of each other. Such epistatic interactions, fundamental in various evolutionary processes, can help elucidating the functional organization of complex genetic architectures. Sensitive identification of epistasis requires unique techniques that we developed for precise microbial growth measurements: bioluminescence detection (Kishony & Leibler, J. Biol 2003); robotically-automated competition between microbial populations differentially tagged with fluorescence markers (Hegreness et al, Science 2006; Deluna et al, Nature Genetics 2008); and time-lapse optical scanning of colony growth coupled with image-analysis (Michel et al, PNAS 2008). We have also used the computational method of flux balance analysis (FBA) to study the epistasis network of yeast metabolism (Segre’ et al, 2005). Our results show that the epistasis network posses a special property, which we term “monochromaticity”, whereby functional gene modules interact with each other with purely aggravating or purely alleviating epistatic links. This property extends the concept of epistasis from the gene-gene level to the system level. The new definition for identifying functional modules is implemented in our Prism algorithm. In drug networks, the same conceptual method allows classification of drugs by their mechanism of actions based only on the properties of their mutual interactions (Yeh et al, 2006).

 Kishony lab