Our research focuses on how mutation and natural selection, together and apart, shape ecologically and medically relevant biological phenomena, such as the accumulation of antibiotic resistance, pathogen diversification, and cancer initiation and progression. Our highly interdisciplinary research combines evolutionary experiments, molecular biology, microbiology, evolutionary theory, genomics, metagenomics and computational biology. Major contributions to date include:

The evolution of antibiotic resistance in the absence of antibiotic exposure – The emergence and spread of antibiotic resistance poses a major and immediate threat to human health. While exposure to antibiotics is undoubtedly a major factor leading to the emergence of antibiotic resistance, increases in resistance frequencies were also observed in the absence of antibiotic exposure. A prevailing assumption was that stress conditions, such as starvation, increase global mutation rates thus leading to increased frequencies of antibiotic resistance. We conclusively demonstrated, that contrary to this long-held assumption, starvation-induced increases in resistance to antibiotics are not the result of global increases in mutation rates. Rather, we demonstrated that specific antibiotic resistance mutations are adaptive under starvation, in a manner that is independent of any antibiotic exposure. We then carried out a broad metagenomic survey and found that specific antibiotic resistance alleles segregate at alarmingly high frequencies within certain environments. These resistance alleles display patterns of segregation that appear to correlate better with their antibiotic-independent fitness effects than with antibiotic exposure. Combined, our results suggest that the antibiotic-independent fitness effects of antibiotic resistance alleles have the potential to significantly influence the accumulation of antibiotic resistance within natural environments.

Bacterial survival under prolonged resource exhaustion – Many bacteria, including the model bacterium Escherichia coli can survive for years in spent media, following resource exhaustion. This very likely reflects a requirement to survive for prolonged periods of resource exhaustion within natural environments as well. We carried out evolutionary experiments, followed by full genome sequencing of hundreds of evolved clones to unravel the machinery and cost of adaptation to prolonged resource exhaustion. Our results revealed that very specific genetic adaptations are required in order to enable bacteria to survive under prolonged resource exhaustion. Yet, despite very strong selection to accumulate specific genetic alterations, resource exhausted bacterial populations maintain within them great genetic heterogeneity. This enables bacterial populations to very rapidly adapt while under resource exhaustion, through temporally precise fluctuations in allele frequencies. The results of this study highlight the striking ability of bacterial populations to rapidly adapt, even under conditions expected to severely limit their growth.

Pathogen diversification through gene loss and gain – Some of the most important bacterial pathogens such as Mycobacterium tuberculosis evolve in the near absence of recombination. Such non-recombining bacterial species are often referred to as ‘clonal’. Clonal bacterial species are characterized by very low levels of gene sequence variation, which raises the question of how they still manage to generate substantial levels of phenotypic variation. We conducted a large-scale study to determine what sources of genetic variation are available for the evolution of both clonal and non-clonal bacterial species. Our study showed that while recombining species tend to evolve via a combination of changes to gene sequences, gene loss and gene gain, gene loss is the predominant source of genetic variation within clonal bacteria. Indeed, gene loss is so prevalent within clonal species as to allow them to develop levels of gene content variation comparable to those of far more diverged recombining species. We next turned to examining at greater depth the dynamics of gene loss and gene gain during the diversification of a large set of pathogenic bacterial species. We found that gene loss tends to be a mostly stochastic process, occurring within a pool of genes that are more readily lost [1]. We also found that genes gained by closely related pathogenic species are very frequently shared between these species in a manner that indicates the existence of a shared pool of frequently horizontally transferred genes.

The evolution of bacterial nucleotide composition – Bacteria display great variation in nucleotide composition, ranging from <25% to >75% GC content. What drives this variation has long been a mystery. In collaboration with Dmitri A. Petrov, we showed that mutation is universally AT-biased in bacteria, and that therefore variation in nucleotide content cannot be explained by variation in mutational biases. In a more recent study, carried out in collaboration with researchers from Drexel University, we demonstrated that the environment exerts selection on bacterial nucleotide composition, demonstrating that natural selection plays a role in driving nucleotide composition variation among bacteria.

Cancer evolution – Cancer is a short-term evolutionary process that occurs within the human body. We have shown that the dynamics of the cancer evolutionary process are very different from those of “normal” organismal evolution. Specifically, we have demonstrated that purifying selection (which removes harmful mutations from the population) is much less prominent within tumors and that positive selection, which favors mutations that are adaptive, is much more pronounced. In the context of cancer evolution, positively selected mutations are precisely those mutations that drive forward cancer progression, often referred to as ‘driver’ mutations. We have shown that within tumors positive selection is particularly strong on genes that are globally expressed across human tissues and that such globally expressed genes include many unidentified cancer drivers. We then expanded this study and showed that it is not just globally expressed genes that tend to be positively selected within tumors. Rather, it seems that adaptation in cancer occurs mostly via modifications to the most central gene functions. The ability of cancer to target the most central gene functions is related to its somatic nature. We are using this insight to develop methodologies for the identification of cancer driving genes and regulatory regions that are positively selected within tumors.

In an additional study, carried out in collaboration with molecular pathologists from Rambam hospital, we disentangled the effects of mutation and natural selection in determining the differential frequency distributions of KRAS cancer-driving substitutions between different types of tumors. This enabled us to identify specific KRAS driver substitutions that are more favored by selection in particular tumor types and demonstrate that such specifically favored driver substitutions tend to associate with worse clinical outcomes specifically within the tumor type in which they are more favored.