Search

The rampant variability in codon bias existing between bacterial genomes is expected to interfere with horizontal gene transfer (HGT), a phenomenon that drives bacterial adaptation. However, delineating the constraints imposed by codon bias on functional integration of the transferred genes is complicated by multiple genomic and functional barriers controlling HGT, and by the dependence of the evolutionary outcomes of HGT on the host’s environment. Here, we designed an experimental system in which codon composition of the transferred genes is the only variable triggering fitness change of the host. We replacedE. coli’s chromosomalfolAgene encoding dihydrofolate reductase, an essential enzyme that constitutes a target for trimethoprim, with combinatorial libraries of synonymous codons offolAgenes from trimethoprim-sensitiveListeria grayiand trimethoprim-resistantNeisseria sicca.The resulting populations underwent selection at a range of trimethoprim concentrations, and the ensuing changes in variant frequencies were used to infer the fitness effects of the individual combinations of codons. We found that when HGT causes overstabilization of the 5’-end mRNA, the fitness contribution of mRNA folding stability dominates over that of codon optimality. The 5’-end overstabilization can also lead to mRNA accumulation outside of the polysome, thus preventing the decay of the foreign transcripts despite the codon composition-driven reduction in translation efficiency. Importantly, the fitness effects of mRNA stability or codon optimality become apparent only at sub-lethal levels of trimethoprim individually tailored for each library, emphasizing the central role of the host’s environment in shaping the codon bias compatibility of horizontally transferred genes.

The stability and dynamics of ecological communities are dictated by interaction networks typically quantified at the level of species.1–10But how such networks are influenced by intra-species variation (ISV) is poorly understood.11–14Here, we use ~500,000 chromosomal barcodes to track high-resolution intra-species clonal lineages ofEscherichia coliinvading mice gut with the increasing complexity of gut microbiome: germ-free, antibiotic-perturbed, and innate microbiota. By co-clustering the dynamics of intra-species clonal lineages and those of gut bacteria from 16S rRNA profiling, we show the emergence of complex time-dependent interactions betweenE. coliclones and resident gut bacteria. With a new approach, dynamic covariance mapping (DCM), we differentiate three phases of invasion in susceptible communities: 1) initial loss of community stability asE. colienters; 2) recolonization of some gut bacteria; and 3) recovery of stability withE. colicoexisting with resident bacteria in a quasi-steady state. Comparison of the dynamics, stability and fitness from experimental replicates and different cohorts suggest that phase 1 is driven by mutations inE. colibefore colonization, while phase 3 is byde novomutations. Our results highlight the transient nature of interaction networks in microbiomes driven by the persistent coupling of ecological and evolutionary dynamics.One-Sentence SummaryHigh-resolution lineage tracking and dynamic covariance mapping (DCM) define three distinct phases during early gut microbiome invasion.

Recent advances in AI-based 3D protein structure prediction could help address health-related questions, but may also have far-reaching implications for evolution. Here we discuss the advantages and limitations of high-quality 3D structural predictions by AlphaFold2 in unravelling the relationship between protein properties and their impact on fitness, and emphasize the need to integrate in silico structural predictions with functional genomic studies.

Natural product methyltransferases (NPMTs) represent an emerging class of enzymes that can be of great use for the structural and functional diversification of bioactive compounds, such as the strategic modification of C-, N-, O- and S-moieties. To assess the activity and the substrate scope of the ever-expanding repertoire of NPMTs, a simple, fast, and robust assay is needed. Here, we report a continuous spectroscopic assay, in which S-adenosyl-L-methionine-dependent methylation is linked to NADH oxidation through the coupled activities of S-adenosyl-L-homocysteine (SAH) deaminase and glutamate dehydrogenase. The assay is highly suitable for a high-throughput evaluation of small molecule methylation and for determining the catalytic parameters of NPMTs under conditions that remove the potent inhibition by SAH. Through the modular design, the assay can be extended to match the needs of different aspects of methyltransferase cascade reactions and respective applications.

Homomers are prevalent in bacterial proteomes, particularly among core metabolic enzymes. Homomerization is often key to function and regulation, and interfaces that facilitate the formation of homomeric enzymes are subject to intense evolutionary change. However, our understanding of the molecular mechanisms that drive evolutionary variation in homomeric complexes is still lacking. How is the diversification of protein interfaces linked to variation in functional regulation and structural integrity of homomeric complexes? To address this question, we studied quaternary structure evolution of bacterial methionine S-adenosyltransferases (MATs)-dihedral homotetramers formed along a large and conserved dimeric interface harboring two active sites, and a small, recently evolved, interdimeric interface. Here, we show that diversity in the physicochemical properties of small interfaces is directly linked to variability in the kinetic stability of MAT quaternary complexes and in modes of their functional regulation. Specifically, hydrophobic interactions within the small interface of Escherichia coli MAT render the functional homotetramer kinetically stable yet impose severe aggregation constraints on complex assembly. These constraints are alleviated by electrostatic interactions that accelerate dimer-dimer assembly. In contrast, Neisseria gonorrhoeae MAT adopts a nonfunctional dimeric state due to the low hydrophobicity of its small interface and the high flexibility of its active site loops, which perturbs small interface integrity. Remarkably, in the presence of methionine and ATP, N. gonorrhoeae MAT undergoes substrate-induced assembly into a functional tetrameric state. We suggest that evolution acts on the interdimeric interfaces of MATs to tailor the regulation of their activity and stability to unique organismal needs.

Methionine S-adenosyltransferases (MATs) are predominantly homotetramers, comprised of dimers of dimers. The larger, highly conserved intradimeric interface harbors two active sites, making the dimer the obligatory functional unit. However, functionality of the smaller, more diverged, and recently evolved interdimeric interface is largely unknown. Here, we show that the interdimeric interface of Ureaplasmaurealiticum MAT has evolved to control the catalytic activity and structural integrity of the homotetramer in response to product accumulation. When all four active sites are occupied with the product, S-adenosylmethionine (SAM), binding of four additional SAM molecules to the interdimeric interface prompts a ∼45° shift in the dimer orientation and a concomitant ∼60% increase in the interface area. This rearrangement inhibits the enzymatic activity by locking the flexible active site loops in a closed state and renders the tetramer resistant to proteolytic degradation. Our findings suggest that the interdimeric interface of MATs is subject to rapid evolutionary changes that tailor the molecular properties of the entire homotetramer to the specific needs of the organism.

S-adenosylmethionine lyase (SAMase) of bacteriophage T3 degrades the intracellular SAM pools of the host E. coli cells, thus inactivating a crucial metabolite involved in plethora of cellular functions, including DNA methylation. SAMase is the first viral protein expressed upon infection and its activity prevents methylation of the T3 genome. Maintenance of the phage genome in a fully unmethylated state has a profound effect on the infection strategy — it allows T3 to shift from a lytic infection under normal growth conditions to a transient lysogenic infection under glucose starvation. Using single-particle Cryo-EM and biochemical assays, we demonstrate that SAMase performs its function by not only degrading SAM, but also by interacting with and efficiently inhibiting the host’s methionine S-adenosyltransferase (MAT) — the enzyme that produces SAM. Specifically, SAMase triggers open-ended head-to-tail assembly of E. coli MAT into an unusual linear filamentous structure in which adjacent MAT tetramers are joined together by two SAMase dimers. Molecular dynamics simulations together with normal mode analyses suggest that the entrapment of MAT tetramers within filaments leads to an allosteric inhibition of MAT activity due to a shift to low-frequency high-amplitude active site-deforming modes. The amplification of uncorrelated motions between active site residues weakens MAT’s ability to withhold substrates, explaining the observed loss of function. We propose that the dual function of SAMase as an enzyme that degrades SAM and as an inhibitor of MAT activity has emerged to achieve an efficient depletion of the intracellular SAM pools.IMPORTANCESelf-assembly of enzymes into filamentous structures in response to specific metabolic cues has recently emerged as a widespread strategy of metabolic regulation. In many instances filamentation of metabolic enzymes occurs in response to starvation and leads to functional inactivation. Here, we report that bacteriophage T3 modulates the metabolism of the host E. coli cells by recruiting a similar strategy — silencing a central metabolic enzyme by subjecting it to phage-mediated polymerization. This observation points to an intriguing possibility that virus-induced polymerization of the host metabolic enzymes might be a common mechanism implemented by viruses to metabolically reprogram and subdue infected cells.

The relationship between sequence variation and phenotype is poorly understood. Here, we use metabolomic analysis to elucidate the molecular mechanism underlying the filamentous phenotype of E. coli strains that carry destabilizing mutations in dihydrofolate reductase (DHFR). We find that partial loss of DHFR activity causes reversible filamentation despite SOS response indicative of DNA damage, in contrast to thymineless death (TLD) achieved by complete inhibition of DHFR activity by high concentrations of antibiotic trimethoprim. This phenotype is triggered by a disproportionate drop in intracellular dTTP, which could not be explained by drop in dTMP based on the Michaelis–Menten‐like in vitro activity curve of thymidylate kinase (Tmk), a downstream enzyme that phosphorylates dTMP to dTDP. Instead, we show that a highly cooperative (Hill coefficient 2.5) in vivo activity of Tmk is the cause of suboptimal dTTP levels. dTMP supplementation rescues filamentation and restores in vivo Tmk kinetics to Michaelis–Menten. Overall, this study highlights the important role of cellular environment in sculpting enzymatic kinetics with system‐level implications for bacterial phenotype., E. coli strains carrying mutations in Dihydrofolate Reductase (DHFR) display a pronounced filamentous phenotype. The origin of this phenotype can be explained by the in vivo kinetics of another downstream enzyme Thymidylate Kinase (TmK).

Nature Ecology & Evolution, 4, ISSN:2397-334X

Evolutionary dynamics in large asexual populations is strongly influenced by multiple competing beneficial lineages, most of which segregate at very low frequencies. However, technical barriers to tracking a large number of these rare lineages in bacterial populations have so far prevented a detailed elucidation of evolutionary dynamics. Here, we overcome this hurdle by developing a chromosomal-barcoding technique that allows simultaneous tracking of approximately 450,000 distinct lineages in Escherichia coli, which we use to test the effect of sub-inhibitory concentrations of common antibiotics on the evolutionary dynamics of low-frequency lineages. We find that populations lose lineage diversity at distinct rates that correspond to their antibiotic regimen. We also determine that some lineages have similar fates across independent experiments. By analysing the trajectory dynamics, we attribute the reproducible fates of these lineages to the presence of pre-existing beneficial mutations, and we demonstrate how the relative contribution of pre-existing and de novo mutations varies across drug regimens. Finally, we reproduce the observed lineage dynamics by simulations. Altogether, our results provide a valuable methodology for studying bacterial evolution as well as insights into evolution under sub-inhibitory antibiotic levels.

Linking the molecular effects of mutations to fitness is central to understanding evolutionary dynamics. Here we establish a quantitative relation between the global effect of mutations on the E. coli proteome and bacterial fitness. We created E. coli strains with specific destabilizing mutations in the chromosomal folA gene encoding dihydrofolate reductase (DHFR) and quantified the ensuing changes in the abundances of 2000+ E. coli proteins in mutant strains using tandem mass tags with subsequent LC-MS/MS. mRNA abundances in the same E. coli strains were also quantified. The proteomic effects of mutations in DHFR are quantitatively linked to phenotype: the standard deviations of the distributions of logarithms of relative (to wild-type) protein abundances anti-correlate with bacterial growth rates. Proteomes hierarchically cluster first by media conditions, and within each condition, by the severity of the perturbation to DHFR function. These results highlight the importance of a systems-level layer in the genotype-phenotype relationship.

Bridging the gap between the molecular properties of proteins and organismal/population fitness is essential for understanding evolutionary processes. This task requires the integration of the several physical scales of biological organization, each defined by a distinct set of mechanisms and constraints, into a single unifying model. The molecular scale is dominated by the constraints imposed by the physico-chemical properties of proteins and their substrates, which give rise to trade-offs and epistatic (non-additive) effects of mutations. At the systems scale, biological networks modulate protein expression and can either buffer or enhance the fitness effects of mutations. The population scale is influenced by the mutational input, selection regimes, and stochastic changes affecting the size and structure of populations, which eventually determine the evolutionary fate of mutations. Here, we summarize the recent advances in theory, computer simulations, and experiments that advance our understanding of the links between various physical scales in biology.

Several genes exhibit gene dosage toxicity yet its molecular underpinnings remain unknown. Here we demonstrate that overexpression of DHFR inE. colicauses toxic metabolic imbalance triggered by interactions with several enzymes involved in 1-carbon metabolism, in particular GlyA and PurH. DHFR overexpression partially inhibits activity of these enzymes, but at physiological concentrations, PurH-DHFR interaction enhances catalytic efficiency of DHFR, implying a functional interactionin vivo. Surprisingly, overexpression of orthologous DHFRs from other bacterial species caused minimal metabolic and fitness perturbations, despite pulling out more interacting partners than overexpressed endogenous DHFR. Orthologous DHFRs were less potent in inhibitingE. coliGlyA and PurH, or gaining a catalytic improvement upon interaction with PurH, indicating a partial loss of interaction specificity due to evolutionary divergence. This study shows how protein overexpression perturbs a dynamic network of weak yet potentially functional PPI with consequences for the metabolic state of cells and their fitness.

We address recent developments in the area of laboratory, or directed evolution, with a focus on enzymes and on new methodologies of generic potential. We survey three main areas: (i) library making techniques, including the application of computational and rational methods for library design; (ii) screening and selection techniques, including recent applications of enzyme screening by FACS (fluorescence activated cell sorter); (iii) new approaches for performing directed evolution, and in particular, the application of 'neutral drifts' (libraries generated by rounds of mutation and selection for the enzyme's original function) and of consensus mutations to generate highly evolvable starting points for directed evolution.

Fitness landscapes of drug resistance constitute powerful tools to elucidate mutational pathways of antibiotic escape. Here, we developed a predictive biophysics-based fitness landscape of trimethoprim (TMP) resistance for Escherichia coli dihydrofolate reductase (DHFR). We investigated the activity, binding, folding stability, and intracellular abundance for a complete set of combinatorial DHFR mutants made out of three key resistance mutations and extended this analysis to DHFR originated from Chlamydia muridarum and Listeria grayi. We found that the acquisition of TMP resistance via decreased drug affinity is limited by a trade-off in catalytic efficiency. Protein stability is concurrently affected by the resistant mutants, which precludes a precise description of fitness from a single molecular trait. Application of the kinetic flux theory provided an accurate model to predict resistance phenotypes (IC50) quantitatively from a unique combination of the in vitro protein molecular properties. Further, we found that a controlled modulation of the GroEL/ES chaperonins and Lon protease levels affects the intracellular steady-state concentration of DHFR in a mutation-specific manner, whereas IC50 is changed proportionally, as indeed predicted by the model. This unveils a molecular rationale for the pleiotropic role of the protein quality control machinery on the evolution of antibiotic resistance, which, as we illustrate here, may drastically confound the evolutionary outcome. These results provide a comprehensive quantitative genotype–phenotype map for the essential enzyme that serves as an important target of antibiotic and anticancer therapies.

Horizontal gene transfer (HGT) plays a central role in bacterial evolution, yet the molecular and cellular constraints on functional integration of the foreign genes are poorly understood. Here we performed inter-species replacement of the chromosomal folA gene, encoding an essential metabolic enzyme dihydrofolate reductase (DHFR), with orthologs from 35 other mesophilic bacteria. The orthologous inter-species replacements caused a marked drop (in the range 10–90%) in bacterial growth rate despite the fact that most orthologous DHFRs are as stable as E.coli DHFR at 37°C and are more catalytically active than E. coli DHFR. Although phylogenetic distance between E. coli and orthologous DHFRs as well as their individual molecular properties correlate poorly with growth rates, the product of the intracellular DHFR abundance and catalytic activity (k cat/KM), correlates strongly with growth rates, indicating that the drop in DHFR abundance constitutes the major fitness barrier to HGT. Serial propagation of the orthologous strains for ~600 generations dramatically improved growth rates by largely alleviating the fitness barriers. Whole genome sequencing and global proteome quantification revealed that the evolved strains with the largest fitness improvements have accumulated mutations that inactivated the ATP-dependent Lon protease, causing an increase in the intracellular DHFR abundance. In one case DHFR abundance increased further due to mutations accumulated in folA promoter, but only after the lon inactivating mutations were fixed in the population. Thus, by apparently distinguishing between self and non-self proteins, protein homeostasis imposes an immediate and global barrier to the functional integration of foreign genes by decreasing the intracellular abundance of their products. Once this barrier is alleviated, more fine-tuned evolution occurs to adjust the function/expression of the transferred proteins to the constraints imposed by the intracellular environment of the host organism., Author Summary Horizontal gene transfer (HGT) is central to bacterial evolution. The outcome of an HGT event (fixation in a population, elimination, or separation as a subdominant clone) depends not only on the availability of a new gene but crucially on the fitness cost or benefit of the genomic incorporation of the foreign gene and its expression in recipient bacteria. Here we studied the fitness landscape for inter-species chromosomal replacement of an essential protein, dihydrofolate reductase (DHFR) encoded by the folA gene, by its orthologs from other mesophilic bacteria. We purified and biochemically characterized 33 out of 35 orthologous DHFRs and found that most of them are stable and more catalytically active than E. coli DHFR. However, the inter-species replacement of DHFR caused significant fitness loss for most transgenic strains due to low abundance of orthologous DHFRs in E. coli cytoplasm. Laboratory evolution resulted in an increase in orthologous DHFR abundance leading to a dramatic fitness improvement. Genomic and proteomic analyses of “naive” and evolved strains suggest a new function of protein homeostasis to discriminate between “self” and “non-self” proteins, thus creating fitness barriers to HGT.

Predicting evolutionary paths to antibiotic resistance is key for understanding and controlling drug resistance. When considering a single final resistant genotype, epistatic contingencies among mutations restricts evolution to a small number of adaptive paths. Less attention has been given to multi-peak landscapes, and while specific peaks can be favored, it is unknown whether and how early a commitment to final fate is made. Here we characterized a multi-peaked adaptive landscape for trimethoprim resistance by constructing all combinatorial alleles of seven resistance-conferring mutations in dihydrofolate reductase. We observe that epistatic interactions increase rather than decrease the accessibility of each peak; while they restrict the number of direct paths, they generate more indirect paths, where mutations are adaptively gained and later adaptively lost or changed. This enhanced accessibility allows evolution to proceed through many adaptive steps while delaying commitment to genotypic fate, hindering our ability to predict or control evolutionary outcomes.

This protocol is used to assay the effect of protein over-expression on fitness of E. coli. It is based on a plasmid expression of a protein of interest from an arabinose-regulated pBAD promoter followed by the measurement of the intracellular protein abundance by Western blot along with the measurement of growth parameters of E. coli cell expressing this protein.

What are the molecular properties of proteins that fall on the radar of protein quality control (PQC)? Here we mutate the E. coli’s gene encoding dihydrofolate reductase (DHFR), and replace it with bacterial orthologous genes to determine how components of PQC modulate fitness effects of these genetic changes. We find that chaperonins GroEL/ES and protease Lon compete for binding to molten globule intermediate of DHFR, resulting in a peculiar symmetry in their action: Over-expression of GroEL/ES and deletion of Lon both restore growth of deleterious DHFR mutants and most of the slow-growing orthologous DHFR strains. Kinetic steady-state modeling predicts and experimentation verifies that mutations affect fitness by shifting the flux balance in cellular milieu between protein production, folding and degradation orchestrated by PQC through the interaction with folding intermediates.

Mutations create the genetic diversity on which selective pressures can act, yet also create structural instability in proteins. How, then, is it possible for organisms to ameliorate mutation-induced perturbations of protein stability while maintaining biological fitness and gaining a selective advantage? Here we used site-specific chromosomal mutagenesis to introduce a selected set of mostly destabilizing mutations into folA —an essential chromosomal gene of Escherichia coli encoding dihydrofolate reductase (DHFR)—to determine how changes in protein stability, activity, and abundance affect fitness. In total, 27 E. coli strains carrying mutant DHFR were created. We found no significant correlation between protein stability and its catalytic activity nor between catalytic activity and fitness in a limited range of variation of catalytic activity observed in mutants. The stability of these mutants is strongly correlated with their intracellular abundance, suggesting that protein homeostatic machinery plays an active role in maintaining intracellular concentrations of proteins. Fitness also shows a significant correlation with intracellular abundance of soluble DHFR in cells growing at 30 °C. At 42 °C, the picture was mixed, yet remarkable: A few strains carrying mutant DHFR proteins aggregated, rendering them nonviable, but, intriguingly, the majority exhibited fitness higher than wild type. We found that mutational destabilization of DHFR proteins in E. coli is counterbalanced at 42 °C by their soluble oligomerization, thereby restoring structural stability and protecting against aggregation.

Protein stability is widely recognized as a major evolutionary constraint. However, the relation between mutation-induced perturbations of protein stability and biological fitness has remained elusive. Here we explore this relation by introducing a selected set of mostly destabilizing mutations into an essential chromosomal gene of E.coli encoding dihydrofolate reductase (DHFR) to determine how changes in protein stability, activity and abundance affect fitness. Several mutant strains showed no growth while many exhibited fitness higher than wild type. Overexpression of chaperonins (GroEL/ES) buffered the effect of mutations by rescuing the lethal phenotypes and worsening better-fit strains. Changes in stability affect fitness by mediating the abundance of active and soluble proteins; DHFR of lethal strains aggregates, while destabilized DHFR of high fitness strains remains monomeric and soluble at 30oC and forms soluble oligomers at 42oC. These results suggest an evolutionary path where mutational destabilization is counterbalanced by specific oligomerization protecting proteins from aggregation.

What changes occur when a natural protein that had been under low mutation rates is subjected to a neutral drift at high mutational loads, thus generating genetically diverse (polymorphic) gene ensembles that all maintain the protein's original function and structure? To address this question we subjected large populations of TEM-1 beta-lactamase to a prolonged neutral drift, applying high mutation rates and purifying selection to maintain TEM-1's existing penicillinase activity. Purging of deleterious mutations and enrichment of beneficial ones maintained the sequence of these ensembles closer to TEM-1's family consensus and inferred ancestor. In particular, back-to-consensus/ancestor mutations that increase TEM-1's kinetic and thermodynamic stability were enriched. These acted as global suppressors and enabled the tolerance of a broad range of deleterious mutations, thus further increasing the genetic diversity of the drifting populations. The probability of a new function emerging (cefotaxime degradation) was also substantially increased in these ensembles owing to the presence of many gene variants carrying the global suppressors. Our findings indicate the unique features of large, polymorphic neutral ensembles generated under high mutational loads and prompt the speculation that the progenitors of today's proteins may have evolved under high mutational loads. The results also suggest that predictable back-to-consensus/ancestor changes can be used in the laboratory to generate highly diverse and evolvable gene libraries.

The distribution of fitness effects of protein mutations is still unknown. Of particular interest is whether accumulating deleterious mutations interact, and how the resulting epistatic effects shape the protein's fitness landscape. Here we apply a model system in which bacterial fitness correlates with the enzymatic activity of TEM-1 beta-lactamase (antibiotic degradation). Subjecting TEM-1 to random mutational drift and purifying selection (to purge deleterious mutations) produced changes in its fitness landscape indicative of negative epistasis; that is, the combined deleterious effects of mutations were, on average, larger than expected from the multiplication of their individual effects. As observed in computational systems, negative epistasis was tightly associated with higher tolerance to mutations (robustness). Thus, under a low selection pressure, a large fraction of mutations was initially tolerated (high robustness), but as mutations accumulated, their fitness toll increased, resulting in the observed negative epistasis. These findings, supported by FoldX stability computations of the mutational effects, prompt a new model in which the mutational robustness (or neutrality) observed in proteins, and other biological systems, is due primarily to a stability margin, or threshold, that buffers the deleterious physico-chemical effects of mutations on fitness. Threshold robustness is inherently epistatic-once the stability threshold is exhausted, the deleterious effects of mutations become fully pronounced, thereby making proteins far less robust than generally assumed.

Antibiotic resistance is an evolving threat to public health. Understanding the evolution of antibiotic resistance at the genetic level is critical to develop novel strategies to diagnose and treat antibiotic resistant infections. We recently developed an automated microbial selection device, the “morbidostat”, which is used to study the evolution of antibiotic resistance in dynamically sustained drug selection. The morbidostat adjusts drug concentrations to maintain nearly constant inhibition of bacterial growth even as evolving bacterial populations acquire higher resistance. using the morbidostat and next generation sequencing, we identified striking features in the evolution of trimethoprim resistance in five E. coli populations evolving in parallel. We found that resistance was acquired in a stepwise manner, through multiple mutations almost exclusively restricted to the gene encoding trimethoprim's target, dihydrofolate reductase (DHFR). Multiple distinct genotypes produced very similar trimethoprim resistant phenotypes, with each highly evolved strain each containing four mutations from a set of six possibilities, that were acquired in a non-random order. Never were more than four mutations acquired, despite sustained selection for further increases in drug resistance, indicating that these genotypes were local adaptive peaks.In order to understand how the adaptive landscape of drug resistance contained multiple peaks, all combinatorial sets of adaptive mutations in the DHFR gene (96 strains) were constructed and characterized. These measurements showed that resistance evolves through an almost maximally rugged adaptive landscape with direct and indirect trajectories leading to distinct peaks. The ruggedness was not explained by pairwise incompatibilities between mutations, instead indicating ‘high-order' genetic interactions between mutations. These high-order interactions were responsible for the existence of multiple adaptive peaks. One mutation was seen to have the power to control the adaptive landscape: its presence or absence largely defined the ruggedness or smoothness of the adaptive landscape.