FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W2155706782> ?p ?o ?g. }

• W2155706782 abstract "Article31 August 2015Open Access Source Data Stress-response balance drives the evolution of a network module and its host genome Caleb González Caleb González Department of Systems Biology - Unit 950, The University of Texas MD Anderson Cancer Center, Houston, TX, USA These authors contributed equally to this study Search for more papers by this author Joe Christian J Ray Joe Christian J Ray Department of Systems Biology - Unit 950, The University of Texas MD Anderson Cancer Center, Houston, TX, USA Center for Computational Biology & Department of Molecular Biosciences, University of Kansas, Lawrence, KS, USA These authors contributed equally to this study Search for more papers by this author Michael Manhart Michael Manhart Department of Physics & Astronomy, Rutgers University, Piscataway, NJ, USA Department of Chemistry and Chemical Biology, Harvard University, Cambridge, MA, USA Search for more papers by this author Rhys M Adams Rhys M Adams Department of Systems Biology - Unit 950, The University of Texas MD Anderson Cancer Center, Houston, TX, USA Search for more papers by this author Dmitry Nevozhay Dmitry Nevozhay Department of Systems Biology - Unit 950, The University of Texas MD Anderson Cancer Center, Houston, TX, USA School of Biomedicine, Far Eastern Federal University, Vladivostok, Russia Search for more papers by this author Alexandre V Morozov Alexandre V Morozov Department of Physics & Astronomy, Rutgers University, Piscataway, NJ, USA BioMaPS Institute for Quantitative Biology, Rutgers University, Piscataway, NJ, USA Search for more papers by this author Gábor Balázsi Corresponding Author Gábor Balázsi Department of Systems Biology - Unit 950, The University of Texas MD Anderson Cancer Center, Houston, TX, USA Laufer Center for Physical & Quantitative Biology, Stony Brook University, Stony Brook, NY, USA Department of Biomedical Engineering, Stony Brook University, Stony Brook, NY, USA Search for more papers by this author Caleb González Caleb González Department of Systems Biology - Unit 950, The University of Texas MD Anderson Cancer Center, Houston, TX, USA These authors contributed equally to this study Search for more papers by this author Joe Christian J Ray Joe Christian J Ray Department of Systems Biology - Unit 950, The University of Texas MD Anderson Cancer Center, Houston, TX, USA Center for Computational Biology & Department of Molecular Biosciences, University of Kansas, Lawrence, KS, USA These authors contributed equally to this study Search for more papers by this author Michael Manhart Michael Manhart Department of Physics & Astronomy, Rutgers University, Piscataway, NJ, USA Department of Chemistry and Chemical Biology, Harvard University, Cambridge, MA, USA Search for more papers by this author Rhys M Adams Rhys M Adams Department of Systems Biology - Unit 950, The University of Texas MD Anderson Cancer Center, Houston, TX, USA Search for more papers by this author Dmitry Nevozhay Dmitry Nevozhay Department of Systems Biology - Unit 950, The University of Texas MD Anderson Cancer Center, Houston, TX, USA School of Biomedicine, Far Eastern Federal University, Vladivostok, Russia Search for more papers by this author Alexandre V Morozov Alexandre V Morozov Department of Physics & Astronomy, Rutgers University, Piscataway, NJ, USA BioMaPS Institute for Quantitative Biology, Rutgers University, Piscataway, NJ, USA Search for more papers by this author Gábor Balázsi Corresponding Author Gábor Balázsi Department of Systems Biology - Unit 950, The University of Texas MD Anderson Cancer Center, Houston, TX, USA Laufer Center for Physical & Quantitative Biology, Stony Brook University, Stony Brook, NY, USA Department of Biomedical Engineering, Stony Brook University, Stony Brook, NY, USA Search for more papers by this author Author Information Caleb González1, Joe Christian J Ray1,2, Michael Manhart3,4, Rhys M Adams1, Dmitry Nevozhay1,5, Alexandre V Morozov3,6 and Gábor Balázsi 1,7,8 1Department of Systems Biology - Unit 950, The University of Texas MD Anderson Cancer Center, Houston, TX, USA 2Center for Computational Biology & Department of Molecular Biosciences, University of Kansas, Lawrence, KS, USA 3Department of Physics & Astronomy, Rutgers University, Piscataway, NJ, USA 4Department of Chemistry and Chemical Biology, Harvard University, Cambridge, MA, USA 5School of Biomedicine, Far Eastern Federal University, Vladivostok, Russia 6BioMaPS Institute for Quantitative Biology, Rutgers University, Piscataway, NJ, USA 7Laufer Center for Physical & Quantitative Biology, Stony Brook University, Stony Brook, NY, USA 8Department of Biomedical Engineering, Stony Brook University, Stony Brook, NY, USA *Corresponding author. Tel: +1 631 632 5414; Fax: +1 631 632 5405; E-mail: [email protected] Molecular Systems Biology (2015)11:827https://doi.org/10.15252/msb.20156185 PDFDownload PDF of article text and main figures. Peer ReviewDownload a summary of the editorial decision process including editorial decision letters, reviewer comments and author responses to feedback. ToolsAdd to favoritesDownload CitationsTrack CitationsPermissions Figures & Info Abstract Stress response genes and their regulators form networks that underlie drug resistance. These networks often have an inherent tradeoff: their expression is costly in the absence of stress, but beneficial in stress. They can quickly emerge in the genomes of infectious microbes and cancer cells, protecting them from treatment. Yet, the evolution of stress resistance networks is not well understood. Here, we use a two-component synthetic gene circuit integrated into the budding yeast genome to model experimentally the adaptation of a stress response module and its host genome in three different scenarios. In agreement with computational predictions, we find that: (i) intra-module mutations target and eliminate the module if it confers only cost without any benefit to the cell; (ii) intra- and extra-module mutations jointly activate the module if it is potentially beneficial and confers no cost; and (iii) a few specific mutations repeatedly fine-tune the module's noisy response if it has excessive costs and/or insufficient benefits. Overall, these findings reveal how the timing and mechanisms of stress response network evolution depend on the environment. Synopsis The evolution of a synthetic gene circuit that trades off costly gene expression for drug resistance is analyzed computationally. The predictions are validated experimentally by adjusting gene expression in the absence or presence of environmental stress. A synthetic gene circuit is integrated into the yeast genome to model the evolution of drug resistance networks with inherent tradeoff. Computational models are constructed to predict the speed and mechanisms of adaptation for various levels of gene expression and stress. The cell population adapts by mutations eliminating the module quickly when the network gratuitously responds in the absence of stress or by mutations that fine-tune the module's suboptimal response and establish slowly in the presence of stress. If the module initially fails to respond to stress, the population adapts by mutations that activate gene expression within the module. Introduction The number of human-designed biological systems has increased rapidly since the inception of synthetic biology (Purnick & Weiss, 2009). Parts and concepts underlying synthetic biological constructs have expanded quickly, feeding on general biological knowledge. Conversely, synthetic biology has enormous but unexploited potential to inform other areas of biology, such as evolutionary biology (Tanouchi et al, 2012b). For example, gene regulatory networks that control the expression of stress-protective genes have emerged through evolution (Lopez-Maury et al, 2008) but can also be built de novo (Nevozhay et al, 2012; Tanouchi et al, 2012a). Depending on the details of gene regulation, cells can survive because they respond to stress (Gasch et al, 2000); diversify non-genetically (hedge bets), independent of the stress (Balaban et al, 2004; Thattai & van Oudenaarden, 2004; Levy et al, 2012); or use a mixture of these two strategies (New et al, 2014). However, stress-protective gene expression can be costly or toxic in the absence of stress (Andersson & Levin, 1999), or even in the presence of stress when the expression level exceeds the requirement for survival (Nevozhay et al, 2012). Overall, the costs and benefits of survival mechanisms create a tradeoff between maximizing growth while also ensuring survival during stress. How mutations alter stress response networks to improve fitness under such circumstances, especially in phenotypically heterogeneous populations (Sumner & Avery, 2002), is an open problem in evolutionary biology. Consider a stress response network module, consisting of a stress-sensing transcriptional regulator and its stress-protective gene target, which has arisen in a cell's genome. Similar modules, such as Tn10 (Hillen & Berens, 1994), toxin-antitoxin systems (Yamaguchi et al, 2011), or bypass signaling (Hsieh & Moasser, 2007), can arise rapidly by recombination, horizontal gene transfer, or inhibitor-mediated alternate pathway activation. Considering their impact on microbial and cancer drug resistance, it is important to know how reproducibly and how quickly such stress defense networks can adapt (Lobkovsky & Koonin, 2012). Yet, we currently lack quantitative, hypothesis-driven understanding of how initially suboptimal stress defense modules evolve inside the host genome, especially in the presence of gene expression noise (Balázsi et al, 2011; Munsky et al, 2012; Sanchez & Golding, 2013). Although network evolution theory (Kauffman, 1993; Mason et al, 2004; Kashtan & Alon, 2005) and laboratory evolution experiments (Lenski & Travisano, 1994; Beaumont et al, 2009; Tenaillon et al, 2012; Toprak et al, 2012; Lang et al, 2013) have generated important insights, they have provided largely descriptive, a posteriori interpretations. Now there is a growing need for predictive, hypothesis-driven, quantitative understanding of gene network evolution, which requires making a priori predictions of mutation effects and evolutionary dynamics that are tested experimentally (Wang et al, 2013). One option could be to study the evolution of small natural regulatory modules (Dekel & Alon, 2005; Hsu et al, 2012; Quan et al, 2012; van Ditmarsch et al, 2013). However, connections of natural regulatory modules with the rest of the genome can be significant (Maynard et al, 2010) and poorly characterized, thus making predictive, quantitative understanding difficult. Synthetic gene circuits (Elowitz & Leibler, 2000; Gardner et al, 2000; Stricker et al, 2008; Moon et al, 2012; Nevozhay et al, 2013) represent a better alternative, since they are small, consist of well-characterized components, and typically lack direct regulatory interactions with the host genome. However, it is unclear whether the evolution of synthetic gene circuits (Yokobayashi et al, 2002; Sleight et al, 2010; Poelwijk et al, 2011; Wu et al, 2014) can be predicted a priori, especially with regard to gene expression heterogeneity. We recently characterized the dynamics and fitness effects of gene expression for a synthetic two-gene “positive feedback” (PF) circuit (Fig 1A) integrated into the genome of the haploid single-celled eukaryote Saccharomyces cerevisiae (Nevozhay et al, 2012). This synthetic gene circuit consists of a well-characterized transcriptional regulator (rtTA) and an antibiotic resistance gene (yEGFP::zeoR). In the presence of tetracycline-analog inducers such as doxycycline, rtTA activates both itself and yEGFP::zeoR by binding to two tetO2 operator sites in two identical promoters (Fig 1A). This positive feedback is noisy, however, and thus, only a fraction of cells switch to high expression of rtTA and yEGFP::zeoR. These cells benefit from high gene expression, which protects them from the antibiotic zeocin. Meanwhile, the same cells experience a cost from rtTA activator expression toxicity, causing a tradeoff when zeocin is present (Nevozhay et al, 2012). The fitness (division rate) of any individual cell is the product of its rtTA expression cost and yEGFP::zeoR expression benefit (Nevozhay et al, 2012), which varies from cell to cell. Thus, quantitative knowledge of dynamics and fitness effects makes the PF gene circuit an excellent model for studying gene network evolution in tradeoff situations. Its design separates stress (zeocin) from its adjustable cellular response (inducible yEGFP::zeoR expression), facilitating predictive, quantitative understanding of how a stress response module adapts inside the host genome. Figure 1. The PF synthetic gene circuit: fitness and gene expression characteristics The PF synthetic gene circuit (Nevozhay et al, 2012) consists of two components. First, the regulator reverse tet-trans-activator (rtTA) (Urlinger et al, 2000) is a reverse-tetR gene fused to three F activator domains (cyan rectangles), which are shorter versions of the VP16 activator (Baron et al, 1997). The target gene yEGFP∷ZeoR consists of the fluorescent reporter yEGFP fused to the drug resistance gene zeoR (Gatignol et al, 1988) that binds and inactivates zeocin, a bleomycin-family antibiotic. Unbound zeocin generates DNA double-strand breaks, causing cell cycle arrest and potentially cell death. Doxycycline added to the growth medium diffuses freely through the cell wall and binds to rtTA dimers. Inducer-bound rtTA undergoes a conformational change that results in strong association with two tetO2 operator sites upstream of each of the two tetreg promoters (Becskei et al, 2001), activating both regulator and target gene expression, while causing toxicity by squelching. Costs and benefits of PF gene circuit components were determined by measuring cell population growth rate (population fitness) versus two environmental factors: inducer doxycycline and antibiotic zeocin. Each point on the population fitness landscape (three-dimensional gray surface on the left) is an average of cellular fitness values (color-shaded slopes in the surrounding plots) as cells stochastically move within gene expression distributions (black histograms in the surrounding plots). Gene expression is measured as log10(fluorescence) (arbitrary units). DxZy denotes the environment (the x and y following D and Z indicate μg/ml doxycycline and mg/ml zeocin concentrations, respectively, with Di = 0.2 μg/ml doxycycline). Cellular fitness (cell division rate) is a function of gene expression for each cell in each environment DxZy. It is inferred from the population fitness, based on a biochemical model (Nevozhay et al, 2012); see the Appendix. The black arrows beneath cellular fitness landscapes illustrate selection pressures pushing the gene expression distribution toward higher fitness. Download figure Download PowerPoint Here, we used our quantitative knowledge of the PF gene circuit to predict a priori the timing and mechanisms of its initial adaptation to several constant environments (squares in Fig 1B) corresponding to various stress-response imbalance scenarios. We tested these predictions with experimental evolution, followed by sequencing to identify the mutations that establish in the population, depending on the imbalance between the environmental stress and the intracellular response. In this way, we tested how different mutations can readjust the response of a network module with inherent tradeoff, to match the stress and minimize the cost in each specific environment. These results could help us understand how fast and through what mechanisms drug resistance emerges or deteriorates in the process of network evolution, and could help the future design of synthetic gene circuits that resist evolutionary degradation. Results The PF gene circuit can mimic various scenarios of stress-response imbalance We considered the following disparities between the external stress and the activity of a stress defense module: (i) the module responds gratuitously to a harmless environmental change; (ii) the module cannot respond to harmful stress when needed; and (iii) the module responds to stress, but suboptimally. To mimic these scenarios using the PF gene circuit in yeast, we relied on the separability of stress and response, adjusting two environmental factors with known fitness effects (Nevozhay et al, 2012): inducer doxycycline and antibiotic zeocin (Fig 1). Hereafter, DxZy will denote environmental conditions, with x and y indicating doxycycline and zeocin concentrations, respectively. The antibacterial compound doxycycline has negligible effect on yeast (Wishart et al, 2005), but causes squelching toxicity in engineered PF cells when bound to rtTA (Gari et al, 1997; Nevozhay et al, 2012). Zeocin is a broad-spectrum DNA-damaging antibiotic (Burger, 1998) that acts on bacteria and eukaryotes. First, the presence of inducer doxycycline alone corresponds to scenario (i): costly, futile response of some (Fig 1B, DiZ0) or most (Fig 1B, D2Z0) cells that start expressing the PF genes. The cost of response slows the cell division rate of responding, high expressor cells compared to non-responding, low expressor cells (Nevozhay et al, 2012). Consequently, the division rate of individual yeast cells can differ drastically from the overall population growth rate. To capture these differences between single cell- and population growth rates, we constructed a population fitness landscape (three-dimensional gray surface in Fig 1B) and cellular fitness landscapes (colored panels in Fig 1B). The population fitness landscape maps the overall population growth against the two environmental variables, doxycycline and zeocin concentrations. Cellular fitness landscapes depict the division rate of single cells versus their gene expression level in a given combination of doxycycline and zeocin. As described in the Appendix, we inferred these landscapes directly from growth rate and gene expression measurements (Appendix Fig S1A) in 13 different combinations of doxycycline and zeocin. Second, the presence of antibiotic zeocin alone (Fig 1B, D0Z2) corresponds to the lack of response when needed, as in scenario (ii). Finally, the presence of both inducer and antibiotic (Fig 1B, DiZ2 and D2Z2) corresponds to scenario (iii) where the fraction of responding, slower-growing cells ensures cell population survival during antibiotic treatment, but the response is in general suboptimal. Altogether, the PF gene circuit is a well-characterized module lacking direct regulatory interactions with the yeast genome. It exemplifies typical tradeoffs between the benefits and costs of gene expression in stress response networks. Importantly, the benefits and costs are independently tunable for the PF gene circuit, making it possible to predict and test their evolution toward optimality. Predicting the first evolutionary steps in constant environments We asked whether the PF cellular and population fitness landscapes (colored squares and panels in Fig 1B; Appendix Fig S1A; Appendix Table S1) could predict evolutionary trends in specific environments. For example, in the D2Z0 environment, most cells are far from their fitness maximum, which is at low expression. If a mutation could push cells downward in expression, toward their fitness maximum (horizontal arrow in Fig 1B, D2Z0 panel), then they should grow faster. Mutations that either abolish or weaken rtTA toxicity could achieve this effect. Let us call these mutation types “knockout” (K) and “tweaking” (T) mutations, respectively (Fig 2A; Appendix Fig S1B). On the other hand, in the D0Z2 environment cells should benefit from mutations that diminish the effect of the antibiotic. This could happen in various ways, for example by upregulation of native stress-response mechanisms; or by increasing yEGFP::zeoR expression. Let us call these latter mutation types “generic” (G) drug resistance mutations (Fig 2A; Appendix Fig S1B). In all these cases, mutant cells can improve their fitness by unidirectionally lowering or increasing PF gene expression. However, in certain conditions (such as DiZ2), when the cells form two subpopulations that flank the cellular fitness peak, a single-directional expression change is not optimal. This is because a one-way expression shift can only move one subpopulation toward the fitness peak, while the other subpopulation must necessarily move away from it. Instead, optimally the two subpopulations should approach each other, both moving toward the fitness peak (horizontal arrows in Fig 1B, DiZ2, D2Z2 panels). Figure 2. Simulation framework predicts evolutionary dynamics Simulating the initial steps of evolution. Three types of potentially beneficial mutations (with an overall rate μ) enter the ancestral population of yeast cells that initially carry the intact PF gene circuit. Each cell can divide and mutate, producing new genotypes with altered fitness that can belong to three different types. The first two types are knockout (K) and tweaking (T) mutations. They eliminate rtTA's regulator activity and toxicity completely or partially, respectively. The third type includes extra-rtTA or generic (G) mutations that cause zeocin resistance independently of rtTA. In the models, we consider exponential growth with random elimination of cells or periodic resuspensions to control population size. Empty circles represent intact PF cells, while blue, magenta, and orange circles represent K, T, and G mutants, respectively. These mutations can arise, be lost, or expand in the population. The speed at which mutants take over the population in each simulated condition is measured as the ancestral genotype's half-life (the time until only 50% of the population carries the ancestral genome). N = 100; mean ± SEM in each simulated condition: D2Z0, DiZ0, D0Z2, D2Z2, and DiZ2. In these plots, we fixed μ−Z = 10−6.2 or μ+Z = 10−5.4/genome/generation (for no zeocin and zeocin, respectively) and P(G) = 0.75. Therefore, P(T) = 0.25 − P(K). On the horizontal axis, we show the probability of T mutations among intra-rtTA mutations: P*(T) = P(T|¬G), which scales P(T) four-fold up such that its maximum is 1 instead of 0.25. The gray bar denotes the value used for time course simulations in subsequent figures. The parameter set for the gray bar on this and the following panels is P(T) = 0.025; P(K) = 0.225; and P(G) = 0.75. Number of established mutations with frequency > 5% at day 20. N = 100; mean ± SEM in each simulated condition: D2Z0, DiZ0, D0Z2, D2Z2, and DiZ2. Parameters, axes, and gray bar: as in (B). Population fractions of T-, K-, and G-type mutations at day 20, for the parameters corresponding to the gray bar, as indicated above. Download figure Download PowerPoint How would the PF cells evolve to adapt in specific combinations of doxycycline and zeocin? Mutations of any type (K, T, G) can arise spontaneously, then establish in the population, and compete with each other depending on two requirements. First, the mutation type must be available (genetic changes causing the phenotype must exist). Second, since we consider large populations, the mutation should be beneficial, improving fitness in the given environment. Despite these intuitive expectations, it is unclear how many mutations of each type will establish in each condition, and how fast. To address these questions in silico, we developed two complementary modeling approaches: a simple mathematical model and a detailed computational simulation framework (see the Computational Models.zip file and the Appendix for detailed descriptions). The two models serve to test the robustness of results to various modeling approaches. The simple model was more general and faster, allowing more extensive parameter scans. On the other hand, the simulation framework allowed testing how specific details of experimental evolution would affect the evolutionary dynamics, and provided more detailed results. We initiated both models with a population of ancestral (wild-type) PF cells, aiming to find out the number and type of mutations that establish and when the ancestral genotype disappears. We modeled 20 days of evolution in each environment indicated by the colored squares in Fig 1B. The simpler model described population dynamics by a system of ordinary differential equations (ODEs), assuming constant population size and mutation rate. We characterized wild-type and mutant cells by a single parameter: their fitness (exponential growth rate), determined from the fitness landscapes in Fig 1B. For example, we assumed that K mutants had cellular fitness corresponding to null expression in Fig 1B. T-type mutant cells altered their fitness randomly to a level corresponding to intermediate expression on the cellular fitness landscapes. Finally, G-type mutants increased their fitness randomly, up to a level they would have without zeocin. This simpler model could predict how fast the wild-type genotype disappears from the population. It could also forecast the mutation type (K, T, G) that predominantly replaces the wild type in each condition. However, it could not predict the number of distinct mutant alleles in the evolving population. Moreover, it lacked potentially important experimental details, such as periodic resuspensions and phenotypic switching. To test the importance of such additional details, the detailed simulation framework captured multiple experimentally relevant aspects of evolution. For example, cells could switch between On and Off states with experimentally inferred rates (Appendix Table S1). K, T, and G mutations with altered switching and growth rates entered the population as single cells at a constant, but adjustable rate μ per cell per generation (Fig 2A; Appendix Fig S1C). K-type mutants could not switch On, and thus had no cellular fitness costs in doxycycline. T-type mutants switched On at a randomly reduced rate, and thus had diminished cellular fitness costs from PF gene expression. G-type mutant cells had randomly increased drug resistance without any change in switching rates. We simulated periodic resuspensions by repeatedly reducing the cell population size to 106. We considered cells to be initially drug- and inducer-free, and allowed them to gradually take up zeocin and doxycycline. This simulation framework could predict the number of distinct mutant alleles, in addition to the characteristics predicted by the simpler model. Both models had three free parameters: the rate of potentially beneficial mutations μ, and the input probabilities P(G) and P(T) of a given mutation being of type G or T, respectively. Once known, these parameters also define the probability of a mutation to be of type K: P(K) = 1 – P(G) – P(T). We note the difference between the rate and probability of a mutation: for example, the probability of P(K) could be equal to 1, while its rate μP(K) is much < 1 per genome per generation. Figure 2A depicts the effect of each mutation type, illustrating the relationships among the free parameters. We extracted the rest of the parameters (Appendix Table S1) from experimental measurements (see the Appendix) and kept them fixed. Using these models, we studied how the three free parameters affected three features of evolutionary dynamics: the ancestral genotype's half-life, as well as the type and number of mutant alleles in each condition (Fig 2; Appendix Figs S2 and S3). We started by studying the ancestral genotype's half-life in each model, scanning each free parameter systematically (Fig 2B; Appendix Figs S2B and S3B). The models consistently indicated (Fig 2B) that the ancestral genotype disappeared fastest in conditions with steep monotone cellular fitness landscapes (Fig 1B, D0Z2 and D2Z0). In contrast, the ancestral genotype remained in the population longer in peaked cellular fitness landscapes (Fig 1B, D2Z2 and DiZ2). Finally, the majority of cells were still genetically ancestral after 20 days in DiZ0, which has the most gradual cellular fitness landscape (Fig 1B, DiZ0). The time when the ancestral genotype disappeared in various environments depended differently on the mutation probabilities P(K), P(T), P(G) (Appendix Fig S3B). For example, the ancestral genotype disappeared later in D2Z2 when we lowered P(T). Likewise, lowering P(G) prolonged the ancestral genotype's presence in the populations in D0Z2. These observations confirmed the expectation that the most beneficial mutation in each condition dictates evolutionary dynamics. Overall, we hypothesized based on these results that the ancestral PF gene circuit should disappear fastest in D2Z0 and D0Z2, followed by DiZ2 and D2Z2, and finally in DiZ0. Making these predictions required quantitatively understanding the fitness properties and genetic structure of the PF gene circuit. Without modeling, it would have been impossible to obtain quantitative estimates of the speeds at which mutants establish and take over the evolving population. In general, K, T, and G allele frequencies at the end of simulated time courses did not match the input probabilities of P(K), P(T), and P(G) mutations. Rather, each condition favored different mutation types as long as they were available (Fig 2D; Appendix Figs S2 and S3). For example, in D2Z0, nearly all mutations were K-type even if K mutations were unlikely to enter the population. T mutations established exclusively in D2Z2, while in DiZ2 they appeared alongside G mutations. In DiZ0, K or T mutations established late and spread slowly, with parameter-dependent relative fractions. Finally, only G alleles could establish in D0Z2. To conclude, both models predicted the environment-specific dominance of various mutation typ" @default.
• W2155706782 created "2016-06-24" @default.
• W2155706782 creator A5032768312 @default.
• W2155706782 creator A5050076723 @default.
• W2155706782 creator A5056958220 @default.
• W2155706782 creator A5067887557 @default.
• W2155706782 creator A5068312877 @default.
• W2155706782 creator A5069451179 @default.
• W2155706782 creator A5072140718 @default.
• W2155706782 date "2015-08-01" @default.
• W2155706782 modified "2023-10-14" @default.
• W2155706782 title "Stress‐response balance drives the evolution of a network module and its host genome" @default.
• W2155706782 cites W1645305843 @default.
• W2155706782 cites W1967330678 @default.
• W2155706782 cites W1970708036 @default.
• W2155706782 cites W1971866274 @default.
• W2155706782 cites W1979159575 @default.
• W2155706782 cites W1983515750 @default.
• W2155706782 cites W1984429906 @default.
• W2155706782 cites W1985561503 @default.
• W2155706782 cites W1988327779 @default.
• W2155706782 cites W1988441309 @default.
• W2155706782 cites W1995989489 @default.
• W2155706782 cites W1996205067 @default.
• W2155706782 cites W1996460711 @default.
• W2155706782 cites W1998688483 @default.
• W2155706782 cites W2005361347 @default.
• W2155706782 cites W2006492816 @default.
• W2155706782 cites W2007418023 @default.
• W2155706782 cites W2010025771 @default.
• W2155706782 cites W2012175205 @default.
• W2155706782 cites W2012977051 @default.
• W2155706782 cites W2016546231 @default.
• W2155706782 cites W2023987782 @default.
• W2155706782 cites W2028979517 @default.
• W2155706782 cites W2038701138 @default.
• W2155706782 cites W2038846160 @default.
• W2155706782 cites W2041985017 @default.
• W2155706782 cites W2054484919 @default.
• W2155706782 cites W2058527659 @default.
• W2155706782 cites W2063563543 @default.
• W2155706782 cites W2068872007 @default.
• W2155706782 cites W2078334027 @default.
• W2155706782 cites W2082165572 @default.
• W2155706782 cites W2083456183 @default.
• W2155706782 cites W2085850813 @default.
• W2155706782 cites W2086516457 @default.
• W2155706782 cites W2087576311 @default.
• W2155706782 cites W2096015757 @default.
• W2155706782 cites W2110232499 @default.
• W2155706782 cites W2110540877 @default.
• W2155706782 cites W2117737648 @default.
• W2155706782 cites W2122492159 @default.
• W2155706782 cites W2123505432 @default.
• W2155706782 cites W2126606094 @default.
• W2155706782 cites W2129901956 @default.
• W2155706782 cites W2132828256 @default.
• W2155706782 cites W2134759932 @default.
• W2155706782 cites W2135783707 @default.
• W2155706782 cites W2137683543 @default.
• W2155706782 cites W2148323109 @default.
• W2155706782 cites W2150207014 @default.
• W2155706782 cites W2150712887 @default.
• W2155706782 cites W2159844490 @default.
• W2155706782 cites W2160803964 @default.
• W2155706782 cites W2163245800 @default.
• W2155706782 cites W2163904534 @default.
• W2155706782 doi "https://doi.org/10.15252/msb.20156185" @default.
• W2155706782 hasPubMedCentralId "https://www.ncbi.nlm.nih.gov/pmc/articles/4562500" @default.
• W2155706782 hasPubMedId "https://pubmed.ncbi.nlm.nih.gov/26324468" @default.
• W2155706782 hasPublicationYear "2015" @default.
• W2155706782 type Work @default.
• W2155706782 sameAs 2155706782 @default.
• W2155706782 citedByCount "79" @default.
• W2155706782 countsByYear W21557067822016 @default.
• W2155706782 countsByYear W21557067822017 @default.
• W2155706782 countsByYear W21557067822018 @default.
• W2155706782 countsByYear W21557067822019 @default.
• W2155706782 countsByYear W21557067822020 @default.
• W2155706782 countsByYear W21557067822021 @default.
• W2155706782 countsByYear W21557067822022 @default.
• W2155706782 countsByYear W21557067822023 @default.
• W2155706782 crossrefType "journal-article" @default.
• W2155706782 hasAuthorship W2155706782A5032768312 @default.
• W2155706782 hasAuthorship W2155706782A5050076723 @default.
• W2155706782 hasAuthorship W2155706782A5056958220 @default.
• W2155706782 hasAuthorship W2155706782A5067887557 @default.
• W2155706782 hasAuthorship W2155706782A5068312877 @default.
• W2155706782 hasAuthorship W2155706782A5069451179 @default.
• W2155706782 hasAuthorship W2155706782A5072140718 @default.
• W2155706782 hasBestOaLocation W21557067821 @default.
• W2155706782 hasConcept C104317684 @default.
• W2155706782 hasConcept C126831891 @default.
• W2155706782 hasConcept C141231307 @default.
• W2155706782 hasConcept C150194340 @default.
• W2155706782 hasConcept C168031717 @default.
• W2155706782 hasConcept C169760540 @default.
• W2155706782 hasConcept C54355233 @default.
• W2155706782 hasConcept C67339327 @default.
• W2155706782 hasConcept C70721500 @default.

• next