Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Adaptation in Escherichia coli : ecological and genetic constraints on diversification Schick, Alana 2013

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Notice for Google Chrome users:
If you are having trouble viewing or searching the PDF with Google Chrome, please download it here instead.

Item Metadata


24-ubc_2014_spring_schick_alana.pdf [ 1.37MB ]
JSON: 24-1.0072140.json
JSON-LD: 24-1.0072140-ld.json
RDF/XML (Pretty): 24-1.0072140-rdf.xml
RDF/JSON: 24-1.0072140-rdf.json
Turtle: 24-1.0072140-turtle.txt
N-Triples: 24-1.0072140-rdf-ntriples.txt
Original Record: 24-1.0072140-source.json
Full Text

Full Text

Adaptation in Escherichia coli :ecological and genetic constraints ondiversificationbyAlana SchickB.Sc., The University of British Columbia, 2007A THESIS SUBMITTED IN PARTIAL FULFILLMENT OFTHE REQUIREMENTS FOR THE DEGREE OFMASTER OF SCIENCEinThe Faculty of Graduate and Postdoctoral Studies(Zoology)THE UNIVERSITY OF BRITISH COLUMBIA(Vancouver)January 2014? Alana Schick 2013AbstractThere is growing evidence that disruptive selection generated by intraspecificresource competition may be a common mechanism for generating biologicaldiversity. Adaptive dynamics models provide a framework describing howfrequency dependent selection drives such diversification, but these modelsdon?t consider the complexities that arise as a result of gene interactions.Here, we explore the relative effects of ecological and genetic constraints ondiversification using an experimental system of Escherichia coli in whichdiversification is driven by frequency dependence based on resource use.Diversified populations consist of ecotypes that consume glucose and ac-etate at different rates, and a mutation in the arcA gene has been identifiedthat has a large effect on this phenotype. By isolating clones of each eco-type from a previously diversified population, we find that the effect of thearcA mutation on rediversification depends on both the ecotype and thegenetic background. While some of these observations are consistent withpredictions made by adaptive dynamics models, others cannot be explainedwithout also accounting for epistasis and genetic constraints, highlightingthe importance of considering both ecological and genetic factors when pre-dicting diversification. Adaptation in this system also provides an exampleof an interaction between ecological and evolutionary processes, adding to agrowing number of studies that exhibit a clear feedback between these twoprocesses.iiPreface? Chapter 3 is based on work completed in UBC?s Biodiversity ResearchCentre, in the lab of Professor Michael Doebeli. I designed the exper-iments, with the advice of Dr. Mickael Le Gac. I was responsible forcarrying out the experiments described here, conducted and/or super-vised all media preparation, bacterial transfers, and data collection.? A version of Chapter 3 has been submitted for publication. Schick, A.,and Doebeli, M. 2013. Adaptation in Escherichia coli : Intraspecificecological interactions and genetic constraints determine diversifica-tion. I conducted all the experiments, statistical analysis and wrotethe manuscript.iiiTable of ContentsAbstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iiPreface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iiiTable of Contents . . . . . . . . . . . . . . . . . . . . . . . . . . . . ivList of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . viList of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . viiAcknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . viii1 General Introduction . . . . . . . . . . . . . . . . . . . . . . . 12 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42.1 Generalizations from evolution experiments . . . . . . . . . . 42.1.1 Repeatability . . . . . . . . . . . . . . . . . . . . . . . 42.1.2 Epistasis . . . . . . . . . . . . . . . . . . . . . . . . . 52.1.3 Rate of adaptation . . . . . . . . . . . . . . . . . . . 62.2 Model system . . . . . . . . . . . . . . . . . . . . . . . . . . 72.2.1 Motivation for sympatric speciation experiments . . . 72.2.2 Previous experiments . . . . . . . . . . . . . . . . . . 83 Adaptation in E.coli : Intraspecific Ecological Interactionsand Genetic Constraints Determine Diversification . . . . 113.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . 113.2 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153.2.1 Bacterial strains . . . . . . . . . . . . . . . . . . . . . 153.2.2 Rediversification . . . . . . . . . . . . . . . . . . . . . 153.2.3 Phenotypic assays . . . . . . . . . . . . . . . . . . . . 183.2.4 Statistical analysis . . . . . . . . . . . . . . . . . . . . 183.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193.3.1 Naming convention . . . . . . . . . . . . . . . . . . . 19ivTable of Contents3.3.2 Phenotypic evolution . . . . . . . . . . . . . . . . . . 193.3.3 Effect of arcA? mutation on diversification . . . . . . 213.3.4 Effect of generation on diversification . . . . . . . . . 233.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . 253.4.1 Genetic and ecological constraints . . . . . . . . . . . 263.4.2 Effect of timepoint . . . . . . . . . . . . . . . . . . . 303.4.3 Conclusions . . . . . . . . . . . . . . . . . . . . . . . 31Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32AppendicesA Supplementary Figures . . . . . . . . . . . . . . . . . . . . . . 38vList of Tables3.1 Summary of proportion of lines of each genotype that diver-sified. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20viList of Figures3.1 Schematic of experimental design. . . . . . . . . . . . . . . . 173.2 Phenotypic evolution within SS types. . . . . . . . . . . . . . 203.3 Phenotypic evolution within FS types. . . . . . . . . . . . . . 213.4 Effect of arcA mutation on diversity by variance (Dv) in bothFS (green circles) and SS (blue circles). . . . . . . . . . . . . 233.5 Effect of generation on diversification for both wild type lines. 243.6 Effect of generation on diversification for both mutant lines. 253.7 Pairwise invasibility plot summarizing underlying ecologicalinteractions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29A.1 Growth curves of clones from gen200-FSarcA?-rep2. . . . . . 38A.2 Growth curves of clones from gen200-FSarcA+-rep2. . . . . . 39A.3 Growth curves of clones from gen800-SSarcA+-rep4. . . . . . 40viiAcknowledgementsCompleting this thesis would not have been possible without the supportof numerous people. First and foremost, I would like to thank my advi-sor, Michael Doebeli, for his encouragement and assistance over the years.Thanks to members past and present of the Doebeli and Otto lab groups fordiscussions and feedback. Thanks to Thomas Prasloski and Sarita Balab-hadra for help collecting data. Thanks to Mickael Le Gac for help designingthe experiments and providing the strains. Thanks to my committee mem-bers, Mike Whitlock and Leticia Aviles for guidance. Lastly, special thanksto Aleeza Gerstein and Sam Yeaman for helping make sense of it all (dataand life included).viiiChapter 1General Introduction?Nothing in evolution or ecology makes sense except in the light of the other.?[40]As evolutionary biologists, our attempts to understand adaptive evolu-tion are driven largely by the desire for foresight; to accurately predict thefuture of populations. For a single population in an isolated and controlledenvironment (like an evolutionary vacuum), this might be possible. In na-ture, however, there is no such thing. We are limited by complexity due toan essentially unlimited number of interdependent variables. For this rea-son, the frontier of evolutionary biology is fundamentally different than thefrontier of other sciences. There are some evolutionary biologists who claimthat all the interesting questions in evolution have been answered. In someways, they have. Certainly many (if not most) of the mechanisms that driveevolution are well understood. The limitation of biology, however, is not alimitation in kind (like in many other sciences), but a limitation in scale.In other words, we may know what the mechanisms are, but our abilitiesto make predictions are still weak due to the infinite interactions betweenthings. These limitations are being broken through. Data is pouring in at arate faster than ever, especially at the molecular level, making this a mostexciting time to be a biologist. In my opinion, the discoveries that advanceevolutionary biology in the next hundred years will not be discoveries of un-known processes or mechanisms, but the prevalence and relative importanceof specific mechanisms.For a long time, evolutionary biologists have made their lives simplerby assuming that ecological and evolutionary processes occur on differenttime scales and can therefore be treated independently. This has led to a di-chotomy in researchers of ecologists and evolutionary biologists. It is becom-ing increasingly clear, however, that these processes do occur on the sametime scale and that one cannot be fully understood without the other. In arecent and frequently cited review of eco-evolutionary dynamics, Schoener(2011) states that evidence for ecological change affecting evolution is abun-dant, but evidence of the reverse is sparse. I suspect that this is due in1Chapter 1. General Introductionlarge part to the complexity and difficulty of mechanistically linking evo-lutionary and ecological dynamics. For example, it may be easier to seehow species interactions drive changes in allele frequencies, but harder tosee how changes in allele frequencies drive species interactions. In bacterialpopulations, however, ecology is much more simple, making it possible tolink these processes mechanistically. The work presented in this thesis addsto a growing number of studies that show the feedback between ecologicaland evolutionary processes on the same time scale.Though advances in genome sequencing technologies are drastically in-creasing the information known about the mutations involved in adaptation,this information is meaningless without ecological context. In other words,the effect of a given mutation depends on the genetic background in whichit arises, but it also depends on the genetic makeup of the population inwhich it arises like the frequencies of alleles involved in resource consump-tion, for example. Many studies have looked at the effects of mutations onphenotype and fitness, sometimes including how this effect depends on envi-ronment and genetic background, but few studies have investigated the effectof an adaptive mutation on subsequent adaptation or diversification. Here,I investigate the role of a single mutation affecting resource consumptionon subsequent diversification. Since this mutation alters the rate at whichresources are consumed in an environment in which individuals competefor resources, there is a direct link between the frequency of this allele andthe selective environment. In other words, the resources available dictatethe selective pressure on the types present, and in turn, the types presentdetermine which resources are available.Diversification in resource exploitation is common in microbes. In thiswork, Escherichia coli adapts to a two-resource environment by diversify-ing in metabolic strategy. Interestingly, diversification is also observed inenvironments supplemented with only one resource ([46]), often due to cross-feeding mechanisms. In these cases, at least one of the other types evolvesto consume a metabolic byproduct of the other type. In this situation, itis easy to see that these types would be maintained by negative frequency-dependent selection. Because of how common this type of selection seemsto be, insights from the experiment presented here could be applied to avariety of other organisms.This thesis contains two core chapters. The next chapter provides thebackground for the main body of the thesis. This part begins with a sum-mary of the contributions of microbial evolution experiments to understand-ing adaptation, followed by a brief description of the experiments that pre-ceded the current experiment, which is described in the last chapter. One2Chapter 1. General Introductionmain finding of this work is that a mutation with no effect on phenotypehas a profound effect on subsequent adaptation, which has important con-sequences for predicting evolutionary trajectories.3Chapter 2Background2.1 Generalizations from evolution experimentsSince there already exist many detailed reviews of the merits and limita-tions of laboratory experimental evolution (including [15], [19], and [27] forexample), I?m not going to provide another one here. Instead, I will brieflysummarize the light that these experiments have shed on our general un-derstanding and predicting of adaptation. Though evolution experimentshave informed a wide array of topics, many key results fall into three verybroad categories: repeatability, epistasis, and the rate of adaptation. Inshort, we have learned that adaptation is somewhat repeatable, epistasis isrampant, and it seems like the rate of adaptation decreases over time afterintroduction to a new environment, but perhaps not.2.1.1 RepeatabilityOne question that evolutionary biologists often hear is ?how repeatable isadaptation?? Any understanding of this question informs the ultimate ques-tion in evolutionary biology: can we predict evolutionary change? In otherwords, how much of the adaptation we observe is constrained to follow a cer-tain path and how much is due to stochasticity? This question can be (andhas been) informed by studies of the parallelism of adaptation, in whichmultiple populations are allowed to evolve and adapt to a given environ-ment or a set of environments and the outcomes are compared. Due to theease of growing microbial populations in a controlled laboratory environ-ment, many of the studies that address this question have been conductedwith microbes (including [32], [51], [47], and [39]). When adaptation iscompared across populations, this question of parallelism can be appliedto either genotypes or phenotypes, as well as the mapping between them.Decades of observations of natural populations have demonstrated that par-allel phenotypic evolution is common (including well known examples suchas anolis lizard ecomorhps ([33]) and threespine stickleback traits ([48]), butit is only recently that evolutionary biologists have been able to address42.1. Generalizations from evolution experimentswhether the genetic changes underlying that phenotypic parallelism is alsoparallel. Microbial evolution experiments are especially well suited to an-swer this question because the ?history of evolution? can be controlled byinitializing populations that are identical genetically, removing any contin-gency that adaptation might have on previous adaptation. So far, microbialstudies that compare the mutations that arise during adaptation betweenreplicate populations (i.e. a measure of genetic parallelism) have revealedthat there is a mix of shared and unique mutations. This includes studies ofadaptation to novel nutrient environments ([25], [37], and reviewed in moredetail in [11]), antibiotics ([17]), and even mutations underlying cancer pro-gression ([23]). In many of these studies, it is unclear which of the mutationsare adaptive and which are neutral due to the arduous task of investigatingthe effects of single mutations on phenotypes. Given what we know aboutmapping genotypes to phenotypes (in particular that many genotypes re-sult in the same phenotype), these results should not be surprising. Whatremains to be seen is what factors determine how parallel the mutationsunderlying adaptation are, and with the ever-increasing ease of acquiringsequence data, these factors will begin to be understood.2.1.2 EpistasisThe second major obstacle in predicting adaptation is the prevalence ofepistasis. Because of epistasis, evolutionary biologists can not simply usea distribution of fitness effects of specific mutations to predict which oneswill increase in frequency over the course of adaptation. An interaction be-tween gene products means that many of the predicted fitness effects onlyapply to a single mutational step. There exists much empirical evidenceshowing that the fitness effect of a mutation (during adaptation) is stronglydependent on the genetic background in which is arises (such as [45], [9],[31], [35], and [41], for example). Due to complex interactions between geneproducts, predicting the effects of any mutation on phenotype becomes nextto impossible. Microbial evolution experiments have begun to investigatehow different combinations of adaptive mutations affect fitness (including[7], [28], and [61]), showing that understanding epistasis is a crucial compo-nent predicting evolutionary trajectories. Since many of these studies haveinvestigated the interactions between mutations that rose to high frequenciesunder strong selection, they may be missing important interactions betweenmutations that arose but were selected against because of negative epistasis.This issue is beginning to be addressed with mutation accumulation exper-iments, which remove clonal interference from the population by reducing52.1. Generalizations from evolution experimentsthe population size to one individual at regular intervals. The idea behindthis is to minimize selection so that spontaneous mutations can be stud-ied without a bias towards those that are beneficial and against deleteriousones (reviewed in [24]). So far, these mutation accumulation experimentsand studies of the fitness effects of different combinations of mutations haveshown that epistasis is rampant and can dramatically alter evolutionary tra-jectories, making it very important for predicting adaptation. How exactlyepistasis impacts adaptation (for example by accelerating or slowing adapta-tion) has been informed by these evolution experiments in microbes, but thelarger-scale experiments that are now becoming possible will provide a muchmore general understanding of the role of epistasis in predicting evolution.2.1.3 Rate of adaptationDue to many convenient properties of microbes, evolutionary biologists havebeen able to answer questions about the rate of adaptation using experimen-tal evolution. In particular, because populations can be frozen and stored forlong periods of time, direct comparisons can be made between ancestral andevolved populations or individuals. Most studies that seek to characterizehow quickly populations adapt to novel environments find that the largestgains in fitness occur upon introduction to the novel environment, and de-crease over time ([5], [3]). This is consistent with the theory that during an?adaptive walk? up a fitness peak, it becomes increasingly difficult to climb asthe population ascends the peak, because mutations in the direction of theoptimum become less common ([43]). One of the major problems with us-ing comparisons between ancestral and evolved populations to quantify therate of adaptation is in defining and measuring fitness. Often, these studiesuse growth rate relative to the ancestor when grown together as the metricfor fitness, but that does not necessarily simulate the environment in whichmutations arose that increased fitness. For example, the observations of de-creasing fitness gains may only be measuring one aspect of fitness and otherpositive fitness effects may go unnoticed. Furthermore, fitness could be fre-quency dependent, which isn?t usually captured in typical fitness assays. Insome cases, certain types have been found to arise and invade ancestral pop-ulations repeatedly without any fitness benefit as typically measured ([22]for example). Fitness aside, rate of adaptation has also been described bythe rate of new mutations being substituted in the population. Sequencinghas revealed that the substitution rate of new mutations usually decreasesover the course of adaptation (reviewed in [11]). One notable exception tothis has been that the rate of genomic evolution in the long-term lines of62.2. Model systemthe Lenski lab has been relatively constant over about 20,000 generations([3]). As with repeatability and epistasis, the rate of adaptation in termsof substitution rate is being explored to new depths with ever-advancingsequencing technology.Evolution experiments have informed the impact that each of these topicshas on predicting adaptation. As genomic data for these experiments beginsto pile up (reviewed in [2], [8], [11]), it will be interesting to see how much ofthe stochasticity in evolutionary trajectories can be understood. In regardsto the model system used in the work presented here, we know that someof the genetic changes underlying adaptation are parallel, and some are not([25]). Interestingly, the degree of parallelism is higher in one ecotype thanthe other. Epistasis is evident; the effects of at least one mutation dependon the genetic background in which it arises ([31]). Lastly, the substitutionrate was found to be higher in some lines than others ([25]), but it?s unclearif it decreases over the course of the experiment.2.2 Model system2.2.1 Motivation for sympatric speciation experimentsFor many decades following the modern synthesis, it was thought that speci-ation occurred most commonly in allopatry; through reproductive isolationbetween subpopulations acquired in geographic isolation ([13]). Theoreti-cally, it was much easier to show how populations that become physicallysubdivided could diverge. On the other hand, theory supporting specia-tion was largely non-existent at the time, leading some scientists ([36] forexample) to remain unconvinced that speciation was even possible withoutgeographical isolation. Near the turn of the millennium, however, a solidtheoretical framework for predicting sympatric speciation was laid down by[21], [38] and [12]. This framework helped to alleviate doubts of sympatricspeciation as an observable process in nature, spurring many researchers toconsider the possibility. Some theoreticians remained skeptical that sym-patric speciation was common, arguing that the conditions under which itwas predicted to occur theoretically were too strict to be found in nature([10]), though some researchers did find that these specific conditions werefound in natural populations ([59]). Following that, some reviews concludedthat sympatric speciation is theoretically plausible and has been observedin multiple instances, but that it is still unknown how common it is ([4]).The experiment presented in this work is based on a near-decade long series72.2. Model systemof experiments that initially sought to understand and explore sympatricdiversification of microbes.2.2.2 Previous experimentsIn the initial experiment, conducted by Friesen et al. (2004), a strain of E.coli was evolved in 12 replicate populations by growing in a liquid mediumcontaining glucose and acetate as carbon sources and diluting by 1:100 every24 hours ([18]). In this experiment, 12 out of the 12 populations were foundto contain two distinct types of colonies after 1000 generations. These twodifferent types of colonies differed mainly in the size of the colony after about12 hours of growth on a solid agar medium. This initial study also foundthat the two types differed in their diauxic growth patterns; one type grewfaster during the first phase of growth (glucose), and the second type hada shorter lag time between switching over to the second phase of growth(acetate). These two types were named FS (for fast-switcher) and SS (forslow-switcher). Furthermore, they found that these types were maintainedby negative frequency-dependent selection.Next, competition experiments were performed between clones isolatedfrom different replicate lines and different evolutionary environments ([57]).In the initial experiment, populations were propagated in single resourceenvironments (glucose only and acetate only) as well as the mixed resourceenvironment. FS and SS clones were isolated from either the same or dif-ferent environments and competed against each other to determine if thecompetitive relationship was dependent on the environment in which theclone had evolved. Clones that had evolved in the same environment (aFS from glucose+acetate vs. a SS from a different glucose+acetate line)demonstrated the same competitive relationships (maintained by negativefrequency dependence), showing that diversification was parallel. Clonesthat had evolved in different environments (a FS from glucose only vs. a SSfrom glucose+acetate, for example) demonstrated varied competitive rela-tionships with no evidence of stable intermediate frequencies, showing thatdiversification was not parallel.Measurements of glucose and acetate concentration in the growth mediumover a 24-hr period showed that acetate concentration increased during thefirst phase of growth of ancestral and SS strains (since acetate is a byproductof glucose metabolism), while acetate did not increase in the first phase ofgrowth of an FS strain ([53]). This indicated that acetate consumption wasnot repressed in the FS strain. To investigate the genetic mechanism behindthis, expression levels of genes known to be involved in acetate metabolism82.2. Model systemwere measured. An increase in expression of the aceB gene was found to beassociated with an insertion mutation in the iclR gene. PCR screening ofthe iclR gene showed that this insertion mutation was present in 8 of 9 FSclones isolated from the same line, but not present in any of the SS clones,and not present in any FS clones isolated from different lines, demonstratingthere are other genetic changes underly the FS phenotype.To further investigate the genetic differences between FS and SS eco-types, microarrays were used to profile global transcription of an FS clone,an SS clone, and the ancestral clone ([30]). The FS and SS clones used inthis study were isolated from the 1000 generation time point of a single pop-ulation. This study found many differentially expressed genes in commonwhen comparing FS to the ancestor and SS to the ancestor. These changeswere thought to be generally adaptive to growth in serial batch culture andincluded genes involved with translational efficiency, glucose uptake capac-ity, and survival during stationary phase. Genes differentially expressedbetween the FS and SS clone were associated with upregulation of the TCAcycle and acetate consumption in FS and with upregulation of genes involvedin acetate excretion in SS. Most importantly, this study strongly supportsthe argument that diversification in this system is driven by competition forcarbon sources, as shown by the metabolic differences between FS and SSclones.To test the theory that populations first undergo a phase of directionalselection before diversification becomes adaptive, the next study isolatedSS clones from three different timepoints (before the branching point) andallowed them to re-evolve ([54]). This was to determine if the likelihood ofdiversification changed over time. The re-evolution period was about 140generations and this experiment found that the clones isolated from 400generations were much more likely to diversify in that time than clones iso-lated from earlier, in line with evolutionary branching models predictingthat populations evolve towards a branching point before diversifying. Thisstudy also describes the diverse ecotypes in terms of specialists and general-ists. Interestingly, instead of populations evolving towards a glucose/acetategeneralist before diversifying into a glucose specialist type and an acetatespecialist type, they evolve towards a glucose specialist before diversifyinginto an ever more specialized glucose specialist and a glucose/acetate gen-eralist.Since the previously described experiment only investigated the like-lihood of diversification before the branching point, and only in a singlepopulation, another ?rediversification? experiment was conducted using mul-tiple FS and SS clones isolated from the endpoint of the original evolution92.2. Model systemexperiment ([56]). This study, lasting for about 200 generations, found thatonly SS-initiated populations diversified into populations containing both FSand SS types. In contrast, none of the FS-initiated populations diversified,suggesting that a much higher proportion of mutations cause a phenotypicchange from an SS to an FS strain as opposed to the reverse direction. Thisstudy highlights the importance of mutational bias in predicting evolution-ary outcomes.Following the experiment that identified differentially expressed genesbetween FS and SS clones, one specific mutation was chosen to investigateits phenotypic and fitness effects ([31]). Since many of the differentiallyexpressed genes are regulated by the global transcription factor arcA, apoint mutation in this gene was thought to underly the phenotypic differ-ences between FS and SS. Using allelic replacement techniques, this muta-tion was inserted into the genomes of both FS and SS clones isolated frommany different timepoints. Interestingly, the arcA? mutation had a largeeffect in the SS background, but not in the FS background. In particu-lar, SSarcA? clones from every timepoint showed a dramatically reducedlag time between switching from glucose consumption to acetate consump-tion. Invasion experiments showed that the mutation was adaptive in SSbackgrounds (SSarcA? outcompetes SSarcA+) but not adaptive in FS back-grounds (FSarcA? does not outcompete FSarcA+). Furthermore, the muta-tion was no longer adaptive in the presence of the FS ecotype; the mutationis only adaptive in SS when the FS type is not present, demonstrating anenvironment-dependent fitness effect.The work presented in this thesis combines these last two studies to in-vestigate the effect of the arcA? mutation on the likelihood of diversification.Simply, this ?rediversification? experiment found that the mutation was as-sociated with reduced diversification in the SS background and increaseddiversification in the FS background.10Chapter 3Adaptation in E.coli :Intraspecific EcologicalInteractions and GeneticConstraints DetermineDiversification3.1 IntroductionDiversity is abundant, and there are many mechanisms that can generatediversity. Some of these mechanisms have been shown to generate intraspe-cific diversity in sympatry, resulting in the formation of new species, or inthe case of microbes, divergent ecotypes ([16], [1], and [29] for example).One such mechanism is disruptive selection that is generated by intraspe-cific competition for resources, resulting in diversification ([12]). With thismechanism, the combination of negative frequency-dependent ecological in-teractions and phenotypic evolution can transform an undifferentiated pop-ulation into a diversified one ([14]). For example, E. coli propagated in awell-mixed resource-limited environment repeatedly diversifies into two dis-tinct types ([18]). While this process has been observed several times, itremains unclear to what degree this divergence is deterministic, and whatfactors affect the likelihood of diversification in populations under this typeof selection. Understanding the potential for these populations to diversifyis critical to predicting evolutionary responses to changing conditions.In populations that are initially isogenic, like a bacterium infecting anovel host, for example, variation for adaptation is supplied by changes inDNA sequence. Because of this, we often use mutation rate as a proxyfor evolutionary potential, predicting that organisms with higher mutationrates will be more successful at adapting to novel environments. Thoughthere does exist experimental evidence to support this relationship (such as113.1. Introduction[6] and [52]), there are other factors influencing the ability of a populationto adapt. We can group these factors into two broad categories: ecologicalfactors that shape selective pressures, and genetic factors that provide adap-tive mutations. Specifically, the evolutionary potential of a population todiversify depends on whether or not selection is disruptive due to the phe-notypic composition of the population (ecological factors) and how likely itis that genetic mutations result in a phenotypic change that is relevant toresource consumption (genetic factors).If we temporarily ignore underlying genetic constraints and assume thatthe probability of diversification depends only on ecology, we can use anadaptive landscape to predict evolutionary trajectories. A monomorphicpopulation occupying a fitness valley between two peaks would have thehighest chance of diversifying, while a population occupying a fitness peakor ridge would have a lower chance ([62], [20]). If the landscape is static,when a population that occupies a valley diversifies, the two diverged pop-ulations would each have a decreased likelihood of further diversifying asthey ascend their respective peaks. Observations in diversifying popula-tions of Pseudomonas fluorescens are consistent with this expectation, asthey show a reduced propensity to diversify after the initial adaptive split([44], [5]). When the shape of the fitness landscape is also determined bythe phenotypic composition of the population (i.e., frequency-dependent),evolutionary change can result in a change in the shape of the landscape.In this case, a dynamic landscape can make predicting diversification muchmore difficult.Adaptive dynamics models account for changing landscapes, describingthe fitness of a given phenotype and subsequent evolution as a function ofthe composition of the population ([21]). While these types of models in-corporate the context dependence of phenotypic fitness, which are typicallyignored in classical population genetic models, they ignore the genetic de-tails that determine which phenotypes are and are not accessible. Microbialevolution experiments have previously been interpreted using the adaptivedynamics framework ([54], [56]), but it remains unclear whether the predic-tions made under this framework are robust to the types of evolutionaryconstraints imposed by mutational landscapes (like those described in [60]).The utility of these models in predicting evolutionary trajectories of naturalpopulations depends on these assumptions, so it is critical to understand howepistasis and genetic background interact to constrain trajectories predictedby these models.Since diversification occurs when new phenotypes arise, the likelihood ofdiversification is constrained by the possible genetic changes that can give123.1. Introductionrise to the necessary phenotypic variation. Typically, models of adaptiveevolution assume that mutation effect sizes are symmetrical around a phe-notype, so that random changes are just as likely to shift a particular traitin one direction as they are in the other ([34]). Due to complex epistaticnetworks, however, this is often not the case in nature ([60], [50]). For ex-ample, there may be many more ways to generate a null mutation than avariant with increased enzyme activity. Furthermore, while the probabilityof a mutation at any point in the genome is approximately uniform, theeffect of that mutation on phenotype, and therefore adaptation, dependsboth on the genomic background in which it arises, and on the environmentthat the organism currently occupies. All of these aspects of the biologyof real organisms could limit the utility of abstract models for predictingevolutionary outcomes.In the current work, we use populations of Escherichia coli to investigatethe interaction of genetic and ecological factors on the likelihood of diversi-fication. This follows a number of experiments exploring the diversificationof E. coli in a laboratory environment. When propagated in a well-mixedenvironment containing two carbon sources, originally isogenic populationsdiversify into two distinct, heritable types ([18]). Typical of bacterial growthin a medium with two resources, populations exhibit diauxic growth in whichthere are two distinct phases of growth separated by a period of little or slowgrowth. The first phase is comprised of the consumption of the preferredresource (glucose), followed by the consumption of the secondary resource(acetate). The two types observed to coexist after many generations differprimarily in the amount of time between phases. One type (referred to asSS for ?slow-switcher?) consumes glucose efficiently, but exhibits a long lagtime between depleting glucose and consuming the secondary resource. Thesecond type (FS for ?fast-switcher?) consumes glucose less efficiently thanSS, but has a much shorter lag between phases.Previous experiments found this diversification to be consistent, arisingin 12 of 12 lines between approximately 200 and 400 generations of dailybatch culture ([18]). Competition experiments determined that the diver-gent types are maintained by negative frequency-dependent selection due totrade-offs in resource consumption ([58]). Sequencing of FS and SS clonesfrom three different divergent populations showed that the metabolic differ-ences were due to some shared mutations and some unique mutations ([25]).Interestingly, SS clones were more similar to each other than FS clones wereto each other, in terms of the mutations that were detected. A single pointmutation in the arcA gene (which encodes a transcription factor for anaer-obic respiration control) was found by [30] to be associated with many of133.1. Introductionthe genes that are differentially expressed between FS and SS. To quantifythe fitness and phenotypic effect of this mutation, it was inserted into thegenome of several SS and FS clones isolated from every 200 generations fromthe original evolution experiment ([31]), generating strains genetically iden-tical except for this point mutation. When introduced into the genome of anFS individual, this mutation has little or no effect on phenotype ([31], thiswork). In the SS background, however, this mutation results in an interme-diate lag time between growth phases (referred to here as MS, for ?mediumswitcher?). Furthermore, between the clones isolated from different timepoints along the original evolution experiment, we found no measureabledifference in the effect of the mutation on switch time (i.e. the arcA? allelereduced the switch time in an SS individual isolated from gen200 by aboutthe same amount as in an individual from gen1000). Because of the clearecotype-dependent phenotypic effect of this particular mutation, these en-gineered strains were used to investigate diversification in this system witha novel set of initial conditions; specifically, the effect of the arcA? muta-tion, initial ecotype, and number of generations previously evolved on thelikelihood of subsequent diversification.Here, we take isogenic colonies of these genotypes (SSarcA+, SSarcA?,FSarcA+, and FSarcA?) from five different time points from the experi-ment conducted by [31] and expose them to a second bout of evolution,lasting approximately 220 generations. To determine if populations under-went diversification, we measure the ?switch-time? of several clones from thepopulation, and from this calculate two parameters: 1) the variance in thisphenotype present in the sample and 2) the range of phenotypes present inthe sample (both measures of diversification). By comparing these statisticsbetween lines, we can determine how the likelihood of diversification changesover time, between ecotypes and what effect the arcA? mutation has on thislikelihood. With a full factorial design, we can determine the effects of thesefactors individually as well as the interactions between them in an attempt tounderstand the relative contributions of ecological and genetic factors to theprocess of diversification. Evidence from this experiment shows that bothof these factors play a role in adaptation and that predicting diversificationrequires a careful integration of ecological and genetic complexity.143.2. Methods3.2 Methods3.2.1 Bacterial strainsThe bacterial strains used in this experiment were isolated from a previouslong-term evolution experiment conducted by [58], in which the originalfounding ancestor was Escherichia coli B strain REL606. Originally, twentyreplicate populations were propagated for approximately 1200 generations ina liquid media supplemented with glucose and acetate, in the same manneras described in [18], and also described in the Rediversification section below.One of these original populations, pop20, was chosen for further experimentsby [31], and all genotypes selected for this work were from this population.This particular population was chosen partly because it diversified relativelyearly, with both SS and FS ecotypes being present in the population by200 generations. [31] isolated both FS and SS clones from several differenttime points along the original evolution experiment. Then, using a pKO3plasmid vector carrying a mutated version of the arcA allele, the arcA locusof the clones was modified (by [31]) to contain a thr81ala point mutation.The mutation was inserted into the genome by homologous recombinationbetween the plasmid and genome in the regions flanking the point mutation,followed by excision of the plasmid DNA by a second recombination event.After sequencing to ensure that no other mutations had arisen, the allelicreplacement resulted in genotypes that differed only at the arcA locus by thepresence or absence of this point mutation. For the current work, to ensurethat our founding populations were isogenic, we isolated individual FS andSS clones from five time points along the frozen fossil record, beginning at200 generations, recurring every 200 generations (illustrated in Figure 3.1,panel A). As well as these 10 genotypes, we use the modified strains with thearcA? mutation, for a total of 20 unique genotypes. Before beginning theevolution experiment, the clones were screened for the presence or absenceof the mutation using restriction enzymes.3.2.2 RediversificationThe 20 founding genotypes, each replicated four times, were propagated inserial batch culture for approximately 220 generations, yielding 80 evolvedlines in total (see Figure 3.1 for a schematic of the experimental design).Since these bacteria have been previously exposed to experimental evolution,we refer to this second bout of evolution as the rediversification period. Theenvironmental conditions we exposed the bacteria to were identical to theone in which the populations were originally evolved; 18-mm diameter tubes153.2. Methodswith 10 mL of Davis Minimal (DM) media supplemented with 0.25 mg/mlglucose and 1.3225 mg/ml sodium acetate trihydrate (DMGA). Every 24(+/- 1) hours, 100 ?L of culture was transferred to fresh media, and thesewere stored in a shaking incubator maintained at 250 rpm and 37oC. Samplesfrom all 80 lines were taken every four days and frozen in 20% glycerol.163.2. MethodsSS	 ?FS	 ??me	 ?(days)	 ? 600	 ?200	 ? 400	 ? 800	 ? 1000	 ?time (hours)ODtime (hours)ODtime (hours)ODanc	 ?=	 ?REL606	 ?time (hours)ODtime (hours)ODtime (hours)ODtime (hours)OD arcA	 ?thr81ala	 ?arcA	 ?thr81ala	 ?FSarcA-??	 ?FSarcA+	 ?SSarcA-??	 ?SSarcA+	 ??me	 ?(days)	 ?time (hours)OD220	 ? C	 ?A	 ?B	 ?diversity	 ?=	 ?DV	 ?,	 ?DR	 ?Figure 3.1: Schematic of experimental design. Individual clones were sam-pled every 200 generations from the frozen fossil record of the original 1000+generation evolution experiment (panel A). When transformed with thethr81ala point mutation in the arcA gene (arcA?), there is a large effect onthe switch time phenotype of the SS type, producing the ?medium-switcher?phenotype (panel B, top row), while having no observable effect on the phe-notype of the FS type (panel B, bottom row). All four types (SSarcA+,SSarcA?, FSarcA+, and FSarcA?) from every 200 generations are subjectedto a secondary bout of evolution (referred to as the rediversification period)lasting for approximately 220 generations, with each genotype evolved infour replicate lines (panel C). After the rediversification period, individualcolonies are assayed for switch time, defined as the amount of time (in hours)between maximum density in the first phase of growth and maximum den-sity in the second phase of growth, and this parameter is used to determinediversity present in a population.173.2. Methods3.2.3 Phenotypic assaysFollowing the rediversification period, all 80 lines were assayed for pheno-types present in the population. As a control, the 20 founding genotypeswere also assayed (i.e. the lines before the rediversification period, or non-evolved). To determine phenotypes, we isolated and measured character-istics of several individuals within each population. First, 5 mL tubesof DMGA were inoculated with frozen samples of culture and incubatedovernight. Then, each population was diluted, plated onto tryptone plates,and incubated overnight. Following that, individual colonies were pickedrandomly and transferred via sterile toothpick to 5 mL tubes of DMGA, tobe incubated overnight. Lastly, after 24 hours of growth, 2 ?L of culturewas used to inoculate 200 ?L of DMGA in a 96-well plate. Using a BiotekEL808 microplate reader, optical density was measured at 600 nm every 15minutes over a period of 36 hours, to generate a plot of population size overtime for each colony. We use this growth curve as the phenotype of a colonyfor further analysis. From all 80 lines following rediversification, we assayed10 colonies per line, and from 20 lines prior to rediversification, we assayed20 colonies, to yield a total of 1200 growth curves.3.2.4 Statistical analysisAll analysis of the growth curve data was done using the statistical programR v 2.14.1 ([42]). We defined ?switch time? as the distance (i.e. time) be-tween the maximum optical density in phase 1 of growth and the maximumoptical density in phase 2 of growth. For SS types, this quantity is relativelylarge (?25 hours), while for FS types it is small (?2 hours, illustrated inFigure 3.1, panel A). For the curves that did not appear to reach a secondmaximum within the 36 hour growth assay, as was the case with a smallproportion of SS types, the end time point was used to ensure a conserva-tive measure of distance between maxima. To quantify diversification, wecalculated two measures from the switch time data. The first, ?diversity byvariance? (Dv), was defined as the variance in switch time between individu-als within a population. The second, ?diversity by range? (Dr), was definedas the difference between maximum and minimum switch time within eachpopulation. Using a threshold value, populations with Dr ? 15.0 hours wereconsidered to have diversified. To compare diversification between geno-types, a multi-factor ANOVA was performed, using generation (200, 400,etc.), ecotype (FS/SS), and mutation (presence/absence) as the explana-tory variables.183.3. Results3.3 Results3.3.1 Naming conventionTo avoid confusion between the evolved phenotypes and the original found-ing genotypes, we adopt the following conventions. Although SS and FShave been used to refer to a phenotype in previous papers, here we use theseterms to refer to the founding genotypes. To refer to a phenotype, we callindividuals slow, medium, or fast, according to the time it takes them toswitch between phases of growth. Since the founding clones isolated fromdifferent time points are also distinct genetically, we use gen200, gen400,etc. to refer to the previous amount of generations they had evolved in theoriginal experiment.3.3.2 Phenotypic evolutionTo assess phenotypic evolution, we sampled random individuals from eachpopulation and quantified their phenotype (switch time), using the 36 hourgrowth profile generated for each clone. To estimate phenotypes before therediversification period, we sampled 20 colonies of each of the four types(SSarcA+, SSarcA?, FSarcA+, and FSarcA?) from all five time points.After the rediversification period, we sampled 10 colonies of each replicatepopulation of each type from all five time points. These data are summarizedin Figures 3.2 and 3.3, though it should be noted that since all timepointsand replicates are pooled together to show general changes in phenotypes,these two figures do not show information about the diversity present in in-dividual populations. We found 12 out of the 20 populations initiated withthe SSarcA+ ancestor contained two distinct ecotypes after 220 generations,while only 1 out of 20 populations initiated with SSarcA? contained morethan one ecotype (Table 3.1, Figure 3.2 bottom panel). Moreover, the pop-ulations initiated with SSarcA+ that diversified, did so into fast and slowtypes, while populations initiated with SSarcA? did not diversify, but in-stead evolved into fast types. The one SSarcA? initiated replicate that diddiversify was a gen1000 line and was found to be clearly dimorphic. Of the20 populations initiated with FSarcA+ and FSarcA?, 8 and 11 populationsdiversified, respectively (Table 3.1, Figure 3.3).193.3. ResultsInitial genotype Lines with Dr ? 15.0SSarcA+ 12FSarcA+ 8SSarcA? 1FSarcA? 11All genotypes 32Table 3.1: Summary of proportion of lines of each genotype that diversified.SS arcA+0 5 10 15 20 25 30 35SS arcA?switch time (hrs)densityFigure 3.2: Phenotypic evolution within SS types. Density distribution ofthe switch time trait in SS arcA+ and SS arcA? both prior to (dotted lines)and following (solid lines) the rediversification period (?220 generations,showing the evolution of switch time within each type. All five timepointsand replicates are grouped together.203.3. ResultsFS arcA+switch time (hrs)0 5 10 15 20 25 30 35FS arcA?switch time (hrs)densityFigure 3.3: Phenotypic evolution within FS types. Density distribution ofthe switch time trait in FS arcA+ and FS arcA? both prior to (dotted lines)and following (solid lines) the rediversification period (?220 generations,showing the evolution of switch time within each type. All five timepointsand replicates are grouped together.3.3.3 Effect of arcA? mutation on diversificationTo quantify diversity present in each population, we calculated both thevariance in switch time of individuals within that population (Dv) as wellas the range of switch times (Dr). Since the distribution of Dr values wasclearly bimodal, we chose the middle point between the two peaks (Dr ? 15.0hours) as the threshold value to consider a population diversified. Giventhis criteria, after 220 generations of evolution, 32 out of the 80 evolutionlines were found to be diversified (Table 3.1). Of these, many populationsshowed a clear dimorphism of fast and slow switcher types, while othersshowed a clear trimorphism of fast, slow, and medium types. See Appendix1 (Figures A.1, A.2, and A.3) for examples of these populations. In a small213.3. Resultsproportion of lines, individual curves varied widely, but did not form distinctclusters, though a lack of clusters could be due to a small sample size. Theeffect of the arcA? mutation on the propensity to diversify was found to behighly dependent on whether it was in an FS or an SS line. To quantifythe effect of the mutation, we took the difference between Dv for arcA+populations and Dv for arcA? populations and called this ?Dv (Figure3.4). Within SS-initiated lines, arcA+ populations were much more likelyto diversify than arcA? populations (Figure 3.4). For all five timepoints, themutation dramatically hindered diversification, with a considerably strongereffect observed for SSgen200. The mutation had the opposite effect in FS-initiated lines; arcA? populations were much more likely to diversify thanarcA+ populations (Figure 3.4). This was the case for all timepoints withthe exception of FSgen200.?Dv = Dv(arcA?) ?Dv(arcA+) (3.1)223.3. Results-150-100-50050100150generation?DV**200 400 600 800 1000 OverallFigure 3.4: Effect of arcA mutation on diversity by variance (Dv) in bothFS (green circles) and SS (blue circles). ?Dv is the difference between typeswith the mutation and types without the mutation, so that a positive valuedenotes an increase in diversification associated with the arcA mutation andvice versa. Starred points denote average effect across all timepoints.3.3.4 Effect of generation on diversificationWe also investigated the effect of timepoint, or the amount of generationspreviously evolved, on diversification. Within populations without the arcA?mutation (i.e. SSarcA+ and FSarcA+ genotypes only), timepoint was foundto be a significant predictor of the evolution of diversity. Specifically, inFSarcA+, we observed a negative correlation between timepoint and Dv(Figure 3.5, second panel), showing that genotypes isolated from earlier inthe original evolution experiment are more likely to diversify than thoseisolated later. Since the founding population diversified around 200 gen-erations, this finding indicates that diversification is more likely for clonesisolated closer to the original branching point. In SSarcA+, however, the233.3. Resultsrelationship between generation and variance was found to be non-linear(Figure 3.5, first panel), showing that genotypes isolated from early andlate in the original evolution experiment are more likely to diversify thanthose isolated from intermediate timepoints. Not surprisingly, we found noeffect of generation on diversification within SSarcA? initiated lines. Sincediversification in these lines was suppressed at all timepoints, no differencesbetween timepoints could be observed (Figure 3.6, first panel). Despite mod-erate amounts of diversification in lines initiated with FSarcA?, we found noeffect of generation on diversification within those lines (Figure 3.6, secondpanel).200 400 600 800 1000050100150200generationvarianceSS arcA+200 400 600 800 1000050100150200generationFS arcA+Figure 3.5: Effect of generation on diversification for both wild type lines.Each point represents a single replicate population and is the variance inthe switch time phenotype present in that population (diversity by variance,Dv). The dotted lines separate populations which are comprised of a rangeof phenotypes (diversity by range, Dr) ? 15.0 hours from those in whichDr < 15.0 hours.243.4. Discussion200 400 600 800 1000050100150200generationvarianceSS arcA -200 400 600 800 1000050100150200generationFS arcA -Figure 3.6: Effect of generation on diversification for both mutant lines.Each point represents a single replicate population and is the variance inthe switch time phenotype present in that population (diversity by variance,Dv). The dotted lines separate populations which are comprised of a rangeof phenotypes (diversity by range, Dr) ? 15.0 hours from those in whichDr < 15.0 hours.3.4 DiscussionUnderstanding the factors that drive and/or constrain diversification caninform how populations will adapt to novel environments. Here, using bac-teria isolated from experimental populations that had previously diverged,we examined the propensity to diversify in a number of ways. We asked howfounding genotype, founding ecotype, and a specific point mutation affecteddiversification. We found that the effect of the arcA? mutation on diversifi-cation was highly dependent on the ecotype in which it was introduced; themutation decreased diversification in SS types and increased diversificationin FS types. We argue that this is evidence that adaptation in this system isconstrained by both genetic and ecological factors and that it is importantto consider the interaction between the two. Recently, investigation of thecomplexity of eco-evolutionary dynamics has been gaining momentum ([40],[49]) and the work presented here adds to a growing number of studies that253.4. Discussionexhibit a clear feedback between ecological and evolutionary processes onthe same time scale.3.4.1 Genetic and ecological constraintsOur data shows that within lines initiated with the intermediate phenotype(SSarcA?), very little diversification is observed and phenotypic evolutionoccurs in one direction; 19 out of the 20 populations that were initiallymedium types evolved to contain only fast types. Since the SSarcA? typeshave an intermediate phenotype, we might expect evolution to proceed ineither direction along that phenotypic axis. We find consistent evolution to-wards a smaller switch time, indicating a constraint on phenotypic evolution,either genetically or ecologically.Genetic factors can constrain evolution if there are differences in theavailability of mutations that increase versus decrease the value of a giventrait. Within the context of switch-time studied here, such differences inmutation availability could arise two ways. First, it could be the case thatgiven the genes involved in the underlying metabolic network, there are moremutations that result in a shorter switch time than a longer switch time (forexample, more loss-of-function than gain-of-function mutations). This hy-pothesis is consistent with results showing that populations initiated by SStypes are more likely to diversify than FS types (Figure 5A), implying agreater availability of mutations that decrease the switch time. This lowerdiversification in FS is unlikely to be constrained by ecology alone, becauseFS types can readily invade SS populations ([18], [54]). The second waythat an asymmetry in mutation effect could explain the results is throughepistasis, whereby the arcA? mutation itself changes the distribution of mu-tation effects available to increase or decrease switch time. Since the proteinmade by arcA regulates at least 30 operons ([55]), it seems plausible thata mutation changing the function of this gene could drastically alter thedistribution of mutation effect. This explanation is consistent with resultsshowing that the FSarcA? types had greatly increased diversification overFSarcA+ types (Figures 3.5 & 3.6), despite having phenotypes that werehighly similar ([31]). Since we know that mutations that result in slow typesare beneficial in populations of fast types, this difference suggests that inan arcA? background, it is more likely that a new mutation will result in aslow phenotype. This finding demonstrates that a seemingly neutral muta-tion can turn out to be adaptive in future generations (like those describedin [61]). Even though the interactions between metabolic genes in E. coliare relatively well understood, since these interactions are environment de-263.4. Discussionpendent, we cannot make predictions about the effects of specific mutations,even given the underlying network of epistasis. For example, [26] found thatLOF (loss-of-function) mutations in any one of three genes (ppsA, sfcA, andmaeB) resulted in a longer transition lag in a mixed environment, but hadno phenotypic effects when grown in either glucose alone or acetate alone.It is this environmental dependence that prevents the mutational landscapefrom remaining constant after a single mutation and provides yet anotherobstacle in predicting evolutionary trajectories.Though the availability of mutations is an important factor in deter-mining adaptation, the selective environment determines which mutationsand phenotypes will increase versus decrease in frequency, which modifiespredictions that might be made based on the above discussion of geneticconstraints. For example, while our results suggest that the arcA? mu-tation causes an increase in the availability of mutations that result in aslower switch time, all of the SSarcA? populations evolved towards a fastphenotype (Figure 3.2). In this experiment, the selective environment ispartly determined by the composition of the current population. The dif-ferent ecotypes have alternative strategies for consuming resources, whichin turn determines which resources (glucose and acetate) will be more orless abundant. Previous experiments have shown that fast types can invadea population of slow types, as well as vice versa, and that these types aremaintained at an intermediate frequency ([18]). From the current experi-ment, it is clear that mutants with a fast phenotype can invade a populationof medium types, but instead of reaching a dimorphic state, medium typesare outcompeted, and the fast types completely replace the ancestral pop-ulation (Figure 3.2). This is indicative of directional selection, as opposedto the disruptive selection experienced by populations of either fast or slowtypes. We hypothesize that since the medium types (initially SSarcA?)have an intermediate phenotype, there is only one niche available to occupy.Initially, the two typical niches (occupied by fast and slow types) are notavailable for colonization until the resident population is far enough awayin phenotypic space from the other potential resident morph. This patternof directional evolution followed by evolutionary branching was observed ina single population (initiated by gen1000-SSarcA? and we hypothesize thatif allowed to evolve for longer, more of the SSarcA? initiated populationswould show this same pattern of diversification.A lack of diversification in SSarcA? initiated lines suggests that mediumtypes cannot coexist with other types. We did, however, observe that someof the lines initiated by other strains did have medium individuals presentafter the rediversification period. It is unlikely that the occurrence of these273.4. Discussionintermediate types is the result of recurring mutations that are selectedagainst, because they reach high frequency (>20%) in certain populations.There are two other possible explanations for the persistence of mediumtypes: either they have diverged into an open niche along a phenotypicaxis other than switch time that we did not measure, or the slow and fasttypes could be phenotypically divergent enough to create an open nicheat an intermediate switch time. We did not find a difference between theswitch time of fast and slow types in populations with medium types presentand the switch time of fast and slow types in populations with no mediumtypes present, suggesting that the second of these possibilities is less likely,though it may be the case that the difference occurs on a finer scale than wemeasured. It would be interesting to isolate and compete different mediumclones to evaluate which of these explanations is most likely.The findings of this experiment, especially the evolution of the mediumtypes, combined with previous experiments suggest that there is a complexunderlying ecology guiding evolution. The interactions between differentphenotypes could be visualized by a pairwise invasibility plot, or PIP ([21]),shown in Figure 3.7. Previously, [54] showed that the ancestral ecotype firstevolved an increase in switch time before diversifying, indicating a phaseof directional selection towards a branching point. This can be interpretedon a PIP as an evolutionary attractor (point B, Figure 3.7), meaning thatthe branching point would always be reached, irrespective of the startingphenotype. The finding that medium types evolve in one direction (towardsa fast phenotype only) allows us to infer that there is another importantfeature in this PIP. We can include this information by adding an evolution-ary repellor somewhere between the ancestral and medium phenotype (pointA, Figure 3.7), meaning that phenotypes will always evolve away from thatpoint in either direction, depending on the starting phenotype. Because weknow that slow types cannot invade populations of medium types, the ini-tial medium phenotype must lie on the fast side of the evolutionary repellor.Though the picture could contain more complexity, this PIP summarizes theinvasion fitnesses for which we have evidence. Understanding how selectionregimes change with evolving phenotypes, together with a clear genetic pic-ture of which phenotypes are attainable is crucial to predicting evolutionarytrajectories.283.4. Discussion++---resident trait valuemutant trait valueFS MS anc SSABFigure 3.7: Pairwise invasibility plot, showing regions of positive invasionfitness (grey) and negative invasion fitness (white) for all combinations of res-ident (x-axis) and mutant (y-axis) phenotypes (in this case, ?switch time?).Invasion fitness is zero on the diagonal (since mutant and resident have thesame fitness), and evolutionary equilibria are the intersection points of thediagonal and the invasion fitness 0-isocline. The slope of this line at theintersection determines the evolutionary dynamics, so that point A is anevolutionary repellor and point B is an evolutionary attractor. Since thereare multiple equilibria, the direction of evolution depends on the initial con-ditions.293.4. Discussion3.4.2 Effect of timepointThe decrease in the propensity to diversify with the number of generationspreviously evolved in the FSarcA+ initiated types shows that, within thistype, as timepoint moves later from the original branching point (whichwas before 200 gens), diversification becomes less likely. This finding is inaccordance with theoretical predictions that attribute this decrease in di-versification to the increasing phenotypic distance between divergent types.Though we do not observe an increase in phenotypic distance (i.e. the fasttypes from gen1000 do not have statistically shorter switch times than thosefrom gen200), it is likely that there are differences in other measures. Fur-thermore, it is possible that the accumulation of mutations has affected thispicture of diversification probability. For example, as evolution proceeds, itcould be the case that the mutations that fix in the population are thosethat are redundant with the mutations that decrease switch time. Underthis hypothesis, what changes over evolutionary time is the distribution ofmutation effect in this particular trait dimension, therefore decreasing thelikelihood that a mutation arising will result in a slow phenotype. It isinteresting to observe the opposite, an increase in diversification in latertimepoints within the SSarcA+ lines, and perhaps this is indicative of asecond branching point further along in evolutionary time, though this iscomplete speculation.Though we observed an effect of timepoint on diversification in bothSSarcA+ and FSarcA+ derived populations, diversification did not dependon timepoint in SSarcA? and FSarcA? derived populations (Figure 3.5 and3.6). While we did not have an a priori prediction for how the effect ofthe arcA? mutation on diversification would depend on the amount of timepreviously evolved, this result could make sense in light of expected epistaticinteractions between arcA? and mutations in the different founding clones.In SSarcA?, an effect of timepoint is not observed, possibly because themutation has such a dramatic effect on phenotype, resulting in reduced di-versification, thus trumping the effect that any genetic differences mighthave on diversification. In FSarcA?, however, while the mutation increasesdiversification on average, there is no clear effect of timepoint, suggestingthat there may be some interaction between the artificially introduced arcA?mutation and the unique mutations that are present in the individuals iso-lated at different timepoints. By inserting arcA? artificially, the lines withthe mutation are exposed to a genetic change that has not been previouslytested in the populations, varying the effect of the timepoint on diversifica-tion.303.4. Discussion3.4.3 ConclusionsThe patterns observed in the experiments described here demonstrate theimportance of the constraints of genetics and the influence of ecology on theprocess of adaptation. We have shown evidence that the diversification ofE. coli in a glucose-acetate batch culture environment depends both on theavailability of mutations as well as the selective pressure determined by theresident population. Since both FS and SS individuals can invade popula-tions of the other type, the asymmetry in likelihood of diversification showsthat certain mutation effects are more likely than others. Furthermore, asingle point mutation that does not affect phenotype (like arcA? in theFS background) can strongly influence diversification rates, suggesting thatepistasis also plays an important role. Diversification in this system is alsoconstrained ecologically because the direction and strength of selection isdetermined by the composition of the population: the fitness of novel typesdepends on the resident types. Both ecological and genetic constraints playan important role, and models of adaptation that consider only one typeare insufficient for predicting evolutionary trajectories. While the effect ofecology can be represented by adaptive dynamics models, predicting theevolutionary outcomes in real-world systems requires a careful integrationof ecology and genetics.The effect of the interaction between ecology and genetics on evolution-ary outcomes shown here highlights the importance of considering both ofthese factors simultaneously. Even though all biologists are familiar withhow common and complicating gene interactions can be, we often conceptu-alize evolution in models as being a linear accumulation of mutations withindependent effects. To simplify the process, we imagine that mutationsarise and are ?tested? in the environment in which they arise, and thenare either removed from the population or increase to fixation. As shownhere, a mutation could change the selective environment as it changes infrequency in a population, creating a feedback loop between changes in al-lele frequencies (evolution) and the environment (ecology). Because theparticular mutation studied here (arcA) has such a clear affect on resourcemetabolism, and therefore on the selective environment, this connection isless obscure than in many well-studied systems of adaptation. Due to thismechanistic link between ecology and evolution, this study provides a clearexample of the eco-evolutionary feedback loops that may become increas-ingly important in understanding evolutionary biology.31Bibliography[1] Marta Barluenga, Kai N Sto?lting, Walter Salzburger, Moritz Muschick,and Axel Meyer. Sympatric speciation in Nicaraguan crater lake cichlidfish. Nature, 439(7077):719?723, February 2006.[2] Jeffrey E Barrick and Richard E Lenski. Genome dynamics duringexperimental evolution. Nature Reviews Genetics, October 2013.[3] Jeffrey E Barrick, Dong Su Yu, Sung Ho Yoon, Haeyoung Jeong,Tae Kwang Oh, Dominique Schneider, Richard E Lenski, and Jihyun FKim. Genome evolution and adaptation in a long-term experiment withEscherichia coli. Nature, 461(7268):1243?1247, October 2009.[4] Daniel I. Bolnick and Benjamin M. Fitzpatrick. Sympatric speciation:Models and empirical evidence. Annual Review of Ecology, Evolution,and Systematics, 38(1):459?487, 2007.[5] Angus Buckling, Matthew A Wills, and Nick Colegrave. Adaptationlimits diversification of experimental bacterial populations. Science,302(5653):2107?2109, December 2003.[6] Lin Chao and Edward C Cox. Competition between high and lowmutating strains of Escherichia coli. Evolution, 37(1):125?134, 1983.[7] H H Chou, H C Chiu, N F Delaney, D Segre, and C J Marx. DiminishingReturns Epistasis Among Beneficial Mutations Decelerates Adaptation.Science, 332(6034):1190?1192, June 2011.[8] Tom M Conrad, Nathan E Lewis, and Bernhard ? Palsson. Micro-bial laboratory evolution in the era of genome-scale science. MolecularSystems Biology, 7:509, 2011.[9] Tim F Cooper, Susanna K Remold, Richard E Lenski, and DominiqueSchneider. Expression profiles reveal parallel evolution of epistatic inter-actions involving the CRP regulon in Escherichia coli. PLoS Genetics,4(2):e35, February 2008.32Bibliography[10] Jerry A Coyne, H Allen Orr, et al. Speciation. Sinauer AssociatesSunderland, MA, 2004.[11] Jeremy R Dettman, Nicolas Rodrigue, Anita H Melnyk, Alex Wong,Susan F Bailey, and Rees Kassen. Evolutionary insight from whole-genome sequencing of experimentally evolved microbes. Molecular Ecol-ogy, 21(9):2058?2077, May 2012.[12] U Dieckmann and M Doebeli. On the origin of species by sympatricspeciation. Nature, 400(6742):354?357, July 1999.[13] Theodosius Dobzhansky and Theodosius Grigorievich Dobzhansky. Ge-netics of the evolutionary process, volume 139. Columbia UniversityPress, 1970.[14] M Doebeli and U Dieckmann. Evolutionary branching and sympatricspeciation caused by different types of ecological interactions. TheAmerican Naturalist, 156(S4):S77?S101, 2000.[15] Santiago F Elena and Richard E Lenski. Microbial genetics: Evolutionexperiments with microorganisms: the dynamics and genetic bases ofadaptation. Nature Reviews Genetics, 4(6):457?469, June 2003.[16] J L Feder, C A Chilcote, and G L Bush. Genetic differentiation betweensympatric host races of the apple maggot fly Rhagoletis pomonella.Nature, 336(6194):61?64, 1988.[17] Anders Folkesson, Lars Jelsbak, Lei Yang, Helle Krogh Johansen, OanaCiofu, Niels H?iby, and S?ren Molin. Adaptation of pseudomonas aerug-inosa to the cystic fibrosis airway: an evolutionary perspective. NatureReviews Microbiology, 10:841?851, 2012.[18] Maren L Friesen, Gerda Saxer, Michael Travisano, and Michael Doe-beli. Experimental evidence for sympatric ecological diversificationdue to frequency-dependent competition in Escherichia coli. Evolution,58(2):245?260, February 2004.[19] T Garland and Rose M R, editors. Experimental Evolution. Universityof California Press, 2009.[20] Sergey Gavrilets. Fitness landscapes and the origin of species (MPB-41). Princeton University Press, 2004.33Bibliography[21] S.A.H. Geritz, E. Kisdi, G. Mesze NA, and J.A.J. Metz. Evolution-arily singular strategies and the adaptive growth and branching of theevolutionary tree. Evolutionary ecology, 12(1):35?57, 1998.[22] Aleeza C Gerstein and Sarah P Otto. Cryptic Fitness Advantage:Diploids Invade Haploid Populations Despite Lacking Any ApparentAdvantage as Measured by Standard Fitness Assays. PLoS ONE,6(12):e26599, December 2011.[23] Mel Greaves and Carlo C Maley. Clonal evolution in cancer. Nature,481(7381):306?313, 2012.[24] Daniel L Halligan and Peter D Keightley. Spontaneous Mutation Accu-mulation Studies in Evolutionary Genetics. Annual Review of Ecology,Evolution, and Systematics, 40(1):151?172, December 2009.[25] Matthew D Herron and Michael Doebeli. Parallel evolutionary dy-namics of adaptive diversification in Escherichia coli. PLoS biology,11(2):e1001490, 2013.[26] Katy C Kao, Linh M Tran, and James C Liao. A global regulatory roleof gluconeogenic genes in Escherichia coli revealed by transcriptomenetwork analysis. The Journal of biological chemistry, 280(43):36079?36087, October 2005.[27] Tadeusz J Kawecki, Richard E Lenski, Dieter Ebert, Brian Hollis,Isabelle Olivieri, and Michael C Whitlock. Experimental evolution.Trends in ecology & evolution, 27(10):547?560, October 2012.[28] Aisha I Khan, Duy M Dinh, Dominique Schneider, Richard E Lenski,and Tim F Cooper. Negative epistasis between beneficial mutations inan evolving bacterial population. Science, 332(6034):1193?1196, June2011.[29] T Kingston and S J Rossiter. Harmonic-hopping in Wallacea?s bats.Nature, 429(6992):654?657, 2004.[30] M Le Gac, M D Brazas, M Bertrand, J G Tyerman, C C Spencer,R E W Hancock, and M Doebeli. Metabolic Changes Associated WithAdaptive Diversification in Escherichia coli. Genetics, 178(2):1049?1060, February 2008.34Bibliography[31] Mickael Le Gac and Michael Doebeli. Epistasis and frequency depen-dence influence the fitness of an adaptive mutation in a diversifyinglineage. Molecular Ecology, 19(12):2430?2438, June 2010.[32] Richard E Lenski and Michael Travisano. Dynamics of adaptation anddiversification: a 10,000-generation experiment with bacterial popula-tions. Proceedings of the National Academy of Sciences, 91(15):6808?6814, 1994.[33] J B Losos. Contingency and Determinism in Replicated Adaptive Ra-diations of Island Lizards. Science, 279(5359):2115?2118, March 1998.[34] Michael Lynch, Bruce Walsh, et al. Genetics and analysis of quantita-tive traits. Sinauer Sunderland, 1998.[35] Guillaume Martin, Santiago F Elena, and Thomas Lenormand. Distri-butions of epistasis in microbes fit predictions from a fitness landscapemodel. Nature Genetics, 39(4):555?560, April 2007.[36] Ernst Mayr et al. Animal species and evolution. Animal species andtheir evolution., 1963.[37] Anita H Melnyk and Rees Kassen. Adaptive landscapes in evolvingpopulations of Pseudomonas fluorescens. Evolution, 65(11):3048?3059,November 2011.[38] Johan AJ Metz, Stefan AH Geritz, Ge?za Mesze?na, Frans JA Jacobs,and JS Van Heerwaarden. Adaptive dynamics, a geometrical study ofthe consequences of nearly faithful reproduction. Stochastic and spatialstructures of dynamical systems, 45:183?231, 1996.[39] J R Meyer, D T Dobias, J S Weitz, J E Barrick, R T Quick, andR E Lenski. Repeatability and Contingency in the Evolution of a KeyInnovation in Phage Lambda. Science, 335(6067):428?432, January2012.[40] F Pelletier, D Garant, and A P Hendry. Eco-evolutionary dynamics.Philosophical Transactions of the Royal Society B: Biological Sciences,364(1523):1483?1489, May 2009.[41] Patrick C Phillips. Epistasis ? the essential role of gene interactionsin the structure and evolution of genetic systems. Nature Reviews Ge-netics, 9(11):855?867, November 2008.35Bibliography[42] R Development Core Team. R: A Language and Environment for Sta-tistical Computing. R Foundation for Statistical Computing, Vienna,Austria, 2011. ISBN 3-900051-07-0.[43] Fisher RA. The genetical theory of natural selection, 1930.[44] P B Rainey and M Travisano. Adaptive radiation in a heterogeneousenvironment. Nature, 394(6688):69?72, July 1998.[45] Susanna K Remold and Richard E Lenski. Pervasive joint influence ofepistasis and plasticity on mutational effects in Escherichia coli. NatureGenetics, 36(4):423?426, April 2004.[46] Daniel E Rozen, Dominique Schneider, and Richard E Lenski. Long-Term Experimental Evolution in Escherichia coli. XIII. PhylogeneticHistory of a Balanced Polymorphism. Journal of Molecular Evolution,61(2):171?180, June 2005.[47] Gerda Saxer, Michael Doebeli, and Michael Travisano. The Repeatabil-ity of Adaptive Radiation During Long-Term Experimental Evolutionof Escherichia coli in a Multiple Nutrient Environment. PLoS ONE,5(12):e14184, December 2010.[48] Dolph Schluter, Elizabeth A Clifford, Maria Nemethy, and Jeffrey SMcKinnon. Parallel evolution and inheritance of quantitative traits.The American Naturalist, 163(6):809?822, June 2004.[49] T W Schoener. The Newest Synthesis: Understanding the Interplayof Evolutionary and Ecological Dynamics. Science, 331(6016):426?429,January 2011.[50] Daniel Segre, Alexander DeLuna, George M Church, and Roy Kishony.Modular epistasis in yeast metabolism. Nature genetics, 37(1):77?83,2004.[51] Pedro Simo?es, Josiane Santos, Ine?s Fragata, Laurence D Mueller,Michael R Rose, and Margarida Matos. How repeatable is adaptiveevolution? The role of geographical origin and founder effects in labo-ratory adaptation. Evolution, 62(8):1817?1829, August 2008.[52] P D Sniegowski, P J Gerrish, and R E Lenski. Evolution of high muta-tion rates in experimental populations of E. coli. Nature, 387(6634):703?705, June 1997.36[53] Christine C Spencer, Melanie Bertrand, Michael Travisano, andMichael Doebeli. Adaptive diversification in genes that regulate re-source use in Escherichia coli. PLoS Genetics, 3(1):e15, January 2007.[54] Christine C Spencer, Jabus Tyerman, Melanie Bertrand, and MichaelDoebeli. Adaptation increases the likelihood of diversification in anexperimental bacterial lineage. Proceedings of the National Academy ofSciences, 105(5):1585?1589, February 2008.[55] Alejandro Toro-Roman, Timothy R Mack, and Ann M Stock. Structuralanalysis and solution studies of the activated regulatory domain of theresponse regulator ArcA: a symmetric dimer mediated by the alpha4-beta5-alpha5 face. Journal of molecular biology, 349(1):11?26, May2005.[56] Jabus Tyerman and Michael Doebeli. Variation in the propensity to di-versify in experimental populations of Escherichia coli: consequences foradaptive radiation. Evolutionary Ecology Research, 12:803?820, 2010.[57] Jabus Tyerman, Naomi Havard, Gerda Saxer, Michael Travisano, andMichael Doebeli. Unparallel diversification in bacterial microcosms.Proceedings Biological sciences / The Royal Society, 272(1570):1393?1398, July 2005.[58] Jabus G Tyerman, Melanie Bertrand, Christine C Spencer, and MichaelDoebeli. Experimental demonstration of ecological character displace-ment. BMC Evolutionary Biology, 8(1):34, 2008.[59] Sara Via. Sympatric speciation in animals: the ugly duckling grow up.Trends in Ecology & Evolution, 16(7):381?390, 2001.[60] Daniel M Weinreich, Nigel F Delaney, Mark A Depristo, and Daniel LHartl. Darwinian evolution can follow only very few mutational pathsto fitter proteins. Science, 312(5770):111?114, April 2006.[61] R J Woods, J E Barrick, T F Cooper, U Shrestha, M R Kauth, and R ELenski. Second-Order Selection for Evolvability in a Large Escherichiacoli Population. Science, 331(6023):1433?1436, March 2011.[62] S Wright. The roles of mutation, inbreeding, crossbreeding and selec-tion in evolution. Proceedings of the Sixth International Congress ofGenetics, 1(6):356?366, 1932.37Appendix ASupplementary Figures0 5 10 15 20 25 30 350. (hours)log(optical density)Figure A.1: Growth curves of clones from gen200-FSarcA?-rep2. Opticaldensity of individual clones measured every 15 minutes over a 36 hour period.38Appendix A. Supplementary Figures0 5 10 15 20 25 30 350. (hours)log(optical density)Figure A.2: Growth curves of clones from gen200-FSarcA+-rep2. Opticaldensity of individual clones measured every 15 minutes over a 36 hour period.39Appendix A. Supplementary Figures0 5 10 15 20 25 30 350. (hours)log(optical density)Figure A.3: Growth curves of clones from gen800-SSarcA+-rep4. Opticaldensity of individual clones measured every 15 minutes over a 36 hour period.40


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items