Download Evolutionary Theory and Experiments With Microorganisms

Evolutionary Theory and Experiments With Microorganisms$
MJ Wiser and RE Lenski, Michigan State University, East Lansing, MI, United States
r 2017 Elsevier Inc. All rights reserved.
Abbreviations
GASP
MOI
Multiplicity of infection
Growth Advantage in Stationary Phase
Glossary
Adaptation A feature of an organism that enhances its
reproductive success and that evolved by natural selection.
Epistasis An interaction among genes, such that the effect
of a mutation depends on the genetic background in which
it occurs.
Evolution Change in the genetic properties of populations
and species over generations, which requires the origin of
variation (by mutation or mixis) as well as the subsequent
spread or extinction of variants (by natural selection and
genetic drift).
Fitness Average reproductive success of a genotype in a
particular environment, usually expressed relative to
another, standard genotype.
Genetic drift Changes in gene frequency caused by the
random sampling of genes during transmission across
generations (rather than by natural selection).
Mixis Production of a new genotype by recombination
of genes from two sources.
Natural selection Changes in gene frequency caused
by specific detrimental or beneficial effects of those
genes.
Population Group of individuals belonging to the same
species and living in close proximity, so that individuals
may potentially recombine their genes, compete for limiting
resources, or otherwise interact.
Defining Statement
Evolution in action can be studied by experiments in the laboratory using bacteria and other microorganisms with suitably rapid
generation. These experiments have confirmed the main principles of modern evolutionary theory, while also providing new
insights into the genetics, physiology, and ecology of microorganisms.
Review of Evolutionary Theory
Evolutionary theory seeks to explain observable patterns of biological diversity in terms of a few fundamental evolutionary
processes. These processes are presumed not only to have operated in the past, but also to continue to operate today. Thus, they
can be studied experimentally in the laboratory. Before discussing a broad range of experiments that have used microorganisms to
examine evolutionary processes, the major elements of evolutionary theory will be reviewed.
Evolutionary Patterns
The three most conspicuous products of organic evolution are (1) the wealth of genetic variation that exists within almost every
species; (2) the divergence of populations and species from one another and from their common ancestors; and (3) the manifest
adaptation, or fit, of organisms to the environments in which they live.
Genetic variation
The existence of extensive genetic variation within species has been demonstrated by a variety of means. Variation in certain traits,
such as seed shape in pea plants and blood type in humans, can be shown to have a genetic basis by careful examination of
pedigrees. For many other traits, such as milk production in cows or body weight in humans, quantitative genetic analyses are
required to partition the phenotypic variation that is due to genetic versus environmental influences. Biochemical and molecular
techniques have also revealed extensive variation in DNA sequences and the proteins they encode.
☆
Change History: August 2016. M.J. Wiser and R.E. Lenski updated Sections “Glossary,” “Genetic and Physiological Bases of Fitness,” “Evolution of New
Metabolic Functions,” and “Evolution of New Genetic Systems.”
Reference Module in Life Sciences
doi:10.1016/B978-0-12-809633-8.13042-0
1
2
Evolutionary Theory and Experiments With Microorganisms
Divergence and speciation
All biological species differ from one another in some respects. It is generally possible to arrange species hierarchically, depending
on the extent and nature of their similarities and differences. This hierarchy is reflected in the classification scheme of Linnaeus
(species, genus, family, and so on). This hierarchical arrangement also suggests a sort of “tree of life” in which the degree of
taxonomic relatedness between any two species reflects descent with modification from some common ancestor in the more or less
distant past. Investigating the origins of groups and their relationships requires an historical approach, which is not amenable to
direct experimentation. Even so, historically based hypotheses can be tested by phylogenetic and comparative methods, which
utilize data on the distribution of traits across various groups and environments, sometimes supplemented with information from
the fossil record.
The extent of evolutionary divergence that is necessary for two groups of organisms to be regarded as distinct species is
embodied in the biological species concept, according to which “species are groups of actually or potentially interbreeding
populations, which are reproductively isolated from other such groups” (E. Mayr, in 1942). Speciation thus refers to the historical
process by which groups of organisms become so different from one another that they no longer can interbreed. However, many
organisms (including most microorganisms) reproduce primarily or exclusively asexually, and the preceding definition is not
applicable. For such organisms, the extent of evolutionary divergence that corresponds to distinct species is somewhat arbitrary
and often more a matter of convenience than of scientific principle.
Adaptation
Many phenotypic features of organisms often exhibit an exquisite match to their environments. For example, the bacteria that live
in hot springs have special physiological and biochemical properties that allow them to survive and grow at very high temperatures, which would kill most other bacteria; often these thermophiles cannot grow at all under the lower temperatures where
most other bacteria thrive. However, organisms are also generally not perfectly adapted to their environments. Evidence for the
imperfection of organisms can be seen when species go extinct, usually as a consequence of some change in the environment to
which they cannot quickly adapt.
Evolutionary Processes
Biological evolution occurs whenever the genetic composition of a population or species changes over a period of generations.
Four basic processes contribute to such change: mutation, mixis, natural selection, and genetic drift. Selection and drift will not
produce evolutionary changes, however, unless there exists genetic variation among individual organisms.
Sources of genetic variation
Genetic variation among individuals is generated by two distinct processes, mutation and mixis. In terms of evolutionary theory,
these processes are usually distinguished as follows: mutation refers to a change at a single gene locus from one allelic state to
another (eg, abcd- Abcd, where each letter indicates a locus), whereas mixis refers to the production of some new multilocus
genotype by the recombination of two different genotypes (eg, abcd þ BCD- aBcD).
There are many different types of mutations, including point mutations, rearrangements, and transposition of mobile elements
from one site in the genome to another. Some mutations cause major changes in an organism’s phenotype; for example, a
bacterium may become resistant to attack by a virus (bacteriophage) as a result of a mutation that alters a receptor on the cell
surface. Other mutations have little or even no effect on an organism’s phenotype: many point mutations have absolutely no effect
on amino acid sequence (and hence protein structure and function) because of the redundancy in the genetic code.
Any number of factors may affect mutation rates, including both environmental agents (eg, UV irradiation) and the organism’s
own genetic constitution (eg, defective DNA repair genes). Evolutionary theory makes no assumptions about the rates of mutations or their biophysical bases, with one exception: mutations are assumed to occur randomly, that is, irrespective of their
beneficial or harmful effects on the organism.
Recombination among genomes can occur by a number of different mechanisms. The most familiar one is eukaryotic sex,
which occurs by meiosis and fertilization. Many eukaryotic microorganisms, including fungi and protozoa, engage in sexual mixis.
Bacteria generally reproduce asexually, but may undergo mixis via conjugation (plasmid-mediated), transduction (virus-mediated), or transformation. Even viruses may recombine when two or more coinfect a single cell.
Unlike mutation, these various mechanisms do not necessarily produce organisms with new genes; instead, they may produce
organisms that possess new combinations of genes. This can have very important evolutionary consequences. In the absence of
mixis, two or more beneficial mutations can be incorporated into an evolving population only if they occur sequentially in a single
lineage. Mixis allows beneficial mutations that occur in separate lineages to be combined and thereby incorporated simultaneously
by an evolving population. Thus, mixis may accelerate the rate of adaptive evolution by bringing together favorable combinations
of alleles.
Natural selection
One of the most conspicuous features of biological evolution is the evident “fit” (adaptation) of organisms to the environments in
which they live. For centuries, this match between organism and environment was taken as evidence for the design of a Creator.
Evolutionary Theory and Experiments With Microorganisms
3
But in 1859, Charles Darwin published “The Origin of Species”, in which he set forth the principle of adaptation by natural
selection. This principle follows logically from three simple premises. First, variation among individuals exists for many phenotypic traits. Second, these phenotypic traits influence survival and reproductive success. Third, phenotypic variation in those
characters that affect survival and reproductive success is heritable, at least in part. (Many phenotypic traits are subject to both
genetic and environmental influences.) Hence, individuals in later generations will tend to be better adapted to their environment
than were individuals in earlier generations, provided that there is heritable phenotypic variation and the environment has not
changed too much in the intervening time. (Environments do sometimes change, of course, and when this happens a population
or species may go extinct if it cannot adapt to these changes.)
Darwin himself did not know about the material basis of heredity (DNA and chromosomes), nor did he even understand the
precise causes of heritable variation among individuals (mutation and mixis). What he clearly understood, however, was that this
heritable variation did exist and its causes could be logically separated from its consequences for the reproductive success of
individuals and the resulting adaptation of species to their environments.
Darwin’s theories were influenced, in part, by his observations on the practices of breeders of domesticated animals and plants.
These practices are now commonly referred to as artificial selection. It is useful to distinguish between artificial and natural
selection, and to relate this distinction to experimental evolution in the laboratory. Under artificial selection, organisms are chosen
by a breeder, who allows some but not all individuals to survive and reproduce. Individuals are thus selected on the basis of
particular traits that are deemed desirable to the breeder. By contrast, under natural selection, no one consciously chooses which
individuals within a population will survive and reproduce and which will not. Instead, the match between organismal traits and
environmental factors determines whether or not any given individual will survive and reproduce.
At first glance, one might regard laboratory studies as examples of artificial selection. Such usage, however, would not reflect the
critical distinction between artificial and natural selection, that is, whether a breeder or the environment determines which
individuals survive and reproduce. In experimental evolution, an investigator typically manipulates environmental factors, such as
temperature and resource concentration, but he or she does not directly choose which individuals within an experimental
population will survive and reproduce. Instead, natural selection in the laboratory, like natural selection in the wild, depends on
the match between organismal traits and environmental factors.
Genetic drift
The frequency of genes within populations, and hence also the distribution of phenotypic traits, may change not only as a
consequence of natural selection, but also as a consequence of the random sampling of genes during transmission across
generations. This random sampling is called genetic drift. In practice, it can be difficult to distinguish between natural selection
and genetic drift, although statistical methods have been developed that may allow one to distinguish between these forces by
comparing DNA sequences among related strains. Alternatively, by using microorganisms to study evolution experimentally, it is
possible to compare the survival and reproductive success of different genotypes that are placed in direct competition with one
another. With proper replication of such experiments, one can distinguish systematic differences in survival and reproductive
success from chance deviations due to drift.
Experimental Tests of Fundamental Principles
Two key principles of evolutionary theory are the randomness of mutation and adaptation by natural selection. According to the
former, mutations occur irrespective of any beneficial or harmful effects they have on the individual organism. According to the
latter, organisms in later generations tend to be better adapted to their environment than were those in earlier generations,
provided that the necessary genetic variation exists and the environment itself does not change.
Random Mutation
For many years, it was known that bacteria could adapt to various environmental challenges. For example, the introduction of
bacteriophage into a population of susceptible bacteria often caused the bacterial population to become resistant to viral infection.
It was unclear, however, whether mutations responsible for bacterial adaptation were caused directly by exposure to the selective
agent, or whether this adaptation was the result of random mutation and subsequent natural selection. Two elegant experiments
were performed in the 1940s and 1950s, which demonstrated that mutations existed prior to exposure to the selective agent, so
that these mutations could not logically have been caused by that exposure.
Fluctuation test
The first of these experiments was published by Salvador Luria and Max Delbrück in 1943, and it relies on subtle mathematical
reasoning. Imagine a set of bacterial populations, each of which grows from a single cell to some large number of cells (N); the
founding cells are identical in all of the populations. If exposure to the selective agent causes a bacterial cell to mutate with some
low probability (p), then the number of mutants in a population is expected to be, on average, pN. Although this probability is the
same for each of the replicate populations, the exact number of mutants in each population may vary somewhat due to chance
4
Evolutionary Theory and Experiments With Microorganisms
(just as the number of heads and tails in 20 flips of a fair coin will not always equal exactly 10). Mathematical theory shows that
the expected variance in the number of mutants among the set of replicate populations is equal to the average number of mutants
under this hypothesis.
Now imagine this same set of populations, but assume that mutations occur randomly, that is, independent of exposure to the
selective agent. During each cell generation, there is a certain probability that one of the daughter cells is a mutant. A mutant cell’s
progeny are also mutants, and so on. According to mathematical theory for this hypothesis, the variance in the number of mutants
among the replicate populations should be much greater than the average number of mutants. This large variance comes about
because mutations will, by chance, occur earlier in some replicate populations than in others, and each of the early mutations will
leave many descendant mutants as a consequence of the subsequent population growth.
Luria and Delbrück designed experiments that allowed them to measure both the average and the variance of the number of
mutants in a set of populations of the bacterium Escherichia coli. The observed variance was much greater than expected if exposure
to the selective agent had caused the mutations. Hence, their results supported the hypothesis of random mutation.
Replica-plating experiment
Joshua and Esther Lederberg devised a more direct demonstration of the random origin of bacterial mutations, which they
published in 1952. In their experiment, thousands of cells are spread on an agar plate that does not contain the selective agent;
each cell grows until it makes a small colony, and the many colonies together form a lawn of bacteria (master plate). By making an
impression of this plate using a pad of velvet, cells from all of the colonies are then transferred to several other agar plates (replica
plates) that contain the selective agent, which prevents the growth of colonies except from those cells with the appropriate
mutation. If mutations are caused by exposure to the selective agent, then there should be no tendency for mutant colonies found
on the replica plates to be derived from the same subset of colonies on the master plate. However, if mutations occur during the
growth of the colony on the master plate (ie, prior to exposure to the selective agent), then those master colonies that give rise to
mutant colonies on one replica plate should also give rise to mutant colonies on the other replica plates. Indeed, Lederberg and
Lederberg observed that master colonies giving rise to mutants on one replica plate gave rise to mutants on the other replica plates.
Moreover, they could isolate resistant mutants from those master colonies, without the cells having ever been exposed to the
selective agent. This experiment thus demonstrates that the mutations had occurred randomly during the growth of the colony on
the master plate.
Adaptation by Natural Selection
In addition to demonstrating the random occurrence of mutations, the fluctuation test and the replica plating experiment both
demonstrate adaptation by natural selection. Two other types of experiments also demonstrate adaptation by natural selection.
One type is very complicated and involves monitoring the dynamics of mutation accumulation at one genetic locus, which is not
under selection, in order to study indirectly what is happening at some other loci, which are under selection.
The other type of experiment that demonstrates adaptation by natural selection involves direct estimation of an evolving
population’s fitness relative to its ancestor, and it is conceptually quite simple. A population is founded by an ancestral clone,
which is also stored in a dormant state (usually at very low temperature). The population is then propagated in a particular
environment, and one or more samples are obtained after many generations have elapsed. These derived organisms are placed in
direct competition with the ancestral clone under the same defined environmental conditions (after both types have acclimated
physiologically to these conditions). If the derived organisms increase their number relative to the ancestral clone, in a systematic
and reproducible fashion, then the evolving population has evidently become better adapted to the environment as the result of
mutation and natural selection.
To distinguish the derived and ancestral types from one another in a competition experiment, it is usually necessary to
introduce a genetic marker that can be scored into one of them. This genetic manipulation necessitates an appropriate control
experiment to estimate the effect of the genetic marker on fitness.
Genetic and Physiological Bases of Fitness
The fact that one strain may be more fit than another in a particular environment usually says little or nothing about the causes of
that difference. It is interesting to know why one strain is more fit than another in terms of their genotypes and their physiological
properties. By using both classical and molecular genetic methods, one can construct genotypes of interest and then determine the
effects of their differences on physiological performance and fitness.
Effects of Single and Multiple Mutations
A study by Santiago Elena and Richard Lenski examined systematically the fitness effects of a large set of random mutations.
Using transposon mutagenesis, they made 225 genotypes of E. coli that each carried one, two, or three insertion mutations.
Evolutionary Theory and Experiments With Microorganisms
5
Each genotype was then placed in competition with the unmutated progenitor strain in order to measure the fitness of the mutant
relative to the progenitor. Single insertion mutations reduced fitness by about 2%, on average, and there was tremendous
variability in the mutational effects, which ranged from approximately neutral to effectively lethal under the conditions where the
competitions were performed. The relationship between average fitness and mutation number was approximately log-linear. That
is, subsequent mutations were neither more nor less harmful, on average, than was the first mutation. However, many pairs of
mutations interacted strongly with one another, so that their combined effect on fitness was different from what was expected
given their separate effects. (The overall relationship between average fitness and mutation number was approximately log-linear
because some interactions were synergistic whereas others were antagonistic.) These data imply that a full understanding of the
genetic basis of any organism’s functional capacity cannot depend entirely on the step-by-step analysis of individual genes and
pathways; instead, a more integrative approach is necessary.
Another approach is to investigate both the actual and possible alternative evolutionary paths that lead from an ancestor to a
derived state that has accumulated multiple mutations. A study by Daniel Weinreich and colleagues examined a set of five point
mutations in E. coli that, in combination, increase resistance to the b-lactam cefotaxime by a factor of B100,000. With five
mutations, there are 120 possible paths to obtain the full complement sequentially, but only 18 of them result in strictly increasing
levels of resistance to this antibiotic. The large fraction of mutational paths along which antibiotic resistance either decreased or
remained the same across one or more steps indicates pervasive epistasis in this system. In contrast, a study by Aisha Khan and
colleagues examined the first five mutations that arose in a population of E. coli as it evolved in a nutrient-limited environment; the
combination of these mutations increased fitness by B30% relative to the ancestor. In this study, 86 of the 120 paths from the
ancestor to the genotype with all five mutations were monotonically increasing in fitness, indicating that epistasis was not a
substantial barrier to adaptation. Which of these two situations is more typical remains an open question in the field.
The study by Khan et al. also provided evidence of diminishing returns as populations adapt to a constant environment. That is,
when they moved beneficial mutations into different genetic backgrounds, they found that the resulting fitness gains were smaller
on the more-fit backgrounds than on the less-fit backgrounds. A similar tendency has been seen in other experiments, including
one in which Hsin-Hung Chou and colleagues engineered a strain of Methylobacterium extorquens to be deficient in growth on
methanol and then allowed it to evolve. They then identified four mutations that led to almost a doubling of fitness, constructed
all of the possible combinations of those mutations, and measured the fitnesses of the constructed genotypes. Here, too, the fitness
gains were usually smaller on the more-fit backgrounds, indicating a pattern of diminishing returns.
Beyond these detailed gene-by-gene analyses of adaptive trajectories, other studies have used the dynamics of adaptation in
evolving populations to make inferences about the tempo and sustainability of evolution. In a recent study, Michael Wiser, Noah
Ribeck, and Richard Lenski examined fitness changes across a set of 12 populations of E. coli as they evolved in and adapted to a
nutrient-limited laboratory environment for 50,000 generations. Consistent with the earlier work by Kahn and colleagues, this
study supported the role of diminishing-returns epistasis in adaptive evolution. However, the diminishing returns were not
sufficient to stop fitness from continuing to increase. Instead, the dynamics of adaptation in this system were well described by a
mathematical function called a power law, according to which improvements scale with the logarithm of time. As a consequence,
fitness can increase indefinitely, but an ever-declining rate of improvement.
Fitness Effects Due to Possession of Unused Functions
A number of studies have used well-characterized bacterial genotypes to examine the effects on fitness caused by the carriage and
expression of superfluous gene functions. These studies have measured the relative fitnesses of (1) bacteria with constitutive (high
level) and repressed (low level) expression of enzymes for catabolism of carbon sources in media where those resources are not
available; (2) prototrophic bacteria (which produce an amino acid or other required nutrient) and auxotrophic mutants (which
cannot produce that nutrient) in media where the required nutrients are supplied; (3) phage-sensitive bacteria and resistant
mutants when the phage are absent; and (4) antibiotic-sensitive bacteria and resistant genotypes in media that contain no
antibiotics.
These studies have often, but not always, demonstrated substantial fitness disadvantages due to possession of unnecessary gene
functions. In some cases where such disadvantages have been detected, they are mu greater than can be explained by the energetic
costs of synthesizing the extra proteins and other metabolites. For example, a study by Daniel Dykhuizen found that the fitness
disadvantage associated with synthesis of the amino acid tryptophan, when it was supplied in the medium, was much greater than
could be explained on the basis of energetic costs alone. Evidently, the expression of superfluous functions can sometimes have
strong indirect effects, which presumably arise through the disruption of other physiological processes.
The idea that microorganisms may have reduced fitness owing to possession of unnecessary functions has two important
practical implications. First, in many bioengineering applications, microorganisms are constructed that constitutively express high
levels of some compound (eg, a pharmaceutical) that can be harvested for its commercial value. However, the microorganisms
themselves do not benefit from producing that compound, and so any mutant that no longer produces the compound may have a
selective advantage. Such a mutation would thus spread through the population and thereby reduce the efficiency of the production process. Second, the spread of antibiotic-resistant bacteria has become a serious concern in public health. It has been
proposed that the prudent use of antibiotics, including even the elimination of their use in certain environments (eg, animal
feeds), might favor antibiotic-sensitive bacteria, thereby restoring the efficacy of antibiotics. This proposal rests, in part, on the
presumption that resistant bacteria are less fit than their sensitive counterparts, in the absence of antibiotic, due to their possession
6
Evolutionary Theory and Experiments With Microorganisms
of the superfluous resistance function. This tradeoff often appears to be the case, but sometimes antibiotic-resistant bacteria evolve
solutions that reduce or even eliminate the cost of resistance, thus complicating efforts to contain their spread.
Effects Due to Variation in Essential Metabolic Activities
It is clear that the expression of unnecessary metabolic functions is often disadvantageous to a microorganism. An equally
important issue concerns the relationship between fitness and the level of expression of functions that are required for growth in a
particular environment. This latter issue is more difficult to address experimentally, because it demands careful analyses of subtle
differences between strains in their biochemical activities. Daniel Dykhuizen, Antony Dean, and Daniel Hartl performed a
pioneering study to examine the relationships between genotype, biochemical activities in a required metabolic pathway, and
fitness. Their study examined growth on lactose by genotypes of E. coli that varied in levels of expression of the permease used for
active uptake of lactose and the b-galactosidase required for hydrolysis of the lactose. (The genotypes were otherwise essentially
identical.) Given that both enzymes are necessary for growth on lactose, how do the activities at each step affect the net flux
through this metabolic pathway? And how does net flux affect fitness?
Metabolic control theory consists of mathematical models that describe the dynamics across multiple steps in a biochemical
pathway. Using this theory, Dykhuizen and colleagues predicted how the net flux through this pathway would depend on the
activities of the permease and p'-galactosidase enzymes, and they measured these activities for several different genotypes using
biochemical methods. They then predicted that the relative fitness of any two genotypes would be directly proportional to their
relative fluxes if lactose was provided as the sole energy source. In order to test the theory and its predictions, they estimated the
relative fitnesses of the various genotypes in a medium in which lactose was the sole source of energy for growth. The observed
fitnesses were very close to the predicted values.
Dykhuizen and Dean have also successfully extended this mechanistic approach to predicting fitness in competition for
mixtures of lactose and glucose. However, the results to date (for both single and mixed sugars) were obtained with genotypes in
which gene regulation was eliminated in order to simplify the analysis. An important challenge for the future is to include the
complex dynamical effects of gene regulation in the models and experiments.
Effects of Genetic Background
It is obvious that the fitness effects caused by particular genetic differences strongly depend on the environment. For example, an
antibiotic resistance gene function that is essential for survival and replication of a bacterium in the presence of antibiotic may
hinder growth in an antibiotic-free environment. Similarly, the fitness effect that is due to a particular gene function may often
depend on the genetic background in which that gene is found.
For example, one study found that different alleles at the 6-phosphogluconate dehydrogenase (6-PGD) locus in E. coli had
similar fitnesses in a gluconate-limited medium, provided that these alleles were present in a genetic background that also encoded
an alternative metabolic pathway for 6-phosphogluconate utilization. In a genetic background where this alternative pathway was
defective, however, these alleles had quite variable fitnesses in that same medium.
In another study with E. coli, it was observed that the selective disadvantage associated with bacteriophage resistance mutations
in a virus-free environment was substantially reduced during several hundred generations of experimental evolution. This fitness
improvement resulted from secondary mutations in the genetic background that compensated for the maladaptive side effects of
the resistance mutations, but which did not diminish the expression of resistance.
Other studies, including one by Stephanie Schrag, Veronique Perrot, and Bruce Levin, have demonstrated similar compensatory
evolution among antibiotic-resistant bacteria. When bacteria resistant to an antibiotic first arise, they are typically less fit in the
absence of the antibiotic, thus helping to control their proliferation. Over time, however, the evolving bacteria become very good
competitors in the absence of antibiotic while retaining their resistance to antibiotic. That is, the bacteria find ways to “have their
cake and eat it, too.” Such compensatory evolution unfortunately makes it more difficult to devise strategies to manage the spread
of antibiotic-resistant pathogens.
Speciation and Genetic Exchange
As populations diverge from one another over time, the potential for gene flow between them lessens. The increased barriers to
gene flow between more distantly related organisms reflect both molecular processes and selective factors. At the molecular level,
highly diverged DNA sequences recombine less efficiently than similar sequences. From the standpoint of selection, genes that
evolved in one lineage may not function as well when they are moved into a different lineage, where they must function within a
different physiological context. Therefore, when two lineages diverge and adapt to different environments for a sufficiently long
time, they become separate species. An experiment by Jeremy Dettman and colleagues examined the effects of adaptation to
different environments on the fitness of progeny in the yeast Saccharomyces cerevisiae. The researchers generated hybrid progeny by
sexually crossing yeast strains, using three different types of strain pairings: (1) two strains that had independently evolved in the
same environment; (2) two strains that evolved in different environments; and (3) one evolved strain and its ancestor. The hybrid
progeny of strains that had both evolved in the same environment were more fit than the other types of hybrids, while the least fit
progeny resulted from crosses between strains that evolved in different environments. These results demonstrate that incipient
Evolutionary Theory and Experiments With Microorganisms
7
speciation can occur rapidly as populations adapt to different environments and diverge from one another, so that their hybrid
progeny are not well adapted to the environment of either parent.
Over time, another process, called reinforcement, may occur that further separates and isolates two incipient species. Reinforcement occurs when organisms evolve mating preferences so that they become less likely to mate with individuals that are
ecologically and genetically distinct, thereby avoiding the production of mal-adapted hybrid offspring. In another experiment with
yeast, Jun-Yi Leu and Andrew Murray showed the rapid evolution of mate discrimination. After only a few dozen rounds of
selection for assortative mating, yeast cells had evolved that were five times more likely to mate with their own type than with the
ancestor.
Genomic Analyses of Experimental Evolution
As technologies change, so do the questions that biologists can address. Our understanding of the extent and patterns of genetic
diversity has grown with each new advance. For many years, geneticists had to rely on differences that were visible to the naked
eye – for example, the round and wrinkled seeds, and white and purple flowers, of the pea plant that Gregor Mendel studied
to discover the basic laws of inheritance in plants and animals. Several decades ago, biochemists discovered they could discern
subtle differences in proteins, caused by mutations in the genes that encode them, by examining how the variant proteins moved
through gels that were subject to electrical fields. This approach revolutionized biology by demonstrating tremendous levels
of genetic diversity in almost every species that was investigated, from bacteria to humans. More recently, the ability to sequence
DNA molecules now allows scientists to examine and compare the hereditary information of organisms at its most fundamental
level. It has become possible to sequence genes isolated directly from the environment, allowing ecologists to examine patterns
of diversity even among microbes that have never been directly observed or cultured. It has also become possible to sequence
the entire genome of any organism. With improving technologies and declining costs, it is becoming feasible to pursue
whole-genome sequencing to discover and investigate all of the genetic changes that occur during laboratory-based evolution
experiments.
A fascinating question in evolution – and one that has benefited from the improving genomic technologies – concerns the
reproducibility of evolutionary outcomes. The late paleontologist Stephen Jay Gould imagined the thought experiment of
“replaying life’s tape” to address this question. Of course, it is impossible to rerun evolution for billions of years on the scale of an
entire planet, but with microbes one can perform careful experiments to ask whether replicated populations that start with the
same ancestor and evolve in the same environment will arrive at the same or different solutions to the challenges they face. A longterm study by Richard Lenski and colleagues has monitored the evolution of 12 initially identical populations of E. coli as they
evolved in and adapted to the same laboratory environment for over 60,000 generations. They identified many parallel, or
repeatable, changes in the phenotypes of the evolving lineages, including improved competitiveness in the glucose-based medium
where they evolved, reduced performance on certain other sugars including ribose and maltose, larger cell volumes, altered
supercoiling of their chromosomes, and so on. Based on the tremendous advances in DNA sequencing technology, a recent paper
by Olivier Tenaillon, Jeffrey Barrick, and colleagues analyzed the complete genomes of 264 isolates from the first 50,000 generations across the replicate populations, allowing them to characterize the long-term dynamics of genome evolution. The authors
found that many genes had acquired mutations in several or even all of the evolved lines, even though most genes acquired no
substitutions in any of the lines. These data indicate a high degree of evolutionary reproducibility even at the genetic level.
Holly Wichman, James Bull and colleagues previously documented even more striking evolutionary reproducibility in the
bacteriophage jX174. In this case, the reproducibility, or parallelism, often extended to the nucleotide level. Christopher Herring,
Bernhard Palsson and colleagues have used new methods for rapidly resequencing entire bacterial genomes to identify several
genes that changed repeatedly in short-term experiments in which E. coli evolved in and adapted to a glycerol-based medium.
Gregory Lang and colleagues also used this approach with the yeast S. cerevisiae, finding many cases in which multiple mutations
arise at similar times and compete to become fixed within a given population. They also found a subset of genes that showed
striking parallelism as mutational targets across dozens of replicate populations. Outside the laboratory, these new sequencing
approaches have been used to study evolutionary dynamics in clinical settings, which sometimes provide a so-called “natural
experiment” in which multiple infected hosts are analogous to replicate culture vessels. For example, Tami Lieberman and
colleagues sequenced over a hundred genomes of pathogenic Burkholderia dolorosa that were isolated over many years from the
lungs of patients with cystic fibrosis, finding a number of genes that underwent parallel evolution during the adaptation to
different individual hosts.
Genome-level resequencing was also used by Gregory Velicer and colleagues to discover mutations that arose during a twostage evolution experiment with the bacterium Myxococcus xanthus. This species is fascinating because cells, when they are starving,
form multicellular aggregations. Some of the cells in these aggregations differentiate into stress-tolerant spores that can be
dispersed and germinated in more favorable environments, but the majority of cells die while forming a mound-like structure that
elevates the spores and helps them disperse. In the first stage of their experiment, Velicer and colleagues evolved M. xanthus strains
that “cheated” on their progenitors during aggregation and differentiation. By themselves, these cheaters were poor at aggregating
and producing spores, but when they were mixed with their cooperative progenitors, the cheaters were disproportionately likely to
end up among the surviving spores. In the second stage of this evolution experiment, Velicer and colleagues found one cheaterderived strain that had recovered the ability to cooperate, although its mechanism of cooperation was somewhat different from the
ancestral cooperator. By resequencing the genomes of both the evolved cheater and the restored cooperator, and comparing them
8
Evolutionary Theory and Experiments With Microorganisms
to the ancestral genome, they found 14 mutations that led to the cheater, but only a single mutation – although obviously an
important one – was responsible for the evolved restoration of the cooperative behavior.
Genetic Variation Within Populations
In nature, there exists abundant genetic variation in most species, including microorganisms. Some of this variation exists within
local populations, while other variation may distinguish one population from another. This section describes some of the dynamic
processes that influence genetic variation within populations.
Transient Polymorphisms
A population is polymorphic whenever two or more genotypes are present above some defined frequency (eg, 1%). A
polymorphism arises whenever an advantageous mutation is increasing in frequency relative to the ancestral allele. This
type of polymorphism is called transient, because eventually the favored allele will exclude the ancestral allele by natural
selection.
Selective Neutrality
At the other extreme, some polymorphisms may exist almost indefinitely precisely because the alleles that are involved have
little or no differential effect on fitness. Such selectively neutral alleles are subject only to genetic drift. Daniel Dykhuizen
and Daniel Hartl sought to determine whether some polymorphic loci in natural populations of E. coli might exist because of
selective neutrality, or whether other explanations are needed. To that end, naturally occurring alleles at certain loci were moved
into a common genetic background, and the fitness effects associated with the various alleles were determined. Even when the
bacteria were grown under conditions where growth was directly dependent on the particular enzymes encoded by these loci, there
were often no discernible effects on fitness due to the different alleles. These results therefore support the hypothesis that random
genetic drift is responsible for some of the genetic variation that is present in natural populations.
Frequency-Dependent Selection
In the course of growth and competition in a particular environment, microorganisms modify their environment through the
depletion of resources, the secretion of metabolites, and so on. When this happens, the relative fitness of genotypes may depend
on the frequency with which they are represented in a population, and selection is said to be frequency-dependent. Frequencydependent selection can give rise to several different patterns of genetic variation.
Stable equilibria
Two (or more) genotypes can coexist indefinitely when each has some competitive advantage that disappears as that genotype
becomes more common. In that case, each genotype can invade a population consisting largely of the other genotype but cannot
exclude that other genotype, so that a stable equilibrium results.
Several different ecological interactions can promote a stable equilibrium. For example, an environment may contain two
different resources. If one genotype is better at exploiting one resource and another genotype is superior in competition for the
second resource, then whichever genotype is rarer will tend to have more resource available to it, thereby promoting their
stable coexistence. In some cases, a resource that is essential for one genotype may be produced as a metabolic by-product of
growth by another genotype; such interactions are often called cross-feeding. Stable coexistence of genotypes in one population
can also occur when the environment contains a population of predators (or parasites); predator-mediated coexistence requires
that one of the prey genotypes be better at exploiting the limiting resource while the other prey genotype is more resistant to the
predator. The evolution of two or more stably coexisting bacterial types from a single ancestral type has been demonstrated in
several experiments involving both cross-feeding and predator–prey interactions.
A striking example of the rapid evolution of several stably coexisting genotypes comes from an experiment on Pseudomonas
fluorescens performed by Paul Rainey and Michael Travisano. The experiment started with a single clone that was placed in a static
(unshaken) flask containing a nutritionally rich liquid medium. Within a few days, the bacteria evolved into three distinct
genotypes that could be distinguished by the appearance of their colonies, and these types then coexisted with one another. By
reconstituting the various combinations of these three types, the authors showed that each type had a selective advantage when it
was rare relative to one or both of the other types. As a consequence, each genotype could invade and coexist with the others, so
that a stable community was formed. However, if the medium and cells were thoroughly mixed by physically shaking the flask,
then this coexistence was disrupted and one genotype prevailed. The three genotypes coexisted in the static flask because they had
evolved different abilities to exploit gradients, such as in oxygen concentration, which were generated by the organisms’ metabolic
activities in concert with the physical environment. When these gradients were eliminated by continually shaking the flask,
stable coexistence of the three genotypes was impossible.
Evolutionary Theory and Experiments With Microorganisms
9
Unstable equilibria
Those ecological interactions that promote the stable coexistence of two or more genotypes contribute to the maintenance of
genetic variation in populations. However, certain ecological interactions give rise to unstable equilibria. An unstable equilibrium
exists when each of two genotypes prevents the other from increasing in number. Such interactions do not promote polymorphisms within a local population. However, they may contribute to the maintenance of genetic differences between populations, because neither type can invade a resident population of the other type.
One form of ecological interaction that can give rise to an unstable equilibrium is interference competition. This interaction
occurs when one genotype produces a toxic substance that inhibits the growth of competing genotypes; it is distinguished from
exploitative (scramble) competition, which occurs by the depletion of resources. Many microorganisms secrete such toxins,
including fungi that produce antibiotics. Similarly, some strains of E. coli produce colicins that kill some competing strains of
E. coli, but to which the producing genotypes are immune. The producing types, when common, make so much toxins that they
can eliminate a sensitive strain that is more efficient in exploitative competition. When the producing cells are rare, however, the
cost of their colicin production is greater than the benefit of the resource that becomes available by the killing of sensitive cells, and
the producing type loses out to the more efficient colicin-sensitive competitor. The outcome of competition between colicinproducing and sensitive strains also depends on the physical structure of the environment, as shown in experiments by Lin Chao
and Bruce Levin. In particular, the advantage shifts to the colicin-producing cells on agar surfaces, even when they are rare, because
the resources made available by the killing action of colicins accrue locally to the producers, rather than being dispersed evenly as
in a well-stirred liquid medium.
Nontransitive interactions
In some cases, the frequency-dependent interactions among three or more genotypes are so complex that they become nontransitive. For example, genotype A out-competes genotype B, and genotype B out-competes genotype C, but genotype C outcompetes genotype A. A familiar example is the game of rock-paper-scissors, in which rock beats scissors, scissors beat paper, and
paper beats rock. Benjamin Kerr, Brendan Bohannan and colleagues studied an example of this game played by three strains of
E. coli on the surface of agar plates. One strain produced a toxic colicin, while the second strain was sensitive to the colicin. The
third strain was resistant to the colicin, although it did not produce any toxin. In pairwise interactions, the colicin-producing strain
out-competes the sensitive strain by poisoning it. The resistant strain out-competes the producing strain by not wasting resources
to produce the colicin, as both strains are resistant to it. The sensitive strain, in turn, out-competes the resistant strain because there
is no colicin around, and the resistant strain pays a fitness cost for its resistance. This research team further showed, both by
experiments and in computer simulations, that the three strains could coexist with one another only in a spatially structured
environment, such as on the surface of an agar plate. By contrast, in a well-mixed liquid medium, one strain dominates, although
the identity of the winner depends on the initial abundances.
As another example, Charlotte Paquin and Julian Adams found nontransitive interactions in populations of the yeast,
S. cerevisiae, that were evolving in chemostats fed with glucose as a sole carbon source. Nontransitive interactions can lead to
situations in which the average fitness of an evolving population declines relative to some distant ancestor, even though each
successive dominant genotype has increased fitness relative to its immediate predecessor. Indeed, Paquin and Adams observed
precisely this phenomenon.
Evolution in a Changing Environment
Most evolution experiments operate within a controlled, defined environment, which facilitates the analysis and interpretation of
these experiments. However, many environments change over time owing to either external processes, such as a change in
temperature, or internal processes, such as resource depletion or coevolution of interacting species. This section examines research
on the evolution of bacteria in response to prolonged resource deprivation, while the next section will address the coevolution of
interacting species.
When bacteria are inoculated into fresh medium, they grow exponentially until they exhaust the available nutrients, at
which point the cells typically enter a “stationary” phase during which they neither grow nor die, at least for many hours or
even days. However, if the bacteria are left in the nutrient-depleted medium indefinitely, they eventually enter a death phase,
although not all the cells necessarily die even after a very long time. Roberto Kolter and Steven Finkel studied the ecological
and evolutionary dynamics in long-term cultures of E. coli that were grown in a nutrient-rich medium and then left for over
5 years without additional nutrients, except that sterile water was added occasionally to compensate for evaporation. Some 99%
of the cells died over the first few weeks of the death phase. If that rate of decline had continued throughout the experiment,
then there would soon have been no survivors at all. However, Kolter and Finkel found that the surviving cells entered a sort
of second stationary phase, in which the population declined only very slowly. During this period, the rates of cell division
and death were nearly equal, so that the population was dynamic rather than static. Moreover, the environment itself
changed continually owing to the buildup and breakdown of metabolic by-products released by both living and dying cells.
This changing environment, in turn, favored mutants that were better able to survive and grow under these challenging conditions
than was the strain used to begin this experiment. The authors called these winners “GASP” mutants because they had a
growth advantage in stationary phase (GASP). Further analyses of this system found multiple waves of GASP mutants that arose
and swept to high frequency in the population, only to be displaced later by other, even tougher, mutants. Thus, many different
10
Evolutionary Theory and Experiments With Microorganisms
mutations can produce GASP phenotypes, and the various mutants differ in the details of their physiological and ecological
advantages.
Coevolution of Interacting Genomes and Species
Microorganisms in nature rarely, if ever, exist as single species, as they are usually studied in the laboratory. Rather, they exist in
complex communities of many interacting populations. Some interactions are exploitative, such that one population makes its
living by parasitizing or preying upon another population. Other interactions are mutualistic, such that each population obtains
some benefit from its association with the other. In many cases, these interactions are plastic both genetically and ecologically. For
example, a single mutation in a bacterium may render it resistant to lethal infection by a bacteriophage. And a plasmid that confers
antibiotic resistance may be beneficial to its bacterial host in an antibiotic-containing environment but detrimental in an
antibiotic-free environment.
As a consequence of this variability, microorganisms have proven useful for investigating questions about the coevolution of
interacting populations. Are there evolutionary “arms races” between host defenses and parasite counter defenses? Why are some
parasites so virulent to their hosts, whereas others are relatively benign? How can mutualistic interactions evolve, if natural
selection favors “selfish” genes that replicate themselves even at the expense of others?
Exploitative Interactions
A number of studies have demonstrated the stable coexistence of virulent bacteriophage (lytic viruses) and bacteria in continuous
culture. In these studies, the virus population may hold the bacterial population in check at a density that is several orders of
magnitude below the density that would be permitted by the available resource if viruses were not present. In most cases, however,
bacterial mutants eventually appear that are resistant to the virus, and these mutants have a pronounced selective advantage over
their virus-sensitive progenitors. The proliferation of bacteria that are resistant to infection by the original virus provides a selective
advantage to host-range viral mutants, which are capable of infecting the resistant bacteria. Thus, one can imagine, in principle, an
endless “arms race” between resistant bacteria and extended host-range viruses.
In fact, there are constraints that often preclude this outcome. Experiments performed by Lin Chao, Richard Lenski, and Bruce
Levin using E. coli and several lytic viruses found that bacterial mutants eventually evolved for which it was very difficult or
impossible to isolate corresponding host-range viral mutants. This asymmetry may arise because bacterial resistance can occur via
mutations that cause either the structural alteration or the complete loss of certain receptors on the bacterial surface, whereas viral
host-range mutations can counter only the former. Despite this asymmetry, these experiments also demonstrated that the virus
population often persisted because the virus-resistant bacterial mutants were less efficient than their sensitive progenitors in
competing for limiting resources. In such cases, the result was a dynamic equilibrium, in which the growth-rate advantage of the
sensitive bacteria relative to the resistant mutants was offset by death due to viral infection. Such tradeoffs between competitiveness and resistance commonly occur, because the same receptors used by viruses to adsorb to the cell envelope often serve to
transport nutrients into the cell or to maintain its structural integrity.
A widely held belief is that a predator or parasite that is too efficient or virulent will drive its prey or host population extinct,
thereby causing its own demise. However, virulent phage often coexists with bacteria, even though successful reproduction of the
virus is lethal to the infected bacterium. Moreover, the process of natural selection does not involve foresight, so the mere prospect
of extinction cannot deter the evolution of more efficient predators or more virulent parasites. Nevertheless, there exist many
viruses (lysogenic and filamentous bacteriophage) that are replicated alongside the host genome, and whose infections, although
deleterious, are not necessarily lethal. These viruses, as well as conjugative plasmids, have life cycles that include both horizontal
(infectious) and vertical (intergenerational) transmission.
At present, the evolutionary forces that favor these alternative modes of transmission are not fully understood. One factor
that is thought to be important is the density of hosts. If susceptible hosts are abundant, then there are many opportunities
for horizontal transmission. In that case, selection favors those parasites that replicate and infectiously transmit themselves
most rapidly, regardless of the consequences of these activities for the host’s fitness. On the other hand, if susceptible hosts
are scarce, then horizontal transmission becomes infrequent. Vertical transmission, by contrast, does not depend upon the
parasite or its progeny finding another host. Instead, the success of a vertically transmitted parasite is determined by the
success of its infected host. The greater the burden that such a parasite imposes on its host, the slower the host can
reproduce its own genome and that of the parasite. Hence, when the density of susceptible hosts is low, selection may favor
those parasites that minimize their replicative and infectious activities, and thereby minimize their deleterious effects on
the host.
Two studies sought to test this hypothesis by manipulating the supply of susceptible hosts. One experiment, by Jim Bull and
colleagues, with a filamentous bacteriophage supported the hypothesis. The other study, by Paul Turner and Richard Lenski, used a
conjugative plasmid, and their experiment did not support that hypothesis, for reasons that are unclear. However, both studies
demonstrated genetically encoded tradeoffs between the parasites’ rates of horizontal and vertical transmission. That is, parasites
that were transmitted between individual hosts at higher rates reduced the host’s growth rate – and hence their own vertical
transmission – more severely than did parasites that were infectiously transmitted at lower rates.
Evolutionary Theory and Experiments With Microorganisms
11
In addition to subtle changes in the interaction between a virus and its current host, evolution can occur when a virus
evolves the ability to infect a different host species. Most viruses are severely constrained in the organisms that they can
infect because they must enter their host via specific proteins or other receptors on the host’s cell surface, and these receptors
typically differ between host species. Two viral genomes can only recombine genetically when they both infect the same host.
Therefore, if a virus evolves the new ability to infect a different host species and, moreover, loses its ability to infect its previous
host, then this host shift may lead to speciation, whereby a single virus population eventually evolves to become two distinct
types of virus.
A study by Wayne Crill, Holly Wichman, and Jim Bull examined the adaptation of a virus to two different host species. They
alternated a bacteriophage population between two bacterial hosts, E. coli and Salmonella enterica. The experiment was designed so
that the virus saw only “naive” bacteria that had not previously been exposed to the virus; hence, the bacteria could not evolve
resistance that might complicate their analysis of the evolving virus. Because the viral genome is small, the authors could
completely sequence several isolates of the virus after each round of their adaptation to the alternating hosts. The authors found a
single base pair in the virus genome that alternated in perfect concordance with the host species to which the virus had most
recently adapted. Owing to this and other mutations, viral growth rate was always higher on the current host than on the alternate
host. However, there was a strong asymmetry in this host-specific adaptation. Specifically, adaptation to S. enterica led to reduce
viral growth on E. coli, but adaptation to E. coli did not cause a comparable reduction in growth on S. enterica. This finding has
potential relevance for the production of vaccines, because it is common practice to produce weakened versions of infectious
viruses by adapting them to new cell types, and then using these attenuated viruses in vaccines. If viral adaptation to a new host
type does not always impair viral growth on its usual host, then it becomes all the more necessary to test carefully the safety of such
putatively attenuated viruses. More generally, these results show that even single mutations can sometimes allow viruses to infect
new host species, a finding that is of interest not only for understanding evolution but also to public health.
Mutualistic Interactions
It has been proposed that many mutualisms evolved from formerly antagonistic interactions. In fact, mathematical models predict
that, at low host densities, genetic elements such as plasmids and phage can persist only if they are beneficial to their host. Many
plasmids encode functions useful to their bacterial hosts, including resistance to antibiotics, catabolic pathways, production of
bacteriocins, and so on. Also, some plasmids are unable to promote conjugation, thus relying exclusively on vertical transmission.
This limitation requires that the plasmid does not harm its host, because any plasmid-free derivative would otherwise out-compete
those cells that carry a costly plasmid. Indeed, several studies have found unexpected competitive advantages for bacteria that are
infected by plasmids, transposons, and even temperate phage, relative to cells that are not infected but otherwise genetically
identical.
Two studies have even demonstrated the evolution of mutualistic interactions from formerly antagonistic associations. Kwang
Jeon showed that the growth of the protist Ameba proteus was initially greatly reduced by a virulent bacterial infection. The harmful
effects of the bacteria were diminished, however, by propagating the infected amoebae for several years. In fact, the amoebae
eventually became dependent on the bacterial infection for their viability. In another study, Judith Bouma and Richard
Lenski found that a certain plasmid initially reduced the fitness of its E. coli host in antibiotic-free medium. However, the
plasmid enhanced the fitness of its host in this same medium after 500 generations of experimental evolution. Interestingly, the
mutation responsible for the newly evolved mutualistic interaction was in the host, not in the plasmid. Both of these experiments
show that hosts can become dependent on, or otherwise benefit from, formerly parasitic genomes, thus giving rise to mutualistic
interactions.
Evolution of New Metabolic Functions
Microbes exhibit a tremendous diversity of metabolic activities, some of which function in degradative pathways (catabolism)
while others work in synthetic pathways (anabolism). How has this diversity evolved? One area of research in the field of
experimental evolution seeks to elucidate the various processes by which microorganisms can acquire new metabolic functions.
This research is timely as humans seek to employ certain microbes that degrade toxic pollutants in the environment, and to
harness others that may be useful in the production of biofuels.
Acquisition by Gene Transfer
Perhaps the simplest way in which a microorganism can acquire some new metabolic function is by gene transfer from another
microorganism that already encodes that function. For example, antibiotic resistance functions are often encoded by plasmids,
which are transmitted from donors to recipients by conjugation. However, acquiring new functions by gene transfer is not always
so simple. Biodegradation of certain recalcitrant compounds may require the complex coordination of several steps in a biochemical pathway, which are encoded by complementary genes from two (or more) different microorganisms. The acquisition of
activities that depend on such pathways may require not only genetic exchange, but also subsequent refinement of the new
function by mutation and natural selection.
12
Evolutionary Theory and Experiments With Microorganisms
Changes in regulatory and structural genes
In a number of studies, microorganisms have been shown to evolve new metabolic functions without any horizontal gene
exchange. The evolution of these new functions often occurs by selection for mutations in existing regulatory or structural genes
that previously encoded some other function. For example, the bacterium Klebsiella aerogenes cannot normally grow on the sugar
D-arabinose, although it does possess an enzyme, isomerase, that is able to catalyze the conversion of D-arabinose into an
intermediate, D-ribulose, which can be further degraded to provide energy to the cell. This isomerase is normally expressed at a very
low level that does not permit growth on D-arabinose. Robert Mortlock and colleagues demonstrated that mutations in regulatory
genes that increase the level of expression of this isomerase are sufficient to enable K. aerogenes to grow on D-arabinose. They also
showed that the ability to grow on D-arabinose could be further improved by certain mutations in the structural gene that change
the amino acid sequence of the isomerase in ways that improve its efficiency in converting D-arabinose into D-ribulose.
Mutations that place existing genes in a new regulatory context can also have profound effects on an organism’s phenotype.
After more than 30,000 generations into the long-term evolution experiment with E. coli discussed elsewhere, one of the 12
populations grew visibly turbid, as bacteria began to grow on citrate that is present in the medium as a chelating agent. At first this
growth was thought to indicate contamination, because E. coli cannot normally grow on citrate in an oxygen-containing environment. However, work led by Zachary Blount showed that a subpopulation had evolved the ability to consume the citrate in the
medium. Blount’s work further showed that a genetic rearrangement – specifically, tandem duplication – had placed a previously
unexpressed gene encoding a citrate transporter protein under the control of a promoter that was expressed in the presence of
oxygen as the cells exhausted the glucose on which they fed. Even with this mutation, efficient growth on citrate was contingent on
one or more mutations that had evolved earlier in this lineage, and subsequent mutations were also important for the bacteria to
effectively exploit its new ecological niche.
These studies, as well as other experiments and comparative analyses, show that the evolution of new metabolic functions often
involves “borrowing” gene products that were previously used for other functions. It is not surprising that this process may
sometimes also encroach upon, and disrupt, the previous function. Such encroachment could, in turn, favor gene duplications,
whereby a single copy of an ancestral gene gives rise to two copies, each of which may then evolve toward different metabolic
capabilities.
Evolution of Genetic Systems
The process of adaptation by natural selection requires genetic variation in traits that influence the survival and reproduction of
organisms. As discussed earlier, the two sources of genetic variation are mutation and mixis. Rates of mutation and mixis depend
not only on environmental factors (eg, ultraviolet irradiation), but also on properties of the “genetic system” intrinsic to the
organism itself. Here, genetic system refers to all those aspects of the physiology, biochemistry, and reproductive biology of an
organism that affect rates of mutation and mixis. For example, organisms have mechanisms of varying efficacy to promote the
accurate replication and repair of their DNA. And while sex is an integral part of reproduction for some organisms, many others
reproduce asexually, so that the progeny are usually identical to their parent and siblings.
Some of the most interesting questions in evolutionary biology concern the significance and evolutionary consequences of
alternative genetic systems. Why do some organisms reproduce sexually, whereas others reproduce asexually? If mutation generates
variation that is necessary for adaptation by a population, but most mutations have deleterious effects on the individual, then
what mutation rate is optimal? Might organisms somehow be able to choose only those mutations that are beneficial to them,
given their present circumstances?
Sexuality and Mixis
Sexual reproduction imposes several costs relative to asexual reproduction. These costs include finding a mate, the risk of disease
transmission, and, in higher organisms, the genetic dilution that occurs because a female produces an offspring that carries only
half of her genes. Therefore, biologists have long sought to understand the advantages for sexual reproduction that could overcome
these disadvantages. Numerous hypotheses have been proposed, and all of them depend, in one way or another, on the genetic
variation that results from mixis. Most efforts to test these hypotheses have relied on comparing distributions of sexual and asexual
organisms to find variables that correlate with reproductive mode. However, several experiments have tested the evolutionary
consequences of mixis using microbes in which one can manipulate the extent of intergenomic recombination.
For example, mixis in viruses can be experimentally manipulated by varying the multiplicity of infection (MOI) of host cells,
since recombination of viral genotypes can occur only if two or more viruses infect the same host cell. An experimental study by
Russell Malmberg compared the rate of adaptive evolution in bacteriophage populations propagated at high and low MOI; the
total number of viruses per population was standardized for both treatments. The average fitness increased more rapidly under the
high-MOI (equal to high recombination) treatment than under the low-MOI (equal to low recombination) treatment. This result
is consistent with the hypothesis that sexual populations can adapt more rapidly than asexual ones because two or more beneficial
mutations can be incorporated simultaneously in the former, but only sequentially in the latter.
Another experiment indicates that the benefit of mixis in accelerating adaptive evolution may depend on the environment.
Matthew Goddard, Charles Godfray, and Austin Burt compared the rate of fitness improvement in evolving sexual and asexual
Evolutionary Theory and Experiments With Microorganisms
13
populations of the yeast S. cerevisiae. By deleting two genes in the yeast, they were able to construct an asexual version that could
not undergo meiosis. They then established separate populations of sexual and asexual yeast in both harsh and benign environments, and allowed the yeast to evolve for several hundred generations. Neither sexual nor asexual strains showed significant
improvement in the benign environment, implying that they were already well adapted to those conditions. However, both strains
adapted to the harsh environment, with the rate of improvement consistently greater for the sexual than the asexual populations.
This experiment suggests, therefore, that genetic recombination can be especially important for adaptation to stressful and
changing environments.
Some experiments have suggested that another advantage of mixis may occur when the rate of deleterious mutation is high and
the effective population size is very small. Such conditions may apply especially to microorganisms with high error rates during
replication (eg, RNA viruses) or those with large genomes (eg, protozoa), if their populations experience periodic “bottlenecks.” In
these cases, deleterious mutations tend to accumulate in asexual lineages, a process called “Muller’s ratchet” (after the geneticist,
H. J. Muller, who first described this phenomenon). However, even occasional mixis can purge lineages of their accumulated load
of deleterious mutations, as demonstrated by Lin Chao and colleagues using a segmented RNA virus that infects bacteria. This
effect occurs because two recombining genomes may each complement the deleterious mutations that are present in the other,
thereby generating some progeny that have a reduced load of deleterious mutations (as well as other progeny with an increased
load, which will be removed by natural selection).
In some cases, mixis might not be an adaptation to recombine genes but rather a coincidental consequence of the movement
between cells of parasitic entities. In many bacteria, for example, mixis occurs only when cells are infected by viruses (transduction) or plasmids (conjugation). The new combinations of chromosomal genes that sometimes result from such infections
may occasionally be advantageous. However, one cannot view phage and plasmids as benevolent agents of bacterial carnal
pleasure, because these infectious agents often harm or even kill their hosts.
Evolutionary Effects of Mutator Genes
“Mutator” genes increase the mutation rate throughout an organism’s genome by disrupting DNA repair functions. Experiments
performed by Lin Chao and others have investigated the effect of mutator genes on bacterial evolution. These studies have revealed
a pattern that seems, at first glance, rather curious. When a mutator gene is introduced in a population above a certain initial
frequency (eg, 0.1%), it tends to increase in frequency over the long term. However, if that same gene is introduced at a frequency
below that threshold, then it typically goes extinct.
What causes this curious effect? In a sense, there is an evolutionary race between two clones, with and without the mutator
gene, to see which one gets the next beneficial mutation. The rate of appearance of new mutations for each clone depends on the
product of its population size, N, and its mutation rate, u. When the ratio of mutation rates for the mutator and nonmutator
clones, u'/u, is greater than the inverse ratio of their population sizes, N/N', then the mutator clone is more likely to have the next
beneficial mutation and thus prevail over the long term. When u'/u is less than N/N', the nonmutator clone, by virtue of its greater
numbers, is likely to produce the next beneficial mutation and thereby exclude the mutator clone.
However, this explanation presents a problem for understanding the evolution of mutators in nature, where they are moderately prevalent in some circumstances. If mutator genes are useful only when they are common, then how do they become
common in the first place? Mathematical models and evolution experiments with bacteria indicate that a process called “hitchhiking” can resolve this paradox. In hitchhiking, a deleterious mutation (such as the one that disrupts DNA repair) gets carried
along to high frequency if it is genetically linked to a beneficial mutation. In bacteria, which usually have a single chromosome and
reproduce asexually, the entire genome is effectively linked. Moreover, a mutator gene is more likely than any other deleterious
mutation to be associated with a new beneficial mutation because of the high mutation rate it causes. It is unlikely that any
particular mutation that produces a mutator will yield a beneficial mutation that allows the mutator to hitchhike. Given enough
time, however, one “lucky” mutator may do so, and the mutator gene can then increase in frequency by hitchhiking with the
beneficial mutation that it caused.
It has also been proposed that mutator genes may be more common in pathogenic bacteria than in their nonpathogenic
counterparts. The idea is that pathogenic bacteria face especially rapidly changing selective conditions owing to immunological
and other host defenses. By having a high mutation rate, pathogens would have a better chance of evolving a counter defense. This
explanation also fits well with the hitchhiking hypothesis, because every change in the host that favors a new mutation in the
pathogen population creates an opportunity for a mutator allele to hitchhike to high frequency. By contrast, for an organism living
in a constant environment to which it is already adapted, beneficial mutations would be much rarer and hence there would be
fewer opportunities for a mutator gene to hitchhike. Antonio Oliver, Fernando Baquero, Jesús Blázquez, and colleagues examined
the frequency of mutator genes in samples of Pseudomonas aeruginosa, a bacterial species that often causes chronic (long-term) lung
infections in people with cystic fibrosis. This same species is common in the environment, and it can cause acute (short-term)
infections in patients with severe burns or otherwise weakened immune systems. The authors documented a much higher
frequency of mutator clones in samples from patients with chronic infections, which supports the hypothesis that frequent changes
in host immunity and antibiotic regimens can promote the evolution of high mutation rates.
An experimental study by Csaba Pal, Angus Buckling, and colleagues examined the effect of viral parasites on the evolution of
mutation rates in bacteria. They found that P. fluorescens that were coevolving in environments with bacteriophage present were
much more likely to become mutators than were bacteria that were evolving in the absence of the bacteriophage. This study also
14
Evolutionary Theory and Experiments With Microorganisms
supports the general hypothesis that rapidly changing environments – including coevolving parasites as well as hosts – can favor
the evolution of elevated mutation rates.
Thus, aspects of genetic systems that increase variation – whether by mutation or mixis – may accelerate adaptive evolution.
On the other hand, mutation and mixis can also break down genotypes that are already well adapted to particular environments.
The evolution of genetic systems may often reflect the balance between these opposing pressures.
Directed Mutations?
The experiments of Luria and Delbrück and the Lederbergs demonstrated that mutations arose before bacteria were exposed to a
selective agent, and thus the mutations were not a response by the bacteria to that agent. However, in 1988, John Cairns and
colleagues called into question the generality of random mutation in bacteria. Their paper and some other studies seemed to show
that certain mutations occurred only (or more often) when the mutants were favored, and such mutations were called “directed.”
However, several subsequent experiments indicated that some of the evidence for directed mutation was flawed or misinterpreted,
and this subject became very controversial. However, there is now widespread agreement that the most radical hypotheses put
forward to explain this phenomenon – for example, a reverse flow of information from the environment through proteins and
RNA back to the DNA – are incorrect.
Nonetheless, this controversy generated renewed interest in bacterial mutation, especially the mechanisms by which a cell
might exercise some control over the mutational process. Attention became focused on understanding the effects of stress, such as
due to starvation, on DNA repair and mutation, as well as the extent of variation among local DNA sequences in their mutability.
For example, numerous studies have documented unusually high mutation rates in short repeated sequences (eg, TTTT); the
mutations are typically frameshift events involving the loss or gain of a repeated element, and they occur via the slippage of DNA
strands during replication. These hypermutable sequences are not distributed randomly throughout bacterial genomes. Rather,
they are found more often in genes encoding products (such as fimbriae and lipopolysaccharides on the cell surface) that are
involved in pathogenicity and evasion of host immune surveillance. This distribution suggests that bacteria have evolved a simple
but effective strategy to increase the mutation rate in genes that help them cope with unpredictable aspects of the environment,
without inflating the load of deleterious mutations in “housekeeping” genes whose products interact predictably with the
environment. These mutations are apparently random insofar as a particular mutation does not occur as a direct response to an
immediate and specific need, but they are nonrandom in their genomic distribution and may thereby promote a more favorable
balance between evolutionary flexibility and conservatism.
Evolving Multicellularity
Experimental evolution is also being used to investigate certain major transitions in evolution. In a pioneering study, William
Ratcliffe, Michael Travisano, and colleagues demonstrated that a simple selection procedure that favored faster settling in the yeast
S. cerevisiae led to multicellular yeast. That is, those individuals that reached the bottom of the culture vessel the fastest were used
to seed the next generation. In principle, this procedure might have favored sticky cells that simply adhered to one another.
However, subsequent analyses revealed that the clusters – shaped almost like snowflakes – were formed by cells that did not
completely separate after division. Moreover, these clusters continue to replicate as multicellular configurations, with daughter
clusters breaking off from an existing cluster, and then growing in size before releasing their own multicellular propagules. The
ability to observe such a major transition in the laboratory demonstrates the power of experimental approaches to addressing
evolutionary questions.
Further Reading
Baquero, F., Nombela, C., Cassell, G.H., Gutiérrez, J.A. (Eds.), 2008. Evolutionary Biology of Bacterial and Fungal Pathogens. Washington, DC: ASM Press.
Barrick, J.E., Lenski, R.E., 2013. Genome dynamics during experimental evolution. Nature Reviews Genetics 14, 827–839.
Bell, G., 1997. Selection: The Mechanism of Evolution. New York, NY: Chapman & Hall.
Bohannan, B.J.M., Lenski, R.E., 2000. Linking genetic change to community evolution: Insights from studies of bacteria and bacteriophage. Ecology Letters 3, 362–377.
Chadwick, D.J., Goode, J. (Eds.), 1997. Antibiotic Resistance: Origins, Evolution, Selection and Spread. Chichester: Wiley.
Chao, L., 1992. Evolution of sex in RNA viruses. Trends in Ecology and Evolution 7, 147–151.
Davies, J., Davies, D., 2010. Origins and evolution of antibiotic resistance. Microbiology and Molecular Biology Reviews 74, 417–433.
Dykhuizen, D.E., Dean, A.M., 1990. Enzyme activity and fitness: Evolution in solution. Trends in Ecology and Evolution 5, 257–262.
Elena, S.F., Lenski, R.E., 2003. Evolution experiments with microorganisms: The dynamics and genetic bases of adaptation. Nature Reviews Genetics 4, 457–469.
Finkel, S.E., 2006. Long-term survival during stationary phase: Evolution and the GASP phenotype. Nature Reviews Microbiology 4, 113–120.
Futuyma, D.J., 2013. Evolution. Sunderland, MA: Sinauer.
Kassen, R., 2014. Experimental Evolution and the Nature of Biodiversity. Roberts: Greenwood Village, CO.
Maynard Smith, J., Szathmáry, E., 1995. The Major Transitions in Evolution. Oxford: Oxford University Press.
Mayr, E., 1942. Systematics and the Origin of Species. New York, NY: Columbia University Press.
Mortlock, R.P. (Ed.), 1984. Microorganisms as Model Systems for Studying Evolution. New York, NY: Plenum.
Moxon, E.R., Rainey, P.B., Nowak, M.A., Lenski, R.E., 1994. Adaptive evolution of highly mutable loci in pathogenic bacteria. Current Biology 4, 24–33.
Rice, W.R., 2002. Experimental tests of the adaptive significance of sexual recombination. Nature Reviews Genetics 3, 241–251.
Travisano, M., Velicer, G.J., 2004. Strategies of microbial cheater control. Trends in Microbiology 12, 72–78.