J Antimicrob Chemother 2011; 66: 1659 – 1672
doi:10.1093/jac/dkr214 Advance Access publication 16 June 2011
The 2010 Garrod Lecture: The dimensions of evolution in antibiotic
resistance: ex unibus plurum et ex pluribus unum
Fernando Baquero *
Department of Microbiology, Ramón y Cajal University Hospital, IRYCIS and CIBERESP, Ctra de Colmenar km 9,100, 28034 Madrid, Spain;
Centre for Astrobiology (CSIC/INTA), National Institute for Aerospatial Technology, Ctra de Torrejón a Ajalvir, km 4, 28850 Torrejón de
Ardoz, Madrid, Spain
*Corresponding author. Tel: +34-91-3368832; Fax: +34-91-3368809; E-mail: [email protected]
Received 15 April 2011; returned 26 April 2011; revised 29 April 2011; accepted 4 May 2011
Antibiotic resistance is not only the result of antibiotic-driven selection, as has been frequently considered, but
rather the consequence of extremely complex evolutionary processes. These processes act on bacterial populations, but also on populations of subcellular (as plasmids) or supracellular (species, communities) units of
evolution. The consideration of the effects of drift and selection provides a first intuition about the dimensions
shaping the evolutionary field. We distinguish two alternative, orthogonal dimensions, respectively pushing
evolutionary units towards diversification (ex unibus plurum) or towards unification (ex pluribus unum). Evolution in each one of these dimensions requires alternative evolutionary functional configurations in the evolving
unit. These configurations are reached under the influence of evolutionary attractors for diversification or unification, presumably with an oscillatory dynamics. This view illustrates the complexity of possible outcomes in
the emergence and evolution of antibiotic-resistant units, and indicates both the absolute need for multilevel
epidemiological surveillance of antibiotic resistance, and the necessity of applying new evolutionary synthesis
approaches to understand and predict human-driven changes in the microbial world.
Keywords: drift, selection, diversification, unification
Introduction
The major aim of my Garrod Lecture and this accompanying
article is to discuss the dimensions used by evolutionary processes involved in antibiotic resistance. The term ‘dimensions’
is used here in the acceptation of coordinates specifying points
in a theoretical space in which biological objects are able to
change positions; i.e. they are able to evolve. The general mechanisms1 by which evolving objects (individuals inside populations) move in the evolutionary space are essentially natural
selection and drift. These so-called ‘mechanisms’ have also
been considered ‘forces’, but I agree with Christopher Stephens2
that they are not real mechanisms or forces acting on populations, but only ‘pseudo-processes’, statistical by-products of
the sequential populations resulting from sequences of events,
essentially birth and death rates. The application to collective
biological objects of selection and drift can shape evolutionary
trajectories within a dimensional space. The central argument
of this work is to present here what we can consider as the
two main dimensional evolutionary axes in which the biological
objects move: along one axis, these objects move towards diversification (ex unibus plurum); along the other, towards unification
(ex pluribus unum), using, as in a recent work,3 the Saint
Augustine (Confessions, IV, 8 –13) classical expression. Even if
we accept the generality of this view in the living world, evolutionary microbiologists are in a much better position than
other scientists to appreciate these fundamental movements.
Many microbial and submicrobial objects retain an evolutionary
plasticity that has been reduced in more complex forms of life.
Once more, we should remark that microbial antibiotic resistance
has provided one of the best possible frames to investigate fundamental mechanisms in evolution. By learning about evolution,
we have learnt about antibiotic resistance, and by studying antibiotic resistance we have easily slipped into the basic themes of
evolution.
The intention of this review was also to honour my esteemed
predecessor, Professor Richard Sykes, who last year devoted the
Garrod Lecture to ‘The evolution of antimicrobial resistance: a
Darwinian perspective’.4 To a certain extent, this lecture can be
considered as a second part of his lecture and therefore an
extension of his aim in recognizing the importance of Charles
Darwin on the 200th anniversary of his birth, and the 150th anniversary of his seminal book On the Origin of Species. The Sykes’
review was more focused on the specific mechanisms involved
in the evolution of antibiotic resistance, whereas the present
one is more directed to discussing the theoretical dimensions
# The Author 2011. Published by Oxford University Press on behalf of the British Society for Antimicrobial Chemotherapy. All rights reserved.
For Permissions, please e-mail: [email protected]
1659
Baquero
in which evolutionary trajectories are tracing their complex
paths, as a necessary exercise for dealing in the future with
the prediction of antibiotic resistance.5 Both reviews are in line
with the feeling that a new evolutionary synthesis is in sight.6
What is and what is not natural selection?
Natural selection is Darwinian selection. According to Lewontin,7
the principles that embody the principle of evolution by natural
selection are the following: (i) if there are variant entities in a
population; (ii) if these variants experience different reproductive
success; and (iii) if the variation is inheritable; then, (iv) the composition of the population will change over time. This can clearly
be applied to the case of antibiotic resistance: (i) if there are
resistant variants in a population; (ii) if differential reproductive
success is experienced in the presence of antibiotics; and (iii) if
the variation is inheritable; then, (iv) the composition of the
population will change over time in favour of resistant variants.
However, not every change in the composition of a population
is necessarily due to natural selection. Of course, natural selection assures the survival and further reproduction of the fittest
(Darwinian selection), but there are other possibilities assuring
the survival and further reproduction of a group of individuals
endowed with specific mechanisms of adaptation to the challenge when others in the same population become extinct.
some characters facilitating their location in particular compartments where the challenge is minimized. For instance, intracellular bacteria might ‘resist’ a number of antibiotics, without any
need for special antibiotic resistance mechanisms. The same is
probably true for necrotizing bacteria, for organisms adhered to
non-vascularized surfaces (e.g. catheters) or for those forming
thick biofilms in altered mucosal surfaces, such as in cystic fibrosis. The possible cross-talk between virulence and antibiotic
resistance has been reviewed elsewhere.8 In some other cases,
bacteria might overcome the antibiotic effect just because, by
chance, they are positioned in the vicinity of other bacteria
that are able to protect them from antibiotic action (e.g. by
excreting detoxifying enzymes).9 These otherwise susceptible
organisms are able to survive without natural selection, acting
as ‘cheats’ of the protective ‘altruist’ population.10,11
Survival of the best connected
High connectivity opens the door to the fertile adaptive field of
horizontal gene transfer. If a bacterial organism belongs to a
community of bacteria forming a preferential ‘genetic exchange
community’,12,13 its possibilities for further adaptation are higher.
Probably, just belonging (being connected) to a complex community should increase the survival possibilities when facing difficult
times, as complexity tends to favour global robustness.14,15
Survival of the luckiest
Survival of the denser
This term refers to individuals that overcome a challenging event
just by chance, as they are randomly located in a physiological
status (such as growth phase) or in a place where the deleterious
agent or condition is not lethal. Let us imagine that some of
these lucky survivors carry a particular inheritable genetic difference devoid of any adaptive value, e.g. a synonymous substitution in a triplet encoding a particular amino acid. This
difference will be amplified in the reconstructed population in
comparison with the ancestral one, even though it could be
totally unrelated to the deleterious event. Reducing the overall
number of cells, antibiotics reduce the diversity of the population,
and, therefore, some of the variants that survive just by chance
and reproduce after crossing the bottleneck (an event decreasing
population size) might appear to have been ‘selected’; however,
they are not the fittest to resist the antibiotic, just the luckiest in
evading its action. For instance, it is tempting to illustrate the
case by mentioning phenotypically antibiotic-tolerant cells that
occur transiently and randomly in a bacterial population (as a
result of a mere asynchrony in the growth cycle), and which
can survive purely by chance.
It should be remembered that the notion of ‘luck’ (either for
microbes or humans) refers to something that implies the total
equality of chances for all individuals. In the natural world,
such deep homogeneity (assuring random sampling) is rare, so
that the notion of ‘survival of the luckiest’ is biased according
to some circumstantial features. Three of these are discussed
here.
High-organism density is one of the preferred strategies of
Nature to assure the non-specific survival of her children. In
reality, an important part of the bacterial struggle for life is
devoted to obtaining resources for multiplication, as an investment for maintenance during time.16 To a certain extent, survival
of the denser is frequently a pre-condition for assuring survival of
the luckiest or survival of the best placed. For instance, high cell
density guarantees stochastic access to a protected condition.
Most importantly in the case of antibiotic resistance, high
density assures availability to random mutations or random
gene duplications eventually able to adapt the cell to external
challenges, and also to increase high connectivity with other
organisms.
Note that in all of these cases the genetic composition of a
bacterial population might change over time, because a part of
it is spared from a deleterious event; however, the survival of
this fraction is not due to the acquisition of any specific adaptive
trait to counteract the challenge. In summary, we could differentiate between a situation in which all individuals in the population have equal probability to survive (survival of the luckiest)
and others in which some of the individuals have, by reasons
unrelated with the change, a higher probability of survival
(such as survival of the better placed). The first case fits with
the canonical concept of evolution by drift; in the second, there
is also drift, but it may be mixed with a certain bit of evolution
by selection; this will be discussed later.
Drift and effective population size
Survival of the better placed
‘Survival of the better placed’ refers to individuals that overcome
a challenging deleterious event because they are endowed with
1660
A population is changing by drift when there is a change in the
frequency of a genetic variant in the population due to random
sampling. Sampling means that a fraction of identical individuals
The Garrod Lecture
in a population is randomly isolated from the rest, as illustrated
by a handful of balls extracted from an urn. Let us imagine that
there are only 2 black balls among 200 white balls in the urn, but
just by chance 1 black ball is part of a handful of 3 balls extracted
from the urn. So, the relative frequency of a hypothetical genotype associated with black balls rises from 0.01 to 0.33. If only
two balls are extracted and one is black, the frequency would
be even higher, 0.5. But, if a large number of balls could be
extracted, let us say 100, and among them was 1 black ball,
the proportion would again be 0.01; in this case, the proportion
in the sample will be identical to that in the whole urn. The
number of balls in the sample required for having a proportion
identical to that in the entire urn illustrates the important
concept of ‘effective population size’ (Ne). Ne is the expected
minimal size of an ideal subpopulation that consistently reflects
the diversity of the whole population. Consequently, sampling of
populations below Ne are at risk of offering genetic variability
that deviates, just through chance, from that of the original
population; in fact, the smaller the size of the sample (3 balls
instead of 100 out of the urn), the higher the deviation in frequency is expected to be. Of course, drift might reduce bacterial
variation; if the number of balls extracted from the urn does not
contain any of the rare (black) genotype, the new population will
be constituted entirely of white balls. A minority of
antibiotic-resistant bacteria might increase in proportion or disappear just by drift. In the first edition of On the Origin of
Species, Charles Darwin was already aware of the possibility of
fluctuations in the frequency of variations that have no adaptive
significance or are otherwise equally fit (and therefore not submitted to natural selection).17
This means that any ‘sampling’ condition resulting in a
reduction in the number of cells below the Ne might produce
changes in the proportion of a particular variant in a bacterial
population that is unrelated with selection, i.e. ‘neutral’. Effective
population sizes in bacteria range from Ne ¼105 to 109.18 In fact,
probably many of the changes in the frequency of particular
genetic (and genomic) changes in bacterial genes, including a
number of changes that might result in antibiotic resistance,
might not be the result of presumed selective events related
with antibiotic consumption, but just occur by drift. Note that,
most probably, the effective population size for most bacterial
populations is very high, in the order of Ne 108 in Escherichia
coli.19 Many circumstances in the microbial lifestyle, including
host-to-host transmission bottlenecks, will produce sampling
with population sizes below Ne.
Host-to-host bacterial transmission in the community, in hospitals or in infective processes20 typically starts with a low
number of bacterial cells and, therefore, small samples are
expected to produce a high degree of drift-based evolution.
Drift, indeed, should be proportional to the number of these bottlenecks. The genetic variability of a highly transmissible epidemic clone whose cells are able to surpass many independent
bottlenecks should be higher than that of a well-established
strain maintained in a stable environment. Similarly, the dilution
or migration of a bacterial population in an environment (e.g.
sewage soil or water) certainly contributes to its ‘sampling’ in
small subpopulations naturally submitted to heavy drift. A
sample population of ‘migrants’ might establish a colony far
from the site of origin, with a genetic composition eventually
biased with respect to that of the ancestral one. For instance,
JAC
resistance might be rare in the original population and frequent
in the derived colony for reasons unrelated to antibiotic selection.
This is a ‘founder effect’, as this population might give rise to new
secondary populations that might follow the diversification
process, but will retain the ‘mark of origin’ of the founder.
Coalescent theory has been applied to model the amount of variation expected from such drift-derived effects on populations
connected by migration.21,22
Lastly, the use of biocides and antimicrobials also reduces
(‘sample’) bacterial populations, which again should increase
drift. Persister cells with phenotypic resistance, able to tolerate
exposure to antimicrobials and to reassume growth in their
absence, apparently occur randomly in bacterial populations.
As previously mentioned, the frequency of genetic traits associated with these persisters will eventually increase locally just
by drift. As we will see later, transmission and migration bottlenecks also occur for suborganismal evolutionary objects, such as
plasmids, phages or conjugative transposons. Of course, that
does not necessarily mean a cumulative and progressive
genetic differentiation from the ancestor population, as drift is
not expected to result in a directional populational movement,
but, most frequently, a kind of random ‘Brownian motion’ in
the evolutionary space, not free from extinction events.
The fuzzy line between drift and selection
Even though evolution by drift and evolution by natural selection
are frequently considered as alternative forces in evolutionary
processes, there are spaces in which both forces are sequentially
or coincidentally acting on evolutionary objects. A number of
steps can be recognized paving the path from drift to selection.
First, we can expect that a single nucleotide variation giving
rise to a synonymous codon should be effectively neutral,
without any consequence in the protein structure. Therefore,
even if this variant could be enriched by drift, nothing will
occur in terms of selective adaptation. Second, the nucleotide
variation might produce an amino acid change influencing a
protein domain but without phenotypic consequences, and so
will not be subjected to natural selection. Note that the
absence of expected phenotypic consequences (such as an
increase in b-lactam MIC) might not necessarily be interpreted
by itself as neutrality. For instance, while the change in
b-lactamase conformation might not influence hydrolytic efficiency, it might affect the stability of the protein and would
therefore comprise a selectable change.23 Third, the nucleotide
change might result in a protein change with all appearances
of neutrality (i.e. without any functional consequence), but
which might influence the effect of other mutations that might
occur later, either increasing or reducing the possibility of
natural selection (positive or negative sign epistasis).24 – 26
Fourth, the variant nucleotide might influence the phenotype,
but in an extremely subtle way, so that the phenotype might
be overlooked by natural selection. This concept was posed (for
b-lactamases) as ‘we do not know how small an effect constitutes a selective advantage’.27 It has, however, been shown that
very small phenotypic differences are indeed selectable along
natural gradients.28,29 In general, the neutrality of the variation
in a particular site can be detected by the calculation of the
ratio of non-synonymous substitutions (dN) to the number of
1661
Baquero
synonymous substitutions per site (dS), (v ¼ dN/dS). When this
calculation gives values of v ¼ 1, the message is that changes
with amino acid replacement are neither selected (positive selection) nor negatively selected (purifying selection), and represent
neutral evolution. Indeed, most statistical methods are based on
single-locus calculations, disregarding gene– gene interactions in
the emergence of phenotypes, a problem that should be urgently
addressed.30
Neutral variation might also occur because of the effect of
phenotypic capacitors. Capacitors are proteins involved in cellular
networks allowing genetic variation to accumulate in a silent
(neutral) state, until it is revealed by environmental stress.31 – 33
Candidate proteins for effectors of evolutionary capacitance are
regulatory genes, networks of chaperones and, in general, proteins with high connectivity with other proteins.
It is of note that in several of the above-mentioned cases
there is a drift-based unselectable variation that becomes selectable upon a particular event. In these cases, drift evolution is to a
certain extent a pre-condition for evolution by natural selection.
In a sense, drift can be considered as pre-adaptive. This raises
the question as to why rare neutral or pre-adaptive random variations do not disappear in bacterial populations. This can be
explained by applying the concept of mutation-selection
balance as it was postulated in the 1920s by Fisher and
Haldane. Because of the huge number of bacterial cells in
most bacterial populations, even very unsuccessful genetic variants are expected to persist for enough time (generations) to
reach an area in the fitness landscape in which they could
provide a new advantage (e.g. a selective antibiotic concentration), resulting in a new increase in frequency.34
A central problem in applying the theoretical definition of drift
to bacterial evolution is the basic heterogeneity of bacterial
populations. In the ‘phenotypically identical balls in an urn’
naive model (see above), one ball cannot be differentiated
from another as they are homogeneously distributed in the
urn, resulting in real random sampling. In living bacterial populations, under normal asynchronic growth, every cell has a particular metabolic stage at different phases of the cell cycle;
subsets of cells are phenotypically different, gene expression
might differ stochastically among cells, mobility might randomly
change a cell’s place in the urn and they are submitted to
random mutational events, eventually leading to pre-adaptive
neutral changes. The problem is that all these random processes
are not ‘randomly distributed’ in the natural ‘urns’, where bacterial organisms are at variable distances from nutrients,
oxygen or colonizable surfaces (e.g. in the periphery or the core
of a biofilm), or are more or less exposed to host immune
responses or the effects of antimicrobial drugs. For instance,
local stress might result in local increases in the stochasticity
(noise) of gene expression.35 – 37 Different samples might thus
contain different types of cells carrying pre-adaptive changes
and, therefore, their possibilities of surviving particular challenges
might differ; in other words, drift might be biased by selection
and vice versa. The relevance of local effects (landscape
ecology) on the bacterial evolution of antibiotic resistance has
been reviewed elsewhere.37
The conclusion from the above is that evolutionary trajectories in bacterial organisms, including antibiotic resistance,
might be frequently shaped by drift, selection, and the interaction of drift and selection. According to the ‘nearly neutral
1662
theory’ of Tomoko Otha,38,39 ‘slightly advantageous’, ‘slightly
deleterious’ and ‘in-between advantageous and deleterious’
changes are really important, or at least as important as
‘strongly selectable’ and ‘really neutral’ changes.40 The intensity
of changes due to natural selection will produce increases in the
standard deviation (s) of the distribution of values in the population and the intensity of the drift will be inversely proportional
to the population size, 1/N, so that in large populations genetic
drift is a weaker force than selection. In conditions of s ≤ 1/N,
both drift and selection are playing a role in evolutionary
trajectories.
What is an evolutionary individual?
Drift and selection, and the interplay of both evolutionary forces,
act on individuals that are members of populations. If the individuals are replicators subjected to inheritable variation, and if
some of these variants should experience different reproductive
success, the composition of the population will evolve over time
by natural selection. It there is no differential reproductive
success of the variants, but the individuals are submitted to
random sampling, the composition of the population will
evolve over time by drift. Individuals can therefore be considered
as evolutionary units.
The critical question now is what is an individual, if ‘lacking
individuality, it cannot be a unit of evolution’.41 Interestingly,
Darwinian thought has used the concept of ‘selection’ to identify
what an evolutionary individual is (expressed in the following
syllogism):
M. Only individuals can be units of evolution
m. A unit of evolution is a unit of selection
C. Individuals are the units of selection
Units and levels of selection in antibiotic
resistance
The units of selection define the evolutionary individuals, but in
the case of antibiotic resistance, what is selected? Possible
units of selection in antibiotic resistance are discrete genetic
sequences, genes, operons, functional genetic modules, mobile
genetic elements such as integrons, integrative conjugative
elements, transposons, plasmids, or genomes, cells (organisms),
clones, clonal complexes, species, communities, and ecosystems. Note that all these possible units belong to different hierarchical levels, ranging from the relatively simple to the complex,
as resistance genes are part of integrons, the integrons part of
transposons, the transposons part of plasmids, plasmids part
of cells, cells part of clones, clones part of species, etc. Each
unit is a ‘vessel’ for the other, affecting its potential dissemination and also its rate of evolution.42 But, are we right in considering each of these ‘particles’ as a potentially selectable unit? At
first sight, selection for higher-level units should force the selection of lower-level units contained in the higher-level particle, but
might the opposite also occur and is there room for ‘independent
selection’ of the different units? This question about the selection
of the different evolutionary individuals is at the root of the
debate on the levels of selection and the multilevel selection
The Garrod Lecture
theory.43 – 45 Notice that in this context, ‘evolutionary individuals’
might be understood not as unique individuals, but as collective
individuals (lineages).46
We will assume from now on that an individual is any entity
that is selected as a unit independently of the number of
elements entering into its composition or its hierarchical complexity. The elements that compose an individual might be, in
their turn, individuals and, therefore, will be able to be selected
as units with independent fitness. ‘Fitness values should be
attributed to both groups of individuals and individuals. . .’.47,48
A simple example will facilitate the reader’s thinking on this
debatable topic. A bacterial cell (B1) acquires a gene (G) that
determines its antibiotic resistance. This gene might be located
in the chromosome or in an indigenous plasmid (P1). As a consequence of this capture and in the presence of a selective agent
(antibiotic exposure), B1 replicates let us say twice, providing
four cells (a sister B cell without G does not multiply or is
killed). We now have four cells, four plasmids and four genes,
independent from the element (chromosome or plasmid) that
has captured the advantageous trait. But, now the plasmid is
transferred by conjugation to another bacterial cell from a different lineage than the first (B2) and subsequently moves to
another location. B2 also multiplies twice as a result of the selective process, so that we have (considering both locations) four B
cells, four B2 cells, but eight P plasmids and eight G genes. Now,
gene G moves (e.g. because of a transposon) from the P1
plasmid to a P2 plasmid inside a new B3 bacterial cell of a different lineage, which multiplies twice. The final count is four B cells,
four B2 cells, four B3 cells, eight P1 plasmids, four P2 plasmids
and 12 G genes. Thus, the frequency of the particles has diverged
in the different units, demonstrating that they are discrete units
of selection. Answering the wise classic question proposed by Elisabeth Lloyd,49 ‘for whom is the major selective benefit?’, apparently the G gene is the most fit entity in this multilevel ensemble.
Of course, in the absence of plasmid and gene transfer, the frequency of all particles will stay stable.
The conclusion is that probably each one of the evolutionary
units involved in antibiotic resistance mentioned above is
capable of both independent and dependent selection. To close
this discussion with a smile, Bruce Levin’s unpublished opinion
is that antibiotic resistance is not only an extraordinary chance
to witness and study adaptive evolution in real time, and a
magnificent career opportunity not only for evolutionary and
other biologists, but also a career opportunity for each one of
the evolutionary units.
JAC
the robustness that such a complex structure provides to bacterial biological systems. We can try to eliminate (e.g. with a
novel antibiotic, a vaccine or an inhibitor of plasmid replication)
parts of the system, but we are fighting against a highly robust
system.
Rvobustness is the maintenance of the system characteristics,
antibiotic resistance in our case, despite fluctuations in the frequency or behaviour of its component parts that might result
from these interventions. In the bacterial world, high robustness
at the clonal or species level is based on the extreme redundancy
resulting from the huge numbers of individuals. At the level of a
single isolated individual cell there is of course complexity and
robustness, but still robustness can be overcome by the disruption of cellular processes by unexpected events (such as
exposure to a new antibiotic). Similarly, at the other side of the
hierarchy, in bacterial communities and possibly in ecosystems,
again the number of ‘individuals’ is small, as a community can
be considered as a ‘unit’ or ‘collective individual’ and robustness
could be overcome by unexpected events affecting all levels (as
in so-called ‘forest-fire models’). In these cases, individuals in low
numbers under isolation have a real ‘robust, yet fragile’ behaviour.14 In contrast, if the ‘unit’ is composed of a very high
number of component individuals with high connectivity
between them (e.g. because of horizontal gene transfer), robustness, tolerance of change, will be high. Inside the conceptual framework of the complexity studies, this behaviour will enter into
the class of systems with ‘highly optimized tolerance’. The
term ‘highly optimized’ means that the components (pieces) of
the system are linked by unique, highly structured, and in a
sense, rare patterns that are evolutionary products.14,51
Risk of being simple, risk of being complex
From the preceding paragraph we can conclude that complexity
provides stability, inertia to change and robustness to a particular system when confronting changes, particularly expectable
(experienced) changes. On the other hand, the internal complexity needed for robustness also constitutes an Achilles’ heel
increasing fragility, particularly when confronted with nonexpectable changes. Of course, the natural history of most biological entities is the result of exposures to both expected and
non-expected changes, acting consecutively or even coincidentally. It is proposed that evolutionary units optimize the risks of
such situations by using (or taking advantage of) two evolutionary dimensions, referred to here as the ‘ex unibus plurum’ and
the ‘ex pluribus unum’ dimensions.
Antibiotic resistance is embedded in individual
Two alternative and orthogonal evolutionary
complex systems
dimensions
We should admit that antibiotic resistance is embedded in
complex systems, which are composed of many components,
and that they themselves are components in larger systems
and networks. The evolution of resistance is therefore the evolution of a complex system.15,50 The complexity is due not only
to the number of components involved in antibiotic resistance,
but to the heterogeneity (self-dissimilarity) of such components,
and their hierarchical organization in highly structured networks
and spreading at multiple rates. An important part of the
difficulties in fighting against antibiotic resistance depends on
The ‘ex unibus plurum’ (from one, many) dimension refers to the
natural events resulting in the diversification of evolutionary
units.3 The ‘ex pluribus unum’ (from many, one) is its alternative
orthogonal (opposite) dimension, leading to the unification of
evolutionary units.
The important concept behind this is that both dimensions
provide alternative fields of action for the development of different (opposite) evolutionary configurations. Here, we are using the
notion ‘field’ and not ‘force’, as the term ‘force’ could represent
1663
Baquero
for the reader a process guided by external causes.2 A diversifying field of action results in collections of evolutionary configurations, which are, in fact, highly related manifestations
occurring in a unique diversifying field. The opposite configurations occur in the unifying field of action. A number of these
alternative evolutionary configurations are listed below. The
first term refers to the diversifying and the second to the unifying
fields of action. Note that all first terms are highly related and
the same applies to the second terms. In fact, we can consider
them as perceptible manifestations or topologies of a unique
basic evolutionary configuration, which we are differentiating
here for heuristic purposes.
(i) Variation versus conservatism. At particular spatio-temporal
frames, evolutionary units could either give rise to variants
(and these variants to other variants in turn) or, in contrast,
they could be stagnant, stably maintaining the previous
configuration. For instance, at particular stages, clones,
plasmids or transposons could be extremely stable or variable. Variation or stability might result from the relative frequencies of genotypes. Note that for population biologists,
conservatism might be understood as preservation under
selection of the original polymorphic structure of a population, a stable status of variability, i.e. coexistence of variants, resulting from stabilizing frequency-dependent
selection. In contrast, the break of this polymorphic structure results from disruptive frequency-dependent
selection.52
(ii) Simplicity versus complexity. Complexity is favoured by
stability and exploitation configurations, and vice versa;
analogously, simplicity favours exploration and variability.
A stable, long-term maintenance of coexistence among
evolutionary units increases the connectivity and hence
the complexity of the system. For instance, the complexity
of the intestinal microbiota favours the spread of resistance
genes among genetic exchange communities (see later),
assuring the stability of the whole system. In contrast,
organisms not integrated (connected) in these networks
might be better suited for interhost spread of resistance.
In general, the efficiency of acquisition of a unit in a
superior level of the hierarchy depends both on the connectivity and the functional compatibility of such a unit with
the members already integrated in the higher level.
These concepts have been developed in the so-called complexity hypothesis.12,53
(iii) Evolvability versus robustness. Highly variant, simple configurations favour evolvability, whereas more complex,
stable configurations tend to perpetuate themselves by
increasing their inertia (robustness) when confronted with
an environmental change, instead of changing the basic
conformation (increasing evolvability).54 This does not
mean that complex systems are not subject to evolution,
but in general that only occurs to further increase its complexity and robustness, and not to create alternative
configurations.
(iv) The fittest versus the flattest. A high-fitness configuration
promotes a high replication rate in a particular environment. Nevertheless, such exquisite adaptation might be
fragile if the environment changes or the evolutionary
unit is altered by new variation. High mutation rates
1664
(v)
(vi)
(vii)
(viii)
could be extremely beneficial to climb an adaptive hill,
but detrimental when the population stays on the top of
the hill. In contrast, populations colonizing flatter areas in
the fitness landscape reproduce less efficiently, but are
more robust to changes and increase their resilience; i.e.
they are able to recover from deleterious variation.55
Exploration versus exploitation. Entering the field of
exploration, an evolutionary unit might acquire a configuration that is well suited to explore the accessible environment in search of novel niches in which higher fitness is
achievable. In contrast, there is an alternative configuration able to optimally exploit a niche in which the evolutionary unit is installed. ‘Explorer’ configurations are
good for probing; ‘exploiter’ configurations for leveraging,
i.e. obtaining the highest profit from the resource. Variability is favoured by exploration and vice versa. Of course, we
can recognize this familiar distinction at all levels of the
biological and sociological scales.56,57 For instance, in antibiotic resistance, at particular stages, some resistant
clones have high epidemigenicity, whereas others are
well established in particular niches or populations.
Niche deconstruction versus niche construction. Niche construction refers to the effect of biological units modifying
their environments, and consequently modifying local
sources of variation and selection.58,59 In reality, this definition implies both niche construction and niche deconstruction. Niche construction could be understood as the
possibility of modifying a niche in such a way that it
adapts perfectly to the genetic possibilities of an invariant
host, resulting in optimal exploitation of the environment.
An example (related with altruism—see later) is the creation of environments ‘protected’ against antibiotics by
the release of detoxifying substances.60 Niche deconstruction is the conversion of a niche in multiple niches, favouring the emergence of genetic variation in the host, as many
niches might host a multiplicity of variants.
r strategy versus K strategy. All biological units are conformed in different configurations that classify them as
acting as K strategists or r strategists. The origin of such
evolutionary distinctions is rooted in ecological and biogeographical research.61 K strategy refers to the carrying
capacity (the maximal number of individuals who can be
stably maintained in a particular habitat) and maintenance, implying the absence of degradation of such an
environment. K strategists invest energy in ‘quality’, specialized colonization and optimal habitat exploitation, i.e. in
efficiency. In contrast, r strategy refers to the maximal
intrinsic rate of increase of a population, expressing the
individual contribution to population growth. r strategists
invest their energy in ‘quantity’, i.e. in productivity.62 r strategists are ‘explorers’, forced to spread, as obviously r
might change with population size, such as in the case of
density-related effects involving, e.g. exhaustion of
resources in the environment.
Selfishness versus altruism. Altruism tends to preserve unification. A number of examples of altruistic traits can be
found in antibiotic resistance. For instance, a limited
number of cells might harbour traits that decrease their
own fitness but favour the fitness of the surrounding
cells, e.g. by releasing detoxifying enzymes,9,11,63 killing
The Garrod Lecture
themselves to release allelopathic agents64 or by harbouring conjugative plasmids. Altruism enhances the overall
population to overcome stressful environments and, most
importantly, tends to preserve the original structure of
the population.65 Selfishness, assuring that everything
needed for survival is available in a single evolutionary
unit, promotes the possibility of dissemination, even in
low numbers, and so increases potential variability, including drift effects.
Of course, other heuristic ‘pairs’ could be added to complete the
intuition about the evolutionary dimensions that I would like to
suggest to the reader, such as freedom versus determinism or
disorder (entropy) versus order, but I hope that the main
message has already been conveyed.
Contingency and oscillatory dynamics across
evolutionary dimensions
I alluded above to the risk of being simple and the risk of being
complex. We should recall here the Law of Evolution proposed
by Herbert Spencer 120 years ago to explain the behaviour of
all (biological or physical) systems.66 The idea is that aggregated
complex systems are unstable (fragile) in the long run, as they
will be unable to maintain the identity and internal relations of
their parts in the face of external perturbations.67 The question
is what might happen with the elements of the disaggregated
ex-complex systems.
The notion of cycles of contingency might be applied here. This
view is grounded on the developmental systems theory, which considers the ontogeny of a system to be based on contingent cycles
of interaction among elements involved in the construction (or
deconstruction) of the systems.68 In my view, contingency
should be considered in the context of the logical term. That is to
say, the orthogonal configurations considered above may or
may not occur, or they might be convenient (true) or deleterious
(false) depending on the value imposed by their own structure
and the environment in which they are located (valuation). This
means that absolute ‘optimal configurations’ are not expected
to exist in evolutionary units. There is not a single ‘truth’, but
rather optimality is obtained by contingent, transitory configurations oscillating between different optima in one or other
evolutionary dimensions, in the path towards diversification
(ex unibus plurum) or in the alternative path towards unification
(ex pluribus unum). Along all scales of nature, evolutionary entities
are organized not in pure fractality (single configurations), but in
allometry (alternative configurations).69
The concept of cycles of contingency also involves the need
for a transition between configurations along the diversification
path and the unification path. This probably occurs via complex
oscillatory dynamics. A simple approach to this idea, which
takes account of pendular dynamics, is that an evolutionary
unit reaches a critical level of complexity (unification) in which
the advantages of being complex start to be increasingly compromised by increasing disadvantages, and from there starts
the path to simplification (diversification). The opposite is also
true, where a simple configuration, having reached all accessible
advantages of variety and simplicity, might start a complexification (unification) path as the only way to persist or to advance.
JAC
Such presumed cyclical behaviour is probably more
complicated in nature, for two main reasons. First, for each configuration we do not have a single cycle (unification –diversification– unification –diversification). In most cases, evolution might
evolve by hypercycles. Unification steps favour other unification
steps, until the critical level that forces diversification is
reached; by the same token, diversifying steps might facilitate
other diversifying steps, until the need for unification is
imposed on the system. Second, most evolutionary units (certainly in the biology of antibiotic resistance) are embedded in
each other, as species contain clones, clones contain plasmids,
plasmids contain transposons and transposons contain integrons, with each of these units potentially evolving in separate
dimensions. For instance, clones could be under diversification
from each other, while simultaneously the plasmid could be
evolving to better unify with its bacterial host. Coming back to
the pendulum metaphor, we should imagine this situation as a
double oscillator (a second pendulum attached to the end of
the first); if we increase the number of evolutionary units (e.g.
transposons or integrons), we have a multioscillator system, in
which the movement of each subpendulum influences the
movement of the other ones, resulting in the dynamics of the
system becoming almost unpredictable (chaotic).70 Only something like ‘circadian rhythms’ might, to a certain extent, contribute to the periodic harmonization of such complexity.71
The evolutionary attractors for diversification
and unification
Evolutionary attractors are considered here as a finite number of
virtual fields in the evolutionary space providing optimal conditions for evolutionary configurations. Such optimization can
be conceived of as an attractor for the evolution of evolutionary
configurations. As such configurations can be established either
in the ‘ex unibus plurum’ or in the ‘ex pluribus unum’ evolutionary
dimensions, we can distinguish attractors for diversification and
attractors for unification. In both cases, these field attractors act
in orthogonal directions and, therefore, as discussed above (‘Two
alternative and orthogonal evolutionary dimensions’ section), we
have pairs of opposite attractors acting on alternative dimensions, the first being a diversifying attractor and the second a
unifying attractor.
Multiple niche occupancy versus optimal niche
exploitation
Both the occupancy of multiple niches and the optimal exploitation of a particular niche might attract the evolutionary configurations. For instance, diversifying configurations, such as variation,
will be attracted by multiple niche occupancy; in contrast, unifying configurations, such as stability, will be attracted by optimal
niche exploitation.
Exploitation of variable versus stable environments
This is a variation of the former attractor, when a multiplicity of
niches occurs in time rather than in the space. Diversifying configurations will favour the exploitation of variable environments
1665
Baquero
and unifying configurations will favour the occupancy of stable
environments.
Exploitation of low/medium versus high fitness peaks
The notion of fitness peaks was presented by Sewall Wright in
1932 during the Sixth Congress of Genetics in his seminal
lecture on ‘The roles of mutation, inbreeding, crossbreeding
and selection in evolution’. Climbing low/medium fitness peaks
provides fewer advantages than climbing high fitness peaks,
but there are more low/medium peaks and the energy investment required to climb high peaks could be too high for the evolutionary unit. To stay in a high fitness peak is a dangerous
situation, as any small variation will push the individual down
in the slope, such a way as enabling the evolutionary unit to
keep its configurations stable, whereas variable configurations
might overcome smaller risks in lower peaks.
Avoidance of competition versus efficient local defence
Avoidance of competition might attract diversifying-type configurations. For instance, the exploratory or niche-deconstruction
configurations facilitate access to a variety of possible niches,
escaping from those in which competition with others might
occur. In contrast, collective defence against potential invaders
can act as an attractor for configurations in the unifying-type
dimension, such as those related with exploitation or niche
construction.
Resistance to unexpected events versus resistance to
expected deleterious events
Although the development of resistance to expected events is an
attractor for highly specific and efficient protection that occurs in
unifying-type configurations, it might lower the possibilities of
adapting to unexpected challenges. In contrast, broad resistance
to unexpected events might attract diversity-type configurations,
priming quantity versus quality, as in r strategy versus K strategy.
Distribution of costly functions within a consortium
versus individual optimality
The consideration of this pair of attractors reflects the possibility
that a particular attractor might influence alternative configurations. On the one hand, the advantage of distributing costly
functions among differentiated members of a consortium constitutes a field of attraction for diversifying configurations. On the
other hand, such an advantage is only possible when the differentiated members are part of a consortium, a typical case
attracting unifying configurations. Individual optimality obviously
attracts unifying configurations, as the evolutionary unit collects
and optimizes the required function.
Mechanisms of diversification and
mechanisms of unification
An indicative collection of mechanisms by which the evolutionary objects enter in particular configurations under the influence
of evolutionary attractors is listed below. A number of items
1666
listed cannot be considered as separate mechanisms, but
rather as reflecting different topologies of related mechanisms.
However, such overlapping may help the reader to identify the
type of action that evolutionary units perform as part of their
entry and progression in the diversifying or unifying dimensions.
Mechanisms of diversification (ex unibus plurum)
High replication
High replication, and hence high density, is frequently but not
necessarily a pre-condition for diversification. High numbers
assure the emergence of fortuitous changes giving rise to variants or increased connectivity, resulting in the possibility of interactions with other elements that might provide access to novel
traits. High replication is the basis of the diversifying r strategy.
High replication facilitates high dispersal and high transmission.
Interestingly, bacteria tend to increase their replication rate at
concentrations of growth-inhibiting substances that are only
slightly lower than those that prevent multiplication, but this
phenomenon, of potential adaptive interest, has as yet been
scarcely explored.
High dispersal, high transmission
The efficient dissemination of evolutionary units is a condition for
the acquisition of exploratory configurations, attracted by the
possibility of multiple niches occupancy. The result is heterogeneous interactions with multiple environments and, therefore,
an increase of diversity.33 As stated earlier, dispersal reduces
competition and assures local low numbers, and low numbers
increase the emergence of variation by random drift. But, apparently, a neutral character in an evolutionary unit in a particular
environment might be converted into a selectable trait in
another one that is only gained after effective dispersal. Differences in dispersal occur in any evolutionary unit, e.g. hyperdispersible bacterial clones or broad-host-range mobile genetic
elements probably increase the possibilities of diversification.
Hypermutability, recombination
Hypermutability essentially depends on an increase in DNApairing mistakes during replication or translation and/or a
failure in correcting these errors. The expected result is an
increase in variability. Nevertheless, hypermutable bacteria
increase the possibility of compensatory evolution, reducing the
possible cost of mistakes. A similar case occurs for hyperrecombination, which might act either as a generator of variability or as a corrector of changes. But, in general, mutation and
recombination should be allocated among generators of diversification.72,73 Hypermutable organisms are frequently hyperrecombinogenic; in both cases, the increase in variation explains
why these strains are more frequently resistant to antibiotics
than normo-mutable ones. Gene duplication, which provides
the possibility for one of the copies to evolve to a new or
expanded function, is a particularly successful way for diversification, once the ‘Ohno dilemma’ (where selection continues
acting on the gene with the ancient function) is overcome.74
JAC
The Garrod Lecture
Competition
Stress responses
Competition frequently forces the increase of dispersal and the
search for novel niches, which tends to enhance diversity. Competition can be observed at cellular levels as competition for the
acquisition of nutrients (as in the ‘iron wars’), for surface
adhesion and colonization, and, eventually, for the use of allelopathic substances, such as bacteriocins, lantibiotics or microcins.75 Interestingly, diversification is particularly enhanced in
closely related evolutionary units (kin diversification), which
share similar biological needs.
Competition assures diversity.76,77 A new aspect of kin diversification is based on ancestor inhibition by progeny, assuring
transgenerational replacements.78 At subcellular levels, an
example of competition is plasmid incompatibility, where two
plasmids sharing similar or identical replication regions are
unable to coexist in the same cell. The net result is that one or
both plasmids are forced to enter different hosts, which might
increase spread and diversification. A similar case occurs for
transposons, an example being Tn3-like elements that cannot
undergo transposition in genomes carrying similar or identical
transposons (transposition immunity), again forcing the spread
of these evolutionary units.79 – 81 Indeed, diversification might
be driven by anti-unification mechanisms; if the restrictionmodification (RM) systems impede the entry of foreign DNA
(see next section), anti-RM systems providing evasion from this
system are frequently hosted by plasmids, phages or
transposons.82
Stress responses, including those acting under nutritional starvation (or stationary phase), and also stress induced by antimicrobial agents, increase mutability and recombination in
bacterial organisms. For instance, stress-induced up-regulation
of error-prone polymerases, such as DinB (DNA polymerase IV),
might increase mutational variation, resulting in increased
genetic diversification.84 – 86 Stress also increases modularizationbased diversification (see below), e.g. new-host or nutrientshortage stress increase the load of modular mobile elements,
such as insertion sequences (ISs). In general, environmental
stress increases evolutionary potential.83
Heterogeneous environments and gradient diversifying
selection
Diversification is frequently limited not by a shortage of variants,
but rather by a narrow spectrum of selectors able to specifically
recognize (select) a particular variant. Environments with a large
qualitative spectrum of selectors facilitate diversification. Importantly, selectors (such as antibiotics) frequently diffuse forming
concentration gradients, offering different selective compartments along the gradient, able to select a diversity of bacterial
variants.28,29
Niche diversification
Evolutionary units have effects on the environment that they are
able to occupy. There is, in fact, a correspondence between the
diversity of units (e.g. bacterial clones) and the diversity of
environments. In a sense, if there is a phylogeny of units, there
is also a phylogeny of environments.83 One of the important concepts here is niche deconstruction or niche diversification,
meaning that a niche might be subdivided by the action of its
occupant population into different ones, which are able to
sustain a derived polymorphic population composed of different
variants with different specialization. Contrary to the classic principle of periodic selection (see below), multiple variant units
might frequently coexist, because those variants maintain reciprocal activities of dependence.3
Modularization
Modularity is an attribute of a system that can be decomposed
into a set of repeated, conserved cohesive entities that are
loosely coupled. That is to say, they might be present or
absent (in this sense, mobile or accessory) in a particular localization of the genome or metagenome architecture and provide
no significant pleiotropy for the bacteria.87,88 Modularity acts as
an increasing-variability process that adds modular units within
a given local genetic structure ‘isolated’ from the functional
core of evolutionary units; hence, such recruitment occurs
almost without cost. Incremental modularization, the addition
of modules to a particular region, might occur because there is
a ‘module-recruiting’ module (e.g. a recombinase), by duplication
of a pre-existing module or by insertion of an incoming module.
In most of these cases of ‘random linkage between modules’,
modularity increases variability.89 The concept of ‘modular evolution’ was first applied to plasmids containing genes encoding
antibiotic resistance early in the 1980s.90,91 The complete
nucleotide sequences of many clones, plasmids, transposons or
integrons are now available and have confirmed that these
elements frequently have a ‘mosaic’ structure, with zones of
high variability (hot spots) where adaptive modules are collected
without harm. Genetic variation frequently depends on ISs, particularly variation seen in resting cells.92 ISs might act as mechanisms for external gene recruitment and integration in host
replicons (such as the chromosome or plasmids), as well as for
promoting recombination between these units; thus, generating
genomic complexity and diversification. The fact that transposases are the most abundant and most ubiquitous genes in
nature indicates the huge biological importance of modularity.93
Mechanisms of unification (ex pluribus unum)
Selection
Variation tends to accumulate in different individual members of
evolutionary populations in situations where new variants are
neutral or quasi-neutral. Such variation is periodically eliminated
from the population when one particular member acquires an
advantageous trait. In such a case, this successful genotype is
selected at the expense of all others (selective sweep of accumulated variation). Sooner or later, individuals of this genotype will
start to diversify in their turn and again collect variation, and yet
again a new successful genotype will arise to purge the newly
accumulated diversity. This phenomenon is expected to occur
periodically (periodic selection).94 In general, selection increases
1667
Baquero
the density of a particular variant, thereby reducing diversification, as in the case of selection-driven epidemics of particular
clones.
Regulation and reduction in mutation rates
The rate of mutation is severely controlled in most living beings
to maintain self-identity (unicity). As hypermutable asexual
organisms, bacteria are expected to reduce their fitness and ultimately collapse as a consequence of the accumulation of deleterious mutations, a situation known as Muller’s ratchet.95
Thus, the regulation of mutation acts in reducing diversity.
A number of effective mechanisms of DNA proof-reading and
repair act in the same direction, such as the methyl-mismatch
repair system or GO system.96 Also, mechanisms probably exist
to reduce or eliminate endogenous oxidative radicals, which
induce DNA damage, eventually triggered by the presence of
antibiotics.97 Indeed, recombinational events might also
restore mutated sequences, in this case acting to reduce
linkage disequilibrium. In fact, ‘sexual’ evolutionary units that
are able to recombine under genetic and ecological compatibility
(as opposed to ‘celibate’ units) might not only diverge,98 but also
converge, and thus possibly increase unification.18,99
Focused mutation and contingency loci
Concentrating the changes on sequences of adaptive interest
might decrease unspecific variation and costs. The mutational
event is focused on ‘contingency sequences’ that increase variability in the specific type of genes required for adaptation to a
recurrent but not continuous environment.100,101 Under the particular ‘contingency’ requiring a particular bacterial adaptation,
there is a high possibility of obtaining adaptive changes at low
cost, i.e. without changes in other loci, thereby reducing
unwanted diversification. Of course, contingency loci increase
‘local variability’.
promoting unification. Competence is induced in Streptococcus
pneumoniae by strain-specific peptide pheromones called
competence-stimulating peptides (CSPs). The primary structure
of CSPs determines separate pheromone types (pherotypes), so
that all bacteria induced to competence by a particular CSP
belong to the same pherotype.104,105 At least one-half of all plasmids (the hallmark of ‘mobile’ genetic units) are not mobilizable,
and all large plasmids are immobile and kept in domestication
(unification) inside the host cell. Domestication means a stable
relation with their hosts, as some of their genes are functionally
intertwined with the host’s functions. The known plasmid toxin–
antitoxin systems programme the death of the bacterial cell that
has lost a particular plasmid, therefore forcing a coexistence
between plasmid and chromosome replicons (unification).106
Moreover, plasmid conjugation depends on a number of core
genes that have evolved in adaptation to specific genetic, physiological and ecological contexts.107 RM systems are able to
detect specific DNA sequences in foreign (not-kin) DNA, which
will be subsequently cleaved by endo- and exonucleases. Interestingly, these barriers might be relaxed in genetic exchange
communities. Antisense RNA regulation has been observed for
RM systems.108 Bacteria with clustered regularly interspaced
short interspaced repeats (CRISPR) regions carrying copies of
the DNA of previously encountered phage and plasmids abort
the replication of phage and plasmids with these sequences,
thereby reducing connectivity and diversification.80 In mobile
elements, not everything is everywhere.
Modularization
Modularization can act as a diversifying mechanism and also as
a unifying mechanism. When modularization acts to influence
random gene ordering, diversification is expected to occur.
However, when modularization enhances epistatic interactions
between specific loci, optimization of these ‘ensembles’ might
happen (epistatic clustering), therefore increasing unification.89
Stress attenuation
Stress potentially increases genetic variability. The possibility of
surviving a stressful situation without stress should conserve
genetic structure (unicity). Stress avoidance could be solved by
down-regulating stress mechanisms such as alarmones,102 but
that is at the expense of reducing a way out of the dangerous
situation. A more effective way is stress attenuation, such as
that seen during bacterial sporulation or tolerance. One of the
reasons for the wide distribution of toxin –antitoxin systems
among bacterial genomes might be the possibility of a toxinmediated entry in a dormant status providing tolerance of
stress at the expense of slow or no replication.103
Defensive allelopathy and colonization resistance
In the ‘fixed habitat hypothesis’ it has been suggested that in
physically structured habitats, where organisms form discrete
colonies rather than individual cells, the production of allelopathic substances, such as antibiotics, bacteriocins or even bacteriophages, might have a defensive role against competitors.109
What happens in particular organisms also seems to occur for
complex microbial consortia, such as microbiomes, that are
extremely refractory to alien invasion.110
Restrictive horizontal gene transfer
Genetic addiction
Diversification is reduced by mechanisms limiting the spread of
genetic material among evolutionary units. For instance, the
horizontal transfer of DNA by recombination is hampered by
specific competence factors and transforming DNA recognition
sequences. This assures that genetic information is kept within
limited genetic exchange communities that form a kind of unit,
with entry-exclusion and shared-environmental strategies
Genetic addiction assures the stable maintenance of an established complex of evolutionary units by programming the
death of the system (e.g. a cell) if member(s) of the complex
are eliminated. Unification of the complex is assured by this
death penalty. Apparently, genetic addiction is particularly
advantageous in the presence of a competitor genetic
element.111
1668
JAC
The Garrod Lecture
Towards understanding the multilevel
epidemiology of antibiotic resistance
Each one of the units of selection, the evolutionary units of antibiotic resistance, should be taken into account when undertaking
surveillance of resistance. A multilayered epidemiological strategy should be applied to deal with the extreme complexity of
the oscillating pathways governing the evolution of antibiotic
resistance. There is a need for simultaneous characterization of
the resistance genes, the genetic platforms in which they are
located (such as integrons), the transposable elements harbouring such platforms, the plasmids eventually carrying them, and,
at the supracellular level, the bacterial clones, species and communities in which the resistance genes occur.112 – 114 Novel metagenomic high-throughput scanning technologies are awaited for
addressing this type of multilevel epidemiology. The first
approaches to this goal are already available.115 Of course,
new mathematical modelling and bioinformatic approaches
should be applied to deal with this target. Developments in
network theory will enable the construction of edge-weighted
networks describing the diversity within and across different
genetic levels, identified in this theory as ‘genetic worlds’.116
However, our knowledge about quantitative links among different evolutionary units across different levels is still in its
infancy. Probably, we should start by evaluating co-variances in
the different evolutionary units, e.g. by applying methods possibly based on Price’s equation43 to data obtained in synthetic
microcosms both challenged or not with antibiotics.117 It
should be noted that these parameters might randomly fluctuate according to environmental changes118 and, therefore, the
possible descriptions of the evolution of such complex systems
will, at best, be of a probabilistic nature. In this respect, evolutionary graph theories might help in designing evolutionary
games on graphs that might at least provide a list of possible
events influencing the structure of the antibiotic resistance
network.5,119 A totally new public health microbiology able to
address the key problem of prediction of antimicrobial resistance
is urgently awaited,120 and this novel approach to medical microbiology should be based on the application of evolutionary thinking. The objective of this review has been to provide a conceptual
framework for the evolutionary concepts that need to be applied
to the study of antibiotic resistance; thus, helping to pave the
path towards this goal.
Acknowledgements
The author owes a large debt to many conversations held over the years
on the topics covered in this review with Juan Carlos Galán, Bruce Levin,
Teresa Coque, José-Luis Martı́nez, Jesús Blázquez, Susanna Manrubia,
Juan-Carlos Alonso, José-Marı́a González-Alba, Rosa del Campo, Steven
Frank, Andrés Moya, Fernando González-Palacios, Antonio Oliver, Marc
Lipsitch and Rafael Canton. Ana Moreno-Bofarull and Mary Harper
provided substantial help with the manuscript.
Transparency declarations
None to declare.
References
1 Bell G. Selection: The Mechanism of Evolution. New York: Chapman &
Hall, 1997.
2 Stephens C. Selection, drift, and the “forces” of evolution. Phil Sci 2004;
71: 550–70.
3 Kinnersley MA, Holben WE, Rosenzweig F. E unibus plurum: genomic
analysis of an experimentally evolved polymorphism in Escherichia coli.
PLoS Genetics 2009; 11: e1000713.
4 Sykes R. The evolution of antimicrobial resistance: a Darwinian
perspective. J Antimicrob Chemother 2010; 65: 1842 –52.
5 Martı́nez JL, Baquero F, Andersson DI. Predicting antibiotic resistance.
Nat Rev Microbiol 2007; 5: 958– 65.
6 Koonin EV. The Origin at 150: is a new evolutionary synthesis in sight?
Trends Genet 2009; 25: 473– 5.
7 Lewontin RC. The units of selection. Ann Rev Ecol System 1970; 1: 1 –18.
8 Martinez JL, Baquero F. Interactions among strategies associated with
bacterial infection: pathogenicity, epidemicity, and antibiotic resistance.
Clin Microb Rev 2002; 15: 647– 79.
9 Baquero F, Vicente MF, Pérez-Diaz JC. b-Lactam coselection of sensitive
and TEM-1 b-lactamase-producing subpopulations in heterogeneous
Escherichia coli colonies. J Antimicrob Chemother 1985; 15: 151–7.
10 Dugatkin LA, Perlin M, Lucas JS et al. Group-beneficial traits,
frequency-dependent selection and genotypic diversity: an antibiotic
resistance paradigm. Proc Biol Sci 2005; 272: 79 –83.
11 Dugatkin LA, Dugatkin AD, Atlas RM et al. Cheating on the edge. PLoS
One 2008; 3: e2763.
12 Jain R, Rivera MC, Moore JE et al. Horizontal gene transfer accelerates
genome innovation and evolution. Mol Biol Evol 2003; 20: 1598 –602.
13 Skippington E, Ragan MA. Lateral genetic transfer and the
construction of genetic exchange communities. FEMS Microbiol Rev
2011; doi: 10.1111/j.1574– 6976.2010.00261.x.
14 Carlson JM, Doyle J. Complexity and robustness. Proc Natl Acad Sci
USA 2002; 99: 2538– 45.
15 Baquero F, Coque TM, Cantón C. Antibiotics, complexity and evolution.
ASM News 2003; 69: 547– 52.
16 Baquero F. Evolution and the nature of time. Int Microbiol 2005; 8:
81– 91.
17 Darwin C. On the Origin of Species: A Facsimile of the First Edition with
an Introduction by Ernst Mayr. Cambridge: Harvard University Press, 1964;
81.
18 Fraser C, Alm EJ, Polz MF et al. The bacterial species challenge: making
sense of genetic and ecological diversity. Science 2009; 323: 741– 6.
19 Berg OG. Selection intensity for codon bias and the effective
population size of Escherichia coli. Genetics 1996; 142: 1379 –82.
20 Margolis E, Levin BR. Within-host evolution for the invasiveness of
commensal bacteria: an experimental study of bacteremias resulting
from Haemophilus influenzae nasal carriage. J Infect Dis 2007; 196:
1068– 75.
Funding
21 Wilkins JF, Wakeley J. The coalescent in a continuous, finite, linear
population. Genetics 2002; 161: 873–88.
This review is in part funded by the FP7 European Commission PAR
(241476) and EvoTAR (HEALTH 2011) Projects.
22 Ray N, Currat M, Excoffier L. Intra-deme molecular diversity in
spatially expanding populations. Mol Biol Evol 2003; 20: 76–86.
1669
Baquero
23 Thomas VL, McReynolds AC, Shoichet BK. Structural bases for
stability–function tradeoffs in antibiotic resistance. J Mol Biol 2010;
396: 47– 59.
45 Sober E. A philosopher looks at the units of selection problem – a
review of Samir Okasha’s Evolution and the Levels of Selection.
BioScience 2008; 58: 78.
24 Weinreich DM, Delaney NF, Depristo MA et al. Darwinian evolution can
follow only very few mutational paths to fitter proteins. Science 2006;
312: 111–4.
46 McMullin B. Replicators don’t! Adv Artif Life 1995; 929: 158–69.
25 Tomatis PE, Fabiane SM, Simona F et al. Adaptive protein evolution
grants organismal fitness by improving catalysis and flexibility. Proc
Natl Acad Sci USA 2008; 105: 20605–10.
48 Okasha S. Levels of selection. Curr Biol 2010; 20: R306– 7.
26 Trindade S, Sousa A, Karina Bivar X et al. I. Positive epistasis drives the
acquisition of multidrug resistance. PLoS Genetics 2009; 5: e1000578.
47 Okasha S. Recent work of selection problem. Human Nat Rev 2003; 3:
349.
49 Lloyd E. Units and Levels of Selection. The Stanford Encyclopedia of
Philosophy (Winter 2010 Edition), Zalta EN (ed.). http://plato.stanford.
edu/archives/win2010/entries/selection-units/ (26 May 2011, date last
accessed).
27 Huang W, Quyen-Quyen L, LaRocco M et al. Effect to
threonine-to-methionine substitution in position 265 on structure and
function of TEM-1 b-lactamase. Antimicrob Agents Chemother 1994;
38: 2266– 9.
51 Baquero F. From pieces to patterns: evolutionary engineering in
bacterial pathogens. Nat Rev Microbiol 2004; 2: 510– 8.
28 Baquero F, Negri MC. Selective compartments for resistant
microorganisms in antibiotic gradients. Bioessays 1997; 19: 731– 6.
52 Levin BR. Frequency-dependent selection in bacterial populations. Phil
Trans R Soc Lond B 1988; 319: 459–72.
29 Negri MC, Lipsitch M, Blázquez J et al. Concentration-dependent
selection of small phenotypic differences in TEM b-lactamase-mediated
antibiotic resistance. Antimicrob Agents Chemother 2000; 44: 2485– 91.
53 Jain R, Rivera MC, Lake JA. Horizontal gene transfer among genomes:
the complexity hypothesis. Proc Natl Acad Sci USA 1999; 96: 3801– 6.
30 Nei M, Suzuki Y, Nozawa M. The neutral theory of molecular
evolution in the genomic era. Annu Rev Genomics Hum Genet 2010; 22:
265–89.
31 Sangster TA, Salathia N, Undurraga S et al. HSP90 affects the
expression of genetic variation and developmental stability in
quantitative traits. Proc Natl Acad Sci USA 2008; 105: 2963– 8.
32 Garcia-Descalzo L, Alcázar A, Baquero F et al. Identification of in vivo
HSP90-interacting proteins reveals modularity of HSP90 complexes is
dependent on the environment in psychrophilic bacteria. Cell Stress
Chaperon 2011; 16: 203–18.
33 Perron GG, González A, Buckling A. Source –sink dynamics shape the
evolution of antibiotic resistance and its pleiotropic fitness cost. Proc R
Soc B 2007; 274: 2351– 6.
34 López-Maury L, Marguerat S, Bähler J. Tuning gene expression to
changing environments: from rapid responses to evolutionary
adaptation. Nat Rev Genet 2008; 9: 583–93.
35 Fraser D, Kaern M. A chance at survival: gene expression noise and
phenotypic diversification strategies. Mol Microbiol 2009; 71: 1333 –40.
36 Snell-Rood EC, Van Dyken JD, Cruickshank T et al. Toward a population
genetic framework of developmental evolution: the costs, limits, and
consequences of phenotypic plasticity. Bioessays 2010; 32: 71– 81.
37 Singer RS, Ward MP, Maldonado G. Can landscape ecology untangle the
complexity of antibiotic resistance? Nature Rev Microbiol 2006; 4: 943–52.
50 Baquero F, Cantón R. Evolutionary biology of drug resistance. In: Mayers
DK, ed. Antimicrobial Drug Resistance. Humana Press, 2009; 1: 9 –32.
54 Lenski RE, Barrick JE, Ofria C. Balancing robustness and evolvability.
PLoS Biology 2006; 4: 2190 –2.
55 Codoñer FM, Daros JA, Sole RV et al. The fittest versus the flattest:
experimental confirmation of the quasispecies effect with subviral
pathogens. PLoS Pathogens 2006; 2: 1187– 93.
56 Bierly P, Chakrabarti A. Generic knowledge strategies in the U.S.
pharmaceutical industry. Strat Management J 1996; 17: 123– 35.
57 von Krogh G, Nonaka I, Aben M. Making the most of your company
knowledge: a strategic framework. Long Range Planning 2001; 34: 421–39.
58 Laland KN, Odling-Smee FJ, Feldman MW. The evolutionary
consequences of niche construction: a theoretical investigation using
two-locus theory. J Evol Biol 1996; 9: 293–316.
59 Laland KN, Odling-Smee FJ, Feldman MW. Evolutionary consequences
of niche construction and their implications for ecology. Proc Natl Acad Sci
USA 1999; 96: 10242–7.
60 Boni MF, Feldman MW. Evolution of antibiotic resistance by human
and bacterial niche construction. Evolution 2005; 59: 477– 91.
61 MacArthur RH, Wilson EO. The Theory of Island Biogeography.
Princeton: Princeton University Press, 1967.
62 Pianka ER. On r- and K-selection. Amer Naturalist 1970; 104: 592– 7.
63 Clark DR, Alton TM, Bajorek A et al. Evolution of altruists and cheaters in
near-isogenic populations of Escherichia coli. Front Biosci 2009; 14:
4815– 24.
38 Otha T. Near neutrality in evolution of genes and gene regulation.
Proc Natl Acad Sci 2002; 99: 16134–7.
64 Cornejo OE, Rozen DE, May RM et al. Oscillations in continuous culture
populations of Streptococcus pneumoniae: population dynamics and the
evolution of clonal suicide. Proc R Soc B 2009; 276: 999– 1008.
39 Otha T. Development of neutral and nearly neutral theories. Theor
Popul Biol 1996; 49: 128–42.
65 Lee HH, Molla MN, Cantor CR et al. Bacterial charity work leads to
population-wide resistance. Nature 2010; 467: 82 –6.
40 Hughes AL. Near neutrality: leading edge of the neutral theory of
molecular evolution. Ann NY Acad Sci 2008; 1133: 162– 79.
66 Spencer H. First Principles, 5th edn. London: William and Norgate,
1890.
41 Szathmàry E. The emergence, maintenance, and transitions of the
earliest evolutionary units. Oxford Surv Evol Biol 1989; 6: 169– 205.
67 McShea DW. Complexity and evolution: what everybody knows. In:
Hull DM, Ruse M, eds. The Philosophy of Biology. Oxford: Oxford
University Press, 1998; 625– 49.
42 Norman A, Hansen LH, Sorensen SJ. Conjugative plasmids: vessels of
the communal gene pool. Phil Trans R Soc Lond B Biol Sci 2009; 364:
2275– 89.
68 Oyama S, Griffiths PE, Russell GD, eds. Cycles of Contingency:
Developmental Systems and Evolution. London: The MIT Press, 2001; 377.
43 Okasha S. Evolution and the Levels of Selection. Oxford and New York:
Clarendon Press, 2006.
69 Gould SJ. The Structure of the Evolutionary Theory. Cambridge, MA &
London: The Beltknap Press of Harvard University Press, 2002; 677.
44 Steven AF. Foundations of Social Evolution. Princeton: Princeton
University Press, 1998.
70 Lai YC. Unpredictability of the asymptotic attractors in phase-coupled
oscillators. Phys Rev E 1995; 51: 2902 –8.
1670
The Garrod Lecture
71 Bell-Pedersen D, Cassone VM, Earnest DJ et al. Circadian rhythms from
multiple oscillators: lessons from diverse organisms. Nat Rev Genet 2005;
6: 544–56.
72 Blázquez J. Hypermutation as a factor contributing to the acquisition
of antimicrobial resistance. Clin Infect Dis 2003; 37: 1201–9.
73 Levin B, Cornejo OE. The population and evolutionary dynamics of
homologous gene recombination in bacterial populations. PLoS Genet
2009; 5: e1000601.
74 Bergthorsson U, Andersson DI, Roth JR. Ohno’s dilemma: evolution of
new genes under continuous selection. Proc Natl Acad Sci USA 2007; 104:
17004– 9.
JAC
93 Aziz RK, Breitbart M, Edwards RA. Transposases are the most abundant,
most ubiquitous genes in nature. Nucleic Acids Res 2010; 38: 4207–17.
94 Atwood KC, Schneider LK, Ryan FJ. Selective mechanisms in bacteria.
Cold Spring Harb Symp Quant Biol 1951; 16: 345–55.
95 Muller HJ. The relation of recombination to mutational advance.
Mutat Res 1964; 106: 2– 9.
96 Sanders LH, Sudhakaran J, Sutton MD. The GO system prevents
ROS-induced mutagenesis and killing in Pseudomonas aeruginosa. FEMS
Microbiol Lett 2009; 294: 89–96.
97 Dwyer DJ, Kohanski MA, Collins JJ. Role of reactive oxygen species in
antibiotic action and resistance. Curr Opin Microbiol 2009; 12: 482–9.
75 Baquero F, Moreno F. The microcins. FEMS Microbiol Lett 1984; 23:
117–24.
98 Enright MC, Spratt BG. The genomic view of bacterial diversification.
Science 2011; 331: 407– 9.
76 Kirkup BC, Riley MA. Antibiotic-mediated antagonism leads to a
bacterial game of rock–paper–scissors in vivo. Nature 2004; 428: 412– 4.
99 Deschamps C, Clermont O, Hipeaux MC et al. Multiple acquisitions of
CTX-M plasmids in the rare D2 genotype of Escherichia coli provide
evidence for convergent evolution. Microbiology 2009; 155: 1656– 68.
77 Linares JF, Gustafsson I, Baquero F et al. Antibiotics as intermicrobial
signaling agents instead of weapons. Proc Natl Acad Sci USA 2006; 103:
19484– 9.
78 Lemmonier M, Baquero F. Generational coexistence and ancestor’s
inhibition in bacterial populations. FEMS Microbiol Rev 2009; 33: 958–67.
79 Bergstrom CT, Lipsitch M, Levin BR. Natural selection, infectious
transfer and the existence conditions for bacterial plasmids. Genetics
2000; 155: 1505– 19.
80 Levin BR. Nasty viruses, costly plasmids, population dynamics, and
the conditions for establishing and maintaining CRISPR-mediated
adaptive immunity in bacteria. PLoS Genet 2010; 6: e1001171.
100 Moxon R, Rainey PB, Nowak MA et al. Adaptive evolution of highly
mutable loci in pathogenic bacteria. Curr Biol 1994; 4: 24– 33.
101 Moxon R, Bayliss C, Hood D. Bacterial contingency loci: the role of
simple sequence DNA repeats in bacterial adaptation. Annu Rev Genet
2006; 40: 307–33.
102 Rowbury RJ. Cross-talk involving extracellular sensors and
extracellular alarmones gives early warning to unstressed Escherichia
coli of impending lethal chemical stress and leads to induction of
tolerance responses. J Appl Microbiol 2001; 90: 677–95.
81 Bouet JY, Nordström K, Lane D. Plasmid partition and
incompatibility—the focus shifts. Mol Microbiol 2007; 65: 1405 –14.
103 Fung DK, Chan EW, Chin ML et al. Delineation of a bacterial starvation
stress response network which can mediate antibiotic tolerance
development. Antimicrob Agents Chemother 2010; 54: 1082– 93.
82 McMahon SA, Roberts GA, Johnson KA et al. Extensive DNA mimicry by
the ArdA anti-restriction protein and its role in the spread of antibiotic
resistance. Nucleic Acids Res 2009; 37: 4887– 97.
104 Havarstein LS. Increasing competence in the genus Streptococcus.
Mol Microbiol 2010; 76: 541–4.
83 Baquero F. Environmental stress and evolvability in microbial systems.
Clin Microbiol Infect Dis 2009; 15 Suppl S1: 5 – 10.
105 Claverys JP, Martin B. Bacterial ‘competence’ genes: signatures of
active transformation, or only remnants. Trends Microbiol 2003; 11:
161–5.
84 Petrosino JF, Galhardo RS, Morales LD et al. Stress-induced b-lactam
antibiotic resistance mutation and sequences of stationary-phase
mutations in the Escherichia coli chromosome. J Bacteriol 2009; 191:
5881–9.
106 Cooper TF, Paixão T, Heinemann JA. Within-host competition selects
for plasmid-encoded toxin– antitoxin systems. Proc Biol Sci 2010; 277:
3149– 55.
85 Galhardo RS, Do R, Yamada M et al. DinB upregulation is the sole role
of the SOS response in stress-induced mutagenesis in Escherichia coli.
Genetics 2009; 182: 55– 68.
86 Pérez-Capilla T, Baquero MR, Gómez-Gómez JM et al. SOS-independent
induction of dinB transcription by b-lactam-mediated inhibition of cell wall
synthesis in Escherichia coli. J Bacteriol 2005; 187: 1515– 8.
87 Baquero F. Modularization and evolvability in antibiotic resistance. In:
Baquero F, Nombela C, Casell GH et al., eds. Evolutionary Biology of
Bacterial and Fungal Pathogens. Washington DC: ASM Press, 2008;
233–47.
88 Rainey PB, Cooper TF. Evolution of bacterial diversity and the origins of
modularity. Res Microbiol 2004; 155: 370–5.
89 Pepper JW. The evolution of evolvability in genetic linkage patterns.
BioSystems 2003; 69: 115–26.
90 Labigne-Roussel A, Witchitz J, Courvalin P. Modular evolution of
disseminated Inc 7-M plasmids encoding gentamicin resistance.
Plasmid 1982; 8: 215–31.
91 Courvalin P. The Garrod Lecture. Evasion of antibiotic action by
bacteria. J Antimicrob Chemother 1996; 37: 855–69.
92 Naas T, Blot M, Fitch WM et al. Insertion sequence-related genetic
variation in resting Escherichia coli K12. Genetics 1994; 136: 721–30.
107 Smille C, Garcillan-Barcia MP, Francia MV et al. Mobility of plasmids.
Microbiol Mol Biol Rev 2010; 74: 434–52.
108 Mruk I, Liu Y, Ge L et al. Antisense RNA associated with biological
regulation of a restriction-modification system. Nucleic Acids Res 2011;
1 –11. doi:10-1093/nar/gkr166.
109 Chao L, Levin BR. Structured habitats and the evolution of
anti-competitor toxins in bacteria. Proc Natl Acad Sci USA 1981; 78: 6324–8.
110 Stecher B, Chaffron S, Käppeli R et al. Like will to like: abundances of
closely related species can predict susceptibility to intestinal colonization
by pathogenic and commensal bacteria. PLoS Pathog 2010; 6: e1000711.
111 Mochizuki A, Yahara K, Kobayashi I et al. Genetic addiction: selfish
gene’s strategy for symbiosis in the genome. Genetics 2006; 172:
1309– 23.
112 Curiao T, Morosini MI, Ruiz-Garbajosa P et al. Emergence of
blaKPC-3-Tn4401a associated with a pKPN3/4-like plasmid within ST384
and ST388 Klebsiella pneumoniae clones in Spain. J Antimicrob
Chemother 2010; 65: 1608 –14.
113 Novais A, Baquero F, Machado E et al. International spread and
persistence of TEM-24 is caused by the confluence of highly
penetrating Enterobacteriaceae clones and an IncA/C2 plasmid
containing Tn1696::Tn1 and IS5075-Tn21. Antimicrob Agents
Chemother 2010; 54: 825– 34.
1671
Baquero
114 Freitas AR, Coque TM, Novais C et al. Human and swine hosts share
vancomycin-resistant Enterococcus faecium CC17, CC5, and Enterococcus
faecalis CC2 clonal clusters harboring Tn1546 on indistinguishable
plasmids. J Clin Microbiol 2011; 49: 925–31.
115 Jones BV, Sun F, Marchesi JR. Comparative metagenomic analysis of
plasmid-encoded functions in the human gut microbiome. BMC Genomics
2010; 19: 11– 46.
116 Halary S, Leight JW, Cheaib B et al. Network analysis structure
genetic diversity in independent genetic worlds. Proc Natl Acad Sci USA
2010; 107: 127–32.
1672
117 Hu B, Du J, Zou RY et al. An environment-sensitive synthetic
microbial ecosystem. PloS One 2010; 5: e10619.
118 DeWitt TJ, Yoshimura J. The fitness threshold model: random
environmental change alters adaptive landscapes. Evol Ecol 1998; 12:
615–26.
119 Lieberman E, Hauert C, Nowak MA. Evolutionary dynamics on
graphs. Nature 2005; 433: 312– 6.
120 Baquero F, Cantón R, Cornaglia G. Public health microbiology, a
challenge for Europe. Clin Microbiol Infect 2010; 16: 123– 5.