Está en la página 1de 8

PERSPECTIVES

drug development and that an evaluation of


OPINION
the potential for antibiotic resistance to be
acquired through HGT should be included
Prediction of antibiotic resistance: inearly preclinicalanalysis.

time for a new preclinical paradigm? Evolution of antibiotic resistance


Assessment of the evolution of resistance
is a key step in the preclinical development
Morten O.A.Sommer, Christian Munck, Rasmus Vendler Toft-Kehler of new antibiotics. Resistance can develop
and Dan I.Andersson through two distinct paths: vertical evolution,
whereby mutations that enhance antibiotic
Abstract | Predicting the future is difficult, especially for evolutionary processes tolerance are selected for and are transmitted
that are influenced by numerous unknown factors. Still, this is what is required of to the progeny (FIG.1a); orhorizontal
drug developers when they assess the risk of resistance arising against a new evolution, whereby antibiotic resistance genes
antibiotic candidate during preclinical development. In this Opinion article, we are acquired from other bacteria, through
conjugation, transduction or transformation,
argue that the traditional procedures that are used for the prediction of antibiotic and are subsequently transmitted to the
resistance today could be markedly improved by including a broader analysis of progeny (FIG.1b). The relative contribution
bacterial fitness, infection dynamics, horizontal gene transfer and other factors. of each of these evolutionary paths to the
This will lead to more informed preclinical decisions for continuing or discontinuing development of clinical resistance is not
the development of drug candidates. yet fully elucidated, but the comprehensive
genomic characterization of human
Antibiotic resistance is increasingly the clinic may be wrongly dismissed, or pathogens during past decades has shown
threatening our ability to treat bacterial antibiotics that may be rendered useless that horizontal evolution contributes to the
infections and to carry out various medical by the rapid acquisition of resistance majority of current resistance problems2.
treatments and procedures, including genes could proceed through the drug The evolution, transmission and
chemotherapy, transplants and surgery, development process. maintenance of antibiotic resistance in
which require effective antibiotics to reduce In this Opinion article, we describe the a population of bacteria are driven by
complications. The emergence of antibiotic- various factors that influence the evolution the complex interplay of several factors,
resistant bacteria limits the clinical use of of antibiotic resistance and discuss these including the rate at which resistance
antibiotics and, as resistant bacteria become factors in the context of early drug discovery, genes and mutations arise, the level of
more prevalent, there is an increasing with the aim of encouraging an updated resistance that is conferred by the particular
concern that existing antibiotics will paradigm for assessing the evolution of resistance mechanism, the relative fitness
become ineffective against these pathogens1. resistance during the development of new of the resistant mutant at different drug
Fortunately, awareness is increasing, and antibiotics. We propose that the use of concentrations and the strength of various
actions to tackle antibiotic resistance invitro mutation rates as a predictor of the selective pressures that drive the evolution
andremedy the funding gaps and market evolution of resistance should be replaced by of resistance (FIG.2). In addition, epistatic
inefficiencies are discussed at international a broader analysis of the major factors that interactions (in which the effect of one gene
levels. In addition, initiatives are being taken can influence the evolution of resistance, locus is influenced by the presence of one
to improve diagnostics and to promote the including bacterial fitness (that is, the relative or more modifier genes), the evolution of
development of new antibiotics that have a ability to survive, reproduce and propagate compensatory mutations, the development
more extended clinical lifetime1. in anenvironment), infection dynamics, of cross-resistance, and coselection of
During the preclinical evaluation of new cross-resistance (that is, when resistance genes that induce drug resistance or drug
antibiotic candidates, the development of to one drug confers resistance to other sensitivity with other genes, can influence
resistance is assessed invitro with the aim drugs), coselection and horizontal gene the evolution of resistance. In this section,
of early detection and termination of drug transfer (HGT). This updated preclinical webriefly describe these processes
candidates with which resistance develops paradigm challenges established procedures and discuss how they contribute to the
too quickly. However, research into the and, by using additional risk-assessment development of antibiotic resistance.
specific factors that drive the evolution of criteria, itmay provide better predictions
antibiotic resistance in the laboratory and in to determine whether a drug candidate Basic factors. The rate at which resistance
the clinical setting has not been adequately should advance to clinical development. In genes and mutations arise (also known as
incorporated into preclinical drug-testing particular, we propose that drug candidates the mutation supply rate) is determined
programmes. Consequently, drug candidates that show a high rate of formation of resistant by population sizes, rates of mutation and
that have the potential to be successful in mutants should not be discarded too early in HGT. The degree of genetic heterogeneity in

NATURE REVIEWS | MICROBIOLOGY ADVANCE ONLINE PUBLICATION | 1



2
0
1
7
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
,
p
a
r
t
o
f
S
p
r
i
n
g
e
r
N
a
t
u
r
e
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
PERSPECTIVES

a Vertical evolution b Horizontal evolution in the absence of antibiotic, which is typically


analysed by measuring growth rates under
Plasmid different conditions3,2123. However, there are
also exceptions in which resistance seems
Bacteria Conjugation neutral or beneficial to the fitness of the
drug-resistant bacterium21,24. On the basis
of the limited studies that have been carried
Transformation out, it seems likely that measurements
of bacterial fitness in the laboratory can
Recipient be used as reasonable predictors for the
cell
evolutionary success of antibiotic-resistant
Phage bacteria in clinical settings. Indeed, the
injecting most common mutations providing
DNA rifampicin and aminoglycoside resistance
Mutation
Transduction to clinical isolates of Mycobacterium
tuberculosis and Staphylococcus aureus are
associated with the lowest fitness costs2529.
Furthermore, the fitness cost of a resistance
Figure 1 | Evolution of resistance. Resistance can evolve through two basic mechanisms: vertical
Nature Reviews | Microbiology mechanism might vary between species
evolution and horizontal evolution. a|Vertical evolution represents the emergence of denovo muta-
tions in the bacterial genome and their subsequent transfer to the progeny. Blue cells are antibiotic- and influence the spread of resistance (for
sensitive bacteria, whereas red cells are resistant to antibiotics. b|Horizontal acquisition of resistance example, resistance to vancomycin). Indeed,
genes, known as horizontal gene transfer, can instead occur through phage transduction, conjugation in enterococci, the fitness cost is very low
or natural transformation. Phage transduction is the process by which a bacteriophage attaches to a and antibiotic resistance dissemination is
recipient bacterial cell and injects its viral DNA, which is then integrated into the chromosome. Bacterial worldwide, whereas in S.aureus the fitness
conjugation occurs through direct contact between the donor and recipient bacterial cells, and medi- cost is high and very little spread has been
ates the acquisition of conjugative plasmids and integrative conjugative elements. Natural transforma- observed30. Thus, fitness costs seem to
tion occurs when naked DNA is released by lysing donor cells and is subsequently taken up by the inversely correlatewith successful spread
recipient bacterial cell. All of these genetic processes mediate the acquisition of mutations and/or
inclinicalsettings.
genes (represented by red DNA tracts) that confer antibiotic resistance to the r ecipient bacterium. Blue
Bacteria in different environments are
cells are recipient bacteria, whereas red cells are donor bacteria.
exposed to a range of various antimicrobial
agents with time-variable and space-variable
a bacterial population in humans is largely resistance levels. For some resistance concentrations and, consequently, the
influenced by the rate at which mutations mechanisms, bacterial growth is unaffected strength of selection in natural settings is
arise and potential bottlenecks during by the use of increasing concentrations of difficult to assess. Furthermore, as shown
intra-host or inter-host transmission3. Our antibiotic, whereas for other mechanisms, in recent invitro evolution experiments, the
knowledge of bacterial population sizes growth decreases progressively until the outcome of evolution varies depending on
in infected hosts is limited but, in cases in minimal inhibitory concentration (MIC) whether it occurs at drug concentrations
which they are known, there generally seem is reached1418. A potential complication of that inhibit pathogen growth (>MIC) or that
to be >108 bacteria per infected host per such measurements is the presence of cases allow the growth of both susceptible and
infection site48. As spontaneous mutation of resistance that is only induced in response resistant bacteria (<MIC)31. Thus, the rate
rates in bacteria are typically around 1010 per to the antibiotic (also known as inducible of emergence of mutations and the type of
nucleotide per generation, it is estimated that resistance)19. Therefore, the relative fitness mutants that are selected for during evolution
resistant mutants are likely to be present in of resistant bacteria might be unaffected or differ between >MIC (lethal) and <MIC
the infecting population911. might vary depending on the concentration (non-lethal) selective pressures. Indeed,
The level of resistance that develops of antibiotic20. Furthermore, several recent at high lethal antibiotic concentrations,
can vary depending on the resistance studies have shown that resistance may be rare mutations that provide a high level of
mechanism and the conditions under which dependent on environmental conditions resistance in a single genetic event have a high
resistance is measured. Some resistance (for example, through collective bacterial probability of being selected and transmitted
mechanisms confer resistance levels that interactions or the presence of specific to progeny, whereas, at non-lethal antibiotic
exceed the solubility limit of the antibiotic12, metabolites or growth conditions that can concentrations, the selected mutants often
whereas other resistance mechanisms modulate and alter the phenotypic expression result from an accumulation of multiple
result in small increases in resistance that of a specific resistance mechanism20,21). mutations that, when combined, result in
can only be detected with sensitive time As shown by theoretical work, a high level of resistance14,3133. Importantly
kill experiments that measure the rate at epidemiology and laboratory experiments, and paradoxically, the weaker the selection
which bacteria are killed during exposure the relative fitness of a drug-resistant for resistance the stronger the enrichmentfor
to an antibiotic13. Measuring the levels of bacterium, both in the absence and more problematic mutants that have a low
resistance is complicated by the nonlinear presence of the drug, is a key parameter fitness cost and mutator phenotypes (that
shape of the pharmacodynamic response in determining its evolutionary success. is, an increased mutation rate owing to
curve (that is, bacterial growth as a function Most experimentally examined resistance defects in DNA repair) that increase the risk
of drug concentration) and how bacterial mechanisms result in decreased bacterial of successful spread and further resistance
growth conditions affect the measured fitness compared with a susceptible ancestor evolution, respectively 31.

2 | ADVANCE ONLINE PUBLICATION www.nature.com/nrmicro



2
0
1
7
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
,
p
a
r
t
o
f
S
p
r
i
n
g
e
r
N
a
t
u
r
e
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
PERSPECTIVES

Additional key factors. Epistasis can also High


affect the outcome of the evolution of Susceptible strain
antibiotic resistance34,35. Epistasis makes
it more challenging to predict potential
High-risk space
antibiotic resistance phenotypes after
mutation or HGT of genes into new
genetic environments, and a greater
understanding of the details of epistatic
interactions could lead to better predictions

Relative tness
of evolutionary outcomes and the resultant
phenotypes. Epistatic interactions could Moderate
affect the level of resistance or the relative
fitness of a resistantmutant. For example,
invitro evolution experiments show that
epistasis can decrease the fitness costs
of mutations and confer resistance to
different classes of antibiotics, including
streptomycin, rifampicin, fusidic acid Low-risk space
and ciprofloxacin3645. It has been shown
that epistatic interactions may constrain Clinical breakpoint for resistance
and direct the order in which antibiotic
Low
resistance mutations arise39,4552. Similarly,
Low Moderate High
epistasis has a role in maintaining plasmids
Relative resistance level
that confer antibiotic resistance in bacteria.
Typically, when a plasmid enters a naive Figure 2 | Factors that determine the risk of evolution of antibiotic resistance. Each circle represents
Nature Reviews | Microbiology
host, it decreases its fitness; however, a hypothetical mutation or a horizontal gene transfer (HGT) event that leads to increased tolerance to a
these costs can be reduced by subsequent given antibiotic. The size of the circles represents the rate by which mutations or HGT events occur
(corresponding to invivo rate population size). Their position on the yaxis reflects the relative fitness
genetic mutations in the plasmid or host
imposed by the genetic change (classified as low, moderate or high), whereas the position on the xaxis
chromosome, or a combination of both, reflects the level of resistance conferred (classified as low, moderate or high). The blue and red shading
thereby enhancing plasmid stability and indicate low and high-risk spaces with regard to development of antibiotic resistance. AHGT or mutation
persistence5358. A specific variant of event in the red space is more likely to get fixed in a population in the presence of antibiotic as the
epistasis known as collateral sensitivity, imposed fitness cost is relatively low and the conferred increase in tolerance is high. The vertical dashed
in which the development of resistance line indicates the clinical breakpoint for resistance, which is the antibiotic concentration that defines
to one antibiotic is associated with higher whether a bacterial species is predicted to be treatable with the antibiotic.
susceptibility to another antibiotic, can also
influence resistance levels5962. Collateral
sensitivity seems to be a relatively common intervention studies in which a reduction Current assessments are limited
phenomenon; however, in most cases, the in the clinical use of a particular class of In the development of new antibiotics,
mechanisms remain poorly understood6367. antibiotics did not result in a decrease the current preclinical assessments of
Another complicating factor in in the frequency of that resistance in the resistance largely focus on the acquisition
predicting resistance evolution, which is community 7173. Generally, coselection of resistanceby denovo mutations in
difficult to quantify, is the coselection of makes prediction very challenging, as the conjunction with assessing the ability of
one resistance gene with another resistance evolutionary success of the acquisition of the drug candidate to kill MDR pathogens.
gene owing to their physical linkage. a resistance gene or a mutation is not only Studies of the evolution of antibiotic
Coselection could occur when two or more influenced by the level of resistance and resistance are typically limited to determining
genes are present on the same chromosome, the fitness it confers but also by its genetic the rate of acquisition of spontaneous
plasmid, transposon or integron. Indeed, context and any selective pressures that act mutations that confer resistance for a specific
most antibiotic resistance plasmids carry on the neighbouring genes. Furthermore, drug candidate and to determining the overall
multiple antibiotic resistance genes and a cross-resistance mechanisms (for example, potential evolution of resistance through
consequence of this genetic linkage is that porin mutations or the activation of repeated exposure of a bacterial population
selection for one of the antibiotic genes antibiotic efflux systems) that confer to constant or increasing concentrations of
then selects for other linked resistance resistance within and between different the drug candidate. These approaches are
genes6870. Therefore, coselection could antibiotic classes or biocides could also beneficial in that they provide a well-defined
decrease the effect of purifying selection complicate predictions by influencing readout of the potential genetic mutations
(that is, selection against mutations and/or resistance levels to many different drugs. that can confer resistance towards a particular
genes that are deleterious) on a resistance Thus, as most pathogens are exposed to drug. However, as outlined above, many
genethat is associated with a high fitness multiple antibiotics, simultaneously or other factors influence the evolution of
cost and, consequently, it could disassociate sequentially, bacteria that are multidrug resistance and therefore these approaches are
the development of resistance to a particular resistant (MDR) will generally be the too simplistic. We propose considering other
antibiotic from the use of that antibiotic. most successful in terms of survival factors that could be of greater importance in
Such disassociation is supported by andmultiplication. predicting future clinical resistance.

NATURE REVIEWS | MICROBIOLOGY ADVANCE ONLINE PUBLICATION | 3



2
0
1
7
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
,
p
a
r
t
o
f
S
p
r
i
n
g
e
r
N
a
t
u
r
e
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
PERSPECTIVES

7.5
cannot be enriched because they grow too
Streptomycin (aminoglycoside)
slowly to remain in the bladder during
Ciprooxacin (quinolone) antibiotic exposure owing to continuous
urine production and normal urination
Time to acquire resistance (years)

Ampicillin (-lactam)
5.0
Prontosil (sulfadrug) patterns. A significant implication of this
is that other parameters, that is, the fitness
Chloramphenicol (phenicol)
costs of resistance, the pharmacodynamic
Erythromycin (macrolide) response and the infection dynamics, are
2.5
Chlortetracycline (tetracycline) more relevant for evaluating the risks of
Rifampicin (rifamycin) developing antibiotic resistance.
Another reason for why current
0.0 Linezolid (oxazolidinone)
assessments of resistance are limited
is because the majority of resistance
mechanisms that are observed in pathogens
2.5 are conferred by HGT of pre-existing
1010 109 108 107 106 105 genes that mediate antibiotic resistance,
Spontaneous resistance mutation (rate per cell per generation) and invitro measurements of the rates
of emergence of mutations cannot
Figure 3 | The invitro rate of resistance mutations does not correlateNature
with the burden
Reviews of resistance
| Microbiology inform us of the antibiotic resistance
in the clinic. Time to first observed resistance from first use of the drug is plotted in the yaxis as a
that is linked to HGT. HGT enables
function of the rate of emergence of spontaneous resistance mutations shown in the xaxis. Time is
shown in years, whereas the rate of mutation emergence is shown as the number of mutations per cell distantly related organisms to exchange
per generation. Note, that for some drugs, emergence of resistance was reported even before its clinical genetic material77 and can be mediated
use. The data points are slightly jittered on both axes to avoid over-plotting. Data on mutation rate was through three mechanisms: transduction,
obtained from REFS15,16,76,9096 and data on time to resistance was obtained from REF.97. transformation and conjugation. The
potential for a pathogen to acquire a
resistance gene through HGT is dependent
There are two main reasons that make of emergence of mutations that confer on the density and the genetic complexity
the prediction of antibiotic resistance based antibiotic resistance are very high (107 of the bacterial community in which it
on the rate of acquisition of mutations of to 105 per cell per generation). Owing to resides and on the selective pressures that
limited use. First, the methods that are this high number of mutations in different are present. For pathogens that typically
currently used for assessing the probability genes, and the resulting high mutational inhabit high-density and high-complexity
of emergence of antibiotic resistance by rate, one would assume that resistance to environments, such as the gastrointestinal
mutation do not typically consider bacterial these antibiotics should rapidly emerge16. tract, HGT is likely to be a significant
population sizes and within-patient In urinary tract infections, the population contributor in the acquisition of genes
dynamics, and they thereby overemphasize sizes range from 104 to 108 bacteria per ml that confer antibiotic resistance. Indeed,
the importance of mutation rates in the of urine, resulting in total population sizes most of the clinically challenging antibiotic
emergence of antibiotic resistance genes. approaching 1010 in the full bladder 15. With resistance genes that are found in common
Indeed, invitro mutation rates that were such large population sizes and great sizeof pathogens, such as lactam resistance
determined for our current repertoire of mutational targets, the expected number in S.aureus and Enterobacteriaceae and
antibiotics span at least six logs (from ~1011 of resistant mutants in the population vancomycin resistance in enterococci,
to 105 per cell per generation) at lethal drug might be as high as 105 (population size are horizontally acquired. The genes that
concentrations (>MIC)16,74. Moreover, the 1010 mutation rate 105) when treatment confer these resistance phenotypes have
rate by which resistance has appeared in is initiated, and mathematical modelling been successful in their spread across the
clinical settings is not obviously correlated suggests that resistance would develop world, as highlighted by the rapid global
to the mutation rates that are observed very often during treatment 15,16,75,76. spread of a specific class of lactamases:
invitro (FIG.3). This finding could imply However, the use of these antibiotics is the CTXMtype extended spectrum
that, for most pathogens, the probability of rarely associated with the development lactamase (ESBL) gene family 78.
selection and fixation of a resistant mutant of antibiotic resistance during treatment. Environmental metagenomic studies have
in a patient is not strongly limited by the rate This can be fully explained by the fitness found that genes that confer antibiotic
of acquisition of a mutation. A reasonable costs that are associated with the resistance resistance are prevalent in all environments,
explanation for this is that resistant mutants mechanisms, the pharmacodynamic which suggests that the evolution of
are generally present in the infecting response and the dynamics of the bladder new resistance genes is not limited by
bacterial population when treatment is environment. Indeed, although resistant the availability of functional antibiotic
initiated, and therefore the time necessary mutants continually appear at a high rate, resistance genes7982. Instead, factors such
for a resistant mutant to appear isshort. the fitness cost that is associated with as mobilization of the potential antibiotic
The examples of mecillinam, resistance reduces the growth rate of the resistance gene, the fitness cost of the gene
fosfomycin and nitrofurantoin resistance bacteria below the threshold that is required and the ecological overlap (that is, sharing
in Escherichiacoli provide an experimental for the pathogen to be maintained in the of niches) between the recipient and donor
illustration of this concept. These antibiotics high flow-rate environment of the bladder. environments seem to be important for the
are used for the treatment of uncomplicated In other words, even though the resistant successful spread of horizontally acquired
urinary tract infections, and the rates mutants are generated at high rates they antibiotic resistance genes82,83.

4 | ADVANCE ONLINE PUBLICATION www.nature.com/nrmicro



2
0
1
7
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
,
p
a
r
t
o
f
S
p
r
i
n
g
e
r
N
a
t
u
r
e
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
PERSPECTIVES

A new preclinical paradigm of resistance, and there are experimental qnrA and tetX, which confer resistance to
As outlined above, the methods that are used tools with which to measure them. Mutants lactams, quinolones, tetracyclines and
for the preclinical assessment of the potential that have a high fitness cost are less likely aminoglycosides, respectively) originate
to evolve antibiotic resistance to new to get fixed (that is, become the dominant from environmental bacteria and have
antibiotics are not adequate, as they largely type in the population) in a competitive recently transferred to pathogens87,88.
disregard the role of HGT in the acquisition environment, despite conferring high levels of Thus, the most important question during
of antibiotic resistance and place too much antibiotic resistance. By contrast, mutations the risk assessment of a new antibiotic
emphasis on the rates of antibiotic resistance that confer low levels of antibiotic resistance is whether genes that could potentially
acquisition that are determined invitro. can be fixed in the population if their fitness mediate resistance to antibiotics already
Consequently, our decisions during drug cost is low. Furthermore, the rate and exist in the resistome (that is, the pool of all
development may be guided by clinically efficiency of the evolution of compensatory resistance genes that are present in a certain
irrelevant parameters. Indeed, mecillinam, mutations, which could restore fitness environment), in particular in environments
a broad-spectrum penicillin drug against without the loss of antibiotic resistance, are in which the relevant human pathogens and
which the frequency of antibiotic-resistant key parameters in determining the long-term potential donor bacteria are likely to interact.
bacteria in clinical settings have remained fate of resistant mutants21. Arepertoire of Using functional metagenomic selection,
low, would probably have been terminated methods is available that enables fitness in which expression libraries of metagenomic
at the preclinical stage on the basis of a very costs and the process of evolution of DNA are transformed into a relevant host
high rate of appearance of resistant mutants compensatory mutations to be analysed and screened for phenotypic resistance, it
in laboratory tests84. Accordingly, we under several different invitro and invivo is possible to investigate whether a specific
should be cautious of ruling out antibiotic conditions, both in the absence and presence environment harbours potential resistance
candidates based solely on mutation of an antibiotic pressure21,85. Similarly, the genes against a new candidate drug 86
rates without further investigation of the infection dynamics (that is, bacterial growth (FIG.4a). This method can be expanded by
properties of these resistant mutants and and killing rates, and population sizes) can using metagenomic libraries from different
therelevant infection dynamics. In addition, be studied in relevant invitro and animal environments, with varying insert sizes
we are currently not adequately assessing infection models together with spontaneous and expression platforms. In addition, the
the potential for horizontal evolution of mutation rates. Experimental methods for screening can be carried out using different
antibiotic resistance, even though this determining spontaneous mutation rates can host bacteria and selection conditions, such
evolutionary process can render last-resort be used to determine the probability that a as different growth media, growth conditions
drugs, such as the carbapenems, ineffective. resistant mutant emerges and becomes the and selection strengths. By using these
Below, we propose how the experimental dominant type in bacterial population in a different screening methods, it is possible to
approaches that are used to assess both the specific host environment. Implementation reduce some of the biases that are associated
vertical and horizontal evolution of resistance of these approaches could revive antibiotic with heterologous gene expression, such as
can be improved to make better-informed candidatesthat were discarded during differences in expression that are dependent
preclinical decisions for continuing or drug development owing to a high rate on genetic background and/or the growth
discontinuing drug development. of emergence of resistance mutations environment used for screening.
invitro9,15,16. Although it is impractical to conduct
Assessing vertical evolution of resistance. metagenomic functional selections for
Several different factors influence the rate Assessing horizontal evolution of resistance. all environments, we suggest that human
of emergence, spread and maintenance of Despite the emergence of clinically relevant microbiomes with which an infecting
any resistance gene, and a key question is antibiotic resistance being most commonly pathogen is likely to interact are investigated
which factors should be included in risk the result of horizontal acquisition of and screened. In addition, microbiomes
prediction. It is clear that mutation rates that resistance genes, assessment of the risk that contain genetic material from diverse
are measured invitro are poor predictors of ofacquisition of genes that confer antibiotic origins, such as those from waste water as
clinical evolution of resistance in particular resistance is rarely carried out for new well as the environments in which antibiotic
for drugs that are associated with high rates of candidate antibiotics. This is surprising producers are found, may be important to
emergence of mutations. Furthermore, even given that the methodology that is required screen. Any identified antibiotic resistance
though selective pressures, coselection and to carry out such testing already exists86. gene represents a potential risk for the
epistatic interactions may markedly influence Several parameters that affect HGT need to evolution and dissemination of clinically
how resistance evolves, they are difficult be considered, including whether potential relevant antibiotic resistance89. Therefore,
to assess in antibiotic resistance prediction resistance genes exist in nature, the possibility metagenomic HGT experiments also
studies during preclinical development, as it is for HGT that is determined by niche provide an opportunity to act as an early
largely unknown how strong a future selective constraints, the rate of HGT and the fitness warning system for downstream surveillance
pressure could be or which coselection effects on the host following HGT (FIG.4). studies. However, the magnitude of this risk
processes or epistatic interactions might Several recent studies have shown that is dependent on several factors (FIG.4bd),
be important when a new antibiotic is all environments, including those without including ecological opportunity (contact
introduced. Instead, we propose that the human influence (for example, tundra and between a potential donor bacterium and
infection dynamics and mutation fitness caves), contain genes that confer resistance to a recipient pathogen), invivo HGT rates
costs in antibiotic-resistant bacteria should the majority of antibiotics that are currently between the relevant species, the likelihood
be considered to increase the relevance of in use79,80. Similarly, it has been shown of fixation of the acquired DNA (which
risk predictions, simply because they are key that several clinically relevant antibiotic largely depends on the fitness costs of the
parameters in determining rates of evolution resistance genes (for example, ctxm, transferred gene), and gene expression

NATURE REVIEWS | MICROBIOLOGY ADVANCE ONLINE PUBLICATION | 5



2
0
1
7
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
,
p
a
r
t
o
f
S
p
r
i
n
g
e
r
N
a
t
u
r
e
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
PERSPECTIVES

occurring at a sufficiently high level to confer Outlook regards to time and cost, these assessments
clinically relevant antibiotic resistance to the On the basis of the knowledge we have today have the potential to add substantial value
pathogen. All of these parameters, except for on the evolution of antibiotic resistance, to antibiotic development programmes
ecological opportunity, can be experimentally we propose that it is time for an updated at the preclinical stage. Indeed, such
determined to provide the basis for a rational preclinical paradigm to assess the risk of an integrative approach to preclinical
risk assessment. Ecological opportunity can resistance development in early preclinical assessment of evolution of resistance should
be partially assessed using bioinformatics development. This change is feasible and enable improved clinical predictions and
approaches that compare metagenomic is not limited by methodology but rather help inform decisions during the drug
datasets from diverse environments to a by tradition. Although the preclinical development process.
selected database comprising the gene or assessments that have been suggested in Moreover, the proposed integrated
genes that, when overexpressed, can cause this Opinion article represent an additional framework may also reactivate shelved
resistance to antibiotics. burden on antibiotic developers, with antibiotic development programmes that
were once considered too prone to develop
resistance based on traditional methods.
a b Indeed, antibiotics with a high emergence of
Opportunity for
Environment gene transfer resistance mutations, as determined invitro,
may not be prone to develop antibiotic
resistance in the clinic. As evidenced by the
example of mecillinam, the fitness cost of
antibiotic resistance mutations and infection
dynamics substantially affect clinically
Resistant relevant evolution of resistance and must be
pathogen considered along with the characterization
Extract
metagenomic of the rate of acquisition of resistance
DNA
mutations. Accordingly, we believe
that the proposed change in preclinical
c Rates of gene transfer characterization of antibiotic resistance
evolution will improve the long-term
Clone into an efficiency of antibiotic drug development.
expression
vector
Morten O.A.Sommer is at AntibioTx A/S, Kemitorvet,
Lyngby DK2800, Denmark; and at The Novo Nordisk
Foundation Center for Biosustainability, Technical
University of Denmark, Lyngby DK2800, Denmark.

Christian Munck is at The Novo Nordisk Foundation


Center for Biosustainability, Technical University of
Transform into Denmark, Lyngby DK2800, Denmark.
expression host
Rasmus Vendler Toft-Kehler is at AntibioTx A/S,
Kemitorvet, Lyngby DK2800, Denmark.

Dan I.Andersson is at Uppsala University, Department


Stable propagation
d of resistance gene of Medical Biochemistry and Microbiology,
Husargatan 3, SE751 23 Uppsala, Sweden.

Correspondence to D.I.A.
Select library Dan.Andersson@imbim.uu.se
on candidate doi:10.1038/nrmicro.2017.75
antibiotics
Published online 31 Jul 2017

1. [No authors listed.] Tackling drug-resistant infections


globally: final report and recommendations (review
onantimicrobial resistance, 2016). AMR Review
https://amr-review.org/sites/default/files/160525_
Loss of Final%20paper_with%20cover.pdf (2016).
resistance 2. Gilmore,M.S., Lebreton,F. & van Schaik,W. Genomic
transition of enterococci from gut commensals to
leading causes of multidrug-resistant hospital infection
Figure 4 | Examining the potential of horizontal gene transfer. The genetic reservoirs across different in the antibiotic era. Curr. Opin. Microbiol. 16, 1016
(2013).
natural and anthropogenic environments are likely to contain genes that can confer
Nature resistance
Reviews to novel
| Microbiology 3. zur Wiesch,P.A., Kouyos,R., Engelstdter,J.,
antibiotics.Functional metagenomic selection enables high-throughput screening to identify potential Regoes,R.R. & Bonhoeffer,S. Population biological
principles of drug-resistance evolution in infectious
resistance genes. For high-throughput screening, after the extraction of DNA, the DNA library that diseases. Lancet Infect. Dis. 11, 236247 (2011).
contains potential resistance-mediating genes is cloned and transformed into suitable expression hosts 4. Smith,M.R. & Wood,W.B. An experimental analysis
and then selected using the candidate antibiotic (part a). Following the identification of potential resist- of the curative action of penicillin in acute bacterial
infections. III. The effect of suppuration upon the
ance genes, it is necessary to assess (part b) the opportunity for gene transfer to the pathogen (that is, antibacterial action of the drug. J.Exp. Med. 103,
interactions between environmental reservoir and pathogen),the rate of gene transfer (part c) and the 509522 (1956).
ability of the acquired resistance gene to be stably propagated in the pathogen (part d), as opposed to 5. Palaci,M. etal. Cavitary disease and quantitative
sputum bacillary load in cases of pulmonary
the possibility of being lost. Blue cells are antibiotic-sensitive bacteria, whereas red cells are resistant tuberculosis. J.Clin. Microbiol. 45, 40644066
to antibiotics. (2007).

6 | ADVANCE ONLINE PUBLICATION www.nature.com/nrmicro



2
0
1
7
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
,
p
a
r
t
o
f
S
p
r
i
n
g
e
r
N
a
t
u
r
e
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
PERSPECTIVES

6. Feldman,W.E. Concentrations of bacteria in inenterococci. Proc. Natl Acad. Sci. USA 107, 54. Loftie-Eaton,W. etal. Evolutionary paths that expand
cerebrospinal fluid of patients with bacterial 1696416969 (2010). plasmid host-range: implications for spread of
meningitis. J.Pediatr. 88, 549552 (1976). 31. Andersson,D.I. & Hughes,D. Microbiological effects antibiotic resistance. Mol. Biol. Evol. 33, 885897
7. Canetti,G. Present aspects of bacterial resistance in of sublethal levels of antibiotics. Nat. Rev. Microbiol. (2016).
tuberculosis. Am. Rev. Respir. Dis. 92, 687703 12, 465478 (2014). 55. San Millan,A., Heilbron,K. & MacLean,R.C. Positive
(1965). 32. Gullberg,E., Albrecht,L.M., Karlsson,C., epistasis between coinfecting plasmids promotes
8. Canetti,G. Dynamic aspects of the pathology and Sandegren,L. & Andersson,D.I. Selection of a plasmid survival in bacterial populations. ISME J. 8,
bacteriology of tuberculous lesions. Am. Rev. Tuberc. multidrug resistance plasmid by sublethal levels of 601612 (2014).
74, 1321 (1956). antibiotics and heavy metals. mBio 5, e0191814 56. San Millan,A. etal. Positive selection and
9. Foster,P.L. Methods for determining spontaneous (2014). compensatory adaptation interact to stabilize non-
mutation rates. Methods Enzymol. 409, 195213 33. Oz,T. etal. Strength of selection pressure is an transmissible plasmids. Nat. Commun. 5, 52085211
(2006). important parameter contributing to the complexity (2014).
10. Drake,J.W., Charlesworth,B., Charlesworth,D. & ofantibiotic resistance evolution. Mol. Biol. Evol. 31, 57. Silva,R.F. etal. Pervasive sign epistasis between
Crow,J.F. Rates of spontaneous mutation. Genetics 23872401 (2014). conjugative plasmids and drug-resistance
148, 16671686 (1998). 34. De Visser,J.A.G.M. & Krug,J. Empirical fitness chromosomal mutations. PLoS Genet. 7, e1002181
11. Lynch,M. Evolution of the mutation rate. Trends Genet. landscapes and the predictability of evolution. (2011).
26, 345352 (2010). Nat.Rev. Genet. 15, 480490 (2014). 58. Porse,A., Schnning,K., Munck,C. &
12. Tubulekas,I., Buckingham,R.H. & Hughes,D. 35. Kondrashov,D.A. & Kondrashov,F.A. Topological Sommer,M.O.A. Survival and evolution of a large
Mutant ribosomes can generate dominant kirromycin features of rugged fitness landscapes in sequence multidrug resistance plasmid in new clinical bacterial
resistance. J.Bacteriol. 173, 36353643 (1991). space. Trends Genet. 31, 2433 (2015). hosts. Mol. Biol. Evol. 33, 28602873 (2016).
13. Lofton,H., Prnting,M., Thulin,E. & Andersson,D.I. 36. Brandis,G. & Hughes,D. Genetic characterization of 59. Alekshun,M.N. & Levy,S.B. Molecular mechanisms
Mechanisms and fitness costs of resistance to compensatory evolution in strains carrying rpoB of antibacterial multidrug resistance. Cell 128,
antimicrobial peptides LL37, CNY100HL and wheat Ser531Leu, the rifampicin resistance mutation most 10371050 (2007).
germ histones. PLoS ONE 8, e68875 (2013). frequently found in clinical isolates. J.Antimicrob. 60. Palmer,A.C. & Kishony,R. Understanding,
14. Gullberg,E. etal. Selection of resistant bacteria at Chemother. 68, 24932497 (2013). predictingand manipulating the genotypic evolution
very low antibiotic concentrations. PLoS Pathog. 7, 37. Brandis,G., Wrande,M., Liljas,L. & Hughes,D. of antibiotic resistance. Nat. Rev. Genet. 14,
e1002158 (2011). Fitness-compensatory mutations in rifampicin- 243248 (2013).
15. Nilsson,A.I., Berg,O.G., Aspevall,O., Kahlmeter,G. resistant RNA polymerase. Mol. Microbiol. 85, 61. Garcia,L.G. etal. Antibiotic activity against small-
& Andersson,D.I. Biological costs and mechanisms of 142151 (2012). colony variants of Staphylococcus aureus: review of
fosfomycin resistance in Escherichia coli. Antimicrob. 38. Lannergrd,J. etal. Genetic complexity of fusidic acid- invitro, animal and clinical data. J.Antimicrob.
Agents Chemother. 47, 28502858 (2003). resistant small colony variants (SCV) in Staphylococcus Chemother. 68, 14551464 (2013).
16. Thulin,E., Sundqvist,M. & Andersson,D.I. aureus. PLoS ONE 6, e28366 (2011). 62. Munck,C., Gumpert,H.K., Wallin,A.I.N.,
Amdinocillin (mecillinam) resistance mutations in 39. Marcusson,L.L., Frimodt-Mller,N. & Hughes,D. Wang,H.H. & Sommer,M.O.A. Prediction of
clinical isolates and laboratory-selected mutants of Interplay in the selection of fluoroquinolone resistance resistance development against drug combinations by
Escherichia coli. Antimicrob. Agents Chemother. 59, and bacterial fitness. PLoS Pathog. 5, e1000541 collateral responses to component drugs. Sci. Transl
17181727 (2015). (2009). Med. 6, 262ra156 (2014).
17. Drusano,G.L., Louie,A., MacGowan,A. & Hope,W. 40. Schrag,S.J., Perrot,V. & Levin,B.R. Adaptation to 63. Imamovic,L. & Sommer,M.O.A. Use of collateral
Suppression of emergence of resistance in pathogenic the fitness costs of antibiotic resistance in sensitivity networks to design drug cycling protocols
bacteria: keeping our powder dry, part 1. Escherichiacoli. Proc. Biol. Sci. 264, 12871291 that avoid resistance development. Sci. Transl Med. 5,
Antimicrob.Agents Chemother. 60, 11831193 (1997). 204ra132 (2013).
(2015). 41. Angst,D.C. & Hall,A.R. The cost of antibiotic 64. Kim,S., Lieberman,T.D. & Kishony,R. Alternating
18. Drusano,G.L., Hope,W., MacGowan,A. & Louie,A. resistance depends on evolutionary history in antibiotic treatments constrain evolutionary paths to
Suppression of emergence of resistance in pathogenic Escherichia coli. BMC Evol. Biol. 13, 163 (2013). multidrug resistance. Proc. Natl Acad. Sci. USA 111,
bacteria: keeping our powder dry, part 2. Antimicrob. 42. Komp Lindgren,P., Marcusson,L.L., Sandvang,D., 1449414499 (2014).
Agents Chemother. 60, 11941201 (2015). Frimodt-Mller,N. & Hughes,D. Biological cost of 65. Lzr,V. etal. Genome-wide analysis captures the
19. Chancey,S.T., Zhner,D. & Stephens,D.S. Acquired single and multiple norfloxacin resistance mutations in determinants of the antibiotic cross-resistance
inducible antimicrobial resistance in Gram-positive Escherichia coli implicated in urinary tract infections. interaction network. Nat. Commun. 5, 4352 (2014).
bacteria. Future Microbiol. 7, 959978 (2012). Antimicrob. Agents Chemother. 49, 23432351 66. Pena-Miller,R. etal. When the most potent
20. Chevereau,G. etal. Quantifying the determinants of (2005). combination of antibiotics selects for the greatest
evolutionary dynamics leading to drug resistance. 43. Trindade,S. etal. Positive epistasis drives the bacterial load: the smilefrown transition. PLoS Biol.
PLoS Biol. 13, e1002299 (2015). acquisition of multidrug resistance. PLoS Genet. 5, 11, e1001540 (2013).
21. Andersson,D.I. & Hughes,D. Antibiotic resistance e1000578 (2009). 67. Prichon,B. & Courvalin,P. Synergism between
and its cost: is it possible to reverse resistance? 44. Bjrkman,J., Samuelsson,P., Andersson,D.I. & -lactams and glycopeptides against VanA-type
Nat.Rev. Microbiol. 8, 260271 (2010). Hughes,D. Novel ribosomal mutations affecting methicillin-resistant Staphylococcus aureus and
22. Andersson,D.I. & Levin,B.R. The biological cost translational accuracy, antibiotic resistance and heterologous expression of the vanA operon.
ofantibiotic resistance. Curr. Opin. Microbiol. 2, virulence of Salmonella typhimurium. Mol. Microbiol. Antimicrob. Agents Chemother. 50, 36223630
489493 (1999). 31, 5358 (1999). (2006).
23. Wiser,M.J., Ribeck,N. & Lenski,R.E. Long-term 45. Hall,A.R. & MacLean,R.C. Epistasis buffers the 68. Brolund,A. & Sandegren,L. Characterization of ESBL
dynamics of adaptation in asexual populations. fitness effects of rifampicin-resistance mutations in disseminating plasmids. Infect. Dis. (Lond.) 48, 1825
Science 342, 13641367 (2013). Pseudomonas aeruginosa. Evolution 70, 11611161 (2016).
24. Hughes,D. & Andersson,D.I. Evolutionary (2016). 69. Mathers,A.J., Peirano,G. & Pitout,J.D.D. The role
consequences of drug resistance: shared principles 46. Rozen,D.E., McGee,L., Levin,B.R. & Klugman,K.P. of epidemic resistance plasmids and international
across diverse targets and organisms. Nat. Rev. Genet. Fitness costs of fluoroquinolone resistance in high-risk clones in the spread of multidrug-resistant
16, 459471 (2015). Streptococcus pneumoniae. Antimicrob. Agents Enterobacteriaceae. Clin. Microbiol. Rev. 28,
25. Brandis,G., Pietsch,F., Alemayehu,R. & Hughes,D. Chemother. 51, 412416 (2007). 565591 (2015).
Comprehensive phenotypic characterization of 47. Vogwill,T., Kojadinovic,M. & MacLean,R.C. Epistasis 70. Molton,J.S., Tambyah,P.A., Ang,B.S.P., Ling,M.L.
rifampicin resistance mutations in Salmonella provides between antibiotic resistance mutations and genetic & Fisher,D.A. The global spread of healthcare-
insight into the evolution of resistance in background shape the fitness effect of resistance associated multidrug-resistant bacteria: a perspective
Mycobacterium tuberculosis. J.Antimicrob. across species of Pseudomonas. Proc. Biol. Sci. 283, from Asia. Clin. Infect. Dis. 56, 13101318 (2013).
Chemother. 70, 680685 (2015). 20160151 (2016). 71. Bean,D.C., Livermore,D.M., Papa,I. &
26. ONeill,A.J., Huovinen,T., Fishwick,C.W.G. & 48. Vogwill,T. & MacLean,R.C. The genetic basis of the Hall,L.M.C. Resistance among Escherichia coli to
Chopra,I. Molecular genetic and structural modeling fitness costs of antimicrobial resistance: a meta- sulphonamides and other antimicrobials now little
studies of Staphylococcus aureus RNA polymerase analysis approach. Evol. Appl. 8, 284295 (2015). used in man. J.Antimicrob. Chemother. 56,
and the fitness of rifampin resistance genotypes in 49. Johanson,U., varsson,A., Liljas,A. & Hughes,D. 962964 (2005).
relation to clinical prevalence. Antimicrob. Agents The dynamic structure of EFG studied by fusidic acid 72. Enne,V.I., Livermore,D.M., Stephens,P. &
Chemother. 50, 298309 (2006). resistance and internal revertants. J.Mol. Biol. 258, Hall,L.M. Persistence of sulphonamide resistance
27. Bottger,E.C., Springer,B., Pletschette,M. & 420432 (1996). inEscherichia coli in the UK despite national
Sander,P. Fitness of antibiotic-resistant 50. Nagaev,I., Bjorkman,J., Andersson,D.I. & prescribing restriction. Lancet 357, 13251328
microorganisms and compensatory mutations. Hughes,D. Biological cost and compensatory (2001).
Nat.Med. 4, 13431344 (1998). evolution in fusidic acid-resistant Staphylococcus 73. Sundqvist,M. etal. Little evidence for reversibility of
28. Sander,P. etal. Fitness cost of chromosomal drug aureus. Mol. Microbiol. 40, 433439 (2001). trimethoprim resistance after a drastic reduction in
resistance-conferring mutations. Antimicrob. Agents 51. Salverda,M.L.M. etal. Initial mutations direct trimethoprim use. J.Antimicrob. Chemother. 65,
Chemother. 46, 12041211 (2002). alternative pathways of protein evolution. PLoS Genet. 350360 (2010).
29. Shcherbakov,D. etal. Directed mutagenesis of 7, e1001321 (2011). 74. Locke,J.B., Hilgers,M. & Shaw,K.J. Novel ribosomal
Mycobacterium smegmatis 16S rRNA to reconstruct 52. Weinreich,D.M., Delaney,N.F., Depristo,M.A. & mutations in Staphylococcus aureus strains identified
the invivo evolution of aminoglycoside resistance in Hartl,D.L. Darwinian evolution can follow only very through selection with the oxazolidinones linezolid
Mycobacterium tuberculosis. Mol. Microbiol. 77, few mutational paths to fitter proteins. Science 312, and torezolid (TR700). Antimicrob. Agents
830840 (2010). 111114 (2006). Chemother. 53, 52655274 (2009).
30. Foucault,M.L., Depardieu,F., Courvalin,P. & 53. Dahlberg,C. & Chao,L. Amelioration of the cost of 75. Gordon,D.M. & Riley,M.A. A theoretical and
Grillot-Courvalin,C. Inducible expression conjugative plasmid carriage in Eschericha coli K12. experimental analysis of bacterial growth in the
eliminatesthe fitness cost of vancomycin resistance Genetics 165, 16411649 (2003). bladder. Mol. Microbiol. 6, 555562 (1992).

NATURE REVIEWS | MICROBIOLOGY ADVANCE ONLINE PUBLICATION | 7



2
0
1
7
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
,
p
a
r
t
o
f
S
p
r
i
n
g
e
r
N
a
t
u
r
e
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
PERSPECTIVES

76. Sandegren,L., Lindqvist,A., Kahlmeter,G. & 85. Huseby,D.L. etal. Mutation supply and relative 93. Buckel,P., Buchberger,A., Bck,A. & Wittmann,H.G.
Andersson,D.I. Nitrofurantoin resistance fitness shape the genotypes of ciprofloxacin-resistant Alteration of ribosomal protein L6 in mutants of
mechanismand fitness cost in Escherichia coli. Escherichia coli. Mol. Biol. Evol. 34, 10291039 Escherichia coli resistant to gentamicin.
J.Antimicrob. Chemother. 62, 495503 (2008). (2017). Mol. Gen. Genet. 158, 4754 (1977).
77. Thomas,C.M. & Nielsen,K.M. Mechanisms of, 86. Moore,A.M., Munck,C., Sommer,M.O.A. & 94. Adler,M., Anjum,M., Andersson,D.I. &
andbarriers to, horizontal gene transfer between Dantas,G. Functional metagenomic investigations of Sandegren,L. Influence of acquired lactamases on
bacteria. Nat. Rev. Microbiol. 3, 711721 (2005). the human intestinal microbiota. Front. Microbiol. 2, the evolution of spontaneous carbapenem resistance
78. Naseer,U. & Sundsfjord,A. The CTXM 188 (2011). in Escherichia coli. J.Antimicrob. Chemother. 68,
conundrum:dissemination of plasmids and 87. Dantas,G. & Sommer,M.O. Context matters the 5159 (2013).
Escherichia coli clones. Microb. Drug Resist. 17, complex interplay between resistome genotypes and 95. Oakberg,E.F. & Luria,S.W. Mutations to sulfonamide
8397 (2011). resistance phenotypes. Curr. Opin. Microbiol. 15, resistance in Staphylococcus aureus. Genetics 32,
79. Allen,H.K., Moe,L.A., Rodbumrer,J., Gaarder,A. 577582 (2012). 249261 (1947).
&Handelsman,J. Functional metagenomics reveals 88. Yoon,E.J. etal. Origin in Acinetobacter gyllenbergii 96. Zurenko,G.E. etal. In vitro activities of U-100592
diverse lactamases in a remote Alaskan soil. ISME J. and dissemination of aminoglycoside-modifying and U-100766, novel oxazolidinone antibacterial
3, 243251 (2008). enzyme AAC(6)-Ih. J.Antimicrob. Chemother. 71, agents. Antimicrob. Agents Chemother. 40, 839845
80. DCosta,V.M. etal. Antibiotic resistance is ancient. 601606 (2016). (1996).
Nature 477, 457461 (2011). 89. Martnez,J.L., Coque,T.M. & Baquero,F. Prioritizing 97. Lewis,K. Platforms for antibiotic discovery. Nat. Rev.
81. Sommer,M.O.A., Dantas,G. & Church,G.M. risks of antibiotic resistance genes in all metagenomes. Drug. Discov. 12, 371387 (2013).
Functional characterization of the antibiotic resistance Nat. Rev. Microbiol. 13, 396396 (2015).
reservoir in the human microflora. Science 325, 90. Jaff,A., Chabbert,Y.A. & Derlot,E. Selection and Acknowledgements
11281131 (2009). characterization of -lactam-resistant Escherichia coli Work in the authors laboratories was supported by grants
82. Munck,C. etal. Limited dissemination of the K-12 mutants. Antimicrob. Agents Chemother. 23, from the Swedish Research Council (to D.I.A.), and the Novo
wastewater treatment plant core resistome. 622625 (1983). Nordisk Foundation, the Lundbeck Foundation and the
Nat.Commun. 6, 8452 (2015). 91. George,A.M. & Levy,S.B. Amplifiable resistance to Danish Free Research Council (to M.O.A.S. and C.M.).
83. Forsberg,K.J. etal. Bacterial phylogeny structures tetracycline, chloramphenicol, and other antibiotics
soil resistomes across habitats. Nature 509, 612616 inEscherichia coli: involvement of a non-plasmid- Competing interests statement
(2014). determined efflux of tetracycline. J.Bacteriol. 155, The authors declare competing interests: see Web version
84. Kahlmeter,G. & Poulsen,H.O. Antimicrobial 531540 (1983). fordetails.
susceptibility of Escherichia coli from community- 92. Heisig,P. & Tschorny,R. Characterization of
acquired urinary tract infections in Europe: the fluoroquinolone-resistant mutants of escherichia coli Publishers note
ECOSENS study revisited. Int. J.Antimicrob. Agents selected invitro. Antimicrob. Agents Chemother. 38, Springer Nature remains neutral with regard to jurisdictional
39, 4551 (2012). 12841291 (1994). claims in published maps and institutional affiliations.

8 | ADVANCE ONLINE PUBLICATION www.nature.com/nrmicro



2
0
1
7
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
,
p
a
r
t
o
f
S
p
r
i
n
g
e
r
N
a
t
u
r
e
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.