PERSPECTIVES

resistance gene in a given ecosystem. This

OPINION
framework could function as the basis for
interpretative guidelines and interventions,
What is a resistance gene? which are urgently required for both scien-
tific and public health reasons. Antibiotic
Ranking risk in resistomes resistance can result from mutations15, as is
the case of resistance to fluoroquinolones,
or by the acquisition of antibiotic resistance
José L. Martínez, Teresa M. Coque and Fernando Baquero genes by horizontal gene transfer (HGT).
Whereas a mutation is relevant only for the
Abstract | Metagenomic studies have shown that antibiotic resistance genes are bacterium that harbours it, the presence of a
ubiquitous in the environment, which has led to the suggestion that there is a high potentially transferable antibiotic resistance
risk that these genes will spread to bacteria that cause human infections. If this is gene in a bacterium might be important
true, estimating the real risk of dissemination of resistance genes from for the dissemination of resistance among
environmental reservoirs to human pathogens is therefore very difficult. In this a population. In this article, we focus on
antibiotic resistance genes that can spread
Opinion article, we analyse the current definitions of antibiotic resistance and among a population, and we do not provide
antibiotic resistance genes, and we describe the bottlenecks that affect the a detailed discussion of the relevance of
transfer of antibiotic resistance genes to human pathogens. We propose rules for mutations for the acquisition of resistance.
estimating the risks associated with genes that are present in environmental
resistomes by evaluating the likelihood of their introduction into human What is antibiotic resistance?
Different antibiotic resistance studies gener-
pathogens, and the consequences of such events for the treatment of infections.
ate different results because the definition of
resistance depends on the objectives of each
The influence of antibiotic resistance on a clear identification of resistance genes. study. This situation makes the adoption of
human and animal health and welfare, as Surprisingly, several resistance gene data- measures that address the global antibiotic
well as its economic consequences, has bases (see Supplementary information S1 resistance crisis difficult and controver-
reinforced the need for a concerted effort (table)) include many genes that encode sial1,16,17. There are three main definitions
to track and control its emergence and dis- regulators of resistance phenotypes, as well of resistance that are currently found in the
semination1,2. Several studies, based on as genes that encode antibiotic targets and literature and each, although adequate for
metagenomic and functional metagenomic transporters that mediate the entry of anti- the objective of a specific type of study, are
techniques3, have determined the presence biotics into cells. Mutation of such genes inappropriate for other purposes.
of antibiotic resistance genes in different can cause resistance; however, the wild-type
ecosystems — from soil and the human version of these genes should not confer Clinical definition of antibiotic resistance.
gut to extreme environments, such as per- resistance following transfer to a suscepti- The clinical definition of resistance is based
mafrost 4–11. Such studies have shown that ble bacterial cell (discussed in more detail on minimum inhibitory concentration
putative antibiotic resistance genes are below). We suggest that finding such genes (MIC) breakpoints, which are fixed because
present in all environments. The authors in a metagenomic analysis should not be only the likelihood of therapeutic failure in
of these studies frequently state that find- considered a real risk. We propose that even human patients is considered. This is the
ing potential resistance genes in a specific in the rare cases in which mutated target definition that is used to inform clinicians
ecosystem poses a risk for human health genes could recombine with wild-type genes about the possible antibiotics of choice for
because these genes can be acquired by bac- in kin recipient cells (for example, genes that therapy. However, it is inadequate for non-
terial pathogens. However, if the resistome in encode topoisomerases in transformable bac- human-associated bacteria and also for
every habitat is found to be a risk for human teria, such as Streptococcus pneumoniae13), antimicrobial compounds that are not used
health, then there are no differential risks it is excessive to classify these housekeeping for therapy, which do not have established
and interventions become unworkable. genes as resistance genes, unless they are breakpoints. In addition, this definition does
Given these considerations, it is neces- ‘decontextualized’ (REF. 14) on mobile genetic not incorporate low-level resistance mecha-
sary to establish rules to define the relative elements (see below), which is an issue that nisms that increase the MIC without reach-
risks for human health that are associated is infrequently addressed in metagenomic ing the breakpoint, which can be a gateway
with each of the candidate genes that are studies. Furthermore, different databases for for the acquisition of high-level clinically
predicted to confer antibiotic resistance antibiotic resistance genes include so-called relevant resistance18.
in studies of environmental resistomes12. novel resistance genes, which are predicted
The relative risks should be based on pub- to confer resistance on the basis of partial Epidemiological definition of antibiotic
lic health criteria and should focus on the sequence homology with classical resistance resistance. Following earlier suggestions19,
likelihood that these genes may be acquired genes; whether these novel genes actually in the mid-1990s, the European Committee
by human pathogens and, hence, limit the have a role in antibiotic resistance is debat- on Antimicrobial Susceptibility Testing
efficacy of classical, and probably also new, able, unless functional studies that verify (EUCAST) established common epidemio-
antimicrobial agents. such a role are performed. logical cut-off values for resistance
However, being able to assess relative In this Opinion article, we present a con- (ECOFFs), which are defined as the MIC
risk requires the re-examination of the ceptual framework to clearly define the risks value that corresponds to the upper limit
definition of antibiotic resistance to provide associated with the presence of a potential of the wild-type population of a particular

116 | FEBRUARY 2015 | VOLUME 13 www.nature.com/reviews/micro

© 2015 Macmillan Publishers Limited. All rights reserved

 PERSPECTIVES

of which were introduced in the absence of

functional validation, and others that are
not real markers of resistance (see below).
As these new predicted resistance genes can

Number of strains
be incorporated into databases under the
antibiotic resistance heading without further
Number of strains

New
ecological functional verification, databases are becom-
breakpoint ing increasingly comprehensive and contain
a large amount of ‘noise’ (BOX 1).
ECOFF breakpoint
Clinical breakpoint Ecological definition of antibiotic resistance
MIC value genes. The ecological function of an
(arithmetic concentrations) antibiotic resistance gene is to protect an
organism from the inhibitory effect of
MIC value (geometric concentrations)
an antimicrobial that is produced by another
Figure 1 | Determination of epidemiological breakpoints of susceptibility to antibiotics. organism. In the case of natural ecosystems,
Nature Reviews | Microbiology
Epidemiological breakpoints are determined from the analysis of minimum inhibitory concentrations this implies an ‘arms-race’ situation, in
(MICs) of a large number of isolates. The number (or proportion) of isolates that have particular MICs
which one microorganism produces an
is plotted to obtain a distribution curve. The common epidemiological cut-off value for resistance
(ECOFF) breakpoint is established as the upper limit of the curve for the majority of the population. antibiotic to outcompete a competing micro-
Those isolates that have MICs above this breakpoint are considered resistant to the antibiotic even organism in the same niche, but the com-
if their MICs do not reach the clinical breakpoint, which predicts clinical success. Breakpoints are petitor is protected because it harbours
usually defined using double dilution methods, which means that the explored concentration range resistance genes. This ecological definition
has a geometric distribution that can hide small changes in MIC. If the MICs of the majority of the forms two categories of resistance genes:
population, as obtained using double dilution tests, are broken down using arithmetic antibiotic those that evolved in nature to counteract
concentrations to establish MICs (see inset graph), subpopulations that have subtle differences in the activity of natural antimicrobials that are
their MICs can be detected, and new (usually lower) ecological breakpoints that define the most produced by neighbouring cells, and those
susceptible population can be established. The double dilution method makes it difficult to detect that have been acquired by HGT since the
populations with low-level antibiotic resistance that can eventually evolve into populations with
use of antibiotics by humans began.
high-level resistance18YJKEJKUCPCURGEVVJCVJCUDGGPQXGTNQQMGFKPVJGRCUV|[GCTUQHTQWVKPG
application of susceptibility testing methods, and which might be the basis of the ‘MIC-creep’ that Genes that are present in organisms that
is observed for different antibiotics79. produce particular antibiotics and encode
proteins that are capable of specifically
detoxifying these compounds have been
species20–22. The main advantages of include predefined breakpoints, it cannot be classically defined as resistance genes36,37.
this definition are that it can be used for used to establish whether a natural isolate However, from an ecological perspec-
any combination of microorganism– is susceptible or resistant unless labour- tive, the function of these genes is not to
antimicrobial — including those biocides for intensive work is performed, as natural resist the action of an antibiotic that is pro-
which a clinical definition of resistance is not isolates do not usually have parental strains duced by a competitor and, thus, cannot
available23 — and that its use can determine for making pairwise comparisons (unless be considered as resistance genes; instead,
low-level resistance. The establishment of pairs of clonal isolates are available, such as they should strictly be considered to be
ECOFF breakpoints requires the analysis in outbreaks, chronic infections or isolates detoxification genes.
of a large number of independent isolates to of globally spread clonal complexes that are Similarly, genes that contribute to the
determine the normal distribution of MICs able to acquire different genetic elements). characteristic resistance phenotype of a
in a given bacterial species (FIG. 1), but it has However, it is the most relevant definition given microorganism (often referred to as
a limited application when a single gene for functional genomics and metagenom- the intrinsic resistome) do not necessarily
or mutation is analysed in a given isolate. ics studies of antibiotic resistance because it have an ecological role as resistance determi-
Consequently, with organisms for which enables researchers to define a specific gene, nants that protect this organism from anti-
databases are scarce, this approach cannot be the expression (or inactivation) of which biotics in nature38. Multidrug efflux pumps
reliably used, which is the case for the majority renders an otherwise susceptible recipient are ubiquitous, have a wide substrate range
of environmental microorganisms. host organism more resistant to a particular and, in most cases, their original function
antibiotic7,27. was not to confer resistance to antibiotics
Operational definition of antibiotic resist- that are currently used in human therapy 39–41;
ance. Differing from previous definitions, What is an antibiotic resistance gene? for example, the AcrAB efflux pump prob-
which are species-centric, this is a gene- The increase in bacterial genomic and ably evolved in Escherichia coli to confer
centric definition of resistance. It is based metagenomic research has generated many resistance to bile salts42. Furthermore, it is
on the pairwise comparison of a parental sequences and genes that are presumptively unlikely that AmpC-type β-lactamases in the
strain with a mutant strain or with a strain associated with antibiotic resistance. Some Enterobacteriaceae evolved to counteract
containing a resistance gene that has been of these studies rely on functional assays of the activity of β-lactams in the gut, as produc-
acquired by HGT. This strain is considered resistance4–7,29,30; however, in several other ers of β-lactam have never been described
resistant if it has a higher MIC value for the studies, function is simply inferred by simi- in this niche; rather, it has been shown that
studied compound than its parental wild- larity to sequences that have been catalogued these enzymes contribute to maintenance of
type strain24–28. As this definition does not as resistance genes in databases8–10,31–35, some the normal morphology of E. coli 43. From an

NATURE REVIEWS | MICROBIOLOGY VOLUME 13 | FEBRUARY 2015 | 117

© 2015 Macmillan Publishers Limited. All rights reserved

PERSPECTIVES

ecological viewpoint, these genes should not Box 1 | Information and noise in databases
be considered as antibiotic resistance genes,
even though the proteins they encode might The implementation of high-throughput sequencing methods has resulted in a massive increase in
protect the individual organism that pro- the number of available sequences in databases. Only a small proportion of these sequences have
duces them in the presence of therapeutic been functionally characterized, whereas for the others, function is predicted by automatic
annotation only. In some instances, newly identified genes are very similar to those that have been
antibiotic concentrations.
functionally characterized and thus are probably true positives, whereas in other cases (and usually
In this context, it is important to distin- in order to be comprehensive), false positives are also annotated as having the same function (in
guish between the ubiquity and abundance the context of the present article, a function in antibiotic resistance).
of a given gene and how these parameters The logic of including these possible false-positive annotations goes against the logic of the
may relate to its function44. Ubiquitous genes current methods that are used for analysing high-throughput sequencing data, whereby statistics
that are classed as resistance genes are prob- to avoid false positives are always implemented. Consequently, false positives are incorporated
ably involved in basic processes in bacterial together with functionally confirmed positives, and true positives are inferred by more detailed
physiology, which contributes to the repro- studies using stringent similarity criteria. As the number of genes for which function is inferred by
ductive fitness of the organism in every eco- homology increases exponentially, this can result in a vicious feedback loop that leads to the
system; but, whether these genes pose a risk incorporation of an increased number of false positives or ‘noise’ in growing databases.
This situation is exacerbated when partial sequences that occasionally contain just a protein
to human health is debatable. By contrast,
motif are included in databases in the same category as whole genes that have been functionally
any gene that is capable of conferring resist- studied. Even though most, if not all, databases explicitly state how the genes have been annotated
ance and which is most highly abundant in (functionally or by homology), this issue is not always taken into consideration when scrutinizing
habitats with a high antibiotic load, such as sequence data sets. Researchers should be more aware of this issue, and stringent search
farms or clinical settings, is more likely to be parameters, manual curation and eventual re-annotation of the results obtained after searching
relevant for resistance development. databases should be performed66.
This does not mean that bona fide
antibiotic resistance genes do not exist in
nature. Indeed, it has been shown that non- not resistance genes because the absence of risk beyond the propagation of their hosts.
producer, resistant microorganisms can these genes makes bacteria more resistant, Consistent with this, recent work has shown
coexist with antibiotic producers, which not more susceptible. Similarly, target genes that the composition of the soil resistome
suggest that there are occasions when anti- with mutations that impair the mechanism correlates with the taxonomic structure of
biotic concentrations in natural ecosystems of action of an antibiotic are not resistance the studied samples29. Metagenomic analyses
can be locally high45. Similarly, it has been genes unless they can be transferred and that search for resistance genes should be
proposed (although not shown) that the dominantly expressed over the wild-type able to differentiate those core genes from
original function of mec methicillin resist- susceptible allele in a recipient host. genes that are captured by mobile genetic
ance genes might be antibiotic resistance46. One specific situation involves global elements.
Nevertheless, it puts the focus on the strong gene regulators, such as the multiple anti-
selective pressure that has driven a change biotic resistance determinant MarA50; Ranking risks
in the ecological function of resistance genes mutations in the gene that encode MarA Considering the previous discussions on
that are acquired by pathogens since the can confer resistance, but the actual resist- definitions of resistance, as well as the bottle-
use of antibiotics by humans began14,47,48. ance genes could be those whose expres- necks for its transfer (discussed in detail in
When these determinants are acquired by sion is regulated by MarA, rather than the BOX 2), we propose the ranking of antibiotic
a human pathogen by HGT, the recipient is regulator itself. resistance genes that are found in metagen-
frequently lacking the regulatory and meta- Many other elements that are identified omic studies into different categories
bolic networks in which these elements were as resistance genes (such as, acrAB42 and depending on the associated public health
originally integrated with in the donor. Such ampC in E. coli 43, mexAB51 in Pseudomonas risks they might pose; that is, the risk that
decontextualization (including changes in aeruginosa, aac(6ʹ)-Ii in Enterococcus they could be acquired and confer resistance
their expression levels) impedes these deter- faecium52 and blaOXY in Klebsiella oxytoca53) to human pathogens (FIG. 2). The first con-
minants from exerting the functions that belong to the core genome of these species, sideration refers to the resistance genes that
they evolved in their original hosts, and the and are responsible for natural, intrinsic belong to the intrinsic resistome, the detec-
only role they have in the new host is resist- resistance phenotypes (rather than resist- tion of which must be considered merely
ance in the presence of the antibiotic14. In this ance phenotypes that are acquired after as a marker of the local abundance of the
regard, the use of antibiotics by humans must anthropogenic use of antibiotics). However, bacterium to which they belong, despite their
be considered as the main force that is chang- the potential public health risks that are contribution to the intrinsic resistance of the
ing the function of some determinants to associated with the detection of these genes organism. The finding of the E. coli efflux
resistance, despite these determinants having in metagenomes must be interpreted with pump gene acrAB in the gut microbiota, for
a different function in their original hosts. caution. Their detection in metagenomes example, should be considered as a normal
only reveals the presence of organisms har- sign of the presence of E. coli in the intestine,
Operational definition of antibiotic resistance bouring these genes in the sample. If these and its detection in water samples as a sign of
genes. From an operational perspective, an genes increase in frequency, it is because the faecal pollution. These determinants should
antibiotic resistance gene confers resistance population size of the organisms in which be considered as a potential risk for the dis-
to antibiotics when it is present or when it they are present has increased. Moreover, semination of antibiotic resistance only if they
increases susceptibility to antibiotics when these genes do not contribute to the com- are found in transferable genetic elements.
it is absent 49. Transporter genes with muta- mon pool of resistance genes that are able to Risk levels should be determined on
tions that impair the uptake of antibiotics are circulate among bacteria, and they are not a the basis of the level of evidence regarding

118 | FEBRUARY 2015 | VOLUME 13 www.nature.com/reviews/micro

© 2015 Macmillan Publishers Limited. All rights reserved

 PERSPECTIVES

Box 2 | Transfer bottlenecks for antibiotic resistance genes

Functional metagenomic analyses have documented the presence of connectivity of the incoming resistance determinant64,75. Some
novel resistance genes that are capable of conferring resistance following investigations have shown that fitness costs that are associated with the
their transfer to a susceptible host. However, the number and variety acquisition of antibiotic resistance genes are not just a consequence of
of antibiotic resistance genes acquired by human pathogens, and the metabolic burden that is imposed by the replication, transcription and
particularly those that lead to therapeutic failures, is extremely low translation of the mobile genetic element that carries the antibiotic
compared to the number of sequences classified as resistance genes resistance gene. On the contray, fitness costs can be gene-specific76,77.
in metagenomic studies4,49,67. This implies the existence of extremely Only those determinants that present affordable fitness costs (including
stringent bottlenecks that modulate the transfer of resistance those that are balanced by the acquisition of compensatory mutations)
determinants from their original hosts to human pathogens68,69. may successfully spread in the bacterial population64,78.
Consequently, these bottlenecks affect the risks that are associated The figure shows the main bottlenecks that determine the spread of
with the presence of a functionally defined resistance gene in a given antibiotic resistance genes. Circles represent bacterial populations that
ecosystem (see the figure). harbour resistance genes; the sizes of the circles correspond to the
The first bottleneck is ecological connectivity: a gene transfer event population densities. Three different compartments are indicated: soil
only occurs when donor and recipient populations come into contact. and water, animals and sewage, and human-associated populations
This usually means that they are able to reach a critical population size in (some of which are pathogenic). Areas of ecological connectivity, where
the same, or neighbouring, ecological space. Genetic transfer ensures the populations from different compartments interact, are shown by a dotted
availability of common goods, which includes antibiotic resistance genes, circle. Substantial interactions only occur at relatively high cell densities,
for the many species within such ecologically connected communities70. which eventually facilitates horizontal gene transfer (HGT). After
Furthermore, connectivity, more than function, seems to be a major acquisition of a new resistance trait, the recipient population might
determinant of the success and selection of gene transfer71. Of course, a decrease in size in the absence of selection owing to the fitness costs that
chain of successive transfer events that link different habitats could occur, are imposed by resistance traits or vehicles (as shown in the animals and
but the efficiency of such a series of transfer events is low unless the sewage compartment). Even if the resistant population maintains its size
recipients in every habitat are under positive selection, such as in and comes into contact with a human bacterial population, the resistance
the presence of antibiotics. determinant might fail to become established. This may happen if the
The second bottleneck is based on the founder effect: when one antibiotic resistance gene is outcompeted by similar, previously
resistance determinant is broadly distributed among bacterial established mechanisms of resistance in the human-associated organisms
populations, it is rare that another resistance determinant that has the (known as the founder effect; marked by the green arrow). Note that the
same or a closely related substrate profile is acquired; bacteria are discovery of a ‘resistance gene’ in a non-human compartment does not
already resistant and, consequently, antibiotics do not exert any necessarily represent a substantial risk for human health.
selective pressure on an organism that already
Soil and water Animals and sewage Humans
harbours a gene with the same function. This
does not mean that different resistance genes
cannot be established in a population because
different founder effects occurred at earlier
stages in different places72, but when these HGT
genes are stably spread in the population, the
chances of recruiting a new gene are lower.
There are also situations in which sequential
events of penetration and extinction of the same
gene may happen, which can be viewed as a form
of short-sighted evolution73. This may include
the acquisition of recombinant-resistant
penicillin-binding proteins (PBPs) by
Streptococcus pneumoniae from commensals that HGT
share the same ecosystem and that are the only
available source of antibiotic resistance genes74.
The third bottleneck that determines the
spread of a given resistance gene is fitness costs,
which reflects the physiological–epigenetic

Nature Reviews | Microbiology

the possibility that genes identified as puta- have previously been reported to reside on glycopeptide resistance. It is important to
tive resistance genes in metagenomic analyses mobile genetic elements that are hosted by note that some of these high-risk resistance
could be involved in creating public health human bacterial pathogens; that is, there genes confer resistance only in some genetic
problems (beyond the individual patient) as a is published evidence that these genes backgrounds54, are heterologously expressed
result of antibiotic resistance. These alert lev- pose a substantial risk for the dissemina- in different hosts (for example, the blaZ
els (resistance readiness condition (RESCon); tion of resistance and treatment failure of β-lactamase gene in Staphylococcus aureus
from RESCon 1, the highest risk, to human infections. This is certainly the case and Enterococcus faecalis 55), and can be
RESCon 7, the lowest risk) are outlined below. for all of the resistance genes that are cur- silenced in specific hosts54,56. In addition,
rently causing problems in hospitals, which any novel antibiotic resistance gene that
RESCon 1. RESCon 1 comprises resistance includes genes that encode β-lactamases inactivates an antibiotic that is currently in
genes that are already known to contribute (for example, TEM, SHV, CTX-M, VIM use and which is present on a mobile genetic
to the failure of antibiotic treatment and and KPC) and vanA genes that encode element in a human pathogen (as happened

NATURE REVIEWS | MICROBIOLOGY VOLUME 13 | FEBRUARY 2015 | 119

© 2015 Macmillan Publishers Limited. All rights reserved

PERSPECTIVES

when New Delhi metallo-β-lactamase 1

Gene of interest
(NDM1) was described the first time57)
immediately falls into this category.

RESCon 2. RESCon 2 comprises novel Genetic identity with

resistance genes that have been functionally known resistance gene NO Not a resistance gene
characterized as inactivating antibiotics that above a given threshold
are used in human therapy and are located
RESCon 7
on mobile genetic elements in non-pathogenic
YES
bacteria. The detection of an antibiotic NO
resistance gene on a mobile genetic element
suggests that it has been under positive
Functional evaluation NO Mobile genetic element
selection (detection is only possible if this
event occurs at an appreciable frequency).
The capture and selection of a novel resist- YES YES
ance gene by a mobile genetic element might
be the first step for its subsequent acquisition Novel mechanism
RESCon 6
by a human pathogen. For example, quino- New antibiotic YES
of resistance
lone resistance determinants, such as qnr,
are encoded by plasmids that are present YES
in non-pathogenic bacteria isolated from NO
RESCon 3
water 58. The association of antibiotic resist-
ance genes with mobile genetic elements (or
the mobilome) can be tracked either by spe- NO NO Antibiotic in use
cifically isolating DNA from mobile genetic
elements59 or by reconstructing the mobile
genetic elements that are detected in general YES
metagenomic studies using specific Intrinsic resistance gene.
bioinformatic approaches. Marker for the presence of
Mobile genetic element Mobile genetic element NO CURGEKȮEOKETQQTICPKUO
in the sample
RESCon 3. RESCon 3 comprises function-
ally characterized genes that confer antibi- NO YES
otic resistance to a new antibiotic in Phase I,
II or III development, just starting clinical
Known and widespread
use, or are no longer widely used. Similar to mechanism of resistance
Human pathogen NO RESCon 2
RESCon 2 genes, the potential of RESCon 3
genes to confer antibiotic resistance has
YES NO YES
been validated by functional assays; how-
ever, they inactivate antibiotics that are not
yet commonly used therapeutically. Even if RESCon 5 RESCon 4 RESCon 1
RESCon 3 genes were not eventually trans-
ferred and spread among human patho- Figure 2 | Ranking the risks of detection of resistance genes inNature resistomes.
ReviewsThe public health
| Microbiology
gens, their detection and characterization risks of finding a given resistance gene in a given ecosystem rank from highest (resistance readiness
enable potential resistance mechanisms to EQPFKVKQP 4'5%QP| KHVJKUIGPGJCUCNTGCF[DGGPUJQYPVQDGCJGCNVJRTQDNGOVQNQYGUV
4'5%QP| KHVJGHWPEVKQPQHVJGIGPGJCUQPN[DGGPKPHGTTGFD[JQOQNQI[6JGHKIWTGUJQYUC
be identified, which could help to inform
decision tree for ranking these risks. The decision tree starts when a resistance gene is detected
the design of new antibiotics or to promote by having substantial homology with known resistance genes. In contrast to simple BLAST (basic
genetic surveillance for RESCon 3 genes local alignment search tool) searches, which have been the usual basis for the classification of a
in Phase II and III clinical trials or in early gene as a resistance gene, the use of hidden Markov models80 and protein structure prediction
clinical use. For example, genes that confer tools are being applied to help improve this first discrimination step. The second step consists of
resistance to glycopeptide antibiotics, which HWPEVKQPCNCPCN[UKU+HVJGHWPEVKQPKUKPHGTTGFD[UGSWGPEGCNQPGVJGTKUMKUNQY 4'5%QP| CPFQPN[
are drugs that have an inhibitory effect on the presence of the putative resistance element on a mobile genetic element increases the risk
Neisseria spp. and are in Phase III trials60, 4'5%QP| (WPEVKQPCNGXCNWCVKQPUJQWNFGPCDNGQPGVQFKUVKPIWKUJDGVYGGPPGYCPFMPQYPTGUKUV-
as well as plasmid-mediated cfr methyl- ance mechanisms. If the mechanism is novel, and confers resistance to antibiotics that are in use
transferases that inactivate the new class of and for which mechanisms of resistance are already present in human pathogens, there are two
UKVWCVKQPUVJGIGPGDGNQPIUVQCPCNTGCF[MPQYPHCOKN[QHTGUKUVCPEGFGVGTOKPCPVU 4'5%QP| QT
synthetic antimicrobals (oxazolidinones)61.
GPEQFGUCEQORNGVGN[PGYOGEJCPKUO 4'5%QP| 4GUKUVCPEGVQPGYCPVKDKQVKEUKUCNUQCPQXGN
Note that target mutations in rrn (rRNA) OGEJCPKUOCPFEQPUVKVWVGUVJG4'5%QP|ITQWRYJKEJKUCECVGIQT[VJCVCNUQKPENWFGUOGEJC-
genes (which provide resistance to oxa- nisms of resistance to antibiotics that have been used in the past but are rarely used today. The
zolidinones) or mutations in other genes, highest risk categories are those that include genes (novel or already described) that confer resist-
such as fmt transformylase or defB deform- ance to antibiotics currently in use and that are on mobile genetic elements. These can be present
ylase (which provide resistance to the new KPPQPRCVJQIGPU 4'5%QP| QTKPJWOCPRCVJQIGPU|t|DQVJRTKOCT[CPFQRRQTVWPKUVKERCVJQIGPU
peptide-deformylase inhibitors) cannot be 4'5%QP| 

120 | FEBRUARY 2015 | VOLUME 13 www.nature.com/reviews/micro

© 2015 Macmillan Publishers Limited. All rights reserved

 PERSPECTIVES

considered as true resistance genes unless compared with that of the widespread genes established in the population. A RESCon 1
their transferability is demonstrated. In this of the family to ascertain the possibilities of gene should have already surpassed the
category, we also include antibiotic resist- selection by a given antibiotic for which the problems associated with high fitness costs,
ance genes for old antibiotics that have only former members of the family do not confer as it has spread among bacterial pathogens;
minimal or residual use; for example, genes resistance. If the gene family is absent in however, some genes are confined to spe-
that mediate resistance to d-cycloserine some geographical areas or specific clonal cific bacterial species (or clones) and the
that were identified in strains of the genus lineages, a risk at the local level is possibile. possibility of species-specific fitness costs
Serratia provide resistance when transferred In this scenario, the possibilities of synergy should be explored. Concerning the other
to E. coli 4. If these antibiotics are eventually with known mechanisms should be tested, genes, which are just predicted to confer
removed from the market, monitoring for as described above for RESCon 4. resistance in the future, studies on fitness
the continued presence of these genes could costs and compensation are mandatory to
help to evaluate whether they are maintained RESCon 6. RESCon 6 comprises genes that clearly define the risk.
in the absence of antibiotic selective pres- are predicted to confer resistance and are The second factor that affects the risk
sure. If these antibiotics are reintroduced for present on mobile genetic elements; the pre- associated with resistance genes that are iden-
therapy, these genes should be categorized as diction of resistance is on the basis that they tified in metagenomic analyses is the original
RESCon 1. share a sufficiently high sequence identity environment where the gene was found. The
with functionally characterized resistance risks rank from highest, in the case of an
RESCon 4. RESCon 4 comprises function- genes and their documented presence on environment that is closely related to human
ally characterized genes that confer resist- mobile genetic elements30. The only differ- pathogens (such as the human microbiota,
ance to an antimicrobial agent in clinical use ence between RESCon 2 and RESCon 6 is food or human sewage), to the lowest, in the
by an unknown mechanism, but resistance that the ability of RESCon 6 genes to confer case of habitats that are never colonized by
to this agent is already known to occur in resistance has not been functionally evalu- human pathogens. It is true that genes that
human bacterial pathogens by other mecha- ated but, given the current possibilities of are capable of conferring antibiotic resist-
nisms. One example of this risk category gene synthesis, a functional evaluation could ance have been found in habitats that are
is the detection of a gene that encodes a be carried out to verify their classification rarely colonized by human pathogens, such
product that inactivates rifamycin in envi- as RESCon 2. As aforementioned, given that as the deep subsurface65. However, although
ronmental and clinically relevant bacteria62. expression of resistance genes can be host- finding potential resistance elements in this
The risk such genes pose is inversely pro- specific54, the testing of different hosts may type of ecosystem may have local ecological
portional to the density and frequency of the be required when expression in a single host relevance, it is unlikely that these elements
known mechanisms of resistance in different does not confer resistance30. will pose a risk for the emergence of resist-
geographical areas and clonal lineages (see ance in the future (and consequently thera-
RESCon 5), which can affect the chances of RESCon 7. RESCon 7 comprises genes that peutic failure) of human pathogens, as they
dissemination of the novel gene because are predicted to confer resistance without are ecologically disconnected.
of the founder effect (BOX 2). The risk might any known association with mobile genetic
be affected by possible additive or synergistic elements. These genes are predicted to con- Summary
effects between the novel mechanism and fer antibiotic resistance because they share a The accumulation of genomic informa-
the known mechanism, or the ability to certain degree of identity with functionally tion relating to the presence of presump-
spread in a wider range of bacteria. characterized resistance genes, but further tive resistance genes in a vast number of
information is not available8. For the most natural environments does not always add
RESCon 5. RESCon 5 comprises genes that promising candidates, a synthetic approach, substantial knowledge to our understand-
belong to a known and widespread gene such as that suggested for RESCon 6 genes, ing of the origin, evolution and spread of
family and can inactivate specific antibiotics, must be considered. antibiotic resistance in human pathogens. On
thus presenting a highly similar antibiotic Of course, low risk does not mean no risk the contrary, similar to what has happened
substrate profile to known resistance genes. and, in fact, genes in RESCon 1 may have with several other comprehensive databases,
One example of this category is the detection first been placed in one of the other catego- the available resistance gene databases are
of a previously unknown β-lactamase with ries. For instance, genes that encode Qnr becoming increasingly populated with
a substrate profile that was already found in proteins that confer resistance to quinolones sequences of putative resistance genes, many
known β-lactamases; for example, the unex- do not resemble any other known resistance of which are unlikely to be involved in
pected array of ‘new’ enzymes that were dis- gene; their detection as an antibiotic resist- clinical antibiotic resistance. As new
covered in metagenomic intestinal samples ance gene in a sequence-based metagenomic sequences are annotated by comparison
in one study 6. These widespread genes are analysis would not be possible before their with existing genes, the number of presump-
expected to produce a founder-effect bar- discovery in clinical strains. tive resistance genes grows accordingly.
rier, which reduces the possibility of selec- Once the potential risks associated with Efforts should be made to clearly define the
tion and fixation of the novel mechanism of each type of gene have been defined by public health risks that are associated with
resistance; that is, if a bacterium is already classifying it in a particular RESCon, two the different types of genes that have been
resistant to a given antibiotic, this antibiotic other factors influence the risk of finding identified as resistance genes in the available
does not exert the selective pressure that is such a gene in a given habitat. The first is databases. In this Opinion article, we have
required for the acquisition, fixation and the fitness cost 63,64; only those determinants proposed a hierarchical risk system to dif-
spread of a novel resistance mechanism that confer low fitness costs or those that ferentiate these genes, which might help to
(BOX 2). The resistance phenotype that is pro- confer fitness costs that are easily amelio- eliminate unwanted noise in this important
vided by the new gene should be carefully rated by compensatory mutations will be field of research.

NATURE REVIEWS | MICROBIOLOGY VOLUME 13 | FEBRUARY 2015 | 121

© 2015 Macmillan Publishers Limited. All rights reserved

PERSPECTIVES

José L. Martínez is at the Departamento 21. Kahlmeter, G. et al. European harmonization of MIC 46. Hiramatsu, K. et al. Genomic basis for methicillin
de Biotecnología Microbiana, Centro Nacional de breakpoints for antimicrobial susceptibility testing of resistance in Staphylococcus aureus. Infect.
bacteria. J. Antimicrob. Chemother. 52, 145–148 Chemother. 45, 117–136 (2013).
Biotecnología, Consejo Superior de Investigaciones
(2003). 47. Martinez, J. L. Natural antibiotic resistance and
Científicas (CSIC), Darwin 3, Cantoblanco, 22. Simjee, S., Silley, P., Werling, H. O. & Bywater, R. contamination by antibiotic resistance determinants:
Madrid 28049, Spain; and the Unidad de Resistencia Potential confusion regarding the term ‘resistance’ in the two ages in the evolution of resistance to
a Antibióticos y Virulencia Bacteriana asociada al epidemiological surveys. J. Antimicrob. Chemother. 61, antimicrobials. Front. Microbiol. 3, 1 (2012).
Consejo Superior de Investigaciones Científicas (CSIC), 228–229 (2008). 48. Martinez, J. L. The role of natural environments in the
23. Morrissey, I. et al. Evaluation of epidemiological cut-off evolution of resistance traits in pathogenic bacteria.
Madrid, Spain.
values indicates that biocide resistant subpopulations Proc. Biol. Sci. 276, 2521–2530 (2009).
Teresa M. Coque and Fernando Baquero are at the are uncommon in natural isolates of clinically-relevant 49. Martinez, J. L., Baquero, F. & Andersson, D. I.
Unidad de Resistencia a Antibióticos y Virulencia microorganisms. PLoS ONE 9, e86669 (2014). Predicting antibiotic resistance. Nature Rev. Microbiol.
Bacteriana asociada al Consejo Superior de 24. Garcia-Leon, G., Salgado, F., Oliveros, J. C., 5, 958–965 (2007).
Sanchez, M. B. & Martinez, J. L. Interplay between 50. Hachler, H., Cohen, S. P. & Levy, S. B. marA, a
Investigaciones Científicas (CSIC), Madrid, Spain; intrinsic and acquired resistance to quinolones in regulated locus which controls expression of
and the Departamento de Microbiología, Hospital Stenotrophomonas maltophilia. Environ. Microbiol. chromosomal multiple antibiotic resistance in
Universitario Ramón y Cajal, Instituto Ramón y Cajal 16, 1282–1296 (2014). Escherichia coli. J. Bacteriol. 173, 5532–5538 (1991).
de Investigación Sanitaria (IRYCIS), 25. Alvarez-Ortega, C., Wiegand, I., Olivares, J., 51. Li, X. Z., Nikaido, H. & Poole, K. Role of MexA–MexB
Hancock, R. E. & Martinez, J. L. Genetic determinants –OprM in antibiotic efflux in Pseudomonas aeruginosa.
Madrid 28034, Spain.
involved in the susceptibility of Pseudomonas Antimicrob. Agents Chemother. 39, 1948–1953
Correspondence to J.L.M aeruginosa to ß-lactam antibiotics. Antimicrob. Agents (1995).
e-mail: jlmtnez@cnb.csic.es Chemother. 54, 4159–4167 (2010). 52. Costa, Y., Galimand, M., Leclercq, R., Duval, J. &
26. Girgis, H. S., Hottes, A. K. & Tavazoie, S. Genetic Courvalin, P. Characterization of the chromosomal
Published online 15 December 2014 architecture of intrinsic antibiotic susceptibility. PLoS aac(6ʹ)-Ii gene specific for Enterococcus faecium.
doi:10.1038/nrmicro3399 ONE 4, e5629 (2009). Antimicrob. Agents Chemother. 37, 1896–1903
27. Fajardo, A. et al. The neglected intrinsic resistome of (1993).
1. World Health Organization. Antimicrobial Resistance: bacterial pathogens. PLoS ONE 3, e1619 (2008). 53. Fevre, C. et al. Six groups of the OXY ß-lactamase
Global Report on Surveillance. (WHO Press, 2014). 28. Fernandez, L. et al. Characterization of the evolved over millions of years in Klebsiella oxytoca.
2. Kesselheim, A. S. & Outterson, K. Fighting antibiotic polymyxin B resistome of Pseudomonas aeruginosa. Antimicrob. Agents Chemother. 49, 3453–3462
resistance: marrying new financial incentives to Antimicrob. Agents Chemother. 57, 110–119 (2013). (2005).
meeting public health goals. Health Aff. (Millwood) 29, 29. Forsberg, K. J. et al. Bacterial phylogeny structures 54. Dantas, G. & Sommer, M. O. Context matters — the
1689–1696 (2010). soil resistomes across habitats. Nature 509, 612–616 complex interplay between resistome genotypes and
3. Baquero, F. Metagenomic epidemiology: a public (2014). resistance phenotypes. Curr. Opin. Microbiol. 15,
health need for the control of antimicrobial resistance. 30. Parsley, L. C. et al. Identification of diverse 577–582 (2012).
Clin. Microbiol. Infect. 18 (Suppl. 4), 67–73 (2012). antimicrobial resistance determinants carried on 55. Zscheck, K. K. & Murray, B. E. Genes involved in the
4. Allen, H. K. et al. Call of the wild: antibiotic resistance bacterial, plasmid, or viral metagenomes from an regulation of ß-lactamase production in enterococci
genes in natural environments. Nature Rev. Microbiol. activated sludge microbial assemblage. Appl. Environ. and staphylococci. Antimicrob. Agents Chemother. 37,
8, 251–259 (2010). Microbiol. 76, 3753–3757 (2010). 1966–1970 (1993).
5. Forsberg, K. J. et al. The shared antibiotic resistome of 31. Zankari, E. et al. Identification of acquired 56. Enne, V. I., Delsol, A. A., Roe, J. M. & Bennett, P. M.
soil bacteria and human pathogens. Science 337, antimicrobial resistance genes. J. Antimicrob. Evidence of antibiotic resistance gene silencing in
1107–1111 (2012). Chemother. 67, 2640–2644 (2012). Escherichia coli. Antimicrob. Agents Chemother. 50,
6. Sommer, M. O., Dantas, G. & Church, G. M. Functional 32. Scaria, J., Chandramouli, U. & Verma, S. K. Antibiotic 3003–3010 (2006).
characterization of the antibiotic resistance reservoir Resistance Genes Online (ARGO): a database on 57. Yong, D. et al. Characterization of a new metallo-ß-
in the human microflora. Science 325, 1128–1131 vancomycin and ß-lactam resistance genes. lactamase gene, blaNDM-1, and a novel erythromycin
(2009). Bioinformation 1, 5–7 (2005). esterase gene carried on a unique genetic structure in
7. D’Costa, V. M., McGrann, K. M., Hughes, D. W. & 33. Modi, S. R., Lee, H. H., Spina, C. S. & Collins, J. J. Klebsiella pneumoniae sequence type 14 from India.
Wright, G. D. Sampling the antibiotic resistome. Antibiotic treatment expands the resistance reservoir Antimicrob. Agents Chemother. 53, 5046–5054
Science 311, 374–377 (2006). and ecological network of the phage metagenome. (2009).
8. Ghosh, T. S., Gupta, S. S., Nair, G. B. & Mande, S. S. Nature 499, 219–222 (2013). 58. Picao, R. C. et al. Plasmid-mediated quinolone
In silico analysis of antibiotic resistance genes in the 34. Liu, B. & Pop, M. ARDB — Antibiotic Resistance resistance in Aeromonas allosaccharophila recovered
gut microflora of individuals from diverse geographies Genes Database. Nucleic Acids Res. 37, D443–D447 from a Swiss lake. J. Antimicrob. Chemother. 62,
and age-groups. PLoS ONE 8, e83823 (2013). (2009). 948–950 (2008).
9. Hu, Y. et al. Metagenome-wide analysis of antibiotic 35. Forslund, K. et al. Country-specific antibiotic use 59. Jones, B. V. & Marchesi, J. R. Transposon-aided
resistance genes in a large cohort of human gut practices impact the human gut resistome. Genome capture (TRACA) of plasmids resident in the human
microbiota. Nature Commun. 4, 2151 (2013). Res. 23, 1163–1169 (2013). gut mobile metagenome. Nature Methods 4, 55–61
10. Durso, L. M., Miller, D. N. & Wienhold, B. J. 36. Davies, J. Inactivation of antibiotics and the (2007).
Distribution and quantification of antibiotic resistant dissemination of resistance genes. Science 264, 60. Marcone, G. L. et al. Novel mechanism of glycopeptide
genes and bacteria across agricultural and non- 375–382 (1994). resistance in the A40926 producer Nonomuraea sp.
agricultural metagenomes. PLoS ONE 7, e48325 37. Benveniste, R. & Davies, J. Aminoglycoside antibiotic- ATCC 39727. Antimicrob. Agents Chemother. 54,
(2012). inactivating enzymes in actinomycetes similar to those 2465–2472 (2010).
11. D’Costa, V. M. et al. Antibiotic resistance is ancient. present in clinical isolates of antibiotic-resistant 61. Gu, B., Kelesidis, T., Tsiodras, S., Hindler, J. &
Nature 477, 457–461 (2011). bacteria. Proc. Natl Acad. Sci. USA 70, 2276–2280 Humphries, R. M. The emerging problem of linezolid-
12. Finley, R. L. et al. The scourge of antibiotic resistance: (1973). resistant Staphylococcus. J. Antimicrob. Chemother.
the important role of the environment. Clin. Infect. Dis. 38. Martinez, J. L. et al. A global view of antibiotic 68, 4–11 (2013).
57, 704–710 (2013). resistance. FEMS Microbiol. Rev. 33, 44–65 (2009). 62. Spanogiannopoulos, P., Waglechner, N., Koteva, K. &
13. Ferrandiz, M. J., Fenoll, A., Linares, J. & De La 39. Martinez, J. L. et al. Functional role of bacterial Wright, G. D. A rifamycin inactivating
Campa, A. G. Horizontal transfer of parC and gyrA in multidrug efflux pumps in microbial natural ecosystems. phosphotransferase family shared by environmental
fluoroquinolone-resistant clinical isolates of FEMS Microbiol. Rev. 33, 430–449 (2009). and pathogenic bacteria. Proc. Natl Acad. Sci. USA
Streptococcus pneumoniae. Antimicrob. Agents 40. Piddock, L. J. Multidrug-resistance efflux pumps 111, 7102–7107 (2014).
Chemother. 44, 840–847 (2000). — not just for resistance. Nature Rev. Microbiol. 4, 63. Andersson, D. I. & Hughes, D. Persistence of antibiotic
14. Baquero, F., Alvarez-Ortega, C. & Martinez, J. L. 629–636 (2006). resistance in bacterial populations. FEMS Microbiol.
Ecology and evolution of antibiotic resistance. Environ. 41. Garcia-Leon, G. et al. A function of SmeDEF, the major Rev. 35, 901–911 (2011).
Microbiol. Rep. 1, 469–476 (2009). quinolone resistance determinant of 64. Andersson, D. I. & Hughes, D. Antibiotic resistance
15. Martinez, J. L. & Baquero, F. Mutation frequencies Stenotrophomonas maltophilia, is the colonization of and its cost: is it possible to reverse resistance?
and antibiotic resistance. Antimicrob. Agents the roots of the plants. Appl. Environ. Microbiol. Nature Rev. Microbiol. 8, 260–271 (2010).
Chemother. 44, 1771–1777 (2000). http://dx.doi.org/10.1128/AEM.01058-14 (2014). 65. Brown, M. G. & Balkwill, D. L. Antibiotic resistance in
16. Laxminarayan, R. et al. Antibiotic resistance — the 42. Thanassi, D. G., Cheng, L. W. & Nikaido, H. Active bacteria isolated from the deep terrestrial subsurface.
need for global solutions. Lancet Infect. Dis. 13, efflux of bile salts by Escherichia coli. J. Bacteriol. Microb. Ecol. 57, 484–493 (2009).
1057–1098 (2013). 179, 2512–2518 (1997). 66. Warinner, C. et al. Pathogens and host immunity in the
17. Bush, K. et al. Tackling antibiotic resistance. Nature 43. Henderson, T. A., Young, K. D., Denome, S. A. & ancient human oral cavity. Nature Genet. 46, 336–344
Rev. Microbiol. 9, 894–896 (2011). Elf, P. K. AmpC and AmpH, proteins related to the (2014).
18. Baquero, F. Low-level antibacterial resistance: a class C ß-lactamases, bind penicillin and contribute to 67. Davies, J. & Davies, D. Origins and evolution of
gateway to clinical resistance. Drug Resist. Updat. 4, the normal morphology of Escherichia coli. antibiotic resistance. Microbiol. Mol. Biol. Rev. 74,
93–105 (2001). J. Bacteriol. 179, 6112–6121 (1997). 417–433 (2010).
19. Baquero, F. European standards for antibiotic 44. Ayala, F. J. Adaptation and novelty: teleological 68. Martinez, J. L. Bottlenecks in the transferability of
susceptibility testing: towards a theoretical consensus. explanations in evolutionary biology. Hist. Philos. Life antibiotic resistance from natural ecosystems to
Eur. J. Clin. Microbiol. Infect. Dis. 9, 492–495 (1990). Sci. 21, 3–33 (1999). human bacterial pathogens. Front. Microbiol. 2, 265
20. Kronvall, G. Normalized resistance interpretation as a 45. Laskaris, P., Tolba, S., Calvo-Bado, L. & Wellington, L. (2011).
tool for establishing epidemiological MIC susceptibility Coevolution of antibiotic production and counter- 69. Aminov, R. I. & Mackie, R. I. Evolution and ecology of
breakpoints. J. Clin. Microbiol. 48, 4445–4452 resistance in soil bacteria. Environ. Microbiol. 12, antibiotic resistance genes. FEMS Microbiol. Lett.
(2010). 783–796 (2010). 271, 147–161 (2007).

122 | FEBRUARY 2015 | VOLUME 13 www.nature.com/reviews/micro

© 2015 Macmillan Publishers Limited. All rights reserved

 PERSPECTIVES

70. Skippington, E. & Ragan, M. A. Lateral genetic transfer genetics of adaptive evolution in bacteria. Genetics
and the construction of genetic exchange communities. 154, 985–997 (2000).
FEMS Microbiol. Rev. 35, 707–735 (2011). 79. Steinkraus, G., White, R. & Friedrich, L. Vancomycin
71. Cohen, O., Gophna, U. & Pupko, T. The complexity MIC creep in non-vancomycin-intermediate
hypothesis revisited: connectivity rather than function Staphylococcus aureus (VISA), vancomycin-susceptible
constitutes a barrier to horizontal gene transfer. clinical methicillin-resistant S. aureus (MRSA) blood
Mol. Biol. Evol. 28, 1481–1489 (2011). isolates from 2001–2005. J. Antimicrob. Chemother.
72. Baquero, F., Tedim, A. P. & Coque, T. M. Antibiotic 60, 788–794 (2007).
resistance shaping multi-level population biology of 80. Gibson, M. K., Forsberg, K. J. & Dantas, G. Improved
bacteria. Front. Microbiol. 4, 15 (2013). annotation of antibiotic resistance determinants
73. Levin, B. R. & Bull, J. J. Short-sighted evolution and reveals microbial resistomes cluster by ecology.
the virulence of pathogenic microorganisms. Trends ISME J. http://dx.doi.org/10.1038/ismej.2014.106
Microbiol. 2, 76–81 (1994). (2014).
74. Sibold, C. et al. Mosaic pbpX genes of major clones of
penicillin-resistant Streptococcus pneumoniae have Acknowledgements
evolved from pbpX genes of a penicillin-sensitive Research in the author’s laboratories is funded by the European
Streptococcus oralis. Mol. Microbiol. 12, 1013–1023 Commission (EvoTAR-282004 for JLM, TMC and FB), the
(1994). Ministry of Economy and Competitiveness (BIO2011-25255
75. Martinez, J. L., Baquero, F. & Andersson, D. I. for JLM, PI12-01581 for TMC, PI10-02588 for FB, and
Beyond serial passages: new methods for predicting NEXTMICRO for FB and TMC), the Regional Government of
the emergence of resistance to novel antibiotics. Madrid in Spain (PROMPT-S2010/BMD2414 for JLM and FB),
Curr. Opin. Pharmacol. 11, 439–445 (2011). and by the Spanish Network for Research on Infectious
76. Olivares, J. et al. Overproduction of the multidrug Diseases (REIPI RD12/0015 for JLM) and the Spanish Network
efflux pump MexEF–OprN does not impair for the Study of Plasmids and Extrachromosomal Elements
Pseudomonas aeruginosa fitness in competition tests, (REDEEX, BFU2011-14145-E for TMC).
but produces specific changes in bacterial regulatory
Competing interests statement
networks. Environ. Microbiol. 14, 1968–1981 (2012).
The authors declare no competing interests.
77. Sanchez, M. B. & Martinez, J. L. Differential epigenetic
compatibility of qnr antibiotic resistance determinants
with the chromosome of Escherichia coli. PLoS ONE 7, SUPPLEMENTARY INFORMATION
e35149 (2012). See online article: S1 (table)
78. Levin, B. R., Perrot, V. & Walker, N. Compensatory ALL LINKS ARE ACTIVE IN THE ONLINE PDF
mutations, antibiotic resistance and the population

NATURE REVIEWS | MICROBIOLOGY VOLUME 13 | FEBRUARY 2015 | 123

© 2015 Macmillan Publishers Limited. All rights reserved

SUPPLEMENTARY INFORMATION In format provided by Martínez et al. (FEBRUARY 2015)

Supplementary information S1 (table) | Antibiotic resistance gene databases

Database Hosted by Repository Characteristics Redundancy Updated Curated Website Refs
ARG-ANOTT The Méditerranée Infection Acquired genes that are able Gene repository No Yes Yes http://en.mediterranee- 1
(Antibiotic resistance Foundation, to confer antibiotic resistance* infection.com/article.
A bioinformatic tool that is able to
gene-annotation) France (complete or partially sequences) php?laref=283&titre=arg-
detect existing and putative new
annot-
A distinction is made between antibiotic resistance genes
acquired genes and mutated genes
A local BLAST program in Bio-Edit
software that enables sequence
analysis without web interface
CARD A Canada–UK joint health Acquired genes that are able Gene repository Yes Yes No http://arpcard.mcmaster.ca/ 2
(The comprehensive research programme on to confer antibiotic resistance*
A bioinformatic tool that is able to
antibiotic resistance antibiotic resistance (complete or partially sequences)
detect existing and putative new
database)
No distinction is made between antibiotic resistance genes
acquired genes and mutated genes
GMOD-based database, which enables
interactive visualization and is analysed
by Gbrowse
Sequences are organized and classified
by sequence ontology, which provides
different levels of organization and
relationships among sequences
RED‑DB University of Siena, Italy Acquired genes that are able Gene repository Yes Yes No http://www.fibim.unisi.it/
(Resistance to confer antibiotic resistance* REDDB/
Funded by the A bioinformatic tool that is able to
determinants (complete or partially sequences)
EU‑FP7/2007‑ detect existing and putative new
database)
2013 -241742, TEMPOTest‑ No distinction is made between antibiotic resistance genes
QC project acquired genes and mutated genes
ResFINDER Center for Genomic Acquired genes that are able Gene repository Yes Yes Yes http://cge.cbs.dtu.dk/ 3
Epidemiology, Denmark to confer antibiotic resistance* services/ResFinder/
A bioinformatic tool that is able to
University, Denmark (complete or partially sequences, as
detect existing and putative new
well as whole genomes and plasmid
antibiotic resistance genes (ResFinder
sequences)
1.4), as well as plasmids replicons
No distinction is made between (PlasmidFinder) and species based on
acquired genes and mutated genes MLST alleles (MLST finder)
Resfam Center for Genome Sciences Protein families and associated Repository of proteins and HMM Yes Yes Yes http://www.dantaslab.org/ 4
and Systems Biology, profile HMMs, which are organized profiles resfams/
Washington University in St. by ontology and confirmed for
Louis School of Medicine, antibiotic resistance function
USA (Dantas’ laboratory)
ARDB Center for Bioinformatics Acquired genes that are able Gene repository Yes No Yes http://ardb.cbcb.umd.edu/ 3
(Antibiotic resistance and Computational Biology, to confer antibiotic resistance* (last update:
A bioinformatic tool that is able to
genes database) University of Maryland, USA (complete or partially sequences) July 3 2009)
detect existing and putative new
No distinction is made between antibiotic resistance genes
acquired genes and mutated genes
Lahey Clinic website Lahey Clinic, USA β-lactamases (genes that encode Focuses on nomenclature. A proposed No Yes No http://www.lahey.org/
Curators: Karen Bush and TEM, SHV, OXA-, CTX‑M-, CMY, numbering scheme in an attempt to Studies/
George Jacoby AmpC, CARB-, IMP-, VIM-, KPC, GES-, avoid duplication
PER- and VEB-, among others) No BLAST tool
Quinolones (qnr genes)

NATURE REVIEWS | MICROBIOLOGY www.nature.com/reviews/micro

SUPPLEMENTARY INFORMATION In format provided by Martínez et al. (FEBRUARY 2015)

LacED Institute of Technical Genes that encode TEM and SHV Focuses on nomenclature and the No No No http://www.laced. 5,6
(The lactamase Biochemistry, University of β-lactamases identification of known and unknown uni-stuttgart.de/
engineering Stuttgart, Germany variants
database)
Provides information on mutations,
sequences and structures of TEM and
SHV β-lactamases
MBLED Institute of Technical Metallo‑β-lactamases Provides information on mutations, Yes No Yes http://www.mbled. 7
(Metallo-beta- Biochemistry, University of sequences and structures of TEM and uni-stuttgart.de/
lactamase Stuttgart, Germany SHV β-lactamases
engineering
No BLAST tool
database)
OXY, OKP, LEN Institute Pasteur, France OXY, OKP and LEN β-lactamases Focuses on harmonization of gene No No Yes http://www.pasteur.fr/
beta-lactamase nomenclature ip/easysite/pasteur/en/
(no acquired antibiotic resistance
protein variation research/plates-formes-
genes) No BLAST tool
Home Page technologiques/pasteur-
genopole-ile-de-france/
genotyping-of-pathogens-
and-public-health-pf8/beta-
lactamase-enzyme-variants/
beta-lactamase-enzyme-
variants
Marylin Roberts University of Washington, Macrolides and tetracyclines Focuses on harmonization of gene No Yes No http://faculty.washington.
personal website USA nomenclature (not edu/marilynr/
systematically
No BLAST tool
updated)
Attaca-RAC Centre for Health Informatics Antibiotic resistance gene cassettes An archive of gene cassettes that Yes Yes Yes http://rac.chi.unsw.edu.au/ 8
(Repository of at UNSW, Australia, in in multidrug resistance integrons includes alternative gene names from rac/
antibiotic resistance collaboration with the Centre multiple nomenclature systems
cassettes) for Infectious Diseases and
Enables consistent annotation and
Microbiology, Westmead, UK
naming of cassettes in submitted
sequences
INTEGRALL University of Aveiro, Portugal Integrons A repository for sequence data Yes Yes Yes http://integrall.bio.ua.pt/ 9
(Repository of and nomenclature that provides
antibiotic resistance researchers with easy and interactive
cassettes) access to DNA sequences and
molecular arrangements of integrons,
as well as their genetic contexts
BLAST, basic local alignment search tool; GMOD, Generic Model Organisms Database; HMM, hidden Markov model; MLST, multilocus sequence typing; UNSW, University of New South Wales.
*Acquired genes that are able to confer antibiotic resistance include: β-lactams, glycopeptides, aminoglycosides, tetracyclines, sulphonamides, macrolides, lincosamides, streptogramins, oxazolidinones and quinolones.

1. Gupta, S. K. et al. ARG-ANNOT, a new bioinformatic tool to discover antibiotic resistance genes in bacterial genomes. Antimicrob. Agents Chemother. 58, 212–220 (2014).
2. McArthur, A. G. et al. The comprehensive antibiotic resistance database. Antimicrob. Agents Chemother. 57, 3348–3357 (2013).
3. Zankari, E. et al. Identification of acquired antimicrobial resistance genes. J. Antimicrob. Chemother. 67, 2640–2644 (2012).
4. Gibson, M. K., Forsberg, K. J. & Dantas, G. Improved annotation of antibiotic resistance determinants reveals microbial resistomes cluster by ecology. ISME J. http://dx.doi.org/10.1038/ismej.2014.106 (2014).
5. Thai, Q. K. & Pleiss, J. SHV Lactamase Engineering Database: a reconciliation tool for SHV ß-lactamases in public databases. BMC Genomics 11, 563 (2010).
6. Thai, Q. K., Bos, F. & Pleiss, J. The Lactamase Engineering Database: a critical survey of TEM sequences in public databases. BMC Genomics 10, 390 (2009).
7. Widmann, M., Pleiss, J. & Oelschlaeger, P. Systematic analysis of metallo-ß-lactamases using an automated database. Antimicrob. Agents Chemother. 56, 3481–3491 (2012).
8. Tsafnat, G., Copty, J. & Partridge, S. R. RAC: Repository of Antibiotic resistance Cassettes. Database (Oxford) 2011, bar054 (2011).
9. Moura, A. et al. INTEGRALL: a database and search engine for integrons, integrases and gene cassettes. Bioinformatics 25, 1096–1098 (2009).

NATURE REVIEWS | MICROBIOLOGY www.nature.com/reviews/micro