OUP user menu

Metabolic regulation of antibiotic resistance

José L. Martínez , Fernando Rojo
DOI: http://dx.doi.org/10.1111/j.1574-6976.2011.00282.x 768-789 First published online: 1 September 2011


It is generally assumed that antibiotics and resistance determinants are the task forces of a biological warfare in which each resistance determinant counteracts the activity of a specific antibiotic. According to this view, antibiotic resistance might be considered as a specific response to an injury, not necessarily linked to bacterial metabolism, except for the burden that the acquisition of resistance might impose on the bacteria (fitness costs). Nevertheless, it is known that changes in bacterial metabolism, such as those associated with dormancy or biofilm formation, modulate bacterial susceptibility to antibiotics (phenotypic resistance), indicating that there exists a linkage between bacterial metabolism and antibiotic resistance. The analyses of the intrinsic resistomes of bacterial pathogens also demonstrate that the building up of intrinsic resistance requires the concerted action of many elements, several of which play a relevant role in the bacterial metabolism. In this article, we will review the current knowledge on the linkage between bacterial metabolism and antibiotic resistance and will discuss the role of global metabolic regulators such as Crc in bacterial susceptibility to antibiotics. Given that growing into the human host requires a metabolic adaptation, we will discuss whether this adaptation might trigger resistance even in the absence of selective pressure by antibiotics.

  • persister
  • biofilm
  • signal transduction
  • bacterial metabolism
  • fitness cost
  • global regulation


The search for antibiotics produced by soil microorganisms was originally based on an ecological hypothesis. This hypothesis arose from the observation that bacterial pathogens are not predominant components of the environmental microbiota, although they have been constantly released from human disposals along history into water and soil. It was thus concluded that microorganisms from natural ecosystems should produce compounds capable of inhibiting the growth of these microbial pathogens (Waksman & Woodruff, 1940). The success of this approach led to the conclusion that the ecological role of naturally produced antibiotics should be to inhibit competitors. Given this role, resistance genes should have evolved to counteract antimicrobial action, so that the function of each of these determinants would be to avoid the activity of one specific antibiotic (or a family of antibiotics with similar structures). Under this scheme, antibiotics and their resistance genes are the task forces in a biological warfare and, given their specificity, their activity should not be necessarily linked to other elements of the bacterial metabolism.

Work with genes acquired by horizontal gene transfer (HGT) confirmed this idea. For instance, plasmid-encoded β-lactamases serve to resist β-lactam antibiotics and are not active against other kinds of drugs. The same applies for mutation-driven resistance; quinolone resistance is the consequence of mutations in topoisomerases genes, which do not alter bacterial susceptibility to other antibiotics. The only impact that the acquisition of resistance is usually supposed to have on bacterial physiology is to produce a general metabolic burden [fitness cost (Andersson & Levin, 1999)]. In spite of the specificity displayed by several resistance determinants (Alekshun & Levy, 2007), others like multidrug resistance (MDR) efflux pumps are rather nonspecific (Saier et al., 1998; Piddock, 2006; Martinez et al., 2009b). MDR efflux pumps are chromosomally encoded elements, highly conserved and present in all living beings (Nikaido, 1998; Vila & Martinez, 2008). They can efflux a large range of compounds including synthetic antibiotics that were not present in natural ecosystems before their invention by humankind (Alonso et al., 1999). This suggests that their original function is not counteracting the activity of these drugs (Martinez et al., 2009a, b).

Together with the proposal that some antibiotics might be signalling molecules at the low concentrations at which they are likely to be present in natural ecosystems (Linares et al., 2006; Yim et al., 2006, 2007; Fajardo & Martinez, 2008; Yergeau et al., 2010), these findings indicate that, at least in some instances, antibiotic resistance elements might have been primarily selected for playing roles relevant for microbial physiology, while their activity in avoiding the action of the antimicrobials is secondary to their original role (Martinez, 2008; Aminov, 2009; Baquero et al., 2009; Fajardo et al., 2009; Martinez et al., 2009a). This possibility is further supported by the recent finding that the building up of the phenotype of intrinsic resistance requires the concerted activity of a large number of elements, several of which play a primary role in the microbial physiology, including elements of the bacterial metabolic networks such as components of the electron transport chain, or of the metabolism of amino acids, fatty acids or nucleotides (Breidenstein et al., 2008; Fajardo et al., 2008; Tamae et al., 2008; Dotsch et al., 2009).

The modulation of antibiotic resistance by bacterial metabolism is also supported by data on phenotypic resistance, which is a transient situation of reduced susceptibility to antibiotics dependent on the metabolic state of the bacterial population (Levin & Rozen, 2006).

Concerning acquired resistance (either by HGT or by mutations), recent data indicate that its effect on bacterial fitness is more specific than supposed previously, in such a way that the acquisition of resistance can produce specific changes in the bacterial physiology rather than a general metabolic burden (Martinez et al., 2009a). These physiological changes include specific alterations in the bacterial metabolism that can even be adaptive for colonizing specific ecosystems. For instance, a Stenotrophomonas maltophilia mutant selected by antibiotic pressure (Alonso & Martinez, 1997), which overexpresses the MDR efflux pump SmeDEF (Alonso & Martinez, 2000), is more proficient that its wild-type counterpart in the use of sugars such as gentibiose, dextrin and mannose, as well as formic acid (Alonso et al., 2004). A similar specific effect of antibiotic resistance is observed for rifampin-resistant mutants of Bacillus subtilis. Resistance to this antibiotic is the consequence of mutations in the rpoB gene, which encodes the β-subunit of the RNA polymerase, and it has been reported that some mutants (but not others) can grow using β-glucosides, which are nutrients present in soil, more efficiently that their parental susceptible strain (Perkins & Nicholson, 2008). These examples illustrate that the metabolic alterations associated with the acquisition of resistance do not just consist of a nonspecific malfunction of the bacterial metabolism, but rather specific changes that derive from the mutations involved.

Understanding the metabolic regulation of bacterial susceptibility to antibiotics and the effect of acquired resistance on bacterial physiology (including the cellular metabolism) is relevant to the development of more accurate strategies for fighting infections. In this review, we will discuss available information on the crosstalk between antibiotic resistance and bacterial metabolism.

Phenotypic resistance

By phenotypic resistance, we refer to those transient situations in which a bacterial population, otherwise susceptible to a given antibiotic, is refractory to its action. This transient resistance does not require a genetic change and thus it is not inheritable (Levin & Rozen, 2006). We will not discuss here the induction of specific mechanisms of resistance [for instance induction of chromosomally encoded β-lactamases by β-lactams (Wiedemann et al., 1998)], although those mechanisms are transient as well, but just those that arise as a consequence of a metabolic shift.

The growth rate is the first parameter that impacts the phenotype of susceptibility to antibiotics of bacterial populations. The relevance of growth rate on the activity of penicillin was already described in 1944, in a study that showed that the activity of this antibiotic was impaired when cells grew slowly (Lee et al., 1944). However, this effect is not restricted to just β-lactam antibiotics. Indeed, it has been shown that resting cells are fully resistant to ampicillin or to tetracycline, whereas streptomycin or ciprofloxacin is still active against cells in the stationary phase, although their activity is lower than that observed for exponentially growing bacteria (Levin & Rozen, 2006). This situation has been named ‘drug indifference’ (Mc Dermott, 1958), and can be relevant for the persistence of bacterial infections even in patients under antibiotic treatment, when bacteria are in a host location that restricts growth or when the microorganisms have consumed the host resources and their growth is impaired. This can be particularly important in the case of long-lasting infections, because it was demonstrated that the concentration of antibiotics required to cure an experimental infection increases with the duration of the infection (Eagle, 1949). This difference might be due either to an increase in the number of bacterial cells, which also increases the probability of acquiring antibiotic resistance mutations (Martinez & Baquero, 2000), or to changes in bacterial metabolism that make bacteria phenotypically resistant.

As stated in Levin & Rozen (2006), the analysis of infections using a mixture of resistant and susceptible bacteria provides clues to resolve this question. Mice infected with a mixture of streptomycin resistant and susceptible Escherichia coli K1 strains were subjected to antibiotic treatment at the time of infection or 8 h after infection. In the first case, there was a large amplification of the population of resistant bacteria, indicating that the antibiotic was killing the susceptible ones as expected. In the delayed treatment, however, this amplification was not observed and susceptible bacteria remained after treatment, suggesting that during infection, bacteria underwent a metabolic shift that made them phenotypically resistant (Smith & Huggins, 1982; Bull et al., 2002). Further work suggested that, at least in this model of infection, bacteria no longer divide after the first 8 h of the infective process (Levin & Rozen, 2006), this resting situation being the reason for the observed ‘drug indifference’.

The existence of resting or slow-growing cells, which are refractory to treatment, is supposed to be the reason for the need to use prolonged regimes for treating infections by organisms such as Mycobacterium tuberculosis (McCune & Tompsett, 1956) or Mycobacterium leprae (Toman et al., 1981), as well as one of the causes of the relapses observed after antibiotic treatment for some bacterial infections (Fitoussi et al., 1997; Clement et al., 2005). In the case of M. tuberculosis, it has been described that hypoxia triggers dormancy and undergoes significant metabolic reprogramming, with upregulation of stress-related genes and downregulation of many central metabolism pathways (Chao & Rubin, 2010). Associated with these changes, bacteria display a lack of susceptibility to drugs that target the cell wall, such as isoniazid (Koul et al., 2011). However, the phenotype of resistance is not just the consequence of the nongrowing state of bacteria. Indeed, the shift from an aerobic to an anaerobic metabolism makes M. tuberculosis susceptible to drugs in use for treating infections by anaerobes (Fig. 1) such as metronidazole (Ginsberg, 2010). This indicates that, at least on some occasions, the susceptibility or lack the susceptibility of bacterial populations to a given antibiotic depends not just on whether bacteria are actively growing or resting, but on their specific metabolic situation at the time of treatment (in the example, aerobic or anaerobic). The relevance of understanding central bacterial metabolism in vivo for predicting the activity of antibiotics has also been highlighted in a recent paper that analyses the in vitro culture conditions required for the rational discovery of new antituberculosis drugs (Pethe et al., 2010). The article analysed a series of compounds that, despite excellent in vitro activity and desirable pharmacological properties, were found to be inactive in a tuberculosis mouse model. In-depth analyses of the reasons for this in vivo resistance to compounds highly active in vitro demonstrated that the differences were explained by the major differences in carbon metabolism between bacteria replicating in a standard tuberculosis broth medium as compared with infected lungs, demonstrating that the actual situation of the bacterial metabolism is highly relevant for bacterial susceptibility to antibiotics.


Examples of the metabolic modulation of phenotypic resistance. Alterations in the bacterial metabolism might alter microbial susceptibility to antibiotics. For instance, biofilms are complex structures that may contain DNA, which can chelate cations (Mulcahy et al., 2008) and hence trigger the PhoP–PhoQ response, thereby inducing resistance to antimicrobial peptides (1). Besides, bacterial biofilms harbour populations in different metabolic situations and with different oxygen availabilities and it has been described that anaerobiosis reduces susceptibility to some drugs, but increases susceptibility to others. This oxygen-dependent susceptibility has also been described for Mycobacterium tuberculosis, which, at low oxygen tension, is resistant to isoniazid (Koul et al., 2011), but is susceptible to metronidazol (Ginsberg, 2010), an antibiotic that is being used for treating infections by anaerobes (4). Carbon sources can also alter the susceptibility to antibiotics. It has been described that a Pseudomonas aeruginosa mutant defective in the Crc global regulator of carbon metabolism is more susceptible to imipenem and to fosfomycin (Linares et al., 2010) because it expresses at higher levels their membrane transporters OprD and GlpT, whose original function is the transport of basic amino acids and glycerol-3-phosphate, respectively (2). In the case of fosfomycin, it has also been demonstrated that Listeria monocytogenes is more susceptible to this antibiotic when growing intracellularly (Scortti et al., 2006), because in this allocation, the hexose phosphates are a good carbon source and their transporter Hpt, which also transports fosfomycin, is expressed at a higher level (2). Finally, it has been described recently that the use of some specific carbon sources can be useful for eradicating persister cells (yellow box). One of the characteristic features of persisters is their low proton-motive force (PMF), which impedes aminoglycosides entrance. Using specific sugars, NADH can be generated without the need for the TCA cycle, thereby fuelling the respiratory chain without growth resumption and allowing the recovery of PMF, the entry of aminoglycosides and the consequent death of persister cells (Allison et al., 2011).

Some evidences indicate that one of the signals relevant for phenotypic resistance is (p)ppGpp (Srivatsan & Wang, 2008; Pesavento & Hengge, 2009). This alarmone is the hallmark of the bacterial stringent response and its accumulation elicits several responses, including the strong downregulation of the expression of genes encoding rRNA and metabolic processes and upregulation of the expression of the routes involved in amino acid biosynthesis (Jain et al., 2006). The accumulation of p(ppGpp) also triggers the expression of determinants involved in stress survival, indicating that the alarmone works as a second messenger that allows a global switch on bacterial metabolism (Srivatsan & Wang, 2008). Besides triggering metabolic changes, (p)ppGpp modulates the bacterial response to antibiotics (Wu et al., 2010). For instance, it has been shown that this alarmone mediates vancomycin tolerance in Enterococcus faecalis (Abranches et al., 2009), and a role of (p)ppGpp in the emergence of persister subpopulations of M. tuberculosis has also been suggested (Warner & Mizrahi, 2006).

Growth rate is a broad, rather nonspecific cause of changes in the susceptibility to antibiotics. A more specific metabolic adaptation to a bacterial behaviour that might confer resistance refers to swarming motility. Swarming is a complex mechanism of adaptation with several genes involved (Overhage et al., 2007, 2008; Yeung et al., 2009). Swarming is the consequence of a metabolic shift induced upon nitrogen limitation and when certain amino acids, such as glutamate, aspartate, proline or histidine, are provided as the sole nitrogen source. It has been shown that swarmer cells can confer phenotypic resistance to antibiotics (Kim & Surette, 2003; Kim et al., 2003), and it was suggested that this reduced antibiotic susceptibility might be due to changes in the cell envelope associated with the metabolic shift that swarmer cells undergo (Overhage et al., 2008). It has been shown recently that the sensor kinase CbrA can provide a linkage between metabolism and antibiotic resistance in swarmer cells (Yeung et al., 2011). CbrA is a sensor kinase that, together with its cognate response regulator CbrB, is involved in the metabolic regulation of carbon and nitrogen utilization in Pseudomonas aeruginosa that controls the expression of a number of catabolic pathways involved in carbon and nitrogen utilization (Li & Lu, 2007). It has been shown that CbrA modulates the susceptibility to antibiotics of P. aeruginosa by regulating the expression of genes such as PA0621, pvdD, PA3784, pchH and pchF, which, besides their role in the bacterial metabolism, have been demonstrated to be relevant for the susceptibility of P. aeruginosa to quinolones (Yeung et al., 2011). CbrA also controls the arn operon, which encodes enzymes involved in the metabolism of bacterial lipopolysaccharide and is important in the susceptibility to polymyxin of P. aeruginosa (Yeung et al., 2011).

Two other situations that might confer phenotypic resistance (in this case just to a bacterial subpopulation) are persistence and growth in biofilms. Because these situations have been reviewed recently (Martinez et al., 2009a), we will just discuss them in brief. Persistence is a phenomenon by which a part of a given bacterial population provides a phenotype of ‘drug indifference’ to several antibiotics (Balaban et al., 2004; Kussell et al., 2005). Whereas it was firstly proposed that this situation might be due to the existence of a fraction of nongrowing cells into any bacterial population that is phenotypically resistant to antibiotics, more recent work indicates that there are several different mechanisms that can lead to persistence (Hansen et al., 2008). Metabolic enzymes, global regulators and toxin–antitoxin systems are among those elements that are relevant for developing a persister phenotype (Keren et al., 2004; Hansen et al., 2008). Among genes coding for metabolic enzymes, it was found that the knockout of ygfA, which codes for a putative 5-formyl-THF cycloligase involved in folate biosynthesis or of yigB, which may block metabolism by depleting the pool of flavin mononucleotide, decreases persistence (Lewis et al., 2010). Folate deficiency impairs the biosynthesis of purines, thymidilate and methionine. Overexpression of YgfA increased tolerance to ofloxacin (Keren et al., 2004; Hansen et al., 2008). These results support the existence of a linkage between some specific metabolic pathways and bacterial persistence. A recent report explores the possibilities that this linkage offers for inducing antibiotic susceptibility in persister cells (Allison et al., 2011). In the article, the authors demonstrate that persisters, although dormant, are primed for metabolite uptake, central metabolism and respiration, and that specific metabolic stimuli that allow the recovery of the proton-motive force without growth resumption enable bacterial killing by aminoglycosides (Fig. 1).

A similar situation might occur in bacteria growing attached to surfaces and forming biofilms. It has been suggested that biofilms are more resistant to several antibiotics. Among the reasons for this reduced susceptibility, it has been suggested that the structure of the biofilm itself precludes the entrance of the drugs (Suci et al., 1994), that MDR efflux pumps can contribute to biofilms resistance to antibiotics (Zhang & Mah, 2008) or that extracellular DNA chelates cations (Fig. 1) and this local low concentration of cations triggers in P. aeruginosa the expression of PhoP–PhoQ and PmrA, which regulate the arn operon involved in lipid A modification and antimicrobial peptide resistance, and hence makes bacteria growing in biofilms more resistant to antibiotics (Mulcahy et al., 2008). It was also suggested that biofilms have different microenvironments with different amounts of oxygen and nutrients, which is reflected in differences in the metabolism of the bacteria inhabiting each microenvironment (Huang et al., 1995; Sternberg et al., 1999). Indeed, it has been shown that biofilms contain different bacterial populations, including dead cells, dormant cells and actively growing cells. In the last case, two populations have been found growing either aerobically or fermentatively, which corresponds to two distinct metabolic states (Rani et al., 2007). Detailed analyses of P. aeruginosa biofilms showed that the response to antibiotics was highly dependent on the metabolic state of each population (Rani et al., 2007). For instance, aerobically growing bacteria were sensitive to ciprofloxacin, but resistant to polymyxin, whereas the opposite was found for cells growing in the deepest part of the biofilm, which present an anaerobic metabolism (Fig. 1). These results indicate that the phenotypic resistance of bacterial biofilms is not the consequence of a general metabolic shut-off. On the contrary, the activity of a particular antibiotic depends highly on the metabolic state of each population.

As discussed above, some metabolic shifts can make bacteria phenotypically resistant to a particular antibiotic. However, the opposite might occur as well, and some specific growing conditions can make bacteria more susceptible during infection than when growing in vitro. This situation has been described for Listeria monocytogenes. This intracellular pathogen grows inside mammal cells using hexose phosphates present in the host cytosol (Ripio et al., 1997; Joseph et al., 2006). For growing into host cells, Listeria upregulates a set of virulence determinants, whose expression is under control of the transcriptional regulator PrfA (Leimeister-Wachter et al., 1990). One of the PrfA-regulated determinants required for intracellular growth is the sugar phosphate transporter Hpt (Chico-Calero et al., 2002). This transporter is highly expressed during Listeria intracellular growth, but its expression is very low when bacteria grow in vitro in the media regularly used for susceptibility tests. It turned out that fosfomycin enters Listeria through this transporter. Under these circumstances, it might be predicted, and indeed it has been demonstrated (Scortti et al., 2006), that intracellularly growing Listeria cells are highly susceptible to fosfomycin as a consequence of their metabolic shift upon entering into their host cells (Fig. 1).

Intrinsic resistance

Intrinsic resistance is the specific phenotype of antibiotics' susceptibility that is common to all members of a given bacterial species (Sanchez et al., 2009). This phenotype has not been acquired recently as a consequence of selective pressure by antibiotics in clinical settings, but is rather an ancient phenotype specific for a given bacterial species (Sheldon et al., 2005). Classically, intrinsic resistance has been attributed to the lack of an effective target or to a reduced permeability to one or several drugs (Hogan & Kolter, 2002; Delcour et al., 2009). Besides these elements that preclude the action of the antibiotics, chromosomally encoded antibiotic-inactivating enzymes such as β-lactamases (Okamoto et al., 2001) or aminoglycoside-inactivating enzymes (Li et al., 2003), MDR efflux pumps (Hogan & Kolter, 2002) and target-protecting determinants such as chromosomally encoded Qnr proteins (Sanchez et al., 2008; Sanchez & Martinez, 2010) are antibiotic-detoxifying mechanisms that contribute to intrinsic resistance.

Whereas the first two elements are passive mechanisms of resistance that avoid either the interaction or the entrance of the antibiotics, the latter are active mechanisms, which detoxify the bacterial cell from one antibiotic that would otherwise be active. This might suggest that chromosomally encoded inactivating enzymes and MDR efflux pumps might have evolved to counteract the activity of antibiotics present in the ecosystems where these bacterial species thrive. Whereas this is likely true on some occasions [for instance antibiotic producers that require resistance elements to avoid the action of the antimicrobials they produce (Benveniste & Davies, 1973; Davies et al., 1997)], the evolutionary role of these determinants is unclear in other cases. For instance, several Enterobacteriaceae harbour chromosomally encoded β-lactamases (Lindberg & Normark, 1986) although the gut is not known to contain β-lactam producers. Because β-lactamases are structurally similar to the enzymes involved in the metabolism of peptidoglycan, which are the targets of β-lactam antibiotics (Massova & Mobashery, 1998; Meroueh et al., 2003), it might be speculated that these chromosomally encoded β-lactamases are involved in peptidoglycan recycling, resistance being a secondary role achieved just because the antibiotic is structurally similar to their natural substrates. Nevertheless, and although it has been described that AmpC contributes to the normal morphology of E. coli (Henderson et al., 1997), the natural role of these chromosomal β-lactamases, in the absence of an antibiotic challenge, remains elusive. A role in the metabolism of bacterial peptidoglycan has been demonstrated for the chromosomally encoded aminoglycoside acetyltransferase of Providencia stuartii, an enzyme involved in peptidoglycan recycling in this bacterial species, which is also capable of recognizing and inactivating gentamicin (Macinga & Rather, 1999).

The natural function of MDR efflux pumps has been reviewed recently in depth (Martinez et al., 2009b), so that we will just mention here that these elements, besides contributing to intrinsic resistance, play relevant roles in the bacterial physiology, including extrusion of metabolic intermediates (Aendekerk et al., 2002, 2005), trafficking of intercellular signal molecules (Evans et al., 1998; Kohler et al., 2001) or mediating plant–cell interactions (Burse et al., 2004; Maggiorani Valecillos et al., 2006). All these functions are relevant for the bacterial behaviour, illustrating that intrinsic resistance is not just a specific adaptive response to the presence of antibiotics in the bacterial natural ecosystem, but rather a consequence of the global bacterial physiology.

This hypothesis is further supported by a series of comprehensive studies on the elements involved in intrinsic resistance in E. coli (Tamae et al., 2008), P. aeruginosa (Breidenstein et al., 2008; Fajardo et al., 2008; Dotsch et al., 2009) or Acinetobacter (Gomez & Neyfakh, 2006), among others, which showed that a large number of genes [up to 3% of the bacterial genome (Fajardo et al., 2008)] can contribute to the characteristic phenotype of susceptibility to antibiotics of these bacterial pathogens. The determinants involved included not just classical resistance elements as those mentioned before, but several elements with a fundamental role in the basic processes of the bacterial metabolism. Furthermore, mutations on some of the genes coding for proteins involved in the cellular metabolism, such as rafDI (encoding ADP-l-glycero-d-mannose-6-epimerase) or gapA (encoding glyceraldehyde 3-phosphate dehydrogenase), made bacteria less susceptible to antibiotics belonging to different structural families (Fajardo et al., 2008), indicating that intrinsic resistance is linked to the bacterial metabolism and has a large degree of unspecificity.

Respiratory chain, oxidative stress, iron metabolism and the susceptibility to antibiotics

A common mechanism of action for different bactericidal antibiotics has been proposed recently (Kohanski et al., 2007). According to this proposal, many bactericidal antibiotics, upon interacting with their targets, would stimulate the production of hydroxyl radicals, which would ultimately contribute to cell death (Fig. 2). The hydroxyl radicals would be generated by means of the Fenton reaction, through a pathway that includes basic elements of the bacterial metabolism such as the tricarboxylic acid (TCA) cycle, the electron transport chain and the metabolism of iron (Kohanski et al., 2010). Because the activity of these metabolic networks is relevant for the activity of the antibiotics, alterations in bacterial metabolism should influence the susceptibility to these antimicrobials. In agreement with this statement, it has been described that E. coli mutants lacking isocitrate dehydrogenase (Helling & Kukora, 1971) or aconitase B are resistant to quinolones (Gruer et al., 1997).


Model of a common cell-death pathway for bactericidal antibiotics. It has been proposed that bactericidal antibiotics challenge bacterial metabolic networks that trigger bacterial death. These alterations begin with the hyperactivation of the electron transport chain, which generates an excess of intracellular superoxide. Alterations in the TCA cycle will influence this hyperactivation and hence modify the susceptibility to antibiotics. Superoxide damages iron–sulphur (FeS) clusters, making Fe2+ available and triggering the Fenton reaction, which leads to the formation of hydroxyl radicals that will damage macromolecules, rendering cell death. The model is presented in more detail in Kohanski et al. (2010).

The fact that alterations in the cell respiratory chain are relevant for the activity of aminoglycosides was demonstrated with the analysis of S. aureus small colony variants (SCVs). These variants have been described in several bacterial species and are slow-growing mutants with reduced susceptibility to aminoglycosides and presenting specific changes in their metabolism (Proctor et al., 2006). There are two groups of SCVs recovered at clinical settings. One group gathers bacteria deficient in thymidine biosynthesis, while the other one includes mutants deficient in electron transport through the respiratory chain (von Eiff et al., 1997; Clements et al., 1999; Bates et al., 2003). One relevant feature of S. aureus SCVs is their enhanced capability to persist inside host cells, which allows them to produce recurrent infections (Proctor et al., 1995).

Several findings support the idea that mutations in elements forming part of the metabolic networks that mediate antibiotic-induced cell death can impact the susceptibility to antibiotics. For example, a comprehensive analysis of an E. coli transposon-tagged mutant library showed that mutations in genes coding for elements of the respiratory chain and for elements relevant in the generation of oxygen radicals created by the Fenton reaction presented changes in the susceptibility to tobramycin (Girgis et al., 2009). Independent studies showed that those genes were also relevant for resistance to other aminoglycosides (Kohanski et al., 2008) and in different bacterial species (Schurek et al., 2008).

The fact that mutations in the respiratory chain are important for the susceptibility to aminoglycosides was described several years ago. However, this effect was attributed to a reduced uptake of the antibiotics by this type of mutants (Bryan & Van Den Elzen, 1977). The finding that mutants impairing the Fenton reaction are also relevant for aminoglycoside resistance favours the hypothesis that those mutations are relevant because they impact the antibiotic-mediated bacterial-death networks. However, it seems that several mutations at elements of the respiratory chain that alter the susceptibility to aminoglycosides do not impact the susceptibility of other bactericidal antibiotics such as quinolones or β-lactams (Girgis et al., 2009), so that more work is still needed to fully understand the basis of these resistance phenotypes.

Because Kohanski's model of antibiotic-mediated cell death implies the generation of hydroxyl radicals through the Fenton reaction, it might be predicted that bacterial iron metabolism is relevant for the activity of antibiotics and also that mutations in elements important for this metabolism might challenge bacterial antibiotic susceptibility. Indeed, it has been shown recently that the deletion of the gene coding for ferric reductase confers resistance to antibiotics in Pseudomonas, whereas its overexpression accelerates antibiotic-induced cell death (Yeom et al., 2010).

The research on antibiotic resistance has been mainly based on the implicit statement that the mechanisms involved are rather specific and somehow ‘independent’. Because of this, the effect of resistance on bacterial metabolism was explained in terms of ‘fitness costs’ usually understood as a general burden on the bacterial metabolism. Whereas the loss of a classical resistance gene (for instance, a chromosomally encoded β-lactamase) makes bacteria more susceptible to antibiotics and its overproduction increases resistance, the situation with metabolic genes might be different, because both their impairment and their overproduction can alter the metabolic networks. For instance, both the absence and the increased expression of the cyanide-insensitive CIO terminal oxidase make Pseudomonas more susceptible to a range of antibiotics including chloramphenicol, β-lactams, quinolones, aminoglycosides and macrolides (Tavankar et al., 2003).

Global metabolic regulators and susceptibility to antibiotics

Whereas the expression of HGT acquired resistance genes is frequently triggered by strong promoters present in gene-capture and gene-transfer elements (Stokes & Hall, 1989; Hall & Collis, 1995; Toleman et al., 2006), the situation for chromosomally encoded resistance genes is different. Those elements have coevolved with the entire bacterial genome in such a way that their expression is integrated within the host regulatory networks and the activity of the proteins they encode is also integrated in the bacterial metabolism. Under these circumstances, it might be predicted that chromosomally encoded antibiotic resistance determinants might form part of global regulatory networks, an issue that has been reviewed recently (Martinez et al., 2009a). Examples of such regulatory networks are the mar (from multiple antibiotic resistance) regulon (George & Levy, 1983; Alekshun & Levy, 1997), which encompasses a large network of elements (including the major MDR efflux pump in Enterobacteriaceae, AcrAB) that produces a global bacterial response to external signals, or the mgrA regulon, which comprises around 350 genes (Luong et al., 2006), including several that encode MDR efflux pumps (Truong-Bolduc et al., 2003, 2005) and virulence factors (Ingavale et al., 2005) in S. aureus.

Because, as described above, changes in bacterial metabolism can modify bacterial susceptibility to antibiotics and some global regulators modulate the expression of resistance determinants, it might be speculated that global regulators of bacterial metabolism might modulate the susceptibility to antibiotics of bacterial pathogens. We have already mentioned that CbrAB, which regulates carbon and nitrogen utilization in P. aeruginosa, also modulates susceptibility to antibiotics in this bacterial species (Yeung et al., 2011). The Crc global regulator provides a further example supporting that this link indeed exists. This protein was initially described in P. aeruginosa as being involved in the carbon catabolite repression of several genes responsible for the transport and metabolism of glucose and mannitol (MacGregor et al., 1991, 1996; Wolff et al., 1991). It should be mentioned here that, in Pseudomonads, glucose is not a preferred carbon source. In this bacterial genus, certain organic acids and amino acids are preferred over glucose, although glucose is in turn preferred to other compounds such as hydrocarbons (reviewed in Collier et al., 1996). Crc was later shown to inhibit the expression of several pathways for the assimilation of aromatic compounds, hydrocarbons and some amino acids when preferred carbon sources are present in the growth medium (Hester et al., 2000; Yuste & Rojo, 2001; Morales et al., 2004a; Moreno & Rojo, 2008; Linares et al., 2010; Rojo et al., 2010).

Crc is an RNA-binding protein that binds to specific sites at some mRNAs, inhibiting their translation (Moreno et al., 2007, 2009b). The activity of Crc is modulated by the CrcZ small RNA, which acts as an antagonist of Crc by binding and titrating it (Sonnleitner et al., 2009). The transcription of crcZ is activated by CbrAB (Yeung et al., 2011), which provides a linkage between this sensor system and the global regulation exerted by Crc (Fig. 3).


Influence of global regulation networks that coordinate metabolism on virulence and antibiotic resistance: the example of the CbrA/CbrB-CrcZ-Crc regulatory cascade. The CbrA sensor kinase autophosphorylates in response to still unclear signals. The phosphoryl group is then transferred to the CbrB response regulator, and perhaps to other still unidentified regulators, modifying the transcription of a wide range of genes. Inactivation of CbrA has pleiotropic effects, finally affecting genes that alter antibiotic resistance and C/N balance (Yeung et al., 2011). CbrB activates the transcription of the CrcZ small RNA (sRNA), which acts as an antagonist of the Crc global regulator by binding to and sequestering it (Sonnleitner et al., 2009). Crc is an RNA-binding protein (Moreno et al., 2007) that inhibits the translation of many mRNAs (Morales et al., 2004b). Its main role is to modulate and optimize metabolism, but has far-reaching effects. By controlling the expression of membrane proteins and genes involved in the composition of the cell envelope, it affects the permeability to several antibiotics. In addition, Crc affects, directly or indirectly, the expression of genes involved in type-III secretion, motility and biofilm formation in Pseudomonas aeruginosa, which has a clear impact on its virulence (Linares et al., 2010).

Direct targets for Crc are present at the mRNAs coding for the transcriptional regulators of specific catabolic pathways, for the transporters of nonpreferred compounds, for the first enzyme of the pathway or combinations of these. The final role of Crc is to optimize metabolism, improving bacterial fitness (Moreno et al., 2009a). Because carbon catabolite repression generates a significant metabolic reprogramming, Crc directly or indirectly affects the expression of many genes (Moreno et al., 2009a; Linares et al., 2010). For this reason, the influence of Crc goes beyond the mere optimization of metabolism.

One of the roles recently described for Crc is the modulation of P. aeruginosa susceptibility to antibiotics. Indeed, inactivation of the crc gene in P. aeruginosa makes this bacterium more susceptible to several antibiotics belonging to different structural families, such as β-lactams, aminoglycosides, fosfomycin or rifampin (Linares et al., 2010). Although the crc-deficient strain has a somewhat lower growth rate in the complete media where susceptibility assays are usually performed, the difference does not account for the observed differences in antibiotic susceptibility, because slow-growing bacteria are less susceptible to antibiotics, while the opposite phenotype was observed for the crc mutant. In addition, the hypersusceptibility effect has certain selectivity as it occurred with only some antibiotics, but not with others.

A likely explanation for the antibiotic hypersusceptibility phenotype of the crc mutant derives from the effect of Crc on the expression of several porins and membrane transporters involved in the uptake of carbon sources, proteins that are also used by distinct antibiotics to gain access into the cell. For example, Crc is known to inhibit the expression of the OprD porin, which mediates the uptake of basic amino acids and peptides (Moreno et al., 2009a; Linares et al., 2010). Interestingly, the antibiotics imipenem and meropenem also use this porin to permeate through the outer-membrane protein in Pseudomonas (Yoneyama & Nakae, 1993). A higher expression of OprD in the crc-deficient strain could explain its higher susceptibility to imipenem (Fig. 1). Similarly, Crc inhibits the expression of several transporters for sugars, among them that of the glycerol-3-phosphate transporter GlpT (Moreno et al., 2009a; Linares et al., 2010). This transporter is also used by fosfomycin to enter the cell (Castaneda-Garcia et al., 2009). As in the former example, a higher expression of GlpT in the crc-deficient strain likely allows for a more efficient transport of fosfomycin into the cell, thereby increasing its antibiotic efficiency.

The Crc protein has been detected in many Pseudomonas species (Rojo et al., 2010) and in some related bacterial genera such as Acinetobacter (Zimmermann et al., 2009). Other bacterial species such as E. coli or B. subtilis lack Crc, but have other catabolite repression control elements that also affect the expression of porins and membrane transporters responsible for the uptake of sugars and other carbon sources, proteins that may also be used by particular antibiotics to enter the cell. It might thus be predicted that other global regulators of bacterial metabolism would modulate bacterial susceptibility to antibiotics, a topic that has not been explored in detail to date, although some results indicate that indeed bacterial susceptibility to antibiotics is under the direct or the indirect control of such global regulators.

One of these global regulators is the histone-like protein H-NS. This nucleoid-associated protein modulates several cellular processes in Enterobacteriaceae (Hommais et al., 2001), including virulence (Harrison et al., 1994; Gomez-Gomez et al., 1996; Nasser et al., 2001; Muller et al., 2006; Olekhnovich & Kadner, 2007), mutation rate (Gomez-Gomez et al., 1997; Palchaudhuri et al., 1998; Shiraishi et al., 2007), responses to stress (Laurent-Winter et al., 1997; Bertin et al., 2001), motility (Bertin et al., 2001; Ghosh et al., 2006), cell envelopes biosynthesis (Hommais et al., 2001) and metabolism (Laurent-Winter et al., 1997; Erol et al., 2006). H-NS modulates resistance in E. coli and in Salmonella enterica serovar Typhimurium by regulating the expression of chromosomally encoded MDR efflux pumps (Nishino & Yamaguchi, 2004; Nishino et al., 2009) and also modulates the phenotype of persistence in E. coli (Hansen et al., 2008), indicating that H-NS simultaneously modulates metabolism, virulence and antibiotic resistance in Enterobacteriaceae. Notably, H-NS also regulates the expression of genes acquired by HGT (Lucchini et al., 2006; Navarre et al., 2006), producing a situation that has been named silencing of antibiotic resistance.

In vivo silencing of antibiotic-resistance genes has been demonstrated for E. coli-resistant strains colonizing pigs (Enne et al., 2006). Strains containing MDR plasmids were inoculated into piglets and recovered on different days after infection. It was found that in vivo bacterial evolution led to the loss of expression of the resistance determinants, a feature that was not observed for bacteria evolving in vitro. Reintroduction of the MDR plasmids into wild-type bacterial strains allowed the recovery of a full resistance phenotype, indicating that the mutation leading to resistance silencing occurred in the bacterial chromosome, either as the consequence of bacterial adaptation for colonizing the pig gut or to overcome the fitness costs associated with the carriage of the resistance plasmid. The molecular basis of in vivo acquired silencing in this study is unknown, although recent work indicates that H-NS plays a relevant role in the silencing of antibiotic resistance genes (and of many other genes) with a low GC content (Lucchini et al., 2006; Navarre et al., 2006) and even in the transfer of F plasmids that might carry resistance genes (Will et al., 2004). Altogether, these results indicate that the broad regulator H-NS modulates, among several other bacterial processes (including the microbial metabolism), the expression of both chromosomally encoded antibiotic resistance genes and those acquired by HGT.

Signal transduction systems and antibiotic resistance

The adaptation of bacterial pathogens to live in different habitats is based in the development of systems capable of sensing key signals from these distinct environments. Two-component signal transduction systems, formed by histidine kinase sensors and their associated response regulators (Stock et al., 2000), play a pivotal role in regulating diverse adaptation processes. A recent review highlights that two-component regulatory systems appear to be instrumental in the regulation of both virulence and resistance in P. aeruginosa (Gooderham & Hancock, 2009). Because this topic has been reviewed in depth, we will just briefly discuss some examples of the regulation of antibiotic resistance by two-component systems. One of the best examples is that provided by PhoP–PhoQ. This system mediates the response of different pathogens to magnesium concentrations (Groisman et al., 1998) and affects the susceptibility to polymyxin B, aminoglycosides and antimicrobial peptides (Macfarlane et al., 2000), mainly by altering the composition of the lipid A of the bacterial lipopolysaccharide. These changes alter not just bacterial susceptibility to cationic antibiotics, but virulence as well (Gooderham et al., 2009), demonstrating that these are interlinked phenotypes dependent on inputs received by bacteria from the environment (Groisman et al., 2001). Recent works indicate that, in addition to lipopolysaccharide composition, PhoP–PhoQ regulates other elements of the bacterial metabolism. Indeed, E. coli glycogen metabolism is controlled by the PhoP–PhoQ (Montero et al., 2009), and several genes encoding basic elements of bacterial metabolism, such as pyruvate kinase, aldehyde dehydrogenase or different cytochrome oxidases, are dysregulated in a P. aeruginosa phoQ mutant (Gooderham et al., 2009). As discussed above, other two-component systems, like CbrAB, also modulate the virulence, resistance and metabolism of P. aeruginosa (Yeung et al., 2011), indicating that these sensor elements mediate a global bacterial response to environmental changes, which includes metabolic shifts and alterations in their susceptibility to antibiotics. Indeed, PhoP–PhoQ forms a part of a regulatory network that includes other two-component systems. In the case of P. aeruginosa, it has been proposed that, at low Mg2+ concentrations, PhoP–PhoQ and PmrA–PmrB regulate the expression of the arn operon, which results in the modification of the lipopolysaccharide and resistance to antimicrobial peptides (Gooderham & Hancock, 2009). At high Mg2+ concentrations, these two systems are not relevant and the ParR–ParS two-component system regulates (Muller et al., 2011) not just arn but also the pnrAB regulator and as well as the MDR efflux pump mexXY and the imipenem transporter oprD (Fig. 4).


Impact of the PhoP–PhoQ/PmrA–PmrB/ParR–ParS regulatory network on the susceptibility to antibiotics of Pseudomonas aeruginosa. Bacteria harbour sensor-regulator two-component systems that sense the environment and trigger adaptive responses. The process is achieved by the transfer of a phosphate group from the sensor to the regulator, which leads to the activation of the latter and the consequent expression of the genes of the regulon. A reduced Mg2+availability leads to the activation of the PhoP–PhoQ and PmrA–PmrB systems, which regulate positively the transcription of their operons as well as of the arn operon. As a consequence, the lipid A is modified and the entry of polymyxins and cationic peptides reduces, leading to phenotypic resistance to these compounds (Gooderham & Hancock, 2009). At high concentrations of Mg2+, these two systems are not active, but ParR–ParS can be activated in response to specific signals. ParR induces the expression of pmrAB, the arn operon and mexXY, which encodes an MDR efflux pump, and represses the expression of oprD, which encodes the P. aeruginosa imipenem transporter (Muller et al., 2011).

Another relevant two-component system is that formed by EnvZ–OmpR. This system is widely distributed among Gram-negative bacteria and allows bacteria to respond to changes of medium osmolarity (Mizuno & Mizushima, 1990; Forst & Roberts, 1994). The E. coli EnvZ–OmpR phosphorelay system plays a role in the metabolism of amino acids, in the synthesis of flagella and in the production of the enterochelin, which is a relevant element for iron homeostasis (Shin & Park, 1995; Park & Forst, 2006). These effects derive, in part, from the regulation exerted by OmpR on the expression of two major outer-membrane proteins of this bacterium, namely OmpF and OmpC. OmpF allows the passage through the outer membrane of solutes such as sugars, ions and amino acids with a molecular size below 600 Da (Cowan et al., 1992), whereas OmpC allows the transport of ions and other hydrophilic solutes with a molecular size below 500 Da. It has also been reported that mutations in OmpR severely compromise the survival of E. coli in habitats like seawater (Darcan et al., 2003), reinforcing the relevance of this system in the bacterial physiology.

Besides their contribution to bacterial metabolism, OmpF and OmpC participate in the transport inside the cell of antibiotics such as β-lactams, quinolones, tetracycline or chloramphenicol (Mortimer & Piddock, 1993). Mutants lacking these porins, or defective in their regulator OmpR, accumulate lower amounts of antibiotics (Mortimer & Piddock, 1993) and confer high-level resistance to antibiotics, alone or in combination with other resistance mechanisms (Reguera et al., 1991; Chenia et al., 2006; Doumith et al., 2009; Kallman et al., 2009). It is important to note that the expression of OmpF and OmpC is subjected to different layers of regulation. For instance, the MarA (derived from multiple antibiotic resistance) global regulator, which induces the expression of the major efflux pump AcrAB in Enterobacteriaceae, downregulates OmpF, without producing major changes in OmpC expression, by a post-transcriptional mechanism mediated by micF (Cohen et al., 1989).

Membrane permeability is the first step in the crosstalk between bacteria and their surrounding environment in order to optimize the cellular metabolism as a function of the environmental conditions. Because of this, the expression of porins and modifications in the lipid composition of bacterial membranes are tightly regulated in response to extracellular inputs. Since antibiotics co-opt bacterial transporters for their entry into bacterial cells, changes in the transporters' composition due to the adjustment of bacterial metabolism to an environmental shift might challenge the susceptibility to antibiotics. Conversely, mutations affecting the expression or the activity of porins, selected under antibiotic selective pressure, might modify bacterial metabolism.

The effect of two-component systems on antibiotic resistance is not just confined to changes in the membrane composition or in the level of expression of transporters. Indeed, among the 32 ORFs annotated as response regulators of two-component signal transduction systems of E. coli, 13 conferred increased β-lactam resistance when overexpressed (Hirakawa et al., 2003). Besides modifications in membrane composition, this increased resistance was due to the overexpression of MDR efflux pumps (mediated by the regulators baeR and evgA) and changes in the level of expression of the chromosomally encoded β-lactamase AmpC (mediated by fimZ). The finding that several of the signal-transduction pathways controlled by two-component systems modulate antibiotic susceptibility through different mechanisms indicates that resistance is under control of the global networks that regulate bacterial physiology in its adaptation to different environments.

In eukaryotes, signal transduction is usually mediated by serine/threonine and tyrosine kinases/phosphatases (Hunter et al., 1995; Schlessinger et al., 2000). Protein phosphorylation is also relevant in prokaryotes. The comparison of the phosphoproteomes of B. subtilis and E. coli has shown that the phosphorylation sites are conserved, indicating that signal transduction through protein phosphorylation is at the root of the evolutionary tree (Macek et al., 2008). It has been described that among the phosphorylated proteins, enzymes involved in carbon metabolism (glycolysis, sugar transport) and several essential proteins were overrepresented, supporting that protein phosphorylation is an important mechanism of signal transduction for bacteria (Macek et al., 2008).

The view that protein phosphorylation is relevant for central bacterial metabolism is also supported by other studies, which showed that phosphoproteins are involved in processes like carbon/protein/nucleotide metabolisms, cell cycle and regulation of cell division (Sun et al., 2009). One of the best-studied bacterial serine/threonine kinases is PknB. This kinase is essential for allowing mycobacterial growth (Fernandez et al., 2006), and plays a fundamental role in maintaining bacterial shape (Kang et al., 2005). PknB is also involved in the regulation of purine biosynthesis, autolysis and central metabolic processes in S. aureus. The loss of PknB makes S. aureus much more susceptible to the cell-wall-active antibiotic tunicamycin (Donat et al., 2009), a feature that supports the linkage between antibiotic susceptibility and metabolism in this bacterial species.

It has been postulated that PknB-like kinases might be key regulators of cell-wall biosynthesis and bacterial response to β-lactams because they share a conserved domain (PASTA) with penicillin-binding proteins (PBPs) (Yeats et al., 2002). Indeed, it has been demonstrated that PknB phosphorylates PBP-A from M. tuberculosis (Dasgupta et al., 2006), further supporting a role for this kinase in the activity of β-lactams. A recent work has shown that PknB is a relevant element regulating bacterial susceptibility to antibiotics targeting the cell-wall metabolism. This kinase is involved in the phenotype of antibiotic resistance and virulence of S. aureus, including laboratory strains and clinical resistant isolates (Beltramini et al., 2009; Tamber et al., 2010). The modulation of antibiotic susceptibility by PknB is more likely an indirect effect of the role of this kinase in the metabolism of the cell wall and the changes of this structure in response to signals or to stress.

Selection of antibiotic resistance as a consequence of bacterial metabolic shifts

Given that antibiotic susceptibility can change as a consequence of alterations in the bacterial metabolism, it is conceivable that bacterial colonization of novel habitats might select antibiotic resistance even in the absence of antibiotic selective pressure. This might be particularly important in the case of chronic infections, because it has been demonstrated that bacteria evolve during the course of such infections to better adapt their physiology to the resources present in the environment of the infected host (Silo-Suh et al., 2005; Smith et al., 2006; Martinez-Solano et al., 2008; Mena et al., 2008; Rau et al., 2010).

One of the best-known chronic infectious diseases is that suffered by patients with cystic fibrosis (CF), which is the most prevalent inherited disease in the Caucasian population (Zielenski & Tsui, 1995). CF patients suffer chronic infections in their lungs by different bacterial species, P. aeruginosa being the most prevalent one (Govan & Nelson, 1992). It has been demonstrated that the same clone of P. aeruginosa can remain in the lung of a patient, evolving during decades (Struelens et al., 1993; Govan & Deretic, 1996; Hoiby et al., 1998; Lyczak et al., 2002; Morales et al., 2004a, b). This evolution presents a specific pattern, in the sense that mutations in some specific loci are frequently selected during the chronic infection (Smith et al., 2006; Mena et al., 2008).

Pseudomonas aeruginosa isolates presenting mutations in lasR are frequently isolated from the lungs of CF patients (Smith et al., 2006; Tingpej et al., 2007; Hoffman et al., 2009). LasR is a transcriptional regulator involved in the quorum-sensing (QS) response of P. aeruginosa, a regulon that includes a large number of determinants (Williams & Camara, 2009). Several hypotheses have been raised for explaining the accumulation of this type of mutants (Heurlier et al., 2005, 2006; D'Argenio et al., 2007; Tingpej et al., 2007). Some of the hypotheses are based on the fact that triggering the QS response is energetically expensive (Haas et al., 2006), so that the colonization of a novel environment in which the QS response is not needed will favour the selection of mutants that do not respond to the QS signal. On some occasions, this type of mutants coexists with other isolates presenting a wild-type phenotype and it is supposed that they cheat the activities (for instance, proteases) produced by the latter (Sandoz et al., 2007). However, on other occasions, the entire population in the lung of the patient is formed by lasR mutants, suggesting that this mutation is adaptive by itself (Smith et al., 2006; D'Argenio et al., 2007). Because of this, it has been suggested that lasR P. aeruginosa mutants have a growth advantage in a habitat such as the lung (D'Argenio et al., 2007), which is rich in amino acids and nitrate and poor in oxygen (Barth & Pitt, 1996). Indeed, the lasR mutants present a profound advantage for growth using the nitrate present in CF lungs (Hoffman et al., 2010), supporting the idea that those mutants might have been selected as a consequence of the metabolic shift that P. aeruginosa should undertake to adapt to the lung environment when producing a chronic infection in CF patients. One characteristic feature of lasR mutants is their growth advantage under conditions of oxidative stress, a feature that, as described above, can be relevant for the activity of antibiotics. In vitro analysis of a lasR mutant demonstrated that this mutation confers tolerance to the aminoglycoside tobramycin and the quinolone ciprofloxacin, both of which are frequently used for treating P. aeruginosa CF infections (Hoffman et al., 2010). Besides, it has also been reported that lasR mutants are less susceptible to β-lactams due to increased β-lactamase activity (D'Argenio et al., 2007).

This situation shows that the metabolic shift that bacteria undergo during the course of their adaptation for colonizing novel environments might be selected for antibiotic-resistant bacteria, even in the absence of selective pressure with antibiotics, highlighting the existence of a tight link between the metabolism of bacterial pathogens and their susceptibility to antibiotics.

Another good example of the selection of antibiotic resistance by metabolic shifts concerns the aforementioned SCVs (Proctor et al., 2006). These mutants were described a 100 years ago (Jacobsen, 1910), before the use of antibiotics for therapy, so that it is unlikely that the primary selective force (at least at that time) consisted of the presence of antibiotics in clinical environments. Indeed, mutants defective in the electron transport chain present an SCV phenotype (Heinemann et al., 2005). As mentioned above, these mutants present changes in their metabolism including a reduced susceptibility to several antimicrobials.

In silico prediction and further biochemical analysis of the metabolic networks of S. aureus SCVs showed that the mutations present in these cells are associated with strong rewiring of the bacterial metabolism (Heinemann et al., 2005; von Eiff et al., 2006), which involve the upregulation of enzymes participating in the glycolytic and fermentative pathways, as well as in the metabolism of purines, arginine and proline (Seggewiss et al., 2006). SCVs are recovered from chronic infections and are considered to be the cause of relapse in some infectious processes, suggesting that this type of mutations is adaptive for some infections and their resistance phenotype is a consequence of the metabolic adaptations more than the adaptive force driving the emergence of SCVs. Indeed, it has been demonstrated that S. aureus SCVs are induced when bacteria grow in the intracellular milieu of endothelial cells (Vesga et al., 1996) or in animal models of chronic infection (Tuchscherr et al., 2011) even in the absence of antibiotics. Selection of SCVs by persistent growth inside a cell host is not just a specific characteristic of S. aureus. It has been shown that the intracellular growth of S. enterica in nonphagocytic eukaryotic cells selects for SCVs (Cano et al., 2003). These results indicate that the generation of SCVs might be a consequence of the adaptation of these bacteria to cellular internalization and in host survival (von Eiff et al., 1997; Proctor et al., 2006).

Effect of acquiring antibiotic resistance on the bacterial metabolism

It is generally accepted that the acquisition of resistance results in a metabolic cost for bacteria so that in the absence of selective pressure, susceptible bacteria should rapidly outcompete their resistant partners. This hypothesis is based on the fact that antibiotic targets and transporters are highly conserved elements extremely relevant for bacterial physiology, so that mutations in these elements that confer antibiotic resistance will impair their functioning. Besides, mutations triggering the overexpression of detoxification elements such as antibiotic-inactivating enzymes of MDR efflux pumps will be physiologically costly because, in the absence of antibiotics, the constant overproduction of the element will lead to a non-needed metabolic load (Andersson & Levin, 1999; Andersson et al., 2006; Martinez et al., 2007; Baquero et al., 2009). The same would apply for the acquisition of antibiotic resistance genes, mainly if they are present in large plasmids, with the associated costs of replication, transcription and translation. If fitness costs consist just of a metabolic burden, the effect would be quite similar for any mechanism of resistance. Because of this, most studies on fitness costs are based on competition experiments between the wild-type strains and their resistant counterparts. This approach has been demonstrated to be useful in several cases (Sander et al., 2002; Balsalobre & de la Campa, 2008; Pranting et al., 2008; Shcherbakov et al., 2010), but the costs associated with resistance are likely more complex than simply a general metabolic impairment, so that different resistance mechanisms might induce different changes in the bacterial physiology (Martinez et al., 2007; Andersson & Hughes, 2010), including changes in the cellular metabolism.

In favour of the specificity of the costs associated with the acquisition of resistance is the finding that the acquisition of resistance might have specific consequences on bacterial virulence (Martinez & Baquero, 2002). For instance, acquisition of chromosomally encoded AmpC impairs the cellular invasivity of S. enterica (Morosini et al., 2000) and the effect of the overproduction of MDR efflux pumps on bacterial virulence is specific to the pump that is overproduced (Lee & Shafer, 1999; Sanchez et al., 2002b; Lee et al., 2003; Alonso et al., 2004; Linares et al., 2005; Warner et al., 2007). Furthermore, the same mutation leading to quinolone resistance can make one strain of Campylobacter jejuni more virulent and impair the virulence of another one (strain specificity; Luo et al., 2005). Finally, the finding that the mutations acquired for compensating the fitness costs are different in vivo than in vitro indicates that fitness costs are highly dependent on the bacterial habitat, and the metabolic adaptations required for colonizing such a habitat (Bjorkman et al., 2000).

Some recent works indicate that the effect of acquiring resistance on bacterial metabolism can have, like for bacterial virulence, some degree of specificity. The analysis of the metabolic profiling of a spontaneous antibiotic-resistant S. maltophilia mutant that overproduces SmeDEF (Alonso & Martinez, 2000), the most relevant MDR efflux pump in this bacterial species (Alonso & Martinez, 2001; Sanchez et al., 2002a), in comparison with that of its isogenic wild-type parental strain, shows that the acquisition of resistance makes S. maltophilia more proficient in the utilization of gentibiose, dextrin, mannose and formic acid (Alonso et al., 2004). On the contrary, the antibiotic-resistant mutant was impaired in the utilization of amino acids such as alanine, serine or proline, among others. These results indicate that resistance due to the overproduction of SmeDEF is associated with a metabolic shift more than with a general metabolic burden in S. maltophilia. The study of a P. aeruginosa nfxB antibiotic-resistant mutant, which overproduces the MDR efflux pump mexCD-oprJ, also supports this notion. Proteomic analyses demonstrated that several proteins were differentially expressed in the mutant as compared with its wild-type isogenic parental strain (Stickland et al., 2010). Among them, many played a role in the amino acid and energy metabolisms. The analysis of secreted metabolites showed that the resistant strain secreted higher levels of fatty acids such as myristic, palmitic and stearic acids, which are major components of P. aeruginosa membranes. The mechanisms for this differential exometabolome observed in the antibiotic-resistant mutant are not fully understood, although it has been suggested that the efflux pump itself might be responsible for the observed increased secretion of fatty acids.

The idea that resistance can produce specific changes in the bacterial metabolism is also supported by the analysis of B. subtilis rifampin-resistant mutants. Resistance to rifampin is acquired as a consequence of mutations in the rpoB gene, which encodes for the RNA-polymerase β-subunit. At first glance, it could be predicted that mutation-driven changes in the RNA polymerase would produce a general metabolic burden, given that it is a key enzyme controlling the information from genotype to phenotype, and that the structure of RpoB is finely adjusted to interact with the other components of the polymerase complex and with bacterial promoters, so that any potential change should be counteradaptive. In spite of this prediction, it has been shown that B. subtilis rifampin-resistant rpoB mutants can present novel metabolic capabilities (fitness gain) when compared with their wild-type susceptible parental strain (Perkins & Nicholson, 2008). Overall, the resistant mutants made less proficient use of strongly utilized substrates, but increased their capability for degrading weakly utilized substrates. Notably, different rifampin-resistance mutations have different effects on the metabolism of B. subtilis, likely because the interactions of RNA polymerase with the different promoters change depending on the mutation involved (Perkins & Nicholson, 2008).

Concluding remarks

Antibiotic resistance has been classically considered as a trait that has been acquired recently by bacterial pathogens as a consequence of the use of antibiotics. Given the extremely short evolution time in comparison with the time required for the evolutionary integration of regulatory and metabolic networks, it has been assumed that resistance will be independent of these networks and the only effect that could be expected from the acquisition of resistance will be the malfunction of bacterial metabolism reflected in an impaired growth (fitness costs). Throughout this review, we have discussed that the situation is far more complex than these simple predictions and the susceptibility to antibiotics is intrinsically linked to the bacterial metabolism. This statement is supported by the finding of several genes, involved in basic metabolic processes, the inactivation of which alters the bacterial susceptibility to antibiotics. Furthermore, global regulators of the bacterial carbon and nitrogen metabolism like CbrAB or Crc also modulate the susceptibility to antibiotics of bacterial populations.

Understanding the crosstalk between bacterial metabolism and the susceptibility to antibiotics can serve to propose novel strategies for treating infections. For instance, the finding that bacterial biofilms present subpopulations showing different metabolic activities and distinct susceptibility to antibiotics (one subpopulation susceptible to polymyxin and resistant to ciprofloxacin and another presenting the opposite phenotype) has served to propose combined therapies for treating biofilm-associated infections (Rani et al., 2007). Understanding bacterial metabolism during infection has also served to propose that fosfomycin will be useful for treating Listeria infections, despite this bacterial species being resistant to fosfomycin using in vitro tests. The reason for this behaviour is that Listeria becomes hypersusceptible to this antibiotic as a consequence of the metabolic adaptation required for using the nutrients present inside its host cell (Chico-Calero et al., 2002; Scortti et al., 2006). The same knowledge of the specific metabolism of bacteria during infection is being used for developing proficient antituberculosis drugs (Ginsberg, 2010) and to propose the use of antibiotics targeting anaerobic bacteria for the treatment of infections by aerobes such as M. tuberculosis, which might have anaerobic metabolism under some specific circumstances during infection (Ginsberg, 2010). If susceptibility to antibiotics depends on the bacterial metabolism, it might be possible that the specific growth conditions afforded during infection can alter antibiotic susceptibility. This is the case of the examples on phenotypic resistance (or hypersusceptibility in the case of fosfomycin or drugs against anaerobes) discussed above. However, inheritable resistance can also be achieved as a consequence of the selection of mutants presenting a metabolism better adapted to the conditions encountered during infection. This is the case of SCVs, selected in the absence of antibiotics during chronic infections or in the intracellular habitat, which are less susceptible to antibiotics as aminoglycosides because of changes in their electron transport chain.

The finding of global regulators that simultaneously modulate bacterial metabolism, virulence and antibiotic resistance provides the possibility of developing new drugs using an antivirulence–antiresistance approach. This would be the case for Crc, whose inactivation makes P. aeruginosa more susceptible to antibiotics and less virulent. Together with high-throughput studies defining the global resistome, which help in unveiling metabolic determinants contributing to intrinsic resistance, the understanding of the links between metabolic networks and antibiotic resistance determinants will serve to implement better therapeutic strategies and novel targets for developing new antibiotics.


Work in our laboratories is supported by grants BIO2008-00090 and BFU2009-07009/BMC from the Spanish Ministry of Science and Innovation, Spanish Network for the Research in Infectious Diseases, REIPI RD06/0008 from the Instituto de Salud Carlos III (cofinanced by ERDF) and KBBE-227258 (BIOHYPO), HEALTH-F3-2010-241476 (PAR) and EVOTAR from the European Union. We wish to acknowledge Dieter Haas for useful suggestions.


  • Editor: Fernando Baquero


View Abstract