Environmental selection of antibiotic resistance genes



  • Ana Alonso,

    1. Departamento de Biotecnología Microbiana, Centro Nacional de Biotecnología, CSIC, Campus Universidad Autónoma de Madrid, Cantoblanco, 28049-Madrid, Spain.
    Search for more papers by this author
  • Patricia Sánchez,

    1. Departamento de Biotecnología Microbiana, Centro Nacional de Biotecnología, CSIC, Campus Universidad Autónoma de Madrid, Cantoblanco, 28049-Madrid, Spain.
    Search for more papers by this author
  • José L. Martínez

    Corresponding author
      *For correspondence. E-mail jlmtnez@cnb.uam.es; Tel. (+34) 91 5854551; Fax (+34) 91 5854506.
    Search for more papers by this author

*For correspondence. E-mail jlmtnez@cnb.uam.es; Tel. (+34) 91 5854551; Fax (+34) 91 5854506.


Acquisition and further spread of antibiotic resistance determinants among virulent bacterial populations is the most relevant problem for the treatment of infectious diseases. Although mutations in antibiotic target genes (Martinez and Baquero, 2000) were supposed to be the primary cause of antibiotic resistance in the early antibiotic era, it soon became evident that acquisition of antibiotic resistance determinants by horizontal gene transfer has a major role on the development and spread of antibiotic resistance among pathogenic bacteria (Davies, 1994; 1997). What is the origin of these genes? The analysis of bacterial isolates from the preantibiotic era demonstrated that the incompatibility groups and the amount of plasmids carried by pathogenic bacteria were essentially the same that can be found today. However, the preantibiotic plasmids did not carry antibiotic resistance genes, so it has been assumed that the acquisition and further dissemination among pathogenic bacterial populations of antibiotic resistance is the consequence of strong antibiotic selective pressure as a result of antibiotic therapy (Datta and Hughes, 1983; Hughes and Datta, 1983). If these genes were not present in the pathogenic bacteria, they must have originated in the environmental bacteria and the most obvious microorganisms in which they might have a functional role are antibiotic producers. In fact, it is widely accepted that antibiotic resistance determinants originated in the antibiotic-producing organisms (Benveniste and Davies, 1973; Webb and Davies, 1993), in which they play an obvious protective role.

Although the origin of some antibiotic resistance genes from the antibiotic producers is clear, this is not the case for some other determinants. For instance, it is difficult to accept that chromosomal beta-lactamases (Bush et al., 1995), multidrug resistance (MDR) determinants (Paulsen et al., 1996; Nikaido, 1998) or some aminoglycoside-inactivating enzymes (Ainsa et al., 1997; Lambert et al., 1999; Macinga and Rather, 1999) that are present in all isolates of a given non-antibiotic producer bacterial species have been selected by antibiotic selective pressure. Another concern resides in the environmental selection of bacteria intrinsically resistant to antibiotics. Some of these bacterial species are relevant opportunistic pathogens and, apart from their environmental origin, they are refractory to treatment by several antibiotics (Quinn, 1998). As those antibiotics are not always present in the environmental habitat of these bacterial species, we might assume that the driving force selecting intrinsic resistance prior to infection must differ from antibiotic selective pressure.

Alternatively, even for those genes with a clear primary role in antibiotic resistance, selection without antibiotic selective pressure might occur if they are present in a replicon that also carries some other ‘selectable elements’ (see below). Selection of antibiotic resistance determinants might then occur in the environment by means of chemical or heavy metal pollution, or because the presence of the determinant accompanying the antibiotic resistance gene gives an ecological advantage to the bacteria for colonizing their environmental habitat (Fig. 1).

Figure 1.

Selection of clinically relevant antibiotic-resistant bacteria. Pathogenic bacteria can be resistant to antibiotics either because they contain the determinants for resistance in their genome or because they acquire antibiotic resistance genes from an exogenous source. The most straightforward selection is antibiotic treatment. Antibiotics will select intrinsically resistant bacterial species, bacteria that have acquired antibiotic resistance genes by horizontal gene transfer on antibiotic-resistant mutants. However, non-antibiotic compounds can also select antibiotic resistance bacteria. Thus, biocides and detergents can select resistant strains as the consequence of the expression of MDR determinants, or antibiotic resistance genes can present in replicons that contain other selectable markers, such as heavy metal resistance or production of siderophores.

Thus, antibiotic resistance genes have an environmental origin, sometimes as an antibiotic protective mechanism and sometimes with a different function. The environment also has a role in their selection that is not always the consequence of antibiotic selective pressure. We will discuss these concepts throughout this review.

The physiological role of antibiotic resistance genes

As previously stated, several antibiotic resistance determinants probably originated in antibiotic producers as bona fide antibiotic resistance genes. This is the case with the tetracycline resistance determinants otrA and otrB (Pang et al., 1994) that are present in mycobacteria and also in the tetracycline-producing bacterium Streptomyces rimosus. However, for most antibiotic resistance genes described to date in pathogenic bacteria, an identical counterpart has not been found in antibiotic producers. It can be argued that less than 1% of environmental species have been isolated to date, so that the antibiotic producers carrying these genes will be found sooner or later. Nevertheless, increasing evidence supports the notion that some antibiotic resistance genes might have a physiological role different to antibiotic resistance, even in the case of antibiotic producers (Piepesberg et al., 1988). In the case of non-antibiotic producers, if all isolates of a bacterial species carry a number of identical antibiotic resistance genes, it might be supposed that these determinants should have a role different from antibiotic resistance because non-producer species are not under constant antibiotic selective pressure in the environment. As previously stated, the most conspicuous examples of those genes are chromosomal beta-lactamases, some chromosomally encoded aminoglycoside-modifying enzymes and MDR efflux pumps.

Chromosomal AmpC beta-lactamases contribute to resistance to beta-lactam antibiotics in Enterobacteriaceae. However, the fact that they share a common ancestor and are present in all members of each bacterial species indicates that they were acquired by Enterobacteriaceae before the evolutionary differentiation of this genus into species, hundreds of thousands of years before the discovery of antibiotics. Beta-lactamases have evolved from transpeptidases involved in cell wall synthesis (Adachi et al., 1992; Knox et al., 1996). Although a role in such process has not been demonstrated, chromosomal beta-lactamases could be involved in peptidoglycan metabolism or be remnant molecules without a clear role in bacterial metabolism. However, chromosomal beta-lactamases are probably house-keeping genes and their activity against antibiotics is a side-effect of their actual (unknown) physiological activity.

Some aminoglycoside-modifying enzymes might have evolved from sugar kinases and acyltransferases (Udou et al., 1989; Macinga and Rather, 1999). In the case of chromosomal aminoglycoside-modifying enzymes present in the isolates of given species, such as Providencia stuartii (Macinga and Rather, 1999), Stenotrophomonas maltophilia (Lambert et al., 1999), Serratia (Shaw et al., 1992) or Mycobacteria (Ainsa et al., 1997), a metabolic role has been suggested. A structural role has been described for the chromosomal acetyltransferase (AAC(2′)-Ia) from P. stuartii. Apart from acetylating aminoglycosides, the enzyme has at least one physiological function, which is the acetylation of peptidoglycan (Payie et al., 1995). As P. stuartii isolates are not always in contact with aminoglycosides, the probable physiological role for this enzyme is cell wall metabolism.

A similar situation must happen with MDR pumps. These determinants are ubiquitously found in all bacterial species (Nikaido, 1998) and also in eukaryotic organisms. The genome of a single bacterial isolate might contain more than 20 putative MDR pumps (Stover et al., 2000). Although they contribute to the intrinsic resistance to antibiotics (Nikaido, 1994), they should also have different functional roles, such as protection against toxic compounds (Alekshun and Levy, 1999) or the involvement in cell/environment signalling pathways. For example, Escherichia coli presents MDR determinants involved in the extrusion of bile salts (Thanassi et al., 1997) and bile salts are part of the natural environment of E. coli, whereas antibiotics have only appeared in this environment during the last 50 years. It is clear, therefore, that these MDR determinants have probably been selected by the presence of bile salts, even if they now contribute to intrinsic antibiotic resistance in E. coli. The most remarkable example of selection of intrinsically resistant microorganisms by the environment is that of Gram-negative opportunistic pathogens with an environmental origin (e.g., Pseudomonas aeruginosa, Burkholderia cepacia and Stenotrophomonas maltophilia). Several of these opportunistic pathogens have their habitat in the soil, in close contact with plants. Soil is an environment that contain several potentially toxic aromatic compounds, derived from degradation processes (Vicuña, 2000) and plant exudates (Canto-Canche and Loyola-Vargas, 1999). It has been demonstrated that MDR determinants can extrude an ample range of substances that include solvents, detergents and aromatic compounds (Isken and de Bont, 1996; Li et al., 1998; Segura et al., 1999). Therefore, at least some of these MDR efflux pumps may have been selected to avoid the effect of toxic compounds present in their natural environment. A role in quorum sensing has been suggested for others (Evan et al., 1998; Pearson et al., 1999). It has been recently described that P. aeruginosa synthesizes a natural 4-quinolone involved in quorum sensing and that the MDR system MexABOprM is capable of extruding this molecule (Pesci et al., 1999). It has also been shown that environmental P. aeruginosa strains isolated before quinolones (a family of synthetic antibiotics) were discovered are capable of extruding these drugs (Alonso et al., 1999). It is possible that the capability for extruding quinolones is a side-effect of the physiological role of this pump, which is the extrusion of this quorum-sensing signal molecule.

We have seen that several antibiotic resistance determinants have a primary physiological role other than antibiotic resistance. In fact, they have been selected for metabolic, biosynthetic or signalling purposes. However, once antibiotic selective pressure is applied, mutants that overproduce these determinants can be selected in this way, reinforcing their adaptive role as antibiotic resistance determinants. Also, the genes can enter an independent replicon and further disseminate among pathogenic bacteria (Davies, 1994). Once these genes enter heterologous hosts outside their physiological context, they are not selected further for their primary physiological role and their function becomes antibiotic resistance alone (Fig. 2). This is a good example of the evolutionary mechanism that Brosius and Gould (1992) named ‘exaptation’.

Figure 2.

Environmental origin of antibiotic resistance genes. Antibiotic resistance genes are present in environmental antibiotic producer and non-producer microorganisms. Some of them serve in the original organism to resist the action of an antibiotic, being the most conspicuous example of the antibiotic resistance genes from antibiotic producers. They have thus been selected by antibiotic pressure and are bona fide antibiotic resistance genes. Other genes can be involved in other cellular processes such as antibiotic biosynthesis, signalling, metabolism or detoxification, and were selected in the original organisms to cover these functions in the absence, therefore, of antibiotic selective pressure. However, once they are introduced into a heterologous host and selected by means of antibiotic selective pressure, they behave as antibiotic resistance genes. Selection of a gene to cover a different function from that it was previously selected for is known as an exaptation mechanism of natural selection.

For example, chromosomal AmpC beta-lactamases are now present in plasmids (Coudron et al., 2000) that disseminate among bacterial populations contributing to the acquisition of an antibiotic resistance phenotype by previously susceptible bacteria. What is the origin of some other plasmid-encoded antibiotic resistance determinants? Some of them are probably native antibiotic resistance determinants and others may be metabolic genes. For example, we cannot know whether prototypic plasmidic beta-lactamases, such as the TEM family (Bush et al., 1995), have originated in the beta-lactam producers as antibiotic resistance genes or are enzymes that were first involved in cell wall metabolism in an unknown environmental bacterial species.

Selection of antibiotic resistance without antibiotic selective pressure

We have seen that intrinsic antibiotic resistance might have been selected in the course of bacterial evolution, without antibiotic selective pressure, for covering functions other than antibiotic resistance. Can this also happen for acquired antibiotic resistance?. The expression of MDR determinants is frequently downregulated under standard laboratory conditions. However, de-repressed mutants can easily be obtained in vitro and are frequently found clinical isolates (Nikaido, 1998) that contribute to acquired antibiotic resistance. As MDR determinants are capable of conferring simultaneous resistance to toxic compounds belonging to several different families, detergents and antiseptics included, an obvious question is whether selection with non-antibiotic compounds might select for antibiotic resistance.

For example, an efflux pump has been described in Listeria monocytogenes that can extrude both antibiotics and heavy metals (Mata et al., 2000). As L. monocytogenes has an environmental habitat, its growth in heavy metal-contaminated soils might also select for antibiotic resistance. It has also been demonstrated that biocides, organic solvents and detergents are capable of selecting mutants with increased expression of MDR determinants. Triclosan and pine oil might select low-level antibiotic-resistant E. coli strains as the consequence of overproduction of chromosomally encoded efflux systems (Moken et al., 1997). It is true that the selective concentrations of both biocides are low compared with those used in clinical practice. However, low concentrations of these biocides can be achieved both in clinical settings and in the home, and might then select low-level antibiotic-resistant bacteria.

Antibiotic selective pressure is not required for the selection of antibiotic resistance genes carried by replicons that contain not only antibiotic resistance determinants but also any other selectable marker (see Table 1). One of the most conspicuous examples of this situation is the linkage between antibiotic resistance and heavy metal resistance genes, which is frequently encountered in environmental bacterial isolates (Davison, 1999). The co-existence of both types of determinants in the same genetic element allows antibiotic resistance to be selected upon heavy metal selective pressure in contaminated environments. This linkage might explain the selection of determinants that would otherwise not be selected. For example, an erythromycin resistance determinant with a Gram-positive origin has been found in the genome of S. maltophilia (Alonso et al., 2000). S. maltophilia, like all Gram-negative bacteria, is intrinsically resistant to macrolides, so that erythromycin resistance genes are not selectable determinants for S. maltophilia. The reason for the presence of the erythromycin resistance determinant in the genome of S. maltophilia must, therefore, rely on accompanying selectable markers. In fact, this erythromycin resistance determinant is flanked by a cadmium resistance efflux pump, so that selection might have occurred as a result of heavy metal antibiotic selective pressure in the environment, prior to infection. Another remarkable example of clinically relevant heavy metal/antibiotic resistance co-selection is the linkage between silver and antibiotic resistance in the same replicon. Silver has historically been used as a biocide in the treatment of burns (Klasen, 2000). Burn wounds are easily colonized/infected by opportunistic pathogens (Pruitt et al., 1998). Treatment with silver ions might then co-select, in the case of a genetic linkage between silver and antibiotic resistance, those bacteria carrying these type of determinants, which are then more resistant to antibiotics (McHugh et al., 1975). It has been suggested that mercury, which is present in the amalgams used in odontology, might select for microorganisms with enhanced antibiotic resistance in the oral cavity and the intestine (Summers et al., 1993; Edlund et al., 1996). However, some recently published work casts doubts on this hypothesis (Osterblad et al., 1995; Leistevuo et al., 2000). More work is needed to clearly establish the actual role of dental fillings in antibiotic resistance.

Table 1.  Examples of selectable determinants carried by antibiotic resistance plasmids.
DeterminantConditions for the selectionReference
AerobactinLow ironDelgado-Iribarren et al. (1987)
Heavy metal resistanceHeavy metal contaminationDavison (1999)
AdhesinAdhesion to surfaces in liquid environmentsFranklin et al. (1981)
CytolysinInfective conditions (environmental interactions with amoebas?)Barja et al. (1990)
MicrocinCompetition with other bacteriaMartinez and Perez-Diaz (1990)
ColicinCompetition with other bacteriaFranklin et al. (1981)
Resistance to disinfectantsDecontamination of clinical settingsRussell (2000)

Selection pressure might be based not only on the toxicity of the potential selector but on stringent growing conditions. In this case, the metabolic capabilities required for growing under such conditions will be good selective markers. The linkage between aerobactin and antibiotic resistance genes in different plasmids has been described (Gonzalo et al., 1988). Aerobactin is a siderophore that allows bacteria to grow in environments where iron is scarcely available (de Lorenzo and Martinez, 1988), so that its presence in antibiotic resistance plasmids might contribute to the dissemination of antibiotic resistance (Delgado-Iribarren et al., 1987). This situation may also happen with antibiotic resistance plasmids containing other determinants of ecological value, such as adhesins or citolysins (Franklin et al., 1981).

A final example resides in the selection of the qac genes present in staphylococci plasmids. qacB specifies resistance to quaternary amines, acridine diamidines and ethidium bromide, whereas qacA additionally encodes resistance to chlorhexidine and is frequently carried on penicillinase plasmids (Rouch et al., 1990). The qacB gene has been found in plasmids isolated 50 years ago (Paulsen et al., 1998), whereas most isolates from 1980 contain qacA (Leelaporn et al., 1994). In the 1980s, chlorhexidine was introduced in hospitals, so it has been suggested that a replacement of one gene by the other occurred in staphylococci populations, as the consequence of the introduction of a biocide (Russell, 2000). As those plasmids also carry beta-lactamase genes, the introduction of chlorhexidine might have contributed to the selection of beta-lactam-resistant staphylococci as the consequence of beta-lactamase production in hospital settings. Interestingly, qac genes are also present in integrons of Gram-negative bacteria carrying multiple antibiotic resistance cassettes (Paulsen et al., 1993; Bass et al., 1999). This may contribute to the successful selection of such determinants in the presence of biocides, without antibiotic selective pressure.

Human intervention in the environment and antibiotic resistance

Chemical pollution

Industrial activities, minery and intensive farming are causing dramatic changes in natural ecosystems. Among the novel selective pressures that face environmental bacterial populations from the industrial revolution, discharges of heavy metals, xenobiotic compounds, antibiotics and organic solvents can have a remarkable role on the environmental selection of antibiotic resistance genes. Also, intensive farming requires the utilization of high amounts of probiotics and antibiotics (as far as 50% of total antibiotic consumption in developed countries) and contributes to the selection of antibiotic resistance genes in bacteria that colonize animals (Piddock, 1996; Witte, 1998). A good example of such a problem is the utilization of the glycopeptide antibiotic avoparcin as a growth promoter. In those countries in which the antibiotic has been extensively used, vancomycin-resistant enterococci are frequently encountered, not only in animals, but also in the human population (Van den Boggard and Stobberingh, 2000). Alternatively, the ban of avoparcin in animal feeding has curbed the development of resistance in European Union countries (Bager et al., 2000), which shows the relevant role that the utilization of antibiotics for animal feeding may have in the selection of antibiotic-resistant bacteria in the clinical setting.

We have to mention here that synthetic antibiotics are xenobiotic compounds that can also be considered as important pollutants. For example, quinolones are extremely stable in the environment (Halling-Sorensen et al., 1998), so their presence might produce dramatic effects on bacterial populations in natural habitats, the most prominent being the selection of antibiotic-resistant bacteria. Of note, quinolones are the most used synthetic antibiotics in aquaculture (Grave et al., 1996) and selection of quinolone resistance in indigenous river water bacterial populations as the consequence of contamination by run-off waters containing quinolones has been suggested (Goñi-Urriza et al., 2000). Nevertheless, in spite of the constant release of these xenobiotic non-degradable compounds in the environment, the effect of quinolones on the environmental bacterial populations has not been properly analysed.

The effects of industrial pollution on environmental bacterial communities have not been extensively studied. Most published work relies on the analysis of heavy metal-contaminated environments. Release of toxic metal species is the most relevant pollution problem since the industrial revolution (Ayres, 1992), mainly because heavy metals cannot be degraded and, therefore, they remain in the environment. In this situation, heavy metal-contaminated environments maintain the selective pressure on indigenous bacterial populations for long periods of time. Natural ecosystems containing high concentrations of heavy metals are also frequent. Not surprisingly, heavy metal resistance genes are easily found in environmental bacteria (Silver and Phung, 1996). It has been documented that heavy metal-contaminated environments also contain a higher percentage of antibiotic-resistant strains than non-contaminated ones, and bacteria isolated from contaminated soils contain more plasmids than those isolated from non-contaminated places (Rasmussen and Sorensen, 1998). Finally, under mercury stress, the gene-mobilizing capacity of soil bacterial populations increases. As heavy metal and antibiotic resistance are frequently linked in the same plasmid (see before), increased mobilization under metal selective conditions might also increase the mobilization of antibiotic resistance genes among environmental bacterial populations.

If contaminated environments might contribute to the selection of antibiotic-resistant bacteria, cleaning of these habitats may contribute to the restoration of an antibiotic-susceptible population. If this was the case, bioremediation of contaminated environments might be of help in reducing environmentally selected antibiotic resistance. Acquisition of an antibiotic-resistant phenotype reduces the fitness of bacteria (Andersson and Levin, 1999) so that replacement of resistant populations by susceptible ones can occur in the absence of selection. Nevertheless, antibiotic-resistant bacteria accumulate mutations that compensate for the effect of antibiotic resistance on fitness (Andersson and Levin, 1999), making the acquisition of antibiotic resistance a non-return evolution. It is thus unclear whether the cleaning of contaminated environments could restore the antibiotic-susceptible populations. The analysis of antibiotic-resistant bacteria in contaminated, non-contaminated and cleaned environments is thus an important topic that should be addressed in the near future.

Introduction of organisms in the environment

In recent years, society is increasingly concerned with the risks of dissemination of antibiotic resistance genes used for the construction of genetically modified organisms. For these reasons, safer systems that avoid the spread of genes (Diaz et al., 1994) and that are based in non-antibiotic markers (Herrero et al., 1990) or even in markers that are eliminated after the organism has been modified (Panke et al., 1998; Zubko et al., 2000) have been implemented. Even for genetically modified organisms that carry antibiotic resistance genes, we do not believe that the potential release of the antibiotic resistance genes currently used for the development of genetically modified organisms constitutes a significant risk for the dissemination of antibiotic resistance genes of clinical importance for two reasons: (i) Current studies indicate that the probability of dissemination of those genes is very low; and (ii) the genes currently used in genetic engineering, such as the beta-lactamase TEM1, are already (unfortunately) widely disseminated among pathogenic and commensal bacteria (Ferber, 1999).

Another concern could be the release of antibiotic resistance genes from non-modified organisms used in the field. It has been shown that the biopesticide Paenibacillus popilliae has a vancomycin resistance gene cluster homologous to the enterococcal vanA vancomycin resistance gene cluster (Patel et al., 2000) and to vancomycin resistance genes present in glycopeptide-producing actinomycetes (Marshall et al., 1998). Biopesticidal powders containing spores of P. popilliae have been used for more than 50 years in the United States for suppression of Japanese beetle populations (Patel et al., 2000). An identical counterpart of the P. popilliae vanA gene in pathogenic bacteria has not been found, so it seems that the use of P. popilliae biopesticidal preparations in agricultural practice have not had (at least at present) an impact on bacterial resistance in the clinical setting. However, this does not mean that it will not have an impact in the near future and this illustrates the need to analyse the effect of the introduction not only of genetically modified microorganisms, but also of ‘natural’ bacterial populations in the field.

The risks for the utilization of intrinsically resistant microorganisms, either genetically modified or not, for bioremediation or biotransformation processes have also been discussed (Holmes et al., 1998; LiPuma and Mahenthiralingam, 1999). Most bacteria currently used in bioremediation/biotransformation belong to the Pseudomonadacea family. Bacterial species belonging to this family are intrinsically resistant to antibiotics and are increasingly isolated from nosocomial infections (Quinn, 1998). Could the release of these intrinsically resistant microorganisms increase the probability of infections owing to antibiotic-resistant bacteria?. The archetype of this situation is Burkholderia cepacia. This bacterial species is being used both for bioremediation and as a promoter of crop growth. B. cepacia is also a relevant antibiotic-resistant opportunistic pathogen. Although the probability of infection by B. cepacia introduced in the field has not been analysed, we believe that it is low because: (i) B. cepacia does not produce infection in the community, but only in immunocompromised, hospitalized or cystic fibrosis patients, so that the number of people at risk of infection by B. cepacia used for agriculture or bioremediation should be low, and (ii) B. cepacia strains are already present in the field, so it is unclear whether introducing some naturally occurring microorganisms might increase the probability of infection. Alternatively, even if the probability of infection is low, it must be evaluated because infection by B. cepacia may have fatal consequences for previously debilitated patients, therefore, the risk might be high. It may be that the utilization of huge amounts of B. cepacia in areas with crowded human populations might increase the probability of raising the number of infections among unhealthy populations (hospitals, AIDS patients, famine situations etc.). As stated by other authors (Holmes et al., 1998; LiPuma and Mahenthiralingam, 1999), this possibility must be carefully evaluated.


The authors wish to thank Fernando Rojo for useful criticism and comments on draft versions of this manuscript. A. Alonso is a recipient of a fellowship from Gobierno Vasco. P. Sánchez is a recipient of a fellowship from Ministerio de Educación y Cultura.