Class 1 Integrons in Environments with Different Degrees of Urbanization

Background Class 1 integrons are one of the most successful elements in the acquisition, expression and spread of antimicrobial resistance genes (ARG) among clinical isolates. Little is known about the gene flow of the components of the genetic platforms of class 1 integrons within and between bacterial communities. Thus it is important to better understand the interactions among “environmental” intI1, its genetic platforms and its distribution with human activities. Methodology/Principal Findings An evaluation of two types of genetic determinants, ARG (sul1 and qacE1/qacEΔ1 genes) and lateral genetic elements (LGE) (intI1, ISCR1 and tniC genes) in a model of a culture-based method without antibiotic selection was conducted in a gradient of anthropogenic disturbances in a Patagonian island recognized as being one of the last regions containing wild areas. The intI1, ISCR1 genes and intI1 pseudogenes that were found widespread throughout natural communities were not associated with urbanization (p>0.05). Each ARG that is embedded in the most common genetic platform of clinical class 1 integrons, showed different ecological and molecular behaviours in environmental samples. While the sul1 gene frequency was associated with urbanization, the qacE1/qacEΔ1 gene showed an adaptive role to several habitats. Conclusions/Significance The high frequency of intI1 pseudogenes suggests that, although intI1 has a deleterious impact within several genomes, it can easily be disseminated among natural bacterial communities. The widespread occurrence of ISCR1 and intI1 throughout Patagonian sites with different degree of urbanization, and within different taxa, could be one of the causes of the increasing frequency of multidrug-resistant isolates that have characterized Argentina for decades. The flow of ARG and LGE between natural and clinical communities cannot be explained with a single general process but is a direct consequence of the interaction of multiple factors operating at molecular, ecological, phylogenetic and historical levels.


Introduction
In addition to the global causes of death by viruses, bacteria and parasites, which are a huge burden on public health, the progressive increase of multidrug resistance in all geographical regions has been identified as a public health priority according to the World Health Organization, 2011 (http://www.who.int/ drugresistance). In recent years, research on the function of antibiotic resistance in non-clinical environments has begun to receive attention [1,2,3,4]. This interest is based on the idea that a better understanding of the diversity of patterns and biological functions of antibiotic resistance mechanisms may eventually help to control its threats towards human and also animal health. The role of the environment as a reservoir of strains that have never before been isolated from humans was demonstrated during the outbreak caused by enteroaggregative Escherichia coli that had acquired the genes to produce Shiga toxins in Germany in May 2011. This episode also stresses the negative consequence of having mechanisms of antimicrobial resistance in these isolates (bla TEM-1 and bla CTX-M-15 ), which most likely helped the bacteria to survive and persist in different habitats [5].
Most of the new research into natural bacterial communities has focused on antimicrobial resistance genes (ARG) that confer resistance to antimicrobial drugs, mainly associated with protection against natural antibiotics or with functional properties among the metabolic pathways of environmental bacteria [2,6,7]. In contrast, scarce research has focused on the natural occurrence and role of the genetic platforms (transposons, integrons) that participate in the capture and dispersion of these genes within and among genomes, usually known as mechanisms of lateral genetic transfer [8]. To our knowledge, there are no prospective studies that analyses the independent occurrence of genetic components of a particular genetic platform, i.e. the lateral genetic transfer determinants such as integrons or transposons with the ARG elements, considering both ecological and molecular parameters.
Lateral genetic transference is a widespread phenomenon that is not only largely responsible for the ability of pathogenic and opportunistic bacteria to resist clinical antibiotic pressures [9], but also enables exchange of the accessory genome, which is a major contributor to bacterial evolution [10]. Of the different mechanisms involved in lateral genetic transfer, the class 1 integrons are one of the most successful elements in the acquisition, abundance, maintenance and spread of antimicrobial resistance gene cassettes among gram-negative bacilli isolated from clinical samples [11,12,13,14]. Although their role has not been yet investigated, class 1 integrons have been also found in gram-positive clinical strains, including methicillin-resistant Staphylococcus aureus and Corynebacterium species [15,16,17,18,19,20], from several hospitals around the world. The basic structure of an integron possesses a gene for an integrase (intI), a recombination site (attI) and a promoter (P c ) that permits the expression of gene cassettes incorporated in the variable region [21]. Several genetic structures have been described at the 39 end of the variable region of class 1 integrons [22,23,24,25,26,27,28]. There are three genetic platforms containing and spreading the class 1 integrons described in clinical samples from Argentina ( Figure 1): (i) the most common one exhibits the well-known 39-conserved segment  at the end of the variable region, which contains the qacED1 gene that is a deleted form of the quaternary ammonium compounds resistance gene cassette, qacE, followed by the sul1 gene that confers resistance to sulphonamides, and finally the orf5 of unknown function [12] ( Figure 1A); (ii) the complete or incomplete module of Tn402, tniC-tniQ-tniB-tniA [29] ( Figure 1B); and (iii) the 39-CS can be invaded by the putative site-specific recombinase ISCR1, which adds a second variable region that can contain an important variety of antimicrobial resistance genes such as bla CTX-M-2 and qnrB10, which are known as unusual or complex class 1 integrons [30,31,32,33] (Figure 1C). Neither class 1 integrons nor unusual class 1 integrons allow intracellular mobilization of the intI1 gene per se. However, almost all clinical members of class 1 integrons harbour the IR of the Tn402 transposon, transforming this genetic platform into a mobile element when the tns genes are provided in trans [25,34]. In addition, Tn402 targets plasmid and transposon resolution sites (res) [35], expanding the range of lateral gene transfer between clinical and natural communities [1]. It is very important from a clinical standpoint that the association of genes in the same genetic platform is co-selected under antibiotic pressure.
The relevant role of natural communities as a reservoir and original source of class 1 integrons was recently identified [1,15,36,37]. Since then, their distribution has been reported in environments with different degrees of human disturbance [1,15,36,37,38,39,40,41,42]. Overall, it is assumed that 2.65% of eubacterial cells in non-clinical samples contain a class 1 integron [41]. However, factors involved in the distribution of non-clinical class 1 integrons within natural communities remain largely unknown. What is known with certainty is that the class 1 integrons confer a benefit to the host cell due to their ability to acquire gene cassettes that could provide advantages for survival in hostile environments [43,44,45,46,47].
Concerning the molecular evolution of these elements, class 1 integrons were found to be chromosomally located, pre-dating the association with the Tn402-like transposon in non-clinical samples, suggesting that the ancestor of the clinical class 1 integron was more like a typical chromosomal integron [1]. The understanding of the molecular and environmental properties that contribute to the global success of class 1 integrons is the first step towards compiling a comprehensive story of how genes, genetic platforms, bacterial populations and selection pressures interact with human activities. However, it is difficult to assess the directionality of the flow of genes among natural environment and human habitats. The different alleles of the intI1 gene from natural communities led to the identification of the sources of both ''environmental'' and ''clinical'' class 1 integrons [48].
The aim of this study was to analyse the relationships among ''environmental'' intI1, its genetic platforms and its distribution with human activities in areas with different levels of urbanization (Table 1). Our methodology was based on two strategies: (i) we worked at two scales of analyses, molecular and ecological levels, and (ii) we evaluated two types of genetic elements from the same samples, ARG (sul1 and qacE1 genes) and genetic elements associated to lateral genetic transfer such as intI1, ISCR1 and tniC genes, called lateral genetic elements (LGE) for convenience in this paper and which comprise the genetic platforms of class 1 integrons. The first strategy allows the simultaneous analyses of ecological patterns and molecular mechanisms. The second strategy is based on the expectation that ARG will respond differently to LGE regarding the geographical variation of urbanization due to the different roles of these elements. We hypothesized that, if LGE serve as a general response mechanism to environmental stress, they should be present in both ''clean'' and urbanized habitats, as far as bacteria meet stressful conditions in ''clean'' habitats (for example, extreme seasonal or daily variations in weather conditions). Thus, they should present a weak relationship with the degree of urbanization. In contrast, ARG should be closely related to antibiotic pressure, and thus will present a strong link to geographical variations in urbanization and human presence. The field work was conducted in Tierra del Fuego, a Patagonian island from Argentina and Chile (Figure 2), which is recognized as being one of the last places on Earth that contains land areas that can still be considered wild or ''clean'', given its great extensions of intact, natural vegetation and large vertebrate assemblages, along with a low human population density [49].

Results and Discussion
Taxonomic Distribution of the Components of the Genetic Platforms of Class 1 Integrons We identified the following bacterial taxa using culturedependent methods: c and b classes within the Proteobacteria phylum (74 and 11 isolates, respectively), the Flavobacteria class within the Bacteroidetes phylum (3 isolates), Arthrobacter, Streptomyces, Microbacterium and Micrococcus genera within the Actinobacteria phylum (9 isolates) and the Paenibacillus genus within the Firmicutes phylum (1 isolate). Whole intI1 genes were identified in 11 isolates of c-proteobacteria (Table 2). Although intI1 genes were more abundant in Pseudomonaceae than among other families of the cproteobacteria (6/11), Enterobacteriaceae intI1 alleles showed high sequence diversity ( Table 2). The intI1 pseudogenes were mostly found in c-proteobacteria, but they were also identified in two bproteobacteria isolates and in one Actinobacteria isolate, evidencing the widespread dissemination of this genetic element. The remaining genetic determinants showed different patterns of taxonomic distribution as ISCR1 and sul1 were found in cproteobacteria and in Actinobacteria, and tniC in c-proteobacteria and Flavobacteria ( Table 2). The qacE1/qacED1 gene showed different frequencies between taxa (Table 3). Its frequency in cproteobacteria (18/74) was almost half of the frequency in the other taxa (10/23).
A previous study from Australia found a prevalence of intI1 genes in non-clinical isolates of b-proteobacteria strains in a similar culture-based method [1]. They also found a great dispersion of qac genes in environmental samples, and proposed that selection for qac resistance before the antibiotic era contributed to the mobilization and widespread of class 1 integrons among the environmental Proteobacteria. They argued that when antibiotics began to be administrated it would be almost inevitable that class 1 integrons would come to play a major role in the dissemination of antibiotic resistance [50].
With the information available from ours and other studies [48,50], it is possible to build a hypothesis on the role of environmental c and b-proteobacteria as sources of clinical intI1 genes. In addition, we found that at each sampling site positive for intI1, ISCR1 was also detected and at least one sul1 and/or qacE1/ qacED1 gene was identified. Although the succession of molecular steps involved in the acquisition of components of the genetic platforms of class 1 integrons circulating in hospitals nowadays ( Figure 1) could not be determined, our study showed a scheme in which the c and b proteobacteria harbouring intI1 genes share habitats with several other genera belonging to c, b-proteobacteria, Actinobacteria and Bacteroidetes, which in turn have the ISCR1, sul1, tniC and/or qacE1/qacED1 genes. These genes could have been co-acquired in one bacterial cell by mechanisms associated with lateral genetic transference and later selected by antimicrobial pressure within clinical settings and/or by human activities.
The Sul1 Gene is the Only Genetic Marker Associated with Urbanization The frequency of occurrence of sul1 was significantly and positively related to the level of urbanization, whereas the other genes, intI1, ISCR1, qacE1/qacED1 and tniC, were not significantly related to this variable (p.0.05) ( Figure 3; Table 3). All 10 sul1 the violet arrow exemplifies the intI1 gene, the blue arrow represents the sul1 gene, the green arrow stands for the qacED1 gene, the red arrow represents the tniC gene, and the yellow arrow represents the ISCR1 gene. The degree of colour intensity indicates different alleles for the corresponding gene. Grey rectangles stand for attCs (light grey) and attIs (dark grey). The dotted lines show the variable region of class 1 integrons (VR-1 is variable region 1). Full lines between the arrowheads above the integron structure show the expected amplification products using the different primer combinations. In order to define the different alleles for the intI1 genes the sequences were compared to AN AM412236 (isolates 1AC4, 4SN1, 9SN1 and 9AL34) and AN DQ247972 (isolates 7AN1, 11601AL, 11602SL, 11603SL and 11604SL). The sul1 and tniC alleles show the percentage of identity to AN JF262166 and GQ857074, respectively. The qacE1/qacDE1 and the ISCR1 sequences (indicated with an asterisk) were 100% identical to the clinical alleles AN HM999792 and EU722351, respectively. The graphic is not drawn to scale. doi:10.1371/journal.pone.0039223.g001 genes obtained by PCR were sequenced and 7 of them exhibited more than two mutations in 581-bp length compared to the sul1 from clinical isolates (accession number JF262166). In addition, high sequence diversity of the sul1 gene (83% identity compared to the sul1 sequence in accession number JF262166) was found in one strain isolated from a site with low-level anthropogenic disturbances (3SC2 isolate, Table 2). In order to analyse the relationship between ''environmental'' and ''clinical'' types of genes we performed a phylogenetic analysis with sul1 alleles from the Genbank and from our work ( Figure 4). More than 30 alleles of the sul1 gene were identified in this analysis. Only the sul1 allele from the 39-CS of the integrons was found in both clinical and nonclinical isolates.
The high frequency of the sul1 genes in urban sites could be the consequence of gene flow of the ''clinical'' sul1 allele from the hospital towards the open environment to which are added the Figure 2. Study area. The geographic sites where the sampling was performed are numbered from 1 to 10 (see Table 1). The circles represent low (white), medium (light blue) and high (dark blue) degrees of urbanization. doi:10.1371/journal.pone.0039223.g002 sul1 alleles of non-clinical strains that could be in turn maintained by the presence of contaminants that co-select for sulphonamide resistance.

A Wide Dissemination of the qacE1/qacED1 Gene was Found Between Taxa and Between Different Degrees of Urbanization in Patagonia
The distribution of the other ARG in this study, the qacE1/ qacED1 gene, was wider among the environmental samples than found for sul1. Although we found that the qacE1/qacED1 gene was present in 28 out of 98 isolates and in 12 bacterial genera, this gene was not significantly related to ''clean'' or to urban sites (p.0.05) (Figure 3). We sequenced all qacE1/qacED1 amplicons, which exhibited 100% identity in 241 bp to the clinical allele (accession number HM999792). The high frequency of qac gene cassettes (qacE, qacG and qacH) found in non-clinical samples from Australia [51] allowed the authors to propose a relevant role for the qacE1 gene cassette in the origin of the most common genetic platform of the clinical class 1 integrons. Following their hypothesis, the common ancestor of class 1 integrons was embedded in a Tn402-like transposon harbouring a complete qacE1 gene cassette within the VR-1; this gene was subsequently deleted by insertion of the sul1 gene and converted into the wellknown qacED1 of the clinical 39-CS of class 1 integrons [1,52,53]. Recently, an environmental permafrost strain which was presumed to date from to 15.000-40.000 years ago with a typical 39-CS of clinical class 1 integrons was found in Siberia [54]. If this finding is not a contamination, it is likely that the 39-CS has been in nature and probably maintained without interaction with human activities before the antibiotic era.
Clearly, no matter which genetic platform contains the qacE1/ qacED1 gene, its dissemination as either a cassette or a pseudogene between clinical and natural communities is widespread around the world since a long time ago [1,51,53,54] [this work], emphasizing the adaptive role that it possess for a large variety of genomes, habitats and possibly different types of stressors.

Multiple Interactions Define the Ecological and Molecular Behaviour of Each ARG
While we found that the sul1 and qacE1/qacED1 genes were usually located separately in our non-clinical isolates, both ARGs are embedded in the widespread 39-CS of class 1 integrons when they are detected in clinical samples [1,53]. We found both ARG together in only one isolate, 1AL4, which corresponds to Streptomyces spp. (phylum: Actinobacteria). The sequence revealed 100% identity over 822 bp with the array of the 39-CS (accession number EU118148). This finding is probably the result of the flow of clinical strains harbouring class 1 integrons with the typical 39-CS from the hospital to site 1. It is likely that a lateral genetic transfer event to the strain 1AL4 Actinobacteria has happened since this species is rarely isolated from human infections.
Previous reports have shown that sulphonamide and quaternary ammonium compound resistances are usually found in bacterial isolates from natural communities [40,51,55,56,57]. Also, very relevant for the evolution of multidrug isolates, both ARGs have been identified as possessing the potential to co-select for multidrug resistance in non-clinical and clinical samples [40,56,57,58,59].
However, both ARGs differ from functional, taxonomic distribution and molecular perspectives when they are analysed separately. It is well known that the qacE1 gene is a mobile element since it has all of the features of a gene cassette, whereas the sul1 is an open reading frame not associated with an attC site. Thus, it is likely that the mobility conferred by the system's integron/cassette could be one reason for the widespread dissemination of qacE1 within natural communities and genomes. The qacE1/qacED1 genes from our non-clinical samples were 100% identical to the 39-CS of clinical class 1 integrons, showing a different molecular pattern to the sul1 gene. For the latter gene it is possible to distinguish its ''clinical'' or ''environmental'' origin on the basis of the different nucleotide sequences, as also shown for alleles of intI1.
The results of this study showed that the sul1 and qacE1/qacED1 genes have a different distribution between sites with different degrees of urbanization in Patagonia and diverse behaviour from a molecular perspective, suggesting that multiple interactions define the abundance of each type of ARG at a particular site. So these factors need to be analysed in individual studies for each antimicrobial resistance gene.

Non-clinical Samples from Patagonia are a Reservoir of ISCR1
The ISCR1 gene was found in c and b-proteobacteria and in Actinobacteria isolates (n = 11) ( Table 2), and it was common in Pseudomonaceae (8 out of 11 positive isolates), which is the first description of this site-specific recombinase gene in non-clinical samples. The sequence of 475 bp from the 11 ISCR1-positive isolates revealed 100% identity with the clinical allele (EU722351). In a previous study from our laboratory on clinical isolates, this gene was present in 40% of 130 Enterobacteriaceae strains and only in 1% of 100 Pseudomonas aeruginosa isolates (data not shown), suggesting a different taxonomic distribution between clinical isolates compared to natural communities. The high frequency exhibited by ISCR1, as well as its distribution in several taxa in Patagonian samples, could be evidence of the important role of the open environment as a reservoir of this gene in our geographical region. The ISCR1 gene was found in bacterial cells in the same The first number of each isolate corresponds to the sampling site (from 1 to 10, see Table1 sampling sites where isolates with intI1, qacE1/qacED1 and/or sul1 were also identified that ensures an encounter among cells and a putative transference of genes. This pool of genes could be the source for the emergence of the first strain harbouring bla CTX-M-2 associated with ISCR1 on the genetic platform of a complex class 1 integron that had emerged in a clinical isolate in Argentina in 1989 [31,60].

Abundance and Flow of the intI1 Gene in Environmental Samples
We found a frequency of 11.2% intI1 genes in 98 isolates by plating on nutritive agar without antibiotics in 5 out of the 10 sampling sites. This frequency is relatively high in comparison to those obtained in previous studies. Rosewarne et al. [42] found only 0.5% of positive strains (4/790 isolates) and Stokes et al. [1] found 2.1% (4/192 isolates) in a similar model of a culture-based isolation of non-clinical strains without antibiotics in Australia. These contrasting results could be the consequence of different methodologies of isolation, types of habitats, bacterial communities, and also phylogenetic patterns could be involved in the abundance of this gene in different geographic regions. When we analysed the molecular features of the intI1 genes from our study, we identified two ''clinical'' genes of the intI1 gene that were found in the sampling site with the highest level of urbanization (site 1, Ushuaia city). These two ''clinical'' intI1 genes could belong to different clinical strains since they were different alleles obtained from different species of bacteria: sample 1SL5 was from a Pseudomonas spp. (accession number DQ247972 with 22.13% of intI1 genes from Genbank) and sample 1AC2 was from an Aeromonas spp. (CP000650 with 2.55% of intI1 genes from Genbank). The remaining nine alleles, which all were ''environmental'' intI1 genes (n = 5 in Pseudomonas spp., n = 1 in Vibrio spp. and n = 3 in Enterobacteriaceae) ( Table 2), showed sequence diversity with novel mutations that have never been described before; neither in non-clinical nor in clinical intI1 genes (Figure 4).
The intI1 gene flow from the hospital to the open environment has been well established [47,48,50]. Based on the large number of intI1 alleles found in non-clinical samples [1,48][this study], in clinical isolates (from GenBank, up until August 2010, Figure 4), and also found in both types of environments (Figure 4), we propose that at least two different routes of the acquisition of class 1 integrons could have interacting in hospitals during the antibiotic era: on one hand, the most common alleles of intI1 (accession numbers DQ247972, AY463797, AM412236 and DQ315789) must have been the first that were introduced in the hospital niche and, thereafter, were continuously selected by the pressure of antibiotics; and, on the other hand, in certain circumstances, ''environmental'' intI1 genes must have been taken from natural communities and thereby began their propagation and circulation in the clinical habitat under antimicrobial pressure. In our understanding, this flow of genes from the open environment to hospitals is also evidenced by the new and unusual 39ends of class 1 integrons that have been described in sporadic isolates worldwide. Examples of this are strains harbouring the IRt of Tn402 [61] or IS440-sul3-orf1-IS26 [26] instead of 39-CS. In fact, this latest genetic platform has been described as harbouring the qacH gene cassette within the VR-1, which has been detected very frequently in non-clinical samples [26,51]. The high frequency of different intI1 alleles found in clinical and nonclinical isolates shown in the phylogenetic tree of the Figure 4 also suggests that intI1 gene flow between human activities and the environment occurs in both directions.

The High Frequency of intI1 Pseudogenes Reveals a Similar Genomic and Ecological Behaviour for Integron Integrases
The intI1 pseudogenes were found in 9 out of the 10 sampling sites of Tierra del Fuego Island. From a total of 30 intI1-positive isolates, the frequency of occurrence of intI1 genes was 11.2% (11/98) if the entire intI1 gene sequence is considered, and it was 19.4% (19/98) if only the amplification of the integron integrase motif [62] is taken into account ( Table 2). While the complete sequence of the intI1 gene was only detected in c-proteobacteria, the intI1 pseudogenes were also found in two b-proteobacteria isolates and in one Actinobacteria isolate (1SC3). Thus, the range of dispersion of intI1 genes was increased between different taxa and between different sampling sites due to the identification of intI1 pseudogenes. However, this genetic marker was not related to urbanization, which demonstrates that the entire intI1 gene and its pseudogenes exhibit a similar ecological behaviour.
A previous bioinformatics study of all families of integron integrase genes found that 1/3 of intI were pseudogenes [63]. The prevalence of intI1 pseudogenes in our non-clinical samples was two times higher than that of the entire intI1 gene. This greater frequency of intI1 pseudogenes observed in this study is evidence of the significant adverse effects produced by the entire gene in many different bacterial genomes. On the other hand, this widespread dissemination also highlights the fact that class 1 integrons possess a successful mechanism for spreading among natural bacterial communities. In addition, the intI1 pseudogenes have a different pattern of distribution if we compare the natural with the clinical communities. The bioinformatics study we performed on DNA sequences from clinical strains in Genbank (up until August 2010) revealed that only Corynebacterium diphtheriae (accession number BX248353) has an intI1 pseudogene. Thus, the low frequency of intI1 pseudogenes and, therefore, the high frequency of entire intI1 genes in the clinical isolates from GenBank revealed that the genomes circulating in clinical communities have possibly been selected because they have a genomic plasticity that facilitates maintenance of the entire intI1 gene.
The intI1 Genes Isolated from the Open Environment were not Related to Urbanization Although a weak trend towards a high occurrence of intI1 genes in urban areas was observed, there were no statistical correlations The phylogram was obtained with the CLUSTALW application in MEGA v 5.05 program with default parameters. Alleles were indicated with either the isolated name (sequences obtained in this work) or the accession number (GenBank sequences). The source of each allele is shown by a coloured square (red is for clinical isolates, green is for environment isolates and blue is for alleles that have been identified in both clinical an environmental isolates. Asterisks indicate the most common clinical alleles of intI1 (A) and the allele of sul1 embedded in the most common 39CS of clinical integrons (B). The outgroup branch has been reduced in order to appreciate better the other branches. doi:10.1371/journal.pone.0039223.g004 Table 3. Wald statistic values obtained from the generalised lineal model analyses that were applied to the relationships between gene frequencies and 3 independent variables. between the mean occurrence per site of the intI1 genes and its pseudogenes and the three degrees of urbanization (p.0.05) ( Figure 3; Table 3). In other words, ''environmental'' intI1 genes were not significantly more abundant in anthropic environments than in remote areas from urban centres. However, previous reports showed that sites close to human activities have a higher frequency of intI1 genes, as a result of the intI1 flow from clinical samples to the open environment [42,64]. The discharge of genes should be maintained by the release of antibiotics at urban sites or by the presence of contaminants as metals [37,42,64]. The process of co-selection, would be involved in the maintenance of intI1 genes [42,47], as it is the case for transposon Tn21, which possesses determinants of resistance to mercury and usually has a class 1 integron embedded in its genetic platform [65]. Several studies evidenced the flow of class 1 integrons from humans to wastewater treatment plants, rivers, soil and domestic and wild animals [64,66,67,68,69,70,71,72]. From a molecular perspective, a similar scenario was found in our study, since the ''clinical'' alleles of the intI1 gene were only found in sites with a high level of urbanization.
When class 1 integrons were searched for in E. coli strains isolated from several animal populations subjected to different degrees of anthropogenic disturbance, the abundance of intI1 was found to correlate with the closeness to humans [64]. Skurnik et al. [64] explained the absence of class 1 integrons in wild animals as a result of never having been exposed to humans. However, the absence of class 1 integrons in E. coli strains not exposed to human disturbances can be expected since it has been suggested that the genome of E. coli is not able to acquire or maintain class 1 integrons without antibiotic pressure [55]. In other study, Rosewarne et al. (2010) [42] compared the abundance of intI1 in catchments with different levels of human disturbance from the Greater Melbourne area of Victoria, Australia, and found a strong positive relationship between the frequency of occurrence of this gene and heavy metal pollution. One explanation for our findings is that the apparently ''clean'' sites of Tierra del Fuego Island in fact receive or have received some source of pollution (flow of intI1-positive clinical strains, antibiotics and/or heavy metals) that is not associated with the level of urbanization. Some sites are visited by tourists, but most only have a very low frequency of visitors in the months when there is no snow. Nor is it likely to be explained by current pollution levels from heavy metals, which are concentrated in areas close to the cities in this region [73,74]. Moreover, the existence of pollution sources in the past is unlikely as the island of Tierra del Fuego has remained largely untapped for decades because of its geographical position and difficulty of access. Therefore, although cases of contamination could have been sporadic in the past, these could not explain the persistence of intI1 in non-urban regions. Another interpretation for our results it is that the lack of correlation found in our study could be due to the small sample size; however, the frequency of occurrence of intI1 in ''clean'' sites of Tierra del Fuego Island was sufficiently large to deserve attention. Another hypothesis for explaining the differences between studies regarding the incidence of intI1 in ''clean'' sites could be that there are regions of the Earth in which some lateral genetic transfer mechanisms might be more abundant, due to historical processes of regional scale. Independent phylogenetic processes may have caused the intI1 gene to be differentially adapted to different bacterial genomes. The consequence would be that, in certain regions, bacterial species maintain the intI1 gene irrespective of anthropogenic pressure. These type of analysis, based on molecular and ecological studies but on a global scale, could be helpful for disentangling the multiple factors that are involved in the flow and maintenance of the intI1 gene between areas of human activities and natural communities.

Conclusions
Simultaneous analyses at ecological and molecular levels appeared to be a successful strategy for elucidating the role of each component of the genetic platforms associated to antimicrobial resistance of class 1 integrons.
We found that both ARGs (sul1 and qacE1/qacED1 genes), which are usually embedded in the most common genetic platform of class 1 integrons within clinical habitats, showed different ecological and molecular behaviours in natural communities. While the presence of the sul1 gene was the only component of the genetic platforms of class 1 integrons related to urbanization, the qacE1/qacED1 gene was found to be widespread in natural communities with different degrees of anthropogenic disturbances, which highlights the adaptive role of this gene to several different habitats. The LGE (intI1, ISCR1 and tniC genes) can exhibit high levels of diversity and different levels of persistence depending on the habitat and regions of the world. Ecological analysis showed that the intI1 gene, as well as the ISCR1 genes, which are relevant mechanisms involved in the spreading of multidrug resistance mechanisms in clinical isolates in Argentina and worldwide, were not associated with urbanization in the Patagonian samples. A total of 30/98 intI1-positive isolates were identified, with a high frequency of intI1 pseudogenes (19/98), which suggests that although intI1 has a deleterious impact within several genomes, it can easily be disseminated throughout natural bacterial communities. We cannot rule out the possibility that the high percentage of intI1 and ISCR1 genes that we found in the natural communities may be one of the factors that contributes to the increasing frequency of antimicrobial resistance isolates that have characterized Argentina for decades.
The main conclusion of this study is that the ability of natural bacterial communities to act as a reservoir and source of multidrug resistance mechanisms cannot be described by a general process but depend on multiple factors operating at molecular, ecological, phylogenetic and historical levels.

Study Area
The study was conducted in the south-eastern portion of Tierra del Fuego Island ( Figure 2). The area lies within the Sub-Antarctic Deciduous Beech Forest, which is characterized by two species of southern beech, Nothofagus pumilio (Lenga) and Nothofagus betuloides (Guindo) [75]. Its climate belongs to the sub-polar oceanic type. Temperatures are cold all year round, with an average annual temperature of 5.7uC and low annual temperature variations, ranging from 20.3uC in July to 9.4uC in January. There are two urban sites: Ushuaia city with 80,000 inhabitants on the southern coast of the island, bathed in the Beagle Channel, and Tolhuin, a town of about 8000 inhabitants. The island was only colonized at the end of the nineteenth century.
All necessary permits for the described field studies were obtained from Clotilde Lizarralde (Director of the Planning Department in the science and technology area of the province) and Laura Malmierca (Tierra del Fuego National Park Management).

Definition of the Degree of Urbanization
Samples were collected in 10 sites that were selected according to three distinct levels of anthropogenic disturbance ( Figure 2 and Table 1). The degree of urbanization was quantitatively estimated by counting the number of buildings and roads that were contained in a 1 km circular area around each site. These estimations were conducted using satellite images provided by Google Earth. We defined a high level of urbanization as being when the number of buildings was greater than 50 and the number of routes greater than 5. We defined a low level of urbanization as being when the number of buildings was less than 5 and the number of routes less than 2. Two sites were located in Ushuaia City, in the mouths of the Pipo (Site 1) and Ushuaia (Site 10) rivers; three sites were in Tierra del Fuego National Park, at a stream that ends in Ensenada Bay (Site 7), at an upper portion of Pipo River (Site 8) and at Ovando River (Site 6); two sites were at Escondido Lake, one on the shore of the lake where a hotel is located (Site 4) and the other one at a stream that finishes in this lake (Site 5); two sites were along the Provincial Road Nu 26 in a mountainous area south-east of Fagnano Lake, one of which was in the intersection of this road with Turbio River (Site 2) while the other was at Turbera Maucasen (Site 3); and site 9 was located in Moat Farm, in a stream close to the sea ( Figure 2). Sites 3 to 9 had no history of clinical or industrial activities. Animal husbandry in the studied areas is minimal, and there is not systematic records of the use of antibiotics on domestic animals. The sites of Turbio River and Moat Farm are occasionally visited, mainly in the summer. Sites 1 and 10 were categorized as being highly disturbed, sites 3 and 4 were categorized as being medium-level disturbed, and the others were categorized as being low-level disturbed sites (Table 1).

Sampling Techniques
Samples were collected at each site between 20 th January and 7 th February 2006. Shallow freshwater sediment and soil samples from the shore were plated on nutritive agar medium (Britania, Argentina) without a selection of antibiotics. The plates were incubated at 4uC for 8 days, after which all individual colonies from each site and from each plate were plated again in nutritive agar and incubated at 4uC for 4 days. Then, each colony was picked out and place onto Luria Bertani broth and incubated at 4uC for 48 h.

Definition of ''Clinical'' and ''Environmental'' intI1 Alleles
We called non-clinical intI1 genes those that were harboured by the strains isolated from water, sediment or soil in this work. The non-clinical intI1 gene can also be an ''environmental'' or a ''clinical'' allele as defined by Gillings et al., suggesting a putative source from natural or clinical communities, respectively [48].

Phylogenetic Analysis
Phylogenetic evolutionary analysis for intI1 and sul1 sequences were conducted using MEGA v 5.05 software [82]. Sequences obtained in this study as well as alleles deposited in GenBank were included. These sequences were aligned using ClustalW application in MEGA with default parameters. The evolutionary history was inferred using the Neighbor-Joining method. The evolutionary distances were computed using the Maximum Composite Likelihood method. All positions containing gaps were eliminated.

Statistical Analysis
The relationship between the environmental characteristics and the distribution of molecular components of antibiotic resistance were analysed using two statistical approaches (STATISTICA package). Generalized linear models (GLM) were applied with the following characteristics: (1) dependent variable: presence/absence of intI1, sul1, tniC and ISCR1, and number of intI1 mutations; (2) independent variables: two categorical variables, substrate (water or soil) and taxa (c-proteobacteria or other taxa) and one ordinal variable, degree of urbanization (1, 2 or 3); (3) assumed distribution of the dependent variable: binomial for presence/ absence data and ordinal multinomial for mutations; (4) link function: logit.
The GLM was complemented with classic non-parametric tests. For the role of substrate and taxa in the presence of resistance genes, we used 262 contingency tables and Fisher's exact tests, applied to frequencies of occurrence. For the effect of degree of urbanization, we applied Spearman's rank correlations to mean occurrences of genes per site as a dependent variable.

Nucleotide Sequence Accession Numbers
IntI1 and sul1 sequences were deposited at GenBank as accession numbers JN870902 to JN870912 and JX048595 to JX048604 respectively.