OUP user menu

Class 1 integrons in benthic bacterial communities: abundance, association with Tn402-like transposition modules and evidence for coselection with heavy-metal resistance

Carly P. Rosewarne, Vincent Pettigrove, Hatch W. Stokes, Yvonne M. Parsons
DOI: http://dx.doi.org/10.1111/j.1574-6941.2009.00823.x 35-46 First published online: 1 April 2010


The integron/gene cassette system contributes to lateral gene transfer of genetic information in bacterial communities, with gene cassette-encoded proteins potentially playing an important role in adaptation to stress. Class 1 integrons are a particularly important class as they themselves seem to be broadly disseminated among the Proteobacteria and have an established role in the spread of antibiotic resistance genes. The abundance and structure of class 1 integrons in freshwater sediment bacterial communities was assessed through sampling of 30 spatially distinct sites encompassing different substrate and catchment types from the Greater Melbourne Area of Victoria, Australia. Real-time PCR was used to demonstrate that the abundance of intI1 was increased as a result of ecosystem perturbation, indicated by classification of sample locations based on the catchment type and a strong positive correlation with the first principal component factor score, comprised primarily of the heavy metals zinc, mercury, lead and copper. Additionally, the abundance of intI1 at sites located downstream from treated sewage outputs was associated with the percentage contribution of the discharge to the basal flow rate. Characterization of class 1 integrons in bacteria cultured from selected sediment samples identified an association with complete Tn402-like transposition modules, and the potential for coselection of heavy-metal and antibiotic resistance mechanisms in benthic environments.

  • intI1
  • lateral gene transfer
  • adaptation
  • heavy-metal pollution
  • freshwater sediment


Bacteria represent a globally ubiquitous domain of life, exhibiting a high degree of functional diversity (Nealson, 1997); however, investigation into the vertical transmission of information via recombination (Milkman, 1997; Coenye & LiPuma, 2003; Suerbaum & Achtman, 2004; Coscolla & Gonzalez-Candelas, 2007) and mutation (Lawrence & Ochman, 1998) has failed to account for this diversity (Syvanen, 1994; Davison, 1999). Comparative analysis of variation in bacterial genomes has provided evidence that adaptive evolution is largely facilitated by lateral gene transfer (LGT). This lateral movement of genetic information has been implicated in the widespread emergence of many new bacterial phenotypes, particularly in response to strong selection pressures (Ochman et al., 2000; van Elsas & Bailey, 2002).

The integron/gene cassette system provides a mechanism for capturing and expressing genes that have been transferred by LGT (Hall et al., 1999). Integrons are genetic elements that contain a site-specific recombination system comprised of a DNA integrase gene (intI) and an attachment site (attI). The intI-encoded integrase protein mediates site-specific recombination events that result in insertion or excision of small mobile units of DNA known as mobile gene cassettes. These most commonly consist of a single ORF and a recombination site known as a 59-base element (Recchia & Hall, 1995) or attC site (Mazel et al., 1998). Experimental observations have demonstrated that ORFs associated with gene cassettes show a high degree of sequence diversity (Stokes et al., 2001; Holmes et al., 2003; Michael et al., 2004; Boucher et al., 2007; Elsaied et al., 2007; Koenig et al., 2008). The encoded proteins that can be assigned putative functions are representative of many different protein families, suggesting that they may encompass a wide range of potentially advantageous traits.

Class 1 integrons are the archetypal example of this type of element. This class was initially discovered in bacteria exhibiting resistance to multiple antibiotics. In such bacteria, class 1 integrons possessed cassette arrays that included genes encoding antibiotic resistance proteins (Stokes & Hall, 1989). Class 1 integrons are prevalent in pathogenic bacteria, particularly those that have been identified as causative agents of nosocomial infections (Hall et al., 1999; Gillings et al., 2008a). Furthermore, clinical class 1 subtypes are becoming increasingly identified in human commensals (Leverstein-van Hall et al., 2002; Labbate et al., 2008) and livestock (Goldstein et al., 2001; Nandi et al., 2004). The clinical class 1 integrons recovered from these environments have similar structural features and are almost exclusively linked with a specific suite of transposition genes similar to that found in transposon Tn402 (Gillings et al., 2008a). The initial linkage to a Tn402-like tni module enabled class 1 integrons to integrate into the res sites common to plasmids and other transposons (Radstrom et al., 1994; Liebert et al., 1999; Sota et al., 2007), facilitating their rapid dissemination in the period since the introduction of antibiotics (Gillings et al., 2008a). This is despite the observation that the tni module has been inactivated by a series of genetic rearrangements in most clinical class 1 integrons, meaning that their mobilization is limited to situations where the tni proteins are provided in trans (Brown et al., 1996).

It is now known that class 1 integrons are widespread in disparate ecosystems (Rosser & Young, 1999; Nield et al., 2001; Nemergut et al., 2004; Gaze et al., 2005; Hardwick et al., 2008; Wright et al., 2008). Those that have been sampled from environments removed from human activities are generally not associated with the Tn402-like transposition genes. This suggests that class 1 integrons were mobilized before the introduction of antibiotics (Gillings et al., 2008b), and that these elements have a more general role to play in bacterial adaptation beyond encoding antibiotic resistance. This is supported by the discovery of novel class 1 integrons in the environment that have both intI1 sequence diversity and gene cassettes encoding novel proteins not associated with drug resistance (Stokes et al., 2006; Gillings et al., 2008a).

Factors influencing the abundance and distribution of class 1 integrons within natural bacterial communities remain largely unknown. Based on the available information, however, it seems reasonable to suggest that bacteria possessing a functional class 1 integron would have an increased selective advantage due to their ability to tap into the environmental gene cassette metagenome, a vast genetic resource important for enhancing bacterial diversity (Stokes et al., 2001; Holmes et al., 2003; Michael et al., 2004; Gillings et al., 2005; Koenig et al., 2008). For example, a cassette found in heavy-metal-contaminated mine tailings showed similarity to a gene encoding a nitroaromatic compound-degrading enzyme (Nemergut et al., 2004). Others show homology to proteins involved in metal scavenging and efflux, DNA modification, synthesis of polysaccharides and binding of sulfates (Mazel, 2006; Elsaied et al., 2007; Koenig et al., 2008; Nemergut et al., 2008), all of which are functions that are potentially advantageous under certain environmental conditions.

A real-time PCR assay was developed to determine the relative abundance of the class 1 integrase gene intI1 (indicative of the presence of class 1 integrons) in freshwater stream sediments taken from the Greater Melbourne Area (GMA) of Victoria, Australia. By targeting conserved regions of the intI1 gene, it is possible to detect all known class 1 integron variants, irrespective of their origin. Previous observations have indicated that intI1 abundance is positively associated with perturbation of aquatic ecosystems (Hardwick et al., 2008; Wright et al., 2008), such that there is an increased potential for LGT to occur within bacterial communities from polluted sites. However, the mechanisms driving these observations remain poorly defined. Sediment quality in many catchments within the GMA is affected by high concentrations of heavy metals and other toxic substances, deposited into waterways by polluted stormwater runoff from industrial and residential areas (Pettigrove & Hoffmann, 2003). Additionally, the flow of water in some catchments is at least partially mediated by discharges from sewage treatment plants (STPs). This has been shown to result in the release of bacteria carrying Tn402-like class 1 integrons into the environment (Tennstedt et al., 2005). The aim of this study was to determine whether there are any associations between these factors and the abundance of intI1 in benthic bacterial communities across the GMA. Additional insights were obtained through sequence analysis of class 1 integrons found in aerobic heterotrophic bacteria cultured from selected sediment samples.

Materials and methods

Sample collection

Sediments were collected from 30 spatially distinct locations within GMA (Victoria, Australia) (Table S1). Samples were collected in March 2006 and between April and July 2007, in a strategy designed to incorporate sedimentary and basaltic sediments from freshwater catchments differing in both the type and extent of pollution. Sediment was scooped from a large area of the creek bed surface (>10 m2) using a 250-μm mesh net and a sweeping motion, in order to obtain a representative sample from each site. Samples were filtered through a 63-μm mesh net into large buckets and allowed to settle, and then excess water was decanted. The sediment was subsequently homogenized in a single bucket before an aliquot was removed into a sterile 50-mL tube and placed on wet ice. The remaining sediment was collected in 4 × 250-mL sterile glass jars, stored on wet ice and transported to the laboratory.

Sediment chemical analysis

Samples were analysed for cadmium, chromium, copper, lead, nickel and zinc using inductively coupled plasma–atomic emission spectrophotometry. Mercury was detected using flow injection mercury system. Semivolatile petroleum hydrocarbons (C10-C36 fraction) were analysed according to US EPA Method 8015B (http://www.epa.gov/osw/hazard/testmethods/sw846/index.htm). The concentration of nitrogen (as NOx and total Kjeldahl nitrogen) and total phosphorus were analysed using the standard published methodology (APHA, 2004), while the concentration of total organic carbon (TOC) was determined according to Australian Standard 1289.4.4.4 (1997). All analyses were conducted by Australian Laboratory Services (ALS Environmental, Melbourne, Australia), accredited by the National Association of Testing Authorities (NATA), Australia. Blanks, checks and spikes were included as part of the standard quality control procedures.

DNA extraction and quantification

DNA was extracted in triplicate from each sediment sample using the PowerSoil DNA extraction kit (MoBio Laboratories Inc.), and quantified using a fluorescence based protocol with some minor modifications (Zipper et al., 2003). Plasmid DNA to be used as a standard (pGeneGrip; encodes GFP, Genlantis) was extracted from Escherichia coli DH5α cells using the Wizard Plasmid Miniprep system (Promega Corporation), and the concentration estimated using a GeneQuant DNA/RNA calculator (Amersham Biosciences). The plasmid was diluted to create a standard curve ranging from 0.1 to 0.7 ng μL−1 in 1 × TE buffer, pH 8.0. 2 μL of each sediment DNA extraction was diluted 1 : 100 in 1 × TE buffer in duplicate. 50 μL of DNA standard, diluted sediment extract or 1 × TE buffer was added to 50 μL of a 1 : 2000 dilution of SYBR Gold (Molecular Probes) in a black MicroClear 96 well plate (Grenier Bio-One). Fluorescence intensity was determined using a SpectraMax M2 microplate reader. GraphPad Prism 5.0 (GraphPad Software) was used to calculate standard curves and interpolate duplicate fluorescence readings to determine the sediment DNA concentration.

Real-time PCR primer design

Primers for real-time PCR amplification of bacterial 16S rRNA genes (BACT-1369F and 1492R) have been described previously (Suzuki et al., 2000). Primers targeting the green fluorescent protein (GFP) gene from Aequorea victoria (qGFP-1 and qGFP-3, expected product size of 209 bp) and the bacterial intI1 gene (qINT-3 and qINT-4, expected product size of 109bp) were designed specifically for real-time PCR (to minimize primer-dimer formation) using the PrimerQuest and Oligo Analyser tools from Integrated DNA Technologies (http://www.idtdna.com). The selected primers were compared with an alignment of all publicly available intI1 sequences (as at 1 September 2006) to ensure all known variants would be detected, before being verified for target specificity by BLAST searching the NCBI database (http://www.ncbi.nlm.nih.gov/blast.cgi). The primer sequences are provided in the Supporting Information.

Construction of plasmid control for real-time PCR

A plasmid containing the 16S rRNA gene and intI1 PCR products was constructed for use as the control for the real-time PCR assays. An overnight culture of the intI1-positive Azoarcus communis strain MUL2G9 (Stokes et al., 2006) was lysed and used as a template for standard PCR using the real-time PCR primers (qINT-3 and 4 for intI1; 1369F and 1492R for 16S). The MUL2G9 intI1 PCR product was cloned into pCR 2.1 TOPO vector (Invitrogen Corporation) and the MUL2G9 16S PCR product was cloned into pGEM T-easy vector (Promega Corporation). Plasmid DNA was extracted from one positive clone for each insert type using the Wizard SV Miniprep Kit (Promega Corporation). 10 μL of each extraction was digested with NotI restriction enzyme (New England Biolabs). The 16S fragment excised from pGEM T-Easy and the linearized pCR 2.1 TOPO vector containing the intI1 gene fragment were purified from a 1.5% agarose gel using the Wizard SV Gel and PCR clean up Kit (Promega Corporation). The NotI sticky ends were ligated using T4 DNA ligase (New England Biolabs) before transformation of chemicompetent E. coli JM109 cells (Promega Corporation). Colony PCR using M13F and M13R was used to select potential positive clones, from which plasmid DNA was subsequently extracted and sequenced in the forward and reverse orientation (Macrogen Inc., Korea). This confirmed the presence of each insert in as a single copy within a plasmid designated pCONTROL-1.

Real-time PCR assay validation

Real-time PCR was performed using a Roche LightCycler® 480 (Roche Applied Sciences). The assays had annealing temperatures of 55 °C (16S rRNA gene) or 60 °C (intI1, GFP) with annealing and extension times of 15 s each. PCR consisted of 40 amplification cycles, followed by an amplicon melting profile step. Each of the three PCR assays consistently showed an amplification efficiency of greater than 1.9 and an error rate of less than 0.02 based on duplicate standards ranging from 3 × 102 to 3 × 108 copies of pCONTROL-1. Background fluorescence observed in negative control reactions was outside the range of these standard curves. Amplicon melting profiles suggested amplification of a single specific product in each standard reaction, with no evidence of primer-dimer formation. Results were confirmed by agarose gel electrophoresis. The specificity of each real-time PCR assay was confirmed through the use of several DNA templates. Genomic DNA was prepared from E. coli JM109 (Promega Corporation) and MUL2G9 and MUL2G11 (different variants of the intI1 gene sequence) (Stokes et al., 2006) using the UltraClean Microbial DNA Isolation Kit (MoBio Laboratories Inc.). Negative control genomic DNA was prepared from HT1080 human fibrosarcoma cells using the UltraClean Tissue DNA Isolation Kit (MoBio Laboratories Inc.). The extracted DNA was used as the template for real-time PCR in each assay as described previously. Results indicated that all templates containing the target gene yielded a positive result, and no false-positive amplification was observed.

Quantitative PCR analysis of sediment DNA

The presence of PCR inhibitors in sediment DNA extractions was investigated using a spiking assay. Twenty-five nanograms of each extraction was added in duplicate to a GFP real-time PCR that also contained 1.5 × 104 copies of pGeneGrip. PCR inhibitors were detected in extractions from five out of the 30 sites (R1, R3, R4, C2 and C3). Diluting the DNA extractions 1 : 5 eliminated the effect of these inhibitors. Target quantification was performed using 5 ng of DNA for samples containing inhibitors and 25 ng for all others. Each extraction was amplified in duplicate in a single experiment, with experiments performed in duplicate for each target, thereby providing four quantifications of each target for each DNA extraction. The resulting quadruplicate quantifications for each target/extraction combination showed no significant intra- or inter-run variation, with crossing points differing by <0.4 cycles in all cases. Furthermore, each crossing point fell within the linear range of the appropriate standard curve, allowing accurate interpolation to determine the mean copy number for each extraction across the two experiments. Results were subsequently calculated as a ratio of the number of intI1 genes per 16S rRNA gene at each sample site, averaged across triplicate DNA extractions. Analysis of a relative abundance calculated in this way ensures that the copy number of either target in any given genome is not assumed.

Statistical analysis

Nonparametric Kruskal–Wallis and Mann–Whitney tests were applied to determine the significance of variation in abundance of intI1 between catchment types (graphpad prism version 5.0, GraphPad Software). Sediment quality data were log-transformed and reduced to factor scores with the principal component analysis (PCA) function of spss for Windows (version 16.0; SPSS, Chicago, IL). Nonparametric Spearman's rank correlations (rs) were computed to compare the intI1 abundance at each site against the PCA factor scores.

Isolation and sequence characterization of intI1-positive strains

A total of 790 aerobic heterotrophic bacterial strains were isolated from selected sampling sites (I1-I5, R6-7, A6 and S8) and screened for carriage of class 1 integrons according to the protocol outlined by Stokes (2006). Four positive isolates were detected and identified by sequence analysis of PCR-amplified 16S and/or rpoB genes. Isolate LMCB014 was classified as Comamonas testosteroni, isolate HMCB026 was classified as Aeromonas media and isolates DCB015 and KCB005 were classified as Pseudomonas alcaligenes and Pseudomonas oryzihabitans, respectively (refer to Supporting Information for further details and trees representing phylogenetic assignments). Fosmid libraries were constructed using the CopyControl fosmid library production kit (Epicentre Biotechnologies), and PCR screened to obtain clones containing intI1 (Stokes et al., 2006). Fosmids were induced and extracted from intI1-positive clones using the Genopure Plasmid Midi Kit (Roche Applied Sciences). The sequence of fosmid clone LMCB014 F153 was determined by shotgun clone sequencing (Macrogen Inc.). Partial sequences from fosmids DCB015 F359 and KCB005 F7 (approximately 5.3 kb each) were deduced by single extension sequencing. Putative ORFs were identified using gene locator and interpolated Markov ModelER (glimmer) v3.02 (http://www.ncbi.nlm.nih.gov/genomes/MICROBES/glimmer_3.cgi). Functions were assigned to each ORF after performing blastn and blastx searches (http://www.ncbi.nlm.nih.gov/BLAST/BLAST.cgi) on 28 July 2008. Gene cassettes and their associated attC sites were identified using the ACID database (http://integron.biochem.dal.ca/ACID/login.php) (Joss et al., 2009).

Nucleotide sequence accession numbers

Fosmid sequences were prepared as GenBank flat files for submission using sequin version 9.20. The complete fosmid sequence from LMCB014 (31 983 bp) has been deposited in GenBank under accession number GQ281704. Partial sequences from DCB015 (5243 bp) and KCB005 (5239 bp) have been deposited in GenBank under accession numbers GQ281702–3. The rpoB and/or 16S rRNA gene sequences for each isolate have been deposited in GenBank under accession numbers FJ824114–120.


Sediment quality data

Sediment heavy metal concentrations varied across each of the sample sites (refer to Supporting Information). The probable impact of this pollution was determined by comparison with the Australian and New Zealand sediment-quality guidelines (SQGs) (ANZECC/ARMCANZ, 2000). Heavy metal concentrations below the SQG-low trigger value are deemed unlikely to cause toxicity to the associated aquatic biota, while concentrations above the SQG-high value are highly likely to have an adverse effect. Sites located in industrial catchments (I1–I7) had relatively high levels of pollution, with zinc being the predominant heavy metal species. With the exception of Site I7, each site had at least two heavy metals in excess of the SQG-low value, and at least one heavy metal in excess of the SQG-high value. Five out of the seven sites located in residential catchments (R1, R3, R4, R5 and R7) and one site located downstream from an STP output (Supporting Information, Fig. S2) were polluted with zinc concentrations in excess of the SQG-high value. However, samples taken from the remaining sites, encompassing clean, unpolluted locations (C1–C3), agricultural pastures (A1–A6) and the majority of sites downstream from STPs (Fig. S1 and Tables S1–S4) had no heavy metal concentrations in excess of the SQG-high value. Sites sampled in this study were also polluted with cadmium, chromium, copper, nickel, lead and mercury to varying degrees. The concentration of total petroleum hydrocarbons, total nitrogen, total phosphorus and the percentage (dry w/w) of TOC was also determined, but no trends could be identified.

Comparison of intI1 abundance in different catchment types

The abundance of intI1 varied across sites (Fig. 1), ranging from 0.020 at Site A1 to 4.294 at Site S3. Overall, the sites located downstream of sewage outputs or in industrial or residential catchments show a higher proportion of intI1 than sites located in agricultural or clean areas. An initial Kruskal–Wallis test confirmed that there were significant differences between sites (K–W U=16.13, P<0.005). Pairwise Mann–Whitney tests indicated that the average intI1 abundance at the agricultural/clean sites was significantly lower compared with the industrial (M–W U=1.00, P<0.001), residential (M–W U=1.00, P<0.0005) and sewage output (M–W U=4.00, P<0.001) sites, but the latter three catchment types were not significantly different from each other.


Abundance of class 1 integrase genes in freshwater sediment samples. Data are plotted as the mean percentage of intI1 to 16S rRNA gene for each sampling location, classified according to the catchment type. Horizontal lines indicate the average intI1 abundance for each catchment type.

Correlations between integrase gene abundance and sediment quality data

The first three principal components accounted for 77.4% of the total variance in the sediment chemical data (44.9%, 18.6% and 13.9%, respectively). Comparison of the factor scores to the abundance of intI1 identified a positive correlation with PC1 (P<0.01). PC1 was strongly correlated with the concentration of copper, lead, zinc and mercury (P<0.001), in addition to chromium and nickel (P<0.05). PC2 showed a strong positive association with cadmium, chromium and nickel (P<0.001), and a negative association with total petroleum hydrocarbons (P<0.001) and TOC (P<0.05), while PC3 was strongly correlated with total nitrogen, phosphorus and organic carbon (P<0.001). No associations between intI1 abundance and PC2 or PC3 were demonstrated.

Abundance of intI1 in sediments receiving treated sewage outputs

The seven sites located downstream of sewage outputs showed the most variation in class 1 integron load, with the highest relative abundance (4.29) found at Site S3. This site is located immediately adjacent to an untreated human effluent output. It was included in the study to represent a ‘worst case scenario’ for release of class 1 integrons from human waste into a watercourse, with the output contributing 100% to the basal flow. Each of the remaining six sites varied in terms of the distance from the output of treated sewage to the sampling site (in km), contribution of the output to the basal flow (as a percentage) and the concentration of nutrients and pollutants. The abundance of intI1 was strongly correlated with the contribution of treated sewage output to basal flow (P=0.0001). This latter relationship was investigated further using linear regression (Fig. 2). A plot of intI1 abundance against the percentage of flow provided by the treated output gave an r2-value of 0.904.


Linear regression analysis of intI1 abundance and percentage contribution of treated sewage output to basal flow.

Characterization of class 1 integrons from cultured bacterial strains

Four aerobic heterotrophic bacteria carrying class 1 integrons were detected from a total of 790 isolates. Pseudomonas alcaligenes DCB015 was isolated from Site I2, P. oryzihabitans KCB005 was isolated from site I3, Aeromonas media HMCB026 was isolated from Site I7 and C. testosteroni LCMB014 was isolated from Site A6. Initial sequencing of one intI1-positive fosmid clone from each strain with primer HS459, which targets the 3′-conserved segment (3′-CS) of clinical class 1 integrons (Holmes et al., 2003), generated data only from A. media HMCB026. The sequence showed 100% identity to the aminoglycoside adenylyltransferase gene cassette designated aadA11 (GenBank accession number AJ567827). Failure to generate sequence data with primer HS459 from the remaining three fosmids suggests that the class 1 integrons in these strains do not contain a 3′-CS.

Class 1 integron from C. testosteroni strain LMCB014

The intI1-positive fosmid insert from C. testosteroni LMCB014, approximately 32 kb in size, consists of 37 putative ORFs. The region from bases 12007 to 28012 is shown in Fig. 3a. A complete 5′-CS, consisting of the inverted repeat IRi, class 1 integrase gene intI1, the gene cassette promoters PC and P2, the integrase promoter PINT and the integrase-associated attachment site attI1, was defined. Immediately adjacent to the attI site are two identical ORFs of unknown function that have been designated as ORF 1 in other class 1 integrons (Tennstedt et al., 2003). Each of these ORFs is followed by an associated attC site, but their structure is unusual in that the ORFs are facing the same direction as the integrase ORF. Adjacent to the ORF1 gene cassettes is a third gene cassette identified as aadA11. It is positioned in the normal orientation for gene cassettes, but the start of the associated attC site is located 51 bp inside the 3′-end of the ORF.


Diagrammatic representation of Comamonas testosteroni LMCB014 F153 partial sequence. (a) Depiction of region spanning bases 12007–28012 of LMCB014 fosmid 153. The class 1 integron contains three gene cassettes and is located inside a complete Tn402-like transposon, flanked by the inverted repeats IRi and IRt. A putative parA gene is located at the IRi end of this transposon. A truncated TnOtChr transposon, missing the tnpR resolvase and associated inverted repeat is located at the IRt end. (b) Alignment of sequences from LMCB014 and TnOtChr from Ochrobactrum tritici strain 5bvl1 (Branco et al., 2008). Homology between the two sequences begins at the end of IRt in LMCB014 as shown with asterisks, with the start codon for tnpR in TnOtChr indicated in bold.

The presence of a complete 5′-CS suggests that the class 1 integron from LMCB014 is a member of the clinical lineage (Stokes et al., 2006; Gillings et al., 2008a), despite the absence of a 3′-CS. A complete putative tni module with 100% similarity to that described in Tn402 was identified adjacent to the aadA11 gene cassette in LMCB014. It encodes four transposition proteins designated TniC, -Q, -B and -A, and an inverted repeat designated IRt. Tn402-like transposons exhibit site-selective recombination activity, inserting into resolvase sites commonly found in plasmids and other transposons (Liebert et al., 1999; Minakhina et al., 1999). Accordingly, the IRi end of the intI1-carrying transposon in strain LMCB014 is located 739 bp downstream from the start codon of a putative parA resolvase gene, suggesting that this element is located on a plasmid. However, no remnant sequences resembling a res site (Sota et al., 2007) could be found in the region between the start codon of parA and IRi, and the sequence at the outer ends of the IRs does not consist of the 5-bp direct target sequence duplication that would typify a recent transposition event (Kholodii et al., 1993, 1995).

The sequence flanking the inverted repeat IRt (bases 21434–28012) shows 99% identity to a transposon previously identified in a strain of Ochrobactrum tritici (Branco et al., 2008). This transposon, designated TnOtChr, consists of a tnpR gene encoding a resolvase, four genes designated chrB, -A, -C and -F that comprise a functional chromate resistance operon and a transposase gene designated tnpA. It is flanked by two inverted repeats. Comparison of TnOtChr with the LMCB014 fosmid sequence indicates that the tnpR gene and its associated inverted repeat are not present in LMCB014. The Tn402-like transposon inverted repeat IRt in LMCB014 is immediately followed by the first 10 bp of the tnpR ORF, but the remaining sequence of this gene and its associated inverted repeat are not present (Fig. 3b). The four genes associated with the chromate resistance operon and the tnpA transposase gene were identified in the adjacent sequence in LMCB014, as was a flanking inverted repeat sequence identical to that found in TnOtChr. Deletions in the genomic regions immediately adjacent to Tn402-like transposons have been observed previously (Sota et al., 2007; S. Petrovski, pers. commun.). We hypothesize that the mobilization potential of ΔTnOtChr would have been lost as a result of the deletion, as both terminal inverted repeats are normally required for TnpA-mediated replicative transposition to occur (Plasterk, 1995).

Class 1 integrons from Pseudomonas strains DCB015 and KCB005

The intI1-positive fosmids from the two Pseudomonas strains were partially sequenced by primer walking. Approximately 5.3 kb of sequence was generated from each strain (Fig. 4). The 5′-CS of these class 1 integrons have the same structure as that found in LMCB014. Two gene cassettes were identified in DCB015. The first encodes a protein of unknown function, while the second encodes a protein previously shown to provide resistance to quaternary ammonium compounds (qacF) (Paulsen et al., 1993). The last 900 bp of sequence from DCB015 shares 100% similarity with the 3′-end of the tniQ gene from Tn402. PCR amplification of DNA from the intI1-positive fosmid using primers targeting the tniA and tniB genes from Tn402 (refer to Supporting Information) was used to confirm that a complete tni module is present in this strain. These primers failed to yield a product when genomic DNA from strain KCB005 was used as the template. Furthermore, no gene cassettes were found in KCB005. The intI1 gene in this isolate is adjacent to a gene encoding a product with weak homology to an AraC-family transcriptional regulator. This is followed by a gene encoding a product with weak homology to a sterol desaturase-like protein. Chromosomal class 1 integrons from members of the Betaproteobacteria are often situated in close proximity to transcriptional regulators. However, the region of the 5′-CS between the start of IRi and 107 bp beyond the intI1 stop codon is notably absent in the clinical predecessors, as it is thought to have been generated through a subsequent recombination event with a Tn402-like transposon (Gillings et al., 2008a).


Diagrammatic representation of DCB015 F359 and KCB005 F7 partial sequences. (a) DCB015 sequence. The class 1 integron contains a two gene cassettes and a complete transposition module. The location of sequencing primers is indicated, which provided information from part of tnpA through to part of tniQ. Primers DCB015-5 and DCB015-6 were used to confirm the presence of the tni module using PCR. (b) KCB005 sequence. The class 1 integron contains no gene cassettes, but is flanked on the left-hand side by the inverted repeat IRi. The location of sequencing primers is indicated, which provided information from part of tnpA through to downstream of orf4.

Sequence data flanking the 5′-CS of the Pseudomonas class 1 integrons suggests that their current genomic position has been mediated by Tn402-like transposition events targeting resident transposons, as they are are adjacent to genes encoding Tn1721-like (DCB015) and Tn5044-like (KCB005) resolvases (refer to Supporting Information). Putative resIII and resII resolvase-binding sites were identified upstream of the tnpR start codons in the Tn1721 and Tn5044 sequences. The inverted repeats IRi associated with the 5′-CS are located in between the resII- and resI-binding sites, indicating that this is the insertion point of the integron-carrying Tn402-like transposon into the resident transposons.


In this study, we have assessed the prevalence of class 1 integrons in freshwater sediment bacterial communities inhabiting a range of catchment types affected by varying degrees of heavy metal pollution. A statistically significant difference (M–W P<0.001) in the abundance of intI1 was identified between sites with considerable anthropogenic disturbance and sites in relatively undisturbed or agricultural catchments, indicating that perturbation of the benthic environment leads to an increase in the number of bacteria that possess class 1 integrons. Similar conclusions have been drawn from previous studies investigating the abundance of class 1 integrons in aquatic ecosystems (Hardwick et al., 2008; Wright et al., 2008). We then investigated potential relationships between sediment chemical composition and the abundance of intI1 within the associated bacterial communities using PCA. The first factor score accounted for 44.9% of the variation in the sediment quality data, and was comprised only of the predominant heavy metal species. A positive correlation between intI1 abundance and this first factor score indicates that increases in the abundance of class 1 integrons occur in conjunction with increases in the concentration of sediment-bound pollutants.

There is compelling support for the notion that exposure to heavy metals can act as an indirect selective pressure for maintaining antibiotic resistance (McArthur & Tuckfield, 2000; Stepanauskas et al., 2006; Wright et al., 2006, 2008; Wardwell et al., 2009). One of the earliest discoveries was made in primates, where it was shown that mercury released from amalgam fillings resulted in an increase in plasmids encoding resistance to both mercury and multiple antibiotics in the normal bacterial flora (Summers et al., 1993). This suggestion is further supported by the existence of complex mobile genetic elements such as plasmid NR1 (R100), first obtained from a Shigella flexneri isolate in the late 1950s (Nakaya et al., 1960). Several transposons have been identified in this multiple drug-resistant plasmid, including Tn21, which contains a class 1 integron with an aminoglycoside adenylyltransferase gene cassette (aadA1) adjacent to a mercury resistance operon (Liebert et al., 1999). Analysis of the flanking sequence of chromosomal class 1 integrons from the Betaproteobacteria provides interesting insights into potential mechanisms for indirect selection of intI1 in aquatic ecosystems. Although relatively few nonclinical class 1 integrons have been characterized to date, at least three examples in which the intI1 ORF is located in close proximity to ORFs encoding products homologous to the heavy metal efflux pump czcA have been discovered (Stokes et al., 2006; Gillings et al., 2008a). In this study, we describe an interesting hybrid mobile genetic element in C. testosteroni strain LMCB014, comprised of a class 1 integron encoded on a functional Tn402-like transposon and a truncated chromate resistance transposon linked by a recombination event on a plasmid. The discovery of such an element provides additional evidence that the presence of heavy metals may lead to retention of class 1 integrons in aquatic environments due to coselection.

The results presented in this study also demonstrate that close proximity to an untreated human effluent output substantially increases the abundance of intI1 in sediments. The concentration of class 1 integrons at sites downstream from treated sewage outputs was strongly correlated with the contribution of the output to the overall flow in the waterway. Results from a previous study identified final effluent from a wastewater treatment plant as a source of these elements in environmental samples (Tennstedt et al., 2005). Class 1 integrons associated with antibiotic resistance gene cassettes are routinely sampled from the environment (Rosser & Young, 1999; Petersen et al., 2000; L'Abee-Lund & Sorum, 2001; Schmidt et al., 2001; Gaze et al., 2005), but there are other potential routes for clinical class 1 integrons to enter freshwater catchments apart from treated sewage outputs. Stormwater drains represent the main source of sediment-bound pollutants within the Greater Melbourne catchment area (Pettigrove & Hoffmann, 2003). Discharges into Melbourne's Yarra River from stormwater drains have been found to contain high levels of E. coli, an indicator organism of faecal contamination (Robinson et al., 2007). It is therefore possible that contaminated stormwater runoff may provide influxes of heavy metals and other pollutants in addition to commensal and pathogenic bacteria carrying class 1 integrons. If this is the case, the correlation between intI1 abundance and sediment heavy metal pollution may have been influenced by colocalization effects. Additional information is required in order to understand the complex relationship between class 1 integron abundance and anthropogenic pollution in freshwater catchments, and to determine how carriage of intI1 may provide a direct adaptive advantage in these environments.

It is difficult to make inferences about the structure and source of class 1 integrons at the sites sampled for this study, as the real-time PCR primers targeted all known intI1 variants. Observations that class 1 integrons can acquire gene cassettes from different integron classes under laboratory conditions (Rowe-Magnus et al., 2002), and that similar qac cassettes are distributed between disparate integron classes (Gillings et al., 2009), suggest that the diverse environmental gene cassette metagenome (Stokes et al., 2001; Holmes et al., 2003; Michael et al., 2004; Gillings et al., 2005; Koenig et al., 2008) is accessible to all class 1 integrons. It is therefore reasonable to suggest that the observed increases in the abundance of intI1 in benthic bacterial communities will enhance the potential for cassette-mediated LGT to occur, irrespective of the lineage of class 1 integrons that are present. In this investigation, four class 1 integrons from cultured bacterial strains were characterized, all of which showed features previously recognized in the evolution of the clinical intI1 subtype. Interestingly, only one isolate (A. media HMCB026) had the features typical of class 1 integrons commonly found in pathogenic and commensal bacteria from humans and animals (Hall & Collis, 1998; Goldstein et al., 2001; Leverstein-van Hall et al., 2002; Nandi et al., 2004). The proposed evolutionary model suggests that clinical class 1 integrons arose from an initial recombination event between a chromosomal class 1 integron and a Tn402-like transposon. This was followed by multiple rearrangements that inactivated the transposition module, thereby eliminating the self-mobilization capability of the clinical lineage (Stokes et al., 2006; Gillings et al., 2008a). Perhaps surprisingly, the existence of class 1 integrons possessing complete transposition modules is relatively rare (Radstrom et al., 1994; Tennstedt et al., 2005; Toleman et al., 2007; Labbate et al., 2008). In this paper, two additional examples are described. Their gene cassette arrays contained ORFs encoding resistance to quaternary ammonium compounds (qacF from P. alcaligenes DCB015) and streptomycin (aadA11 from C. testosteroni LMCB014). Detection of class 1 integrons in aquatic ecosystems that are associated with functional transposition modules and clinically relevant gene cassettes is of interest due to their increased dissemination potential and the intersection of such environments with the food chain. It suggests that future investigations into the continuing evolution of class 1 integrons should not be solely focused on isolates from clinical settings.

Supporting Information

Additional Supporting Information may be found in the online version of this article:

Table S1. Sample sites aand heavy metals in excess of recommended SQGs. awhere I: industrial; S: downstream from sewage output; R: residential; A: agricultural; and C: clean.

Table S2. List of primers used in this study.

Fig. S1. Molecular phylogenetic reconstruction of the Comamonadaceae, based on 16S rRNA gene sequences.

Fig. S2. Molecular phylogenetic reconstruction of the Aeromonas genus, based on rpoB gene sequences.

Fig. S3. Molecular phylogenetic reconstruction of the Pseudomonas genus, based on rpoB gene sequences.

Fig. S4. Targeting of res sites by class 1 integron-carrying transposons in DCB015 and KCB005.

Please note: Wiley-Blackwell is not responsible for the content or functionality of any supporting materials supplied by the authors. Any queries (other than missing material) should be directed to the corresponding author for the article.


We thank Kelly Ewen-White and Matthew Tinning (Roche Applied Science) for helpful discussions and technical support, Maurice Labbate (Department of Medical and Molecular Biosciences, University of Technology Sydney) for assistance with fosmid library construction, Stephen Doyle (Department of Genetics, La Trobe University) for provision of HT1080 human fibrosarcoma cells and critical review of this manuscript and Kirby Ellis for helpful discussions. This work was supported by an Australian Research Council Linkage Grant.


  • Editor: Patricia Sobecky


View Abstract