Regional dissemination of a trimethoprim-resistance gene cassette via a successful transposable element.

BACKGROUND: Antimicrobial resistance is a growing international problem. We observed a 50% increase in the prevalence of trimethoprim resistance among fecal Escherichia coli from healthy Nigerian students between 1998 and 2005, a trend to increase that continued in 2009. METHODS AND
FINDINGS: A PCR-based screen revealed that 131 (43.1%) of isolates obtained in Nigeria in 2005 and 2009 carried integron-borne dfrA cassettes. In the case of 67 (51.1%) of these isolates, the cassette was a class 1-integron-borne dfrA7 gene, which has been reported at high prevalence from E. coli isolates from other parts of Africa. Complete sequencing of a 27 Kb dfrA7-bearing plasmid from one isolate located the dfrA7 gene within a Tn21-type transposon. The transposon also contained an IS26-derived bla/sul/str element, encoding resistance to β-lactams, sulphonamides and streptomycin, and mercury resistance genes. Although the plasmid backbone was only found in 12 (5.8%) of trimethoprim-resistant isolates, dfrA7 and other transposon-borne genes were detected in 14 (16.3%) and 32 (26.3%) of trimethoprim resistant isolates collected in Nigeria in 2005 and 2009, respectively. Additionally, 37 (19.3%) of trimethoprim-resistant E. coli isolates collected between 2006 and 2008 from Ghana were positive for the dfrA7 and a transposon marker, but only 4 (2.1%) harbored the plasmid backbone.
CONCLUSIONS: Our data point to transposition as a principal mechanism for disseminating dfrA7 among E. coli from Nigeria and Ghana. On-going intensive use of the affordable broad-spectrum antibacterials is likely to promote selective success of a highly prevalent transposable element in West Africa.

Introduction

The broad-spectrum antibacterial trimethoprim is heavily used worldwide, most commonly in combination with sulphonamides. Following the evolution of resistance trimethoprim has, in many cases, been replaced by newer antimicrobials. But because of its low cost, broad spectrum, excellent safety profile, high stability and oral bioavailability, trimethoprim-sulphamethoxazole is still used intensively in many African countries [1], [2]. In the last two decades trimethoprim-sulphamethoxazole has been recommended as a prophylactic in HIV/AIDS patients on top of extensive curative use. This medicament has prevented potentially life-threatening opportunistic infections, is a useful prelude to antiretroviral therapy for many patients and averts loss of life that would otherwise accompany the slow roll-out of antiretrovirals [3], [4].

Trimethoprim inhibits the enzyme dihydrofolate reductase (Dfr). E. coli is most commonly becomes resistant by acquiring one of over 30 known dfr genes encoding resistant variants, most of which belong to the dfrA category [5]. Many dfr gene cassettes, including about half of the known dfrA alleles, lie within gene capture and expression systems known as integrons [6]. Integrons contain an integrase gene, the product of which catalyzes the integration of circular cassettes at an adjacent integration site, attI. Each antimicrobial resistance cassette consists of a promoterless gene and an attachment site, attC, which recombines with attI upon integration (Figure 1). A given integron can have from zero to several cassettes and cassette transcription is facilitated by a promoter at the 5′ end of the integration site. Four classes of integrons have been well-characterized, although others do exist in nature [7]. Class 1 and class 2 integrons have been implicated in resistance gene dispersal, particularly in E. coli and other enteric bacteria.

Figure 1

Structure of generic integron.

Each integron is comprised of 5′ and 3′ conserved ends separated by a variable region (black), which contains zero to several cassettes. The figure illustrates an integron bearing a single cassette. Cassette incorporation occurs by recombination between an attC attachment site at the 3′ end of the the gene cassette (hatched) at the integron’s attachment site (attI), located at the 5′ conserved end. This reaction is catalyzed by a site-specific recombinase encoded by the 5′ intI gene. Class 1 and Class 2 integrons, which are the sub-categories most commonly associated with drug resistance contain a promoter at the 5′ conserved end, from which cassette genes are transcribed.

For three decades, the class 2 integron-bearing transposon Tn7 has been implicated in global dissemination of the dfrA1 gene across a wide range of bacterial species [8], [9], [10], [11], [12], [13], [14], [15]. There have since been many studies cataloguing the relative distribution of different dfr cassettes across the globe but, in spite of increasing interest in the forces underlying evolution to trimethoprim resistance [16], [17], [18], very little is known about the context in which more than a handful of the most promiscuous alleles are spread. Trimethoprim resistance rates among commensal and pathogenic enteric bacteria reported in the literature from Africa are very high, typically approaching or exceeding 60% [19], [20], [21], [22], [23], [24]. In 2005, we observed an increase in trimethoprim resistance among fecal E. coli from healthy adults in Nigeria compared to similar strain collections procured in 1986–1998. We sought to determine the contributions of integrons to multiple- and trimethoprim resistance. Trimethoprim resistance was most commonly attributed to a class 1 integron-borne dfrA7 cassette. We locate this promiscuous cassette to a transposable element and demonstrate that dfrA7 cassettes on mobile elements may be regionally disseminated.

Materials and Methods

Strains

Permission to conduct this study was provided by the Institutional Review Board of Obafemi Awolowo University, Ile-Ife, Nigeria. The study was also approved by the Institutional Review Board of the University of Ghana Medical School. All volunteers gave informed consent. E. coli isolates were recovered from stool specimens collected from apparently healthy undergraduates at the Obafemi Awolowo University and from apparently healthy adults presenting for physical check-ups at the Korle Bu Hospital Teaching Hospital as described previously [20], [25]. We collected and processed specimens in 2005 and 2009 using identical protocols to those used in 1986–1998 in Nigeria and 2006–2008 in Ghana [20], [25]. Lactose-fermenting colonies isolated on MacConkey agar were confirmed as E. coli by biochemical testing. (Lactose-negative E. coli were excluded). Colonies from the same specimen with identical biochemical and susceptibility profiles were treated as identical isolates. A total of 128 isolates from Nigeria in 2005 and 176 in 2009 were included along with 130 isolates from Ghana in 2006, 73 isolates from Ghana in 2007, and 88 isolates from Ghana in 2008.

Susceptibility Testing

The Clinical and Laboratory Standards Institute (CLSI, formerly NCCLS) disc diffusion method was used to determine susceptibility of the isolates to eight antibacterials [26]. Discs used contained ampicillin (10 µg), streptomycin (10 µg), trimethoprim (5 µg), tetracycline (30 µg), nalidixic acid (30 µg), chloramphenicol (30 µg), ciprofloxacin (5 µg) and sulphonamide (300 µg) (Oxoid/Remel). E. coli ATCC 35218 was used as control strain. Inhibition zone diameters were interpreted in accordance with CLSI guidelines using WHONET software version 5.3 [27].

General Molecular Biology Procedures

Genomic DNA was extracted using the Promega Wizard kit. DNA amplification was performed using Platinum PCR Supermix (Invitrogen) and 1 µM oligonucleotide primer in each reaction. Oligonucleotide primer sequences are listed in Table S1. Sensitivity and specificity of each primer pair for its target sequence was tested in silico, using Genbank and BLAST, and experimentally using strains of known genotype. All amplifications began with a two-minute hot start at 94°C followed by 30 cycles of denaturing at 94°C for 30 s, annealing at 5°C below primer annealing temperature for 30 s, and extending at 72°C for one minute per kilobase of DNA. When the target PCR product was over 3 Kb, we used Pfx polymerase in accordance with manufacturer’s instructions, with annealing at 5°C below the temperature used for Taq PCRs. Where necessary, for sequencing, PCR amplicons were TA cloned into the pGEMT vector (Promega) according to manufacturer’s directions and plasmids were transformed into chemically competent E. coli K-12 TOP10 cells. Large, naturally occurring plasmids were electroporated into DH5αE electrocompetent cells (Invitrogen) using a Biorad micropulser according to manufacturers’ instructions, and extracted by a modified boiling protocol as described previously [28], [29]. All other molecular biology operations were performed using standard procedures [30].

Integron Cassette Amplification and Identification

The variable, cassette-containing portions of class 1 and class 2 integrons were amplified using conserved-end primers and protocols described by Levesque et al [31] and White et al [32] respectively. Restriction fragment polymorphism analyses of all amplicons were performed in separate RFLP reactions using AluI and MboI. At least three representatives of unique profiles were cloned and sequenced.

Plasmid Replicon Typing

Three multiplex panels comprised of 18 primer pairs were used to identify plasmid replicons by PCR as described by Johnson et al [33]. Strains carrying well-characterized plasmids pMAR-7, pB171, pHCM1 and pED204 were used as controls [34], [35], [36], [37].

fliC Genotyping and Multilocus Sequence Typing

fliC PCR-restriction fragment length polymorphism (RFLP) typing was performed as described by Fields et al [38]. The E. coli fliC gene was amplified using the primers F-FLIC1 and R-FLIC2 (Table S1). Amplicons were digested with RsaI and restriction profiles were compared after electrophoresis on 2.5% agarose gels. Multilocus sequence typing was performed as described by Wirth et al [39]. PCR primers listed in Table S1 were used to amplify gene fragments from the adk, fumC, gyrB, icd, mdh, purA and recA and amplified DNA products were sequenced from both ends. Allele assignments were made at the publicly accessible E. coli MLST database at http://www.mlst.net/.

Shot-gun Sequencing and Sequence Analysis

Whole-replicon shotgun library preparation, sequencing and assembly of a large plasmid were performed by SeqWright DNA Technology Services (Houston, TX). Sequence analyses and annotation were performed in Artemis [40]. Open reading frames were initially defined by Glimmer. Annotations were made where BLAST e-values equaled or approached zero and there was 98% or greater identity at the DNA and amino acid levels. Direct and inverted repeats were identified by dot-plot analysis of pairwise FASTA alignments made using the BLAST suite. Open reading frame and feature plots were prepared using Artemis and DNAPlotter [41].

Results and Discussion

Increase in Trimethoprim Resistance Rates in South-western Nigeria from 2005 is Associated with a Predominant, Class 1-integron-borne dfrA7 Cassette

Between 1986 and 1998, we followed resistance trends for eight antimicrobials among commensal bacteria belonging to the species E. coli from apparently healthy Nigerian students [25]. We observed a temporal increase in resistance for most heavily used agents with the notable exception of trimethoprim, a drug commonly administered as a fixed combination with sulphamethoxazole. During this period, resistance to trimethoprim, remained at a stable high of 35–45% [25]. In this study, deliberately designed to replicate the 1986 to 1998 protocols, stool specimens were collected from 83 and 101 consenting adults in 2005 and 2009 respectively [1]. Susceptibility testing of the 304 E. coli isolates obtained revealed that resistance to most agents commonly used in Nigeria was high (Table S2). As shown in Figure 2, trimethoprim resistance has become much more prevalent than in previous decades, showing an 80% increase in 2005 on 1998 data, with the number and percentage of trimethoprim-resistant strains increasing again, from 86 (67.2%) to 122 (70.8%), between 2005 and 2009.

Figure 2

Trimethoprim resistance trends among E. coli isolates from healthy adults in Ile-Ife Nigeria 1988–2009.

We used primers that anneal to the conserved ends of class 1 and class 2 integrons [31], [32] to amplify the interim cassette regions of all 128 E. coli isolates collected in Ile-Ife, Nigeria in 2005 and 176 isolates from 2009. These primers identify strains carrying complete and intact integrons but exclude those that may be truncated at the 3′ end or otherwise modified [42]. We identified specific cassettes by sequencing and restriction fragment length polymorphism analyses. Based on data from similar studies elsewhere in the world, we hypothesized that dfrA genes would be detected and that dfrA1, the most commonly reported allele worldwide [5], [10], [16], [18], [43], [44] would feature predominantly. We did indeed identify only dfrA alleles, including several strains harboring dfrA1. There were only two varieties of class 2 integron cassettes, both carrying dfrA1 cassettes. The more common variant was dfrA1-sat-aadA, known to be globally disseminated via transposon Tn7 [16], [45]. However, in contrast to recent studies performed outside Africa [46], [47], dfrA1 was not the most common allele in class 1 integrons. There was low diversity overall and 67(56.8%) of isolates with a class 1 integron, carried a single dfrA7 cassette identical to that in Tn5086, originally isolated from an E. coli strain in Colombo, Sri Lanka [48].

Class 1 Integron-borne dfrA7 Cassettes Occur Widely Among E. coli from Sub-Saharan Africa

We examined the literature for studies that sought class 1 integron cassettes from intestinal E. coli. dfrA7 alleles in the class 1 integron single-cassette conformation have been reported previously from E. coli isolates, but not commonly. Of the over 100 studies that characterized integron cassette contents, most found dfrA7 relatively uncommon, if geographically widespread. The cassette was identified at low levels in community-acquired E. coli in the following countries: Finland (0–2%) [45], [49], Korea (4%) [50], Bolivia (2.6%) [47] Spain (3.6%) [46], Taiwan (3.8%) [51], Greece (0.9%) [52], the USA (4.4%) [17], Lebanon (2%) [53], Australia (1/20) [54] as well as Sweden and Sri Lanka (<7%) [16], [48]. dfrA7 is also more commonly reported from Salmonella [55], [56]). By contrast, the few, generally small studies from Africa, reported dfrA7 from E. coli in much larger proportions: 18.7% in Nigeria [57], 33.3% in Senegal [58], 38% in South Africa [59], and more recently, 49% in the Central African Republic [60]and 40% in Tunisia [61]. Descriptive genetic studies of E. coli isolates from Nigeria have invariably found dfrA7 [62], [63]. In other recent studies across Africa where the specific dfr alleles have been uncharacterized, rates of resistance to trimethoprim are high [22], [64]. These data suggest that dfrA7 may be disseminated sub-regionally.

To test the hypothesis that class 1 integrons bearing dfrA7 cassettes were also highly prevalent among E. coli outside our Ile-Ife, Nigeria study area, we screened 291 isolates collected in Accra, Ghana between 2006 and 2008 [20]. As shown in Table 1, a few cassette combinations were identified repeatedly at one geographic location but not at the other. Future studies could reveal that these combinations represent locally disseminated elements or clones. They included dfrA5 and dfrA15 from Nigeria and dfr2d in Ghana. However, eight cassette arrays were seen at both sites. In addition to dfrA1, these include dfrA15-aadA1 (reported from Vibrio spp in Africa [65], [66]) as well as, dfrA17-aadA5, which is associated with uropathogenic clonal group A strains known to be present in Nigeria [24], [67]. However, all of other cassette combinations detected at both sites, combined, were less common than dfrA7, which was identified in 54 (52.4%) class 1 integron-bearing isolates.

Table 1

Integron cassettes in E. coli from Nigeria and Ghana between 2005 and 2009.

	Nigeria 2005 (n = 128)	Nigeria 2009(n = 176)	Ghana 2006 (n = 130)	Ghana 2007 (n = 73)	Ghana 2008 (n = 88)
Integrons in E. coli isolates (%)
Single class 1 integron only	37 (28.9)	70 (39.8)	35 (26.9)	23 (31.5)	45 (51.1)
Multiple class 1 integrons	–	4 (2.3)	–	1 (1.3)	–
Class 2 integron only	8 (6.3)	14 (8.1)	6 (6.6)	1 (1.3)	2 (22.7)
Class 2 integron and single class 1 integron	1 (0.8)	13 (7.6)	5 (3.8)	2 (2.7)	9 (10.2)
Class 2 integron and multiple class 1 integron	-	1 (0.6)	–	–	–
Gene Cassettes in Class 1 Integrons in isolates bearing one integron (%)	(n = 37)	(n = 74)	(n = 35)	(n = 23)	(n = 45)
dfrA7	22 (59.5%)	45 (60.8%)	19 (54.3%)	11 (4.8%)	24 (53.3%)
dfrA1 aadA1	4 (10.8%)	11 (14.9%)	3 (8.6%)	5 (21.7%)	12 (26.7%)
dfrA5	3 (8.1%)	6 (8.1%)	–	–	–
dfrA15 aadA1	3 (8.1%)	–	9 (25.7%)	1 (4.3%)	–
dfrA17 aadA5	2 (5.4%)	2 (2.7%)	3 (8.6%)	2 (8.7%)	3 (6.7%)
aadA5	1 (2.7%)	–	–	–	–
dfrA1	1 (2.7%)	–	–	–	–
dfrA2d	–	–	1 (2.9%)	4 (17.4%)	5 (11.1%)
aadA1	–	1 (1.4%)	2 (5.7%)	3 (13%)	–
blaOXA-1 aadA1	–	3 (4.1%)	1 (2.9%)	–	4 (8.9%)
dfrA17 aadA4	–	–	–	–	2 (4.4%)
dfrA15		5 (6.8%)
aadA2		1 (1.4%)
dfrA12 orfF aadA2		1 (1.4%)
Gene Cassettes in Class 2 Integrons (all)	(n = 9)	(n = 27)	(n = 11)	(n = 3)	(n = 11)
dfrA1, sat1	1 (11.1%)	4 (14.8%)	1 (9.1%)	-	3 (27.2%)
dfrA1, sat1, aadA1	8 (88.9%)	23 (85.2%)	10 (82.6%)	3 (100%)	8 (72.7%)

Independent Isolates from Nigeria and Ghana Bearing dfrA7 Cassettes are not Clonal

Clonal expansion has accounted for dissemination of some dfrA alleles elsewhere on the globe and we hypothesized that it could account for dfrA7 spread in Nigeria and Ghana [24], [68], [69], [70]. When we examined the genetic background of 12 strains by multi-locus sequence typing (MLST) by the scheme of Wirth et al [39], we identified nine different sequence types (Table S3). Similarly, flagellin typing [38] of 35 other dfrA7-positive isolates identified at least 15 different flagellin types among 23 typeable strains (Figure S1), confirming that the genomic background of isolates bearing the dfrA7 gene is diverse.

Isolate 05/01a from Nigeria Bears dfrA7 on a Plasmid Encoding Resistance to Multiple Antimicrobials

We next sought to determine whether the dfrA7 cassette could be transferred to a nalidixic acid-resistant derivative of E. coli K-12 strain C600 by conjugation. We employed seven dfrA7-bearing strains as conjugative donors, five from Nigeria and two from Ghana. All were resistant to ampicillin, streptomycin, sulphonamides, tetracycline and trimethoprim, five strains were additionally resistant to chloramphenicol and five showed low-level nalidixic acid resistance. All the isolates tested were ciprofloxacin-sensitive. As shown in Table 2, the dfrA7 gene could be conjugated from only three of these isolates. Therefore, although we are able to conclude that the dfrA7 cassette can be conjugated in vitro in some instances, it is present in a non-conjugable form in other strains.

Table 2

Results of conjugation experiments using dfrA7-positive isolates as donors.

Strain	Source (year)	Cassette(s)	Donor Resistance Profile	Trans-conjugant Resistance Profile
dfrA7-positive strains
05/01a	Nigeria (2005)	dfrA7	A (N)SLTR	None*
05/9c	Nigeria (2005)	dfrA7	AC (N)SLTR	None*
05/23a	Nigeria (2005)	dfrA7	AC (N)SLTR	AC NSLTR
05/25a	Nigeria (2005)	dfrA7	AC (N)SLTR	None*
05/33a	Nigeria (2005)	dfrA7	AC (N)SLTR	AC NSLTR
046	Ghana (2006)	dfrA7	A SLTR	A N LTR
116	Ghana (2006)	dfrA7	AC SLTR	AC NSLTR
Strains bearing cassettes other than dfrA7
115	Ghana (2006)	aadA1 dfrA15	AC L R	AC N LTR
05/27a	Nigeria (2005)	dfrA17 aadA5	AC SLTR	AC NLTR
Control strains
R100.1	[92]	aadA1	C SLT	C SLT
SM10	[93]	ND	K	None*

Indicates that no transconjugants were isolated;

Antibiotic code key: A, ampicillin; C, chloramphenicol; N, nalidixic acid (high level); (N), Nalidixic acid (lower level); S, streptomycin; L, sulfonamides; T, tetracycline; R, trimethoprim; K, Kanamycin.

As shown in Figure 2, dfrA7-positive strain 05/01a from Nigeria harbors at least four plasmids. This strain is unable to conjugate dfrA7 in vitro (Table 2), but could serve as a transformative donor of the gene (Table 3). Strain 05/01a is susceptible to ciprofloxacin and chloramphenicol but resistant to the other six tested antimicrobials, including low-level resistance to nalidixic acid. Electroporation of a Kado and Liu [28] plasmid extract into DH5α-E cells (Invitrogen) yielded transformants on trimethoprim plates that are co-resistant to ampicilin, streptomycin and sulphonamides but not to nalidixic acid or tetracycline (Table 3). While having different plasmid profiles, each of these transformants carried a single, large, low-copy number plasmid on which the dfrA7 gene is located (Figure 3). As shown in Table 3, the plasmid in question could not be incompatibility-typed with the primers of Carattoli et al [71], which represent the most common plasmid incompatibility groups.

Table 3

Plasmid replicons detected in select dfrA7-bearing strains isolated from Nigeria in 2005.

Strain	Resistance profile	Integron-borne trimethoprim resistance gene	Plasmid replicons detected by PCR
05/01a	A (N)SLTR	dfrA7	T, Y
05/09c	AC (N)SLTR	dfrA7	T, Y
05/23a	AC (N)SLTR	dfrA7	FIC, T, Y
05/30a	AC (N)SLTR	dfrA7	FIB, T, Y
05/31a	AC (N)SLTR	dfrA7	–
05/32c	AC (N)SLTR	dfrA7	B/O, FIB, Y
05/33a	AC (N)SLTR	dfrA7	B/O, T
DH5αE (pASL01a)	A SLR	dfrA7	–
DH5αE	–	None	–

Antibiotic code key: A, ampicillin; C, chloramphenicol; (N), Nalidixic acid (low level); S, streptomycin; L, sulfonamides; T, tetracycline; R, trimethoprim.

Figure 3

Plasmid profiles of E. coli strain 05-01a and its plasmid transformants.

Lane 2: Plasmid-free DH5α; Lane 3: 05-01a, dfrA7-positive 2005 isolate from Nigeria; Lanes 4–7: four independent transformants produced by electroporating DH5α with a plasmid preparation from 05-01a and selecting on plates containing 50 µg/ml trimethoprim. The arrow indicates the position of chromosomal DNA, as inferred from the DH5α profile. Lane 1: Hyperladder 1 marker.

We sequenced the dfrA7-bearing plasmid from isolate 05/01a, and designated it pASL01a. The complete sequence of this plasmid has been deposited in Genbank (Accession number JQ480155). As shown in Figure 4, pASL01a is a 27,072 bp plasmid largely comprised of a Tn21-family transposable element. pASL01a contains 29 predicted open reading frames and has an overall G+C content of 56.67%. The relatively low G+C content 5,168 bp plasmid backbone contains only three open reading frames: the replicon, a putative mob gene and an orf of unknown function. Although a precise origin of replication could not be identified, a sharp change in GC skew just upstream of repAAS01a indicates the location of the likely origin. Analysis using BLAST revealed that the backbone region is most similar to several unpublished or recently published sequences of small ColE-related plasmids. The best matches (98% or greater identical at the DNA level) include the 5,146 bp E. coli plasmid pIGJC156 (gb|EU090225.1|), the 7,462 bp E. coli plasmid pMG828-4 (gb|DQ995354.1|) and multiple plasmids from recently genome-sequenced pathogens [72], [73]. Plasmid pIGJC156 is 97% identical to the backbone of pASL01a. The only dissimilar region between pIGJC156 and the pASL01a backbone is the replicon. The replicon of plasmid pMG828-4 is also not significantly similar to that of pASL01a and 1 kb of downstream pASL01a sequence is replaced by a putative NADH oxidase gene not present in pASL01a. The replicons of the pathogen plasmids are over 98% identical to that of pASL01a [72], [73]. None of these plasmids carry any virulence genes and all are present in strains that carry other, larger plasmids.

Figure 4

Circular map of 27,072bp pASL01a.

Circles display (from the inside) (a) GC skew ([G + C]/[G - C]) in a 1,000-bp window (b) GC content in a 1,000-bp window (purple representing above average and yellow below average in both cases) (c) plasmid backbone in beige (d) predicted coding sequences and (e) kilobase scale. Open reading frames predicted to encode drug resistance genes are colored red and those that encode resistance to other chemical entities are gold. All other predicted open reading frames are marked in blue. Dark gray lines on the open-reading frame and backbone tracks mark the position of inverted repeats.

pASL01a is principally comprised of a transposon with the basic transposition module and mercury resistance genes of Tn21. With the exception of the In2-associated tni module, which is truncated in pASL01a, all the transposition components of Tn21 are present and conserved with the transposon, suggesting that the element may be mobile. Between these modules, as in Tn21, a class 1 integron(carrying dfrA7, sulI and qacEΔ1 genes) is inserted. Between the In2 and tni modules of the Tn21-like element is inserted an IS26-derived resistance element, bearing widely disseminated genes encoding resistance to beta lactams (blaTEM), sulphonamides (sul2) and aminoglycosides (strAB). The gene content and order of this module is identical to that of plasmid pAKU_1 and pHCM1 from Salmonella enterica Paratyphi and Typhi respectively [55], [74]. This 9 Kb bla/sul/str module has also been encountered on a number of other enterobacterial multidrug resistance plasmids and our data add to the evidence that suggests that the element is highly promiscuous [74], [75], [76] and, as Holt et al [74] speculate, moves as a block. A separate study from Australia proposes that the IS26-flanked sul/str portion of this transposon, originally described in plasmid RSF1010, may itself be a separable module [77]. The locations of repeats and alterations in G+C content depicted in Figure 4, suggest that any or all of these possibilities may account for our encountering this resistance transposon on a plasmid of unusual backbone from Nigeria. The transposon described here lacks a Tn9-derived portion encoding chloramphenicol resistance often seen in resistance plasmids bearing Tn21-derived elements and could share an ancestor with those that have mediated rapid acquisition of multiple resistance genes in Salmonella [36], [74], [78]. The dfrA7-bearing transposon from pASL01a is also in many ways similar to transposon Tn5086, a Tn21-family transposon restriction mapped in 1993 [48]. Tn5086 possesses a Tn21 backbone, with dfrA7 and mercury genes included, but lacks the bla/sul/str element.

Close examination of the predicted DfrA protein encoded by pASL01a reveals that the gene, and its flanking sequence, is identical to that of Tn5086 [48]. Sundström et al [48] demonstrated experimentally that the start codon was a TTG, and that the upstream ATG codons that are annotated as starts for many dfrA7 and dfrA17 alleles in the Genbank database were in fact not required for translation and overlapped the ribosomal binding site. As there are some dfrA7 alleles in the database that have an ATG codon at this locus, we resequenced the dfrA7 gene alone, obtaining 6X sequence coverage from three independent pASL01a subclones, and confirming the TTG codon. TTG start codons are uncommon in E. coli and it is therefore unusual that a resistance gene that begins with this codon has been so widely and successfully disseminated. A variant dfrA7 allele, this time with polymorphisms at the C-terminal end, has been previously reported from Nigeria [63]. The existence of multiple dfrA7 alleles in this population is likely linked to the high frequency of the gene. However, conservation of the functional regions strongly suggests that the gene remains under selection, consistent with intensive local trimethoprim use [1].

pASL01a is Only One of Multiple Vectors for TnASL01a

We used PCR to screen the isolates from Nigeria for repC, a gene that lies within the predicted transposon as well as repASL01a and the putative mob locus on pASL01a, which are part of the plasmid backbone. Using primers Lev5’CS (located immediately upstream of dfrA7) and JSMrepCR (within the transposon-borne repC gene), we verified by long PCR that the distance between these two genes was 6 Kb for all 22 2005 isolates from Nigeria that carry the dfrA7 and the repC genes as well as for the three dfrA7-positive isolates from Ghana that were employed in conjugation experiments.

Using primers for intI1 (TnASL01a class 1 integron gene), repC from TnASl01a and the region of pASL01a encompassing the plasmid origin of replication and putative mob genes (Table S1), we screened the 2005 and 2009 isolates for different regions of pASL01a. As shown in Table 4, only 12 (18.5%) of 65 dfrA7-positive isolates from Nigeria possessed the pASL01a rep-mob backbone region together with repC and intI1. However, 36 (55.4%) strains were positive for repC and intI1 but negative for the pASL01a backbone markers. Similarly, we detected repC in 37 (64.9%) dfrA7-positive strains from Ghana but only four (7.0%) dfrA7-positive strains also had the pASL01a backbone genes. Interestingly, the pASL01a plasmid backbone was significantly less common in dfrA7-positive isolates from Ghana, compared to isolates from Nigeria, where it was originally identified (p = 0.038, Yates-corrected Chi-squared test). However the repC was identified at statistically similar levels at both sites (p = 0.9).

Table 4

Number and percentage of isolates bearing TnASL01a-associated sequences intI1 and repC, and pASL01a backbone associated sequences rep and mob.

Country, Year of isolation	Class 1 integron cassettes	repC, intI1 (without repASL01a, mob)	repC, intI1andrepASL01a, mob*	Neither repC norrepASL01a, mob
Nigeria, 2005	dfrA7	14 (63.6)	7 (31.8)	1 (4.5)
	Other class 1 integron cassettes	5 (33.3)	2 (13.3)	8 (53.3)
	No class 1 integron	8 (9.8)	0 (0)	73 (89.0)
Nigeria, 2009	dfrA7	32 (71.1)	5 (11.1)	8 (17.8)
	Other class 1 integron cassettes	19 (65.5)	5 (17.2)	9 (31.0)
	No class 1 integron	30 (42.9)	1 (1.42)	39 (55.7)
Ghana, 2006	dfrA7	15 (78.9)	2 (10.5)	0 (0)
	Other class 1 integron cassettes	3 (18.8)	0 (0)	13 (81.3)
	No class 1 integron	19 (22.6)	0 (0)	65 (77.4)
Ghana, 2007	dfrA7	9 (81.8)	1 (9.1)	1 (9.1)
	Other class 1 integron cassettes	5 (41.6)	0 (0)	7 (58.3)
	No class 1 integron	8 (17.0)	0 (0)	38 (80.9)
Ghana, 2008	dfrA7	13 (54.2)	1 (4.2)	10 (41.7)
	Other class 1 integron cassettes	13 61.9)	0 (0)	8 (38.1)
	No class 1 integron	11 (25.6)	1 (2.3)	31 (72.1)

The data demonstrate that while presence of dfrA7 is strongly associated with TnASL01a, there are other class 1 integron-borne genes that are likely to be disseminated by Tn21-like transposons that have the repC gene (Table 4). These include aadA1, dfrA1-aadA1, blaOXA and dfrA5, which have previously been reported in the literature and showed similar strong associations in this study (Table 4) [76], [78], [79]. These findings emphasize that transposon-borne integrons are flexible platforms that can evolve to carry different genes, depending on cassette availability, integrase activation and selective pressure. Such a model could account for the origin of the dfrA7-bearing transposon we report here and suggests that other cassettes may be incorporated when new antibacterials are introduced.

Plasmids bearing dfrA7 genes in the context we describe have been reported from Salmonella enterica Typhi and Paratyphi serovars [36], [74]. Although integrons are more commonly sought in E. coli reports of class 1-integron-borne dfrA7 alleles are comparatively less common. In Salmonella, dfrA7-bearing transposons are typically found on large IncH1 plasmids, which have spread through Asia [55], [74]. However we did not identify IncH plasmids in this study and the plasmid isolated from strain 05/01a was relatively small, carrying only the resistance transposon, and a small 5 Kb core. Resistance genes can gain success by hitchhiking on successful self-transmissible elements or entering epidemic bacterial clones, both mechanisms of which have been the focus of most published studies. Resistance genes, such as the dfrA7 allele we have mapped in this study, can equally achieve prominence by trans-mobilization, either because they are carried on elements that can be mobilized [80] or because they can transpose between mobile and non-mobile elements [81].

Conclusions

Our work demonstrates that a small repertoire of widely disseminated mobile cassettes account for most of the trimethoprim resistance among E. coli isolates in Nigeria and Ghana. Many dfrA alleles associated with integrons, transmissible plasmids, or both, are globally disseminated. For example, detection of dfrA1-containing integrons in our study is not surprising and indicative of a globally disseminated resistance pool, from which organisms in West Africa are not isolated [82], [83]. We identified a dfrA7 allele that is extraordinarily over-represented in Nigeria, Ghana and possibly elsewhere in sub-Saharan Africa. The allele lies within a class 1 integron borne on a potentially mobilizable Tn21-like transposon.

It is not clear when the dfrA7–bearing transposon entered the E. coli commensal population in Nigeria because strains isolated prior to 2005 were unfortunately not archived. dfrA7 was reported as early as 1995 from South Africa when Adrian et al [59] described it as the most common dfr allele. Interestingly, they were unable to transfer the dfrA7 marker in that study by conjugation and southern hybridization demonstrated that the gene was, in most cases, located on the chromosome. Lamikanra and coworkers [57], [84], suggest that the predominant dfr allele in the 1980s and early 1990s was a class 2 integron borne dfrA1, which we still detect significantly in both Nigeria and Ghana. dfrA7 was detected in those earlier studies, but less commonly, and was transferable by conjugation [57], [84]. Thus it is possible that the transposon we describe has been in circulation in parts of Africa for some time. However it is tempting to view its introduction, or selective success, as a recent event because we have seen trimethoprim resistance increase rapidly after 2005.

Minimal data on resistance elements from Africa are available because few molecular studies are performed there. Kingsley et al (2009) report that strains belonging to the recently evolved invasive non-typhoidal Salmonella lineage isolated in Malawi and Kenya carry resistance genes on Tn21-type transposons very similar to that identified in this study [78]. The resistance genes on those transposons largely account for the treatment failure associated with these infections in recent years. Our study opens questions as to how antimicrobial use patterns, or other risk factors, may have provided enormous selective advantage for multi-drug resistance plasmids. Much of Africa has seen unprecedented selective pressure from trimethoprim and sulphamethoxazole in recent times and Skurnik et al (2009) propose that integron-associated resistance is linked to high levels of selective pressure [85]. Since resistance rates remained stable until 1998, we hypothesize that a recent change in antimicrobial use patterns would most likely account for the spread of pASL01a and consequent upsurge in trimethoprim resistance.

Rapid spread of HIV has been coupled with severe resource constraints for managing infected people in Africa. In 1999, the results of a clinical trial revealed that trimethoprim-sulphamethoxazole prophylaxis decreased morbidity and mortality from opportunistic infections in patients in Abidjan, Ivory Coast [86], [87]. These findings resulted in trimethoprim-sulphamethoxazole prophylaxis policies that have been applied almost continent-wide. Much-cited ‘cost effectiveness’ of long-term trimethoprim-sulphamethoxazole prophylaxis for AIDS patients, who, but for slow roll outs, could be on more effective but more expensive antiretrovirals [4], does not factor in the losses from declining trimethoprim and sulphonamide susceptibility due to resistance. In parallel with prophylactic use, this and other recent studies from Africa have reported trimethoprim-resistance rates in clinical isolates of E. coli and related organisms that exceed 70% [88], [89], [90]. Studies that specifically examined carriage of resistant E. coli and Streptococcus pneumoniae among individuals on trimethoprim-sulphamethoxazole prophylaxis, in comparison with those not receiving these drugs, document a significantly greater level of resistance among patients receiving prophylaxis [88], [90], [91]. Trimethoprim- sulphamethoxazole prophylaxis did not select for trimethoprim and sulphonamide resistance alone, but resistance to multiple antimicrobials. In E. coli, this is consistent with a role for multiply-resistant elements, such as pASL01a, and the transposon it carries.

In Finland, the rates of trimethoprim resistance among urinary E. coli isolates increased from about 10% to 40% between 1978 and 1984, at a time when trimethoprim was used intensively for empiric treatment of urinary tract infections [45]. This rapid increase in resistance was largely due to epidemic dissemination of mobile elements bearing dfrA1 genes in a scenario similar to dfrA7-mediated resistance in West Africa today. When trimethoprim was withdrawn, resistance rates not only failed to decline, they continued to rise because the resistance gene in question was borne on a transposon that could stably integrate into the chromosomes as well as into plasmids [45]. Ultimately, dfrA1-bearing transposons were disseminated globally. Because we found that trimethoprim resistance is linked to resistance to other commonly used antimicrobials (penicillins, sulphonamides and streptomycin), we cannot directly implicate trimethoprim use alone in the selection of the transposon identified in this study. In the same vein, we predict that, as was demonstrated in the Finland case and experimentally tested in Kronberg Country Sweden, withdrawal of trimethoprim-sulphamethoxazole is unlikely to be sufficient to produce a return to susceptibility [16], [45]. Our study reveals that gene context is an important determinant of evolutionary success. Since the structure of the transposon demonstrates flexible incorporation of resistance genes, it is essential to control selective pressure to newer agents to prevent their resistance being incorporated into this or other successful platforms.

I-based PCR-RFLP of amplicons from fourteen independent -bearing isolates demonstrating eleven unique restriction profiles. Lanes 1, 9 and 18: I Kb ladder plus (Invitrogen).

(PPTX)

Click here for additional data file.

Oligonucleotide primers for PCR.

(DOC)

Click here for additional data file.

Resistance to antibiotics 2005 and 2009.

(DOC)

Click here for additional data file.

Allele profiles of bearing strains that were multilocus sequenced typed.

(DOC)

Click here for additional data file.