SGI2, a Relative of Salmonella Genomic Island SGI1 with an Independent Origin

Abstract

Multiply antibiotic-resistant Salmonella enterica serovar Emek strains isolated in Australia and the United Kingdom had similar features, suggesting that they all belong to a single clone. These strains all contain SGI2 (formerly SGI1-J), an independently formed relative of Salmonella genomic island SGI1. In SGI2, the complex class 1 integron which includes all of the resistance genes is not located between tnpR (S027) and S044 as in SGI1 and SGI1 variants. Instead, tnpR was found to be adjacent to S044, and the integron is located 6.9 kb away, within S023. In both SGI1 and SGI2, the 25-bp inverted repeats that mark the outer ends of class 1 integrons are flanked by a 5-bp duplication of the target, indicating that incorporation of the integron was by transposition. A small number of differences between the sequences of the backbones of SGI1 and SGI2 were also found. Hence, a class 1 integron has entered two different variants of the SGI backbone to generate two distinct lineages. Despite this, the integron in SGI2 has a complex structure that is very similar to that of In104 in SGI1. Differences are in the cassette arrays and in the gene which encodes the chloramphenicol and florfenicol efflux protein. The CmlA9 protein, encoded by InEmek, is only 92.8% identical to FloRc (also a CmlA family protein) from SGI1. A variant form of SGI2, SGI2-A, which has lost the tet(G) and cmlA9 resistance determinants, was found in one strain.

The Salmonella genomic island (SGI) known as SGI1 was first identified in multiply antibiotic-resistant strains of Salmonella enterica serovar Typhimurium DT104 (5), which is currently the most common type of multiply antibiotic-resistant serovar Typhimurium strain found in most countries (27). SGI1 is a 42.4-kb mobilizable integrated element (Fig. 1) that contains five antibiotic resistance genes (4, 13). The resistance genes are all located within the boundaries of a complex class 1 integron, designated In104 (22), which belongs to the In4 family (32). In104 has two attI1 sites, with an aadA2 gene cassette (conferring resistance to streptomycin and spectinomycin) in the left attI1 site and a blaP1 gene cassette (conferring resistance to beta-lactam antibiotics) in the right attI1 site. Though these cassettes are both flanked by sequence from the 5′-conserved segment (5′-CS) and 3′-CS, only the 5′-CS on the left is complete and a complete sul1 sulfonamide resistance gene, which is normally part of the 3′-CS, is found only on the right. The two class 1 integron-derived regions are separated by a segment (boxed in Fig. 1) that includes a chloramphenicol and florfenicol resistance determinant, which is a member of the cmlA gene family (38) but known as floR, and the tetRA(G) tetracycline resistance determinant (4, 7, 27). A copy of the small mobile element known as CR3 (31) is also found in this segment. SGI1 can be mobilized by an IncC plasmid (13), and consistent with such horizontal gene transfer, SGI1 or variants of it have also been found in several other S. enterica serovars, including serovars Agona, Albany, Cerro, Derby, Dusseldorf, Infantis, Kentucky, Kiambu, Meleagredis, Newport, Paratyphi B dT⁺, and Tallahassee (2, 4, 15-17, 22-24, 27, 40). Recently, SGI1 variants have also been found in Proteus mirabilis (1, 6).

FIG. 1.

Integrons in SGI1 and SGI2. (A) SGI1 is a 42.4-kb mobilizable integrated element. The map of SGI1 was generated from the sequence under GenBank accession no. AF261825. (B) Complete sequence of the integron in SGI2 and its immediate surrounds (this study) (GenBank accession no. AY963803). The surrounding SGI1 backbone and the adjacent chromosomal region are shown below the integron. Narrow vertical bars represent the inverted repeats (IRi and IRt) of class 1 integrons, and the 5′-CS, 3′-CS, and tni regions are indicated by lines of different thicknesses. The attI1 site is shown as a tall open box, and gene cassettes are shown as open boxes with a black bar at one end, indicating the attC sites (59-be). IS6100 and the central, non-integron-derived region are shown as open boxes. Arrows indicate the position and orientation of genes and open reading frames, which are named or numbered.

SGI1 is usually found in the Salmonella chromosome in one end of the thdF gene and supplies a replacement for the last bases of that gene, which thus remains functional (27). In Proteus mirabilis, SGI1 is also found in the thdF gene (6). This site-specific integration is mediated by an integrase (tyrosine recombinase family) encoded by the int gene at the left end of the SGI (13). In all Salmonella serovars except for serovar Typhimurium, the yidY gene lies on the other side of the SGI (Fig. 1). In Salmonella serovar Typhimurium strains, 99 bp of sequence found between the end of the thdF gene and yidY in the other serovars (23) is replaced by a second integrated element (called a “retron phage”) (4, 5). The In104 integron is found near the right end of the SGI backbone, between the tnpR (S027) and S044 genes, and is flanked by a 5-bp duplication of the target sequence, indicating that it took up this position by transposition. Since tnpR encodes a resolvase (7), the integron is likely to be in or near a resolution (res) site, and this is the observed location for most class 1 integrons (20, 28), which are members of a family of transposons that have been shown to target res sites (20, 25).

Variants of SGI1, SGI1-A to SGI1-O, containing different sets of resistance genes, have also been found (Table 1). Most of them appear to have simply gained, lost, or exchanged resistance genes. Some variants have lost the central region of In104 and hence include only one attI1 site containing a gene cassette(s), e.g., SGI1-B and -C, and others have exchanged the gene cassettes at one of the attI1 sites in SGI1, replacing them with the aadB or dfrA15 gene cassette or with the dfrA1-orfC or aacCA5-aadA7 pairs. In one case, SGI1-K, the central region is missing and the cassettes in the remaining attI1 site have been exchanged. Further variants have gained an additional partial copy of the 3′-CS together with the dfrA10 trimethoprim resistance gene, which is associated with the small mobile element known as CR1 (formerly called orf513). CR1 and CR3 are related elements that draw a variety of resistance genes into the complex class 1 integrons that contain them (31). All of these deletion, insertion, and exchange events could have occurred by homologous recombination (3, 15, 22, 23), as the wild-type strains are usually recombination proficient and such events have been shown to occur (28, 30, 31).

TABLE 1.

SGI1 variants

Variant	Cassette		Presence of CR1-dfrA10	Resistance phenotypeb	Reference
	Left attI1 site	Right attI1 sitea
SGI1	aadA2	blaP	−	Ap Cm Fl Sm Sp Su Tc	7
SGI1-A	aadA2	blaP	+	Ap Cm Fl Sm Sp Su Tc Tp	3
SGI1-B	blaP	—	−	Ap Su	3
SGI1-C	aadA2	—	−	Sm Sp Su	3,9
SGI1-D	aadA2	—	+	Sm Sp Su Tp	3
SGI1-Ec	aadA2	blaP	−	Ap Sm Sp Su Tc	3
SGI1-F	dfrA1-orfC	blaP	−	Ap Cm Fl Su Tc Tp	15
SGI1-G	blaP	—	+	Ap Su Tp	14
SGI1-H	aacCA5-aadA7	blaP	−	Ap Cm Fl Gm Sm Sp Su Tc	16
SGI1-I	aadA2	dfrA1-orfC	−	Cm Fl Sm Sp Su Tc Tp	22
SGI1-Jd	dfrA1-orfC	Δ	−	Cm Fl Su Tc Tp	22
SGI1-Ke	aacCA5-aadA7	—	−	Gm Sm Sp Su	23
SGI1-L	dfrA15	blaP	−	Ap Cm Fl Su Tc Tp	1,12
SGI1-M	aadB	blaP	−	Ap Cm Fl Gm Su Tc Tm	40
SGI1-O	dfrA1-orfC	—	−	Su Tp	6
SGI1-AΔ1Rf	aadA2	blaP	−	Ap Cm Fl Sm Sp Su Tc	14
SGI1-AΔ2Rf	aadA2	blaP	−	Ap Cm Fl Sm Sp Su Tc	14
SGI1-AΔ3Rf	aadA2	blaP	−	Ap Cm Fl Sm Sp Su Tc	14
SGI1Δg	aadA2	blaP	−	Ap Cm Fl Sm Sp Su Tc	34

^a—, only one attI1 site is present, and hence the floR (Cm^r Fl^r) and tet(G) (Tc^r) genes are missing; Δ, a 5′-CS/3′-CS fusion identical to that in Tn610. A deletion has removed part of the attI site and part of the 3′-CS.

^bAp, ampicillin; Cm, chloramphenicol; Fl, florfenicol; Gm, gentamicin; Sm, streptomycin; Sp, spectinomycin; Su, sulfonamides; Tc, tetracycline; Tm, tobramicin; Tp, trimethoprim.

^cA rearrangement has occurred. See reference for details.

^dSGI1-J was renamed SGI2 here.

^eContains an internal IS6100-mediated deletion of the 3′-CS, and S044 is not found adjacent to the outer IRt of the integron. See the reference for details.

^fDerivatives of SGI1-A where deletions have removed all SGI1 sequences to the right of CR1 and extending various distances into the adjacent chromosome.

^gAn IS6100-mediated deletion has removed all SGI1 sequences to the right and extending into the yieE gene in the chromosome.

Because the resistance phenotype changes as the resistance genes carried by SGI1 vary (Table 1), the most reliable method for detection of SGI1-containing strains is to use PCR to detect the linkages between SGI1 and the flanking chromosomal thdF and yidY genes and between the SGI1 backbone and the integron (see Fig. 1 for positions of these boundaries) (3). Generally, all four PCR products are obtained (22). However, further variations that have lost parts of SGI1 have also been detected because one or more of these reactions does not yield a product. In a few cases, the chromosome-SGI1 and SGI1-In104 boundaries on the left are present, but those on the right are not. One type of variant has lost all of the sequence to the right of the small mobile element CR1 in the CR1-containing SGI1-A variant, with the deletion extending into the adjacent chromosomal genes (14), and hence appears to have been formed by the activity of CR1. Deletion of the right end of In104 and SGI1, caused by IS6100 in the integron, has also been reported (34). In the SGI1-K variant, only the right junction of In104 with the SGI1 backbone (S044) cannot be detected (23), and this appears to be due to the incorporation of a large DNA segment that includes a mercury resistance module and other, as yet unidentified sequence.

We have previously described a further unusual SGI type, called SGI1-J (22). This SGI was found in a Salmonella enterica serovar Emek strain isolated from sewage effluent. In SGI1-J, though a class 1 integron was present, neither of the PCR products specific for the boundaries of the integron with the SGI1 backbone were detected. The integrase-encoding end (IRi) of the class 1 integron was localized to a different part of the SGI1 backbone, within open reading frame S023 (22). This suggested that a large deletion may have occurred at that end, but an alteration at the other end would also be needed to explain the findings. Here we have further analyzed the structure and location of the integron in the original serovar Emek strain in order to examine its origin. Additional serovar Emek strains isolated in Australia and the United Kingdom from recently returned travelers were also shown to carry an SGI that has the same features.

MATERIALS AND METHODS

Bacterial strains.

The properties of the Salmonella strains used in this study are listed in Table 2. SRC19 (serovar Emek) and the Salmonella serovar Paratyphi B dT⁺ strain SRC49 and serovar Typhimurium strain DT104-2 used as control SGI1-containing strains were described previously (22). SRC239, kindly supplied by Luke Randall, was isolated and partially characterized at the Laboratory of Enteric Pathogens, Health Protection Agency, London, United Kingdom, and further characterized in a published study (35). The remaining strains were isolated and initially characterized in Australia and were supplied by Diane Lightfoot (Melbourne Diagnostic Unit, Melbourne University, Victoria, Australia). Resistance profiles were determined as described previously (22) and were confirmed and extended according to the CDS method (calibrated dichotomous sensitivity test [http://web.med.unsw.edu.au/cdstest]), using antibiotic discs (Oxoid, New Hampshire, England) containing ampicillin (25 μg), chloramphenicol (30 μg), florfenicol (30 μg), gentamicin (10 μg), kanamycin (50 μg), neomycin (30 μg), streptomycin (25 μg), spectinomycin (25 μg), sulfafurazole (300 μg), tetracycline (30 μg), tobramicin (10 μg), trimethoprim (5 μg), nalidixic acid (30 μg), and ciprofloxacin (5 μg). Pulsed-field gel electrophoresis of XbaI-digested whole-cell DNA and IS200 analysis were carried out as described previously (21).

TABLE 2.

Salmonella enterica strains used for this study

Strain	Serovar	Origin (year of isolation)	Source	Travel history	Resistance phenotypea	Resistance genesb
DT104-2	Typhimurium	Australia (2001)	Human	Noc	Ap Cm Fl Sm Sp Su Tc	aadA2, blaP1, floRc, sul1, tet(G)
SRC49	Paratyphi B dT+	Australia (2001)	Human	No	Ap Cm Fl Sm Sp Su Tc	aadA2, blaP1, floRc, sul1, tet(G)
SRC19	Emek	Australia (1999)	Effluent	Not applicable	Cm Fl Su Tc Tp Nx	dfrA1, cmlA9, sul1, tet(G)
SRC234	Emek	Australia (2000)	Human	Overseas	Su Tp Nx	dfrA1, sul1
SRC235	Emek	Australia (2002)	Human	Thailand	Cm Fl Su Tc Tp Nx	dfrA1, cmlA9, sul1, tet(G)
SRC239d	Emek	United Kingdom (1999)	Human	Unknown	Cm Fl Su Tc Tp Nx	dfrA1, cmlA9, sul1, tet(G)

^aAp, ampicillin; Cm, chloramphenicol; Fl, florfenicol; Nx, nalidixic acid; Sm, streptomycin; Sp, spectinomycin; Su, sulfamethoxazole; Tc, tetracycline; Tp, trimethoprim.

^bThe genes confer resistance as follows: aadA2, Sm and Sp; blaP, Ap; floR, Cm and Fl; dfrA1, Tp; sul1, Su; and tet(G), Tc.

^cLinked to consumption of imported food.

^dIsolated at the Laboratory for Enteric Pathogens, Health Protection Agency, London, United Kingdom.

PCR mapping.

Cultures were grown overnight at 37°C on MacConkey agar (Becton Dickinson and Company, MD). Whole-cell DNA was isolated using standard methods (36) and used as the template for PCR amplification reactions, using various combinations of the primers listed in the relevant tables or described elsewhere (3, 9, 10, 22, 23). Amplification was carried out in PCR buffer (Roche Molecular Biochemicals, Mannheim, Germany) containing each deoxynucleoside triphosphate at 160 μM, 20 pmol of each primer, approximately 10 to 50 ng of template, and 1 unit of Taq DNA polymerase (Roche). Reaction conditions, described in detail elsewhere (22), were generally 94 to 96°C for 3 to 5 min followed by 30 to 40 cycles of denaturation (94 to 96°C for 30 s), annealing (52 to 63°C for 30 to 60 s), and extension (72°C for 30 s to 2.5 min), with a final incubation at 72°C for 10 to 15 min. Products were separated in agarose gels, and sizes were estimated using 100-bp Plus (Fermentas, Vilnius, Lithuania) or Hyperladder I (Bioline, London, United Kingdom) and known amplification products from SGI1 as standards.

DNA analysis.

Southern hybridization to separated fragments of whole-cell DNA digested with restriction enzymes was performed as described previously (23). The probe, spanning an XbaI site in S024, which was generated by PCR using primers RL-D3 and S024-RV (23), is equivalent to p1-9 (4, 5).

For sequencing, PCR products were purified using a QIAquick PCR purification kit (Qiagen Inc., Valencia, CA) following protocols supplied by the manufacturer. A 10-kb HindIII fragment that included the floR and tet(G) resistance determinants was cloned into pUC19, and plasmid DNA for sequencing was prepared using a Qiagen miniprep kit. The sequence was determined, assembled, and analyzed as described previously (31).

The sequence of the region of the Acinetobacter baumannii genome (GenBank accession no. CT025832) that contains a complex class 1 integron comparable to InEmek was not accurately described previously (18), with several genes misidentified and others missing. It was therefore reanalyzed here. The various discrete regions of DNA (e.g., gene cassettes, 5′-CS, 3′-CS, IS6100, etc.) were delineated by being matched with known examples of these regions in pairwise comparisons.

Nucleotide sequence accession number.

The nucleotide sequence of the SGI2 integron and flanking regions from strain SRC19, reported in this paper, as well as the sequences of the left and right boundaries of SGI2 (the thdF-int and S044-yidY junctions), have been added to GenBank under accession no. AY963803.

RESULTS

Location of the SGI in the chromosome of serovar Emek strains.

DNA from SRC19 had previously yielded products of the expected size (500 bp), using primers U7-L12 (in thdF) with LJ-R1 (in int of SGI1) and 104-RJ (in S044 of SGI1) with 104-D (in yidY), indicating that the standard junctions between an SGI1-like region and the thdF and yidY genes on the chromosome are present (Fig. 1A). Three further serovar Emek strains, listed in Table 2, yielded the same products (Table 3). The sequences of the PCR amplicons from all four strains were identical. These sequences were compared to those recently reported for SGI1-K in serovar Kentucky (GenBank accession no. AY463797) (23) and to relevant parts of the SGI1 sequence (GenBank accession no. AF261825). Assuming that the SGI-derived region commences at the beginning of the left direct repeat, the segments internal to the SGI were the same in the SGI1 and SGI1-K sequences but differed by 4 bp of 334 bp (>1%) at the left end in the Emek strains. A small number of base changes relative to the corresponding region in serovar Kentucky were also found in the serovar Emek chromosomal sequences flanking the SGI on the right. A 99-bp segment found adjacent to the right direct repeat is found in SGI-containing Emek and Kentucky strains, and also in two serovar Derby strains (2), but is replaced by the “retron phage” in serovar Typhimurium DT104 strains. Apart from the few differences characteristic of each serovar, the chromosomally derived sequences were also identical to a continuous region extending from within the thdF gene to within the yidY gene in genome sequences of serovars Typhi, Paratyphi A, and Choleraesuis (GenBank accession no. AL627280, AE014613, CP000026, and AE017220) that lack the island, as described previously for SGI1-K (23).

TABLE 3.

PCR linkage in S. enterica strains containing SGI1 or SGI2^a

Strain	SGI	Amplification of PCR product
		thdF-int	S044-yidY	S026-intI1	IS6100-S044	S023-intI1	IS6100-S024	S026-S044	S023-S024
SRC49	SGI1	+	+	+	+	−	−	−	+
SRC19	SGI2	+	+	−	−	+	+	+	−
SRC234	SGI2	+	+	−	−	+	+	+	−
SRC235	SGI2	+	+	−	−	+	+	+	−
SRC239	SGI2	+	+	−	−	+	+	+	−

^aPrimer pairs linking thdF-int (U7-L12/LJ-R1), S044-yidY (104-RJ/104-D), S026-intI1 (S026-FW/int-RV), and IS6100-S044 (DB-T1/MDR-B) and their product sizes have been reported elsewhere (3, 9, 22). When no integron is present, S026-S044 (S026-FW/MDR-B) yields a product of 1.82 kb and S023-S024 [SGI19(F3)/S024-RV] yields a product of 2.92 kb. Primer pairs linking S023-intI1 [SGI19(F3)/int-RV] and IS6100-S024 (IS6100-F/S024-RV) yield products of 1.37 and 2.56 kb when the integron is in S023. Primer SGI19(F3) is from reference 22, and primer IS6100-F (5′-AAG GGA TTC GAA GTC ATG C-3′; matches positions 41322 to 41340 in IS6100 [GenBank accession no. AF261825]) is from this study.

Context of the integron.

Previously, PCR products were not obtained using SRC19 DNA and the primer pairs that detect the boundaries of the In104 integron with the SGI1 backbone, namely, S026-FW (in S026) with int-RV (in intI1 of the integron) and DB-T1 (in IS6100 in the integron) with MDR-B (in S044), indicating that the integron is not in the same position as in SGI1 (22). All of the SGI-containing serovar Emek strains examined here gave the same result (Table 3). Furthermore, a product was detected using S026-FW (in S026) with MDR-B (in S044), and its size (1.8 kb) was as predicted if the integron is excised from the published SGI1 sequence (GenBank accession no. AF261825), bringing tnpR next to S044 (Fig. 1B). The sequence contained four base changes relative to SGI1 and only one copy of the 5-bp duplication flanking In104.

Using a series of primers specific for other SGI backbone genes with the standard primer in intI1 to search for the location of the integron, the IRi end of the integron was previously shown to lie 6.9 kb to the left of In104, where it abuts part of the S023 gene (22). To determine if the integron was inserted precisely into S023, PCR using IS6100-F (in IS6100) with a second primer in S024 (S024-RV) was used. The product was of the predicted size (2.5 kb), and its sequence revealed a 5-bp duplication of the target sequence flanking the integron. Hence, the integrons in SGI1 and in the Emek strains are in completely different locations (Fig. 1B), indicating that the two SGIs have independent origins. The differences detected in the sequence of the SGI backbone support this conclusion. Because this configuration cannot be said to be a simple variant arising from SGI1, SGI1-J was renamed SGI2.

Structure of the integron in SGI2 and the cmlA9 gene.

SGI2 was previously shown to contain the dfrA1-orfC cassette array in the left attI1 site and a specific deletion that removes part of the attI1 site and part of the 3′-CS on the right (22). Similar results were obtained here for SRC235 and SRC239, but SRC234, which is susceptible to chloramphenicol, florfenicol, and tetracycline, had lost all sequences located between the left and right 3′-CS segments, and this variant was named SGI2-A.

Further differences internal to the integrons in SGI1 and SGI2 were indicated by the finding that a floR gene could be detected using primers StCM-L and StCM-R (4), while a multiplex PCR which detects the five resistance genes of SGI1 (10) failed to detect a floR gene in SGI2. The sequence of the 13.8-kb integron (hereafter referred to as InEmek) extending from IRi to IRt was completed for SRC19 (GenBank accession no. AY963803). Except for the exchanged cassettes on the left and the deletion spanning the cassette integration site on the right (Fig. 1B), this sequence is very closely related to that of In104. However, a segment of 1.4 kb that overlaps the floR gene in SGI1 is replaced by one that is only 86% identical to the corresponding region in SGI1. The encoded protein is 404 amino acids long and is identical to one from a multidrug-resistant Acinetobacter baumannii strain (18) (GenBank accession no. CT025832) but only 93.1% identical to FloR encoded by SGI1 (GenBank accession no. AF261825). The protein encoded by InEmek, designated CmlA9, is a member of the CmlA chloramphenicol efflux protein family, being 47% identical to the progenitor of this family, CmlA1 (38). Likewise, the set of proteins known as FloR proteins, including the variant from SGI1 (designated FloRc for clarity), are CmlA family proteins. CmlA9 exhibits further differences from the first FloR protein identified (GenBank accession no. AF231986), designated FloRa for clarity, to which it is 91.8% identical. The floRa gene is widely distributed and usually associated with the small mobile element CR2 (31, 37). Comparison of all of the sequences revealed that floRc in SGI1 appears to be a hybrid of the floRa and cmlA9 genes. FloRc is thus a hybrid of FloRa and CmlA9, with the first 335 to 348 amino acids derived from FloRa (amino acids 335 to 348 are the same in both proteins) and the last amino acids (residues 335 to 404) derived from CmlA9 (data not shown). CmlA9 confers resistance to florfenicol, as evidenced by a zone size for the three chloramphenicol-resistant Emek strains of 5 mm (annular radius) with a 30-μg disc, compared to an annular radius of 11 mm for SRC234, which lacks the cmlA9 gene. However, CmlA9 confers a lower level of resistance to florfenicol than that conferred by FloRc in DT104-2 and SRC49, which yield an annular radius of 1 mm.

A further difference between In104 and InEmek is found in the short segment of 152 bp derived from the IRt end of the tni module and found to the right of IS6100 in In4-type integrons (Fig. 1). While in all of the In4-type integrons sequenced to date, the sequence of this segment is identical to that of the IRt outer end of Tn402 (32), the sequence of the IS6100-F-S024-RV PCR product from both SRC19 and SRC239 that includes this region exhibited two alternate bases of approximately equal peak sizes at several positions. One of the two bases corresponded to the expected Tn402-In4 sequence, and the second base was that found in the corresponding segment of the mercury resistance transposon Tn5058 (GenBank accession no. Y17897) (26). Since this finding suggests that there may be two copies of SGI2 in these strains, primers that would detect the left end of the SGI adjacent to the right end (as in a duplication or a circular intermediate) were used, but no product was detected. Furthermore, whole-cell DNA digested with BamHI or HindIII and hybridized with a probe in S044 revealed only one band, consistent with the presence of only one copy of the SGI.

The SGI2 backbone is closely related but not identical to the SGI1 backbone.

Hybridization, PCR, and partial sequencing were all used to establish that the backbone of SGI2 is the same overall as the 27.4-kb backbone of SGI1. SGI1 has seven XbaI sites, five of which are located in a short span in SGI2 that includes the S026, tnpR, and S044 genes (Fig. 2). Four of these five sites lie within the S026-FW (in S026) to 104-D (in yidY) amplicon, which was sequenced and differs by only 5 bp (of 2,491 bp) from the equivalent parts of SGI1. In SGI1, the fifth XbaI site in S026, together with the remaining sites in S024 and S011, generates restriction fragments of 4.0 kb and 8.9 kb that are usually detected by hybridization using a probe that spans the XbaI site in S024 (12, 23). In SGI2, the integron lies within the 8.9-kb fragment, increasing its size to 22.7 kb, and the probe detected the 4-kb fragment together with a fragment of over 20 kb (data not shown). Thus, the region to the right of InEmek in SGI2 is essentially the same as the corresponding parts of the SGI1 backbone.

FIG. 2.

Map of the SGI backbone. The complete SGI backbone is shown at the top, with the sites of insertion of In104 and InEmek indicated by vertical arrows. The positions of XbaI sites are indicated by X or X₂ (for two close sites). Lines below show the positions and extents of PCRs used to map the backbone. The primers used to generate these products and their lengths are tabulated at the bottom. Primers were made in this study, except for DR-FW, S004-RV, S011-FW, S014-FW, S014-RV, S022-FW, S024-RV, and S026-RV, which are from reference 9, and 104-D, from reference 3.

To examine the rest of the SGI backbone, a set of overlapping fragments generated by PCR amplification (Fig. 2) were used. The lengths of these fragments, excluding ones interrupted by an integron, were all as predicted from the SGI1 sequence and were identical to those of the products produced by serovar Typhimurium strain DT104-02 and serovar Paratyphi B dT⁺ strain SRC49, indicating that the remainder of the SGI1 backbone is present. The fragments generated by digestion of these amplicons from each of the three serovars with restriction enzyme RsaI were also identical for all but one of the PCR amplicons. In PCR 1, the four Emek strains all had the same digestion profile, but it differed from that of the control strains, which had one less RsaI site.

A total of 5.7 kb, derived from the regions at the left and right ends of SGI2 and at the boundaries of the integron with the backbone and representing over 20% of the backbone, was sequenced and was >99.7% identical to the sequence of SGI1. The differences indicate that the SGI2 backbone is closely related but not identical to the SGI1 backbone.

Globally disseminated serovar Emek clone.

Though serovar Emek strains are usually isolated very rarely (<10 per annum in Australia), the isolates from Australia had properties which were the same as or closely related to those of an isolate from the United Kingdom. This suggests that they may be members of a single clone that has been disseminated globally. The fact that the sequences of the segments overlapping the site of integration of SGI2 were identical in all cases supports this conclusion. Pulsed-field gel electrophoresis of XbaI macrorestriction digests of whole-cell DNA (data not shown) revealed identical profiles for three of the Emek strains. SRC19 had an additional band, possibly due to the presence of a plasmid. IS200 profiles were identical for all four strains (data not shown).

DISCUSSION

The genomic island SGI2 (formerly SGI1-J) found in the serovar Emek strains described here exhibits a number of differences from SGI1 that indicate it has an independent origin. First, the integron is located in a different position, and second, the sequence of the SGI backbone differs at 10 positions in the 5.7 kb sequenced here (approximately 0.2%). Though the sequenced portion represents only a little over about 20% of the SGI1 backbone, the same level of variation probably extends over its full length, as differences in the digestion pattern were observed for one of the 13 PCR products used to map the backbone. Since similar numbers of changes were not observed in the shared regions within the integrons, it seems likely that the backbones found in SGI1 and SGI2 had already diverged prior to acquisition of an integron. A similar situation was observed with Tn21 and Tn1696, where related mercury resistance transposons independently acquired a class 1 integron (29).

Acquisition of a class 1 integron normally involves a transposition event, with the transpositional machinery supplied in trans, because it is missing from the most commonly occurring class 1 integron configurations (8, 32). Class 1 integrons are members of a family of transposons that have been described as res site hunters because they target the res sites associated with the resolvases of plasmids and class II transposons, where they integrate in an orientation-specific manner (20, 25). Hence, the expected location for an integron is near the resolvase-encoding tnpR gene in the SGI backbone, and this is the location of In104 in SGI1. However, though there is a strongly preferred position for insertion, the target specificity is not absolute, with rare insertions found at some distance from the preferred site (20, 25). Previously, integrons have been found even further from the res site (23, 29, 31), with the furthest being about 350 bp away, in Tn21 (29). Such insertions have the advantage of not disrupting the res site and hence inactivating the resolution system (39). In SGI2, the integron is 6.7 kb from the putative res site, which was assumed to be the site of In104 in the SGI1 backbone. However, it remains to be established that this confers any advantage, as it is not known if the resolution system encoded by the SGI1 backbone is active or required for any specific function relating to SGI1 acquisition or maintenance.

Until the precise location of the preferred target in the SGI backbone for insertion of class 1 integrons is determined experimentally, the possibility that the SGI1 group includes those that represent independent integron acquisition events cannot be ruled out completely. In contrast, variants of SGI2 with differences located within the integron are most likely to have arisen by changes that occurred when the integron was already in situ. Three of the serovar Emek strains examined here were resistant to trimethoprim, sulfonamides, chloramphenicol, and tetracycline and contained all four of the resistance genes found in SGI2. One isolate which exhibited resistance to only trimethoprim and sulfonamides includes only two of them and is a variant of SGI2 that has lost the central region containing the cmlA9 and tet(G) determinants but retains the dfrA1-orfC cassette array. This configuration can arise simply by homologous recombination within the duplicated 3′-CS segments of SGI2, leading to loss of the central portion (Fig. 3B). The circular nonreplicating molecule generated can, via the reverse reaction, be acquired by any class 1 integron carrying a 3′-CS and present in the same cell. This is likely to explain how both SGI1 and SGI2 contain this region, as it can be acquired after the integron has been incorporated into the SGI backbone.

FIG. 3.

Map of integrons containing floR and tetA(G) genes. (A) In104 from SGI1. (B) InEmek from SGI2. (C) Integron in Acinetobacter baumannii (GenBank accession no. CT025832). The core In4-type integrons are shown, with additional regions above and below. The region related to SGI2 in panel C is boxed. Features of integrons are as described in the legend to Fig. 1, except that a long array of cassettes in panel C is indicated simply by the names of the genes in the cassettes it contains. Vertical arrows indicate the positions of additional CR-associated segments within the integron that have been incorporated by homologous recombination and the position of IS1999. The small mobile elements CR1 and CR3 are indicated by an open box with a vertical bar, marked ori, at the origin end. The open reading frames labeled rcr1 and rcr3 (rolling circle replication) correspond to orf513 of CR1 and orf2 (Fig. 1) of CR3 and encode the proposed initiators of rolling circle replication responsible for movement of CR1 and CR3, respectively. DNA segments associated with the CR3 element in the SGIs are represented as smaller open boxes, with a horizontal line indicating the extent of the floR region that is substituted in SGI1. Various cross-hatching patterns indicate other CR-associated regions.

Homologous recombination is also likely to be involved in generating the differences between SGI1 and SGI2 in the cmlA9/floR region. Three versions of the floR gene, all of which produce proteins that are quite closely related (>97% pairwise identity), have been reported (37), and to distinguish them we have assigned letters in the order of discovery. However, the cmlA9 gene in InEmek is distinct, with the encoded CmlA9 protein being only about 90 to 93% identical to the FloR variants and conferring a somewhat lower level of resistance to florfenicol. An identical cmlA9 gene region is found in the same context in a large antibiotic resistance gene cluster from a multidrug-resistant Acinetobacter baumannii strain (18) (GenBank accession no. CT025832). This cluster was reanalyzed here and shown to have a large region of similarity to InEmek that extends from IRi to the groEL-intI1 fusion (Fig. 3C). The version of FloR produced by SGI1 (FloRc) is identical to FloRa until it switches to CmlA9 near the N terminus (data not shown). A second crossover is found upstream of the cmlA9/floRc genes, after which both sequences are the same, as in SGI1 (Fig. 3A). Thus, it appears that a segment from a floRa gene region has become incorporated, presumably via a homologous recombination exchange, into the cmlA9 region found in InEmek to generate In104. The floRb gene found in plasmid R55 (11) (GenBank accession no. AF332662) also differs from floRa in a short region, and this is reflected in differences in the protein sequence that lie near the center of the protein. However, the new segment is not derived from cmlA9 (not shown). Hence, it is clear that recombination is at least partly responsible for the generation of diversity in the cmlA (floR) genes, as previously found for the aadA1 and aadA2 genes (19).

It seems likely that the unusual SGI2-containing serovar Emek strains isolated in Australia and in Great Britain are all derived from a single cell which acquired SGI2. This new clone, which may have emerged only recently, is likely to be found to be prevalent in a single country, possibly Vietnam, where serovar Emek strains have been found in retail meats (33). It has also spread from there to other countries via travelers, though there is no evidence that it has become established in either the United Kingdom or Australia. This highlights again how readily new multidrug-resistant strains can spread around the globe with travelers or imported food. Given the similarities between SGI1 and SGI2, it also seems likely that SGI2 will eventually spread to different Salmonella serovars and to different bacterial species. The fact that an extended region identical in sequence to a large segment of InEmek is also found in a clinical A. baumannii strain also illustrates how readily established resistance gene clusters can spread across species boundaries.