Molecular Analysis of Antibiotic Resistance Gene Clusters in Vibrio cholerae O139 and O1 SXT Constins

Abstract

Many recent Asian clinical Vibrio cholerae E1 Tor O1 and O139 isolates are resistant to the antibiotics sulfamethoxazole (Su), trimethoprim (Tm), chloramphenicol (Cm), and streptomycin (Sm). The corresponding resistance genes are located on large conjugative elements (SXT constins) that are integrated into prfC on the V. cholerae chromosome. We determined the DNA sequences of the antibiotic resistance genes in the SXT constin in MO10, an O139 isolate. In SXT^MO10, these genes are clustered within a composite transposon-like structure found near the element's 5′ end. The genes conferring resistance to Cm (floR), Su (sulII), and Sm (strA and strB) correspond to previously described genes, whereas the gene conferring resistance to Tm, designated dfr18, is novel. In some other O139 isolates the antibiotic resistance gene cluster was found to be deleted from the SXT-related constin. The El Tor O1 SXT constin, SXT^ET, does not contain the same resistance genes as SXT^MO10. In this constin, the Tm resistance determinant was located nearly 70 kbp away from the other resistance genes and found in a novel type of integron that constitutes a fourth class of resistance integrons. These studies indicate that there is considerable flux in the antibiotic resistance genes found in the SXT family of constins and point to a model for the evolution of these related mobile elements.

The intercellular spread of the genetic determinants of resistance to antimicrobial agents is facilitated by mobile genetic elements, such as conjugative plasmids and conjugative transposons. The antibiotic resistance genes in these elements are often located within transposons and/or integrons, elements that facilitate the intracellular movement of genes. Two types of transposons have been found to contain resistance genes. Class I transposons, also known as composite transposons, consist of two insertion sequence (IS) elements that flank additional DNA sequences, such as resistance genes. Class II transposons do not contain recognizable IS elements; instead, the genetic information for their transposition and other phenotypes (including antibiotic resistances) is bordered by 35- to 110-bp inverted repeats (reviewed in reference 10). Integrons also play a major role in the spread of antibiotic resistance genes in gram-negative bacteria (32). Integrons are gene-capturing systems that incorporate gene cassettes and convert them to functional genes (31, 32). Integrons characteristically encode an integrase (intI) that mediates recombination between a sequence in the gene cassette (attC) and an integron-associated sequence (attI). This results in integration of the cassette downstream of a resident promoter to permit expression of the encoded protein. While integrons often are found in plasmids and usually contain antibiotic resistance genes, they can also be located on the chromosome and can contain genes that do not specify resistance to antibiotics (4, 26). To date, three classes of resistance integrons have been described based on similarities in the integrase sequences. Class I integrons usually contain the gene sulI, encoding sulfamethoxazole resistance, at their 3′ end (32). Recently, a new type of integron, collectively called chromosomal superintegrons, has been found in the chromosomes of several species belonging to the gamma proteobacteria, including Vibrio cholerae (18, 26, 34).

V. cholerae is the causative agent of the severe and sometimes lethal diarrheal disease cholera. While the genetic bases of resistance to antibiotics in V. cholerae have not been extensively characterized, antibiotic resistance determinants are usually found on plasmids in this organism (13, 17, 40). Historically, only the O1 serogroup of V. cholerae has been associated with epidemic cholera. However, in late 1992 in India and Bangladesh, a novel serogroup designated V. cholerae O139 emerged and gave rise to major cholera outbreaks. Initially, V. cholerae O139 replaced V. cholerae El Tor O1 as the predominant cause of cholera on the Indian subcontinent (5). Microbiologic and genetic characterization of V. cholerae O139 revealed that this serogroup arose from V. cholerae O1 El Tor by horizontal gene transfer and substitution of the genes encoding the O139 serogroup antigen for the genes encoding the O1 serogroup antigen (3, 9, 38, 42). Besides the novel serogroup antigen, the initial O139 isolates could be distinguished from the O1 strains they replaced by characteristic resistances to the antibiotics sulfomethoxazole (Su), trimethoprim (Tm), chloramphenicol (Cm), and low levels of streptomycin (Sm). In MO10, a 1992 clinical O139 isolate, the genes encoding these resistances were found to be located on a novel transmissible genetic element designated the SXT element (referred to here as SXT^MO10) (44).

Though it is self-transmissible, an autonomously replicating extrachromosomal form of SXT^MO10 has not been isolated; instead, this ∼100-kbp element is always integrated into the 5′ end of the chromosomal gene prfC. SXT^MO10 encodes an integrase related to the λ family of site-specific recombinases, and we have shown that the integrase mediates the element's integration and its chromosomal excision, which generates a circular episome (21). This circular but apparently nonreplicating form of the element is believed to be a requisite intermediate for its conjugative transfer between V. cholerae strains, as well as between other gram-negative bacteria. We proposed a new term, constin, an acronym for the element's properties (conjugative, self-transmissible, and integrating) to describe SXT^MO10 and other elements with similar features.

After the extensive cholera outbreaks caused by V. cholerae O139 strains, El Tor O1 V. cholerae strains reemerged in 1994 as the predominant cause of cholera on the Indian subcontinent. In contrast to the El Tor O1 strains before the O139 outbreak, these reemerged El Tor strains, like the initial O139 isolates, were resistant to Su, Tm, Cm, and Sm (48). The corresponding resistance genes were found to be located in a constin (designated here SXT^ET) that is closely related but not identical to SXT^MO10 (21, 44). Variation is also evident in more recent O139 isolates from India, as these are generally no longer resistant to Su and Tm (28). However, molecular analyses have revealed the presence of an SXT^MO10-like element integrated into prfC in these strains, indicating that they still harbor constins related to SXT^MO10 (21).

SXT-like elements are not unique to V. cholerae O139. For example, the IncJ element R391 that mediates kanamycin (Kn) and mercury resistance, originally derived from a South African Providencia retgerii isolate (8), is functionally and genetically related to SXT^MO10 (20). Analysis of these two elements suggested that they consist of similar basic building blocks—modules encoding integration and transfer functions—to which have been added genes encoding defining features, such as antibiotic resistance genes (20).

In this study, we determined the sequence and organization of the antibiotic resistance genes in SXT^MO10 and compared them to those of other SXT constins. The SXT^MO10 resistance genes are embedded in a ∼17.2-kbp composite transposon-like element that interrupts the SXT-encoded rumAB operon. A deletion event, likely mediated by recombination between duplicated sequences in this region, accounts for the Su and Tm sensitivity of recent O139 isolates. In SXT^ET, unlike in SXT^MO10, resistance to Tm is encoded outside the cluster of resistance genes; instead, the Tm resistance determinant is found in a novel class of integrons located far away from the remainder of the antibiotic resistance genes within SXT^ET. By comparison, the Kn resistance gene in R391 is found to be part of a transposon containing IS26 elements that is located ∼3 kbp 5′ to the R391 rumAB operon. Overall, these studies indicate that the antibiotic resistance determinants on constins are often part of dynamic genetic structures that allow relatively rapid alteration of the properties encoded by a constin.

MATERIALS AND METHODS

Bacterial strains and media.

The bacterial strains used in this study are described in Table 1. Bacterial strains were routinely grown in Luria-Bertani (LB) broth (2) at 37°C and stored at −70°C in LB broth containing 20% (vol/vol) glycerol. Antibiotics were used at the following concentrations: ampicillin (Ap), 100 mg liter⁻¹; Kn, 50 mg liter⁻¹; Su, 160 mg liter⁻¹; Tm, 32 or 250 mg liter⁻¹; tetracycline, 10 mg liter⁻¹; and Cm, 2 mg liter⁻¹ for V. cholerae and 20 mg liter⁻¹ for Escherichia coli.

TABLE 1.

Bacterial strains and plasmids used in this study

Strain or plasmid	Relevant genotype or phenotype	Reference or source
V. cholerae O139
MO10	Toxigenic 1992 clinical isolate from India, SXTMO10+, Sur Tmr Cmr Smr	41
AS207	Toxigenic 1997 isolate from Calcutta, Sur TmS Cmr Smr	36
E712	Nontoxigenic 1994 isolate from Sri Lanka, Sur Tmr Cmr Smr	30
2055	1998 clinical isolate from Bangladesh, SuS TmS CmS SmS	21
HKO139-SXTS	Clinical isolate from Hong Kong, Sus TmS CmS Sms	47
V. cholerae O1
CO943	El Tor 1994 clinical isolate from India, Sur Tmr Cmr Smr	44
1811/98	El Tor 1998 clinical isolate from Bangladesh, Sur Tmr Cmr Smr	21
C10488	El Tor 1999 clinical isolate from Bangladesh, Sur Tmr Cmr Smr	This study
E. coli K-12
TOP10	F′ mcrA Δ(mrr-hsdRMS-mcrBC) φ80lacZΔM15 ΔlacX74 recA1 endA1	Invitrogen
DH5α	F−thi-1 ΔlacU196 φ80lacZΔM15 hsdR17 recA1 endA1
Plasmids
pSXT1	pSuperCos1 containing a part of the SXT element encoding Cmr Sur Tmr Smr	44
pWKS30	Apr pSC101 derivative	45
pATMP1	pWKS30 + 14 kbp from SXTMO10, Cmr Tmr Smr	This study
pSUL1	pWKS30 + 1.7 kbp from SXTMO10, Sur	This study
pYL1	pWKS30 + 2.77 kbp from SXTET, dfrA1+	This study
pYL8	pWKS30 + 3.8 kbp from SXTET, dfrA1+	This study
pGB2	Spcr pSC101 derivative	6
pRLH422	pGB2 + 11 kbp from R391, Knr	19, this study

Molecular biology procedures.

Plasmid DNA was prepared using the Qiaprep Spin Miniprep Kit (Qiagen, Valencia, Calif.), and chromosomal DNA was isolated with the Genome DNA Kit (Bio 101, Vista, Calif.) as described by the manufacturer. Recombinant DNA manipulations were carried out with standard procedures (2). Automated DNA sequencing was carried out as described previously (43) at the Tufts Medical School DNA Sequencing Core Facility. Computer analysis of DNA sequences was performed with the MacVector and AssemblyLIGN programs (Oxford Molecular Group, Campbell, Calif.), the Vector NTI program (InforMax, North Bethesda, Md.), and the BLAST programs (1) available on the web site of the National Center for Biotechnology Information (Bethesda, Md.). Protein sequences were analyzed for the presence of motifs with the SMART program (http://smart.embl-heidelberg.de).

Cloning and sequencing of antibiotic resistance genes of V. cholerae O139 MO10.

The previously described cosmid pSXT1 contains a ∼40-kbp insert of SXT^MO10 DNA and mediates resistance to Su, Cm, Tm, and Sm (44). A library of pSXT1 EcoRI fragments was constructed in pWKS30 (45). Subsequently, plasmids mediating resistance to Su, Cm, and Tm were isolated by plating the library on L-agar plates containing the respective antibiotics. One such plasmid, pATMP1, contained a 14-kbp insert that conferred resistance to Cm and Tm; another, pSUL1, contained a 1.7-kbp insert that conferred resistance to Su. Overlapping BamHI, PvuII, and PstI fragments of pATMP1 were subcloned into pUC18, and the DNA sequences of the inserts were determined by primer walking. Additional primer walking using pSXT1 as a template was carried out to determine the sequences flanking the inserts in pATMP1 and pSUL1 on SXT^MO10.

Cloning and sequencing of dfrA1 from V. cholerae O1 C10488.

Chromosomal DNA from C10488 was partially digested with Sau3AI, and then fragments of ∼2 to 5 kbp were isolated and ligated with BamHI-digested pWKS30. The ligation mixture was electroporated into E. coli DH5α and plated on L-agar plates containing Tm (250 mg liter⁻¹) and Ap. Two plasmids mediating Tm resistance, pYL1 and pYL8, were isolated. The inserts in these two plasmids (2.77 and 3.8 kbp, respectively) were sequenced and found to overlap.

Cloning and sequencing of aphAI from R391.

As described previously (19, 20), EcoRI fragments of R391 mediating Kn resistance were subcloned into pGB2 (6). One plasmid, called pRLH422, contained a single ∼11-kbp EcoRI fragment and was used for our present studies. The DNA sequence of the ∼11-kbp EcoRI fragment was obtained by nebulizing 20 μg of pRLH422, so as to randomly shear the DNA into fragments of 1 to 2 kbp. These fragments were blunt ended and subsequently cloned into SmaI-digested pUC19; 288 clones were picked and arrayed into three 96-well plates. The DNA sequence of the inserts was obtained using an Applied Biosystems ABI377 sequencer using standard sequencing protocols and primers that were designed to extend from both the 5′ and 3′ ends of the vector into the insert. The sequence data obtained were aligned into a contiguous sequence using the PhredPhrap program, and the correct alignment of the compiled sequence was confirmed by restriction mapping based on the compiled sequence.

PCR amplification.

The primers used in this study are listed in Table 2 and were synthesized by the Tufts Medical School DNA Sequencing Core facility. PCRs were performed using standard reaction conditions in total volumes of 20 μl.

TABLE 2.

DNA sequences of the oligonucleotides used in this study

Primer	Locus (direction)a	Nucleotide sequence (5′ to 3′)
INT1	int (+)	GCTGGATAGGTTAAGGGCGG
INT2	int (−)	CTCTATGGGCACTGTCCACATTG
FLOR-F	floR (+)	TTATCTCCCTGTCGTTCCAGCG
FLOR-B	floR (−)	TCGTCGAACTCTGCCAAATG
SUL2-F	sulII (+)	AGGGGGCAGATGTGATCGAC
SUL2-B	sulII (−)	TGTGCGGATGAAGTCAGCTCC
STRA-F	strA (+)	TTGATGTGGTGTCCCGCAATGC
STRA-B	strA (−)	CCAATCGCAGATAGAAGGCAA
TMP-F	dfr18 (+)	TGGGTAAGACACTCGTCATGGG
TMP-B	dfr18 (−)	ACTGCCGTTTTCGATAATGTGG
TMP3	orfA (+)	CATGCTGTTTCTCGACGGTG
TMP4	orf6 (−)	GATCCGATCTGTTTGTTCAG
LEND4	orf1 (+)	CCTTTGGTTACACATTCGC
LEFTF3	rum′B (−)	GGTGCCATCTCCTCCAAAGTGC
RUMA	Intergenic	CGAGCAATCCCCACATCAAG
YL6	orf73 (+)	TGTGGAACGGCTTTCTGACG
YL3	orfC5A (+)	CGTTGGTTTGGGGTAACACC

^a+, oligonucleotides corresponding to the coding strand (forward primer), −, oligonucleotides corresponding to the noncoding strand (backward primer). 

Nucleotide sequence accession numbers.

The sequence of the antibiotic resistance gene cluster of SXT^MO10 has been deposited in GenBank under accession no. AY034138. The sequence of the integron of SXT^ET has been deposited under accession no. AY035340. The sequence of the Kn resistance transposon found in R391 has been deposited under accession no. AF375956.

RESULTS AND DISCUSSION

Arrangement of antibiotic resistance genes in V. cholerae O139 strain MO10.

We previously constructed a cosmid library with chromosomal DNA derived from O139 strain MO10, a 1992 clinical isolate from Madras, India (44). pSXT1, one of the cosmids from this library, was found to confer resistance to Su, Cm, Tm, and Sm, indicating that the genes mediating these resistances were not randomly distributed in SXT^MO10. Isolation of subclones of the ∼40-kbp insert from pSXT1 (as described in Materials and Methods) revealed that these resistance genes were in fact clustered together in a region of about 9.4 kbp (Fig. 1). Detailed analysis of the DNA sequence of this region along with that of flanking sequences resulted in two major findings. First, the antibiotic resistance genes appear to be part of a large transposon-like element. This element is itself a mosaic composed of other transposon-like elements and DNA sequences found in other mobile elements. Second, SXT^MO10 contains previously identified genes (floR, sulII, strA, and strB) encoding resistance to Cm, Su, and Sm, respectively, and a novel gene encoding resistance to Tm.

FIG. 1.

Organization of the antibiotic resistance gene cluster in SXT^MO10. The SXT^MO10 genes mediating resistance to antibiotics, dfr18, floR, strA, strB, and sulII, are represented by gray arrows, and genes with similarity to transposases (orf1, orf2, and orfA) are represented by hatched arrows. Genes encoding hypothetical proteins similar to known proteins are shown as horizontal hatches, and genes encoding hypothetical proteins dissimilar to known proteins are shown in white. Genes rumA and rumB are in black. The rumAB operon of R391 is presented above the SXT^MO10 antibiotic resistance gene region. The sequence in rumB which is repeated in SXT^MO10 is in bold and underlined; the flanking imperfect repeat (IR) sequences in SXT^MO10 are marked by arrows. Also indicated are the EcoRI sites (E) used for construction of pATMP1 and pSUL1. Regions of nucleotide sequence identity to other published nucleotide sequences are represented by boxes.

SXT^MO10 appears to have acquired its antibiotic resistance genes and some adjacent sequences via a transposition event(s). This event introduced a 17.2-kbp region containing all five resistance genes into rumB, the second gene of the rumAB operon. This is likely to have been a multistep process, as outlined below. Consistent with this hypothesis, the 17.2-kbp sequence is flanked both by an 8-bp direct repeat (corresponding to amino acids [aa] 76 to 78 of rumB) and by 16-bp imperfect inverted repeats, structures often found at the boundaries of transposons.

A role for transposition is also suggested by the presence of open reading frames (ORFs) with similarity to previous described transposases at the left end of these 17.2 kbp (Fig. 1 and Table 3). The deduced amino acid sequence of orf1 has 39% similarity to the C terminus of a transposase found in Pseudomonas putida (Table 3), and the deduced amino acid sequence of orf2 has 29% identity and 47% similarity to a transposase found in Tn5501 and Tn5502, two cryptic transposons located in P. putida (25). The 5′ end of orf2 is repeated downstream of sulII. However, despite its transposon-like features, the 17.2-kbp sequence is apparently not (or no longer) an autonomously mobile genetic element; all our attempts to mobilize the resistance genes independent of the remainder of SXT^MO10 have failed.

TABLE 3.

Gene products with sequence similarity to the antibiotic resistance gene cluster in SXT^MO10

Coding regiona	Gene name	Length (aa)	Closest similarity	Accession no.	% Identity/rangeb
1–1044 (−)	rum′B		RumB R391 (C terminus)	862633
1457–2047	orf1	197	Transposase, P. putida	4754812	39/188 of 530
2164–5061	orf2 (tnpA)	966	Transposase, P. putida	7465523	29/956
5559–7049	orfA	497	Putative transposase, E. coli	10312101	99/496
7416–7967 (−)	dfr18	184	Dihydrofolate reductase type VIII	2833495	37/157
7984–8526 (−)	orf3	181	No homology	NAc	NA
8529–9692 (−)	orf4	388	No homology	NA	NA
9710–10435 (−)	orf5	294	Deoxycytidine triphosphate deaminase, P. aeruginosa	9949625	44/208
10695–11024	orf′A	110	Putative transposase, E. coli (3′ end, aa 388–497)	10312101
11058–11939	orf6	294	No homology	NA	NA
12159–13370	floR	404	FloR (florfenicol exporter), E. coli	10312100	99/404
13401–13703	orf7	101	Putative transcriptional regulator (LysR family), P. aeruginosa	11352177	51/74
13818–14354	orfA′	179	Putative transposase, E. coli (5′ end, aa 1–135)	10312101
14332–15165 (−)	strB	278	Aminoglycoside phosphotransferase	420965	100/278
15168–15968 (−)	strA	267	Aminoglycoside phosphotransferase	420964	100/267
16032–16844 (−)	sulII	271	Dihydropteroate synthase transposase, P. putida (5′ end)	1075456	100/271
17315–18674	orf2′ (tnpA′)			7465523	24
18399–19118 (−)	orfB	240	Mutl, V. cholerae	127554	51/173
					(217–345 of 563)
19308–19540 (−)	rumB′		RumB R391 (N terminus)	862633
19551–19997 (−)	rumA	149	RumA R391	862632	98/149

^aGenes encoded on the minus strand are indicated with (−). 

^bThe number of amino acids in a contiguous stretch from which the identity was calculated is shown. The length of the similar protein is also presented. 

^cNA, not applicable. 

The Tm resistance gene of SXT^MO10 was mapped to subclones of pSXT1 that included a 551-bp ORF. As this ORF's deduced amino acid sequence had 37% identity and 52% similarity (Table 3) to a type VIII dihydrofolate reductase found in some E. coli strains (39), it was named dfr18, for a new gene encoding a Tm-resistant dihydrofolate reductase. dfr18 is preceded by three ORFs, orf3, orf4, and orf5, with the same orientation as dfr18. The deduced amino acid sequences of orf3 and orf4 do not have similarities to any known proteins, but the deduced amino acid sequence of orf5 has 44% identity and 60% similarity (Table 3) to a chromosomal Pseudomonas aeruginosa deoxycytidine triphosphate deaminase (37). Whether orf5 encodes a functional deaminase remains to be studied. These four genes are bracketed by the previously described orfA (7). A complete copy of orfA lies downstream of dfr18, while a 5′-truncated copy of orfA lies upstream of orf5 (Fig. 1 and Table 3). An identical full-length orfA was found by Cloeckaert et al. in a plasmid from an E. coli isolate (7). The predicted OrfA amino acid sequence has been noted to have some similarity to a putative transposase from Pseudomonas pseudoalcaligenes (12). It seems likely that orfA plays some role in promoting the acquisition and loss of antibiotic resistance genes, since orfA or fragments of orfA are closely linked to antibiotic resistance genes in several instances (7, 22, 35). The molecular mechanism(s) by which orfA acts to promote gain or loss of genes remains to be explored.

In two prior cases, orfA or orfA fragments have been found associated with floR (7, 22). This is the case in SXT^MO10 as well (Fig. 1). In SXT^MO10, floR is found close to the 3′ end of the 5′-truncated orfA, preceded by a putative ORF (orf6) of unknown function. FloR is thought to be an export protein which mediates resistance to Cm and florfenicol. This gene has been found in plasmids derived from E. coli isolates from cattle (7), in the chromosome of the multidrug-resistant Salmonella enterica serovar Typhimurium phagetype DT104 (4) and in an R-plasmid derived from the fish pathogen Photobacterium damselae subsp. piscida (22). As expected, in-frame deletion of floR from SXT^MO10 resulted in cells that were no longer resistant to Cm (J. Beaber and B. Hochhut, unpublished observations), confirming that floR is required for resistance to Cm. In SXT^MO10, floR is followed by a short putative ORF (orf7) that includes a region with similarity to the helix-turn-helix (HTH) motif of LysR family transcriptional regulators, and another incomplete copy of orfA that is deleted in its 3′ end. The two incomplete copies of orfA that bracket floR together do not constitute a full-length orfA. Comparative DNA sequence analysis revealed extensive nucleotide identity to the floR loci in E. coli isolates and P. damselae (Fig. 1). The genes strA, strB, and sulII, which follow orfA′, are identical to previously described resistance genes and are found on several plasmids, including RSF1010 (35). They encode a sulfonamide-resistant dihydropterate synthase (sulII) and an aminoglycoside phosphotransferase (strAB).

Distribution of SXT^MO10 antibiotic resistance genes in related SXT elements from V. cholerae O1 and O139 strains.

Since the discovery of SXT^MO10 in isolates from the initial O139 outbreak in 1992, closely related constins have been detected in many V. cholerae isolates of both the O1 and O139 serogroups. These related constins, like SXT^MO10, are integrated into prfC (21); however, these elements do not confer the same antibiotic resistances as SXT^MO10. For example, many recent O139 clinical isolates from Asia were found to be sensitive to Su and Tm (28). We analyzed the genetic basis for this sensitivity in two O139 clinical isolates, strain 2055 from Bangladesh and strain HKO139-SXT^s from Hong Kong. PCR assays designed for amplification of internal sequences of dfr18, floR, strA, and sulII from these strains failed, whereas a PCR amplification of int_SXT, a signature sequence of an SXT-related constin, was successful (Table 4). PCR assays utilizing primers that flank the antibiotic resistance genes in SXT^MO10 facilitated the mapping of the borders of the DNA missing in strains 2055 and HKO139-SXT^S. Using chromosomal DNA from either strain as the template, with primer pair LEND4 and RUMA, we obtained a product of ∼3.3 kbp, and with primer pair LEFTF3 and RUMA, we amplified a 4-kbp product (Fig. 2). In contrast, in MO10 these primer pairs flank sequences of 18.5 and 19.2 kbp.

TABLE 4.

Distribution of antibiotic resistance genes in V. cholerae isolates containing int_SXT^a

Strain	Serogroup	Detectionb
		dfr18	dfrA1	floR	strA	sulII
MO10	O139	+	−	+	+	+
HKO139-SXTS	O139	−	−	−	−	−
2055	O139	−	−	−	−	−
AS207	O139	−	−	+	+	+
E712	O139	−	+	+	+	+
CO943	O1	−	+	+	+	+
1811	O1	−	+	+	+	+
C10488	O1	−	+	+	+	+

^aAll strains tested were positive for int_SXT in a PCR assay. 

^bSymbols: +, listed gene detected by a PCR assay and by a Southern hybridization experiment; −, no detection of the gene. 

FIG. 2.

Organization of the region containing antibiotic resistance genes in SXT^MO10 and in V. cholerae O139 strains sensitive to Tm, Su, and Cm. The gene order found in strains 2055 and HKO139-SXT^S (bottom) is compared to that of SXT^MO10 (top). Homologous recombination between the identical sequences in orf2 and orf2′ may have resulted in loss of the antibiotic resistance genes. Also shown are the primers (LEFTF3, LEND4, and RUMA) used to amplify this region in 2055 and HKO139-SXT^S.

The DNA sequence of the 3.3-kbp fragment was partially determined. As in MO10, the reading frame of rumB is interrupted in these strains, but by a much smaller insert encompassing orf1, orf2′, and orf8. This genetic structure suggests that deletion mediated by homologous recombination between the two identical 5′ ends of orf2 that bracket the resistance gene cluster in SXT^MO10 may have rendered these strains sensitive to antibiotics (Fig. 2). An alternative possibility is that these antibiotic-sensitive O139 strains never carried any of the resistance genes and that their constins represent a precursor of SXT^MO10. Since the rum operon is interrupted in both types of elements, the latter possibility seems less likely. In either case, the lack of the ∼15.2-kbp fragment from these antibiotic-sensitive O139 strains has not rendered their SXT-like elements deficient for transfer (data not shown).

Other recent int_SXT-containing O139 isolates, such as the 1996 Calcutta isolate AS207, have been found to be resistant to Cm, Su, and Sm but sensitive to Tm (21, 27). Using AS207 DNA as the template, we were able to amplify floR, strA, and sulII by PCR (Table 4). Southern hybridization indicated that the arrangement of these genes was similar in AS207 and in MO10 (data not shown). However, both a PCR assay (Table 4) and a Southern hybridization assay (not shown) indicated that AS207 lacked dfr18. The precise borders of the deletion including dfr18 in the AS207 constin are discussed below.

After the initial spread of V. cholerae O139 on the Indian subcontinent in 1993, clinical isolates of V. cholerae O1 El Tor from this region were found to be resistant to the same antibiotics, Su, Sm, Tm, and Cm, as O139 strains. We analyzed the genes encoding these resistances in three El Tor strains, CO943, 1811, and C10488, isolated in different years and from different locations on the Indian subcontinent (Table 1). As in O139 strain MO10, the resistance determinants in these strains were part of a constin designated SXT^ET, that is very similar but not identical to SXT^MO10 (21, 44; data not shown). Using chromosomal DNA from these strains as templates for PCR, products corresponding to internal regions of floR, strA, and sulII were amplified (Table 4). Southern hybridization experiments indicated that the organization of these genes in SXT^ET is identical to that in SXT^MO10 (data not shown). To our surprise, despite their resistance to Tm, these El Tor isolates were found by PCR (Table 4) and Southern hybridization (not shown) not to harbor dfr18.

PCR primers (TMP3 and TMP4, Fig. 3) which anneal to sequences that flank dfr18 in SXT^MO10 were used to define the extent of the region missing from SXT^ET. Using these primers and MO10 chromosomal DNA as the template, a PCR product with the expected size of 5.35 kbp was obtained, whereas with C10488 chromosomal DNA as the template, a product of 1.3 kbp was obtained. The DNA sequence of this 1.3-kbp PCR product revealed that in addition to dfr18, orf3, orf4, and orf5 were also absent in C10488 (Fig. 3). Furthermore, in C10488, a complete copy of orfA is followed by orf6 and floR, whereas in MO10, a complete copy of orfA is located next to dfr18 and only a 5′-end-truncated copy of orfA is found next to orf6 (Fig. 3). The 3.34-kbp “insert” that includes the genes dfr18, orf3, orf4, and orf5 and that distinguishes SXT^MO10 from the constin present in C10488 is flanked by a 640-bp duplication (Fig. 3). This repeated DNA sequence encompasses the 3′ end of orfA and the first 205 bp of orf6. A PCR showed that the same sequences were also missing from the constins in the other two El Tor Tm^r strains, CO943 and 1811/98, as well as in the constin in the Tm^s O139 strain AS207 discussed above. We have no direct evidence of the mechanism by which these additional genes were acquired by SXT^MO10 or, alternatively, lost from the C10488 constin. However, given the presence of the duplicated 640-bp sequence, homologous recombination probably played some role in the loss or acquisition of these four genes.

FIG. 3.

SXT^ET lacks dfr18, orf3, orf4, and orf5. In El Tor O1 strain C10488, floR is preceded by a complete copy of orf6 and orfA (top). In contrast, in SXT^MO10, there is a duplication of 640 bp (dark gray boxes) that flanks the genes dfr18, orf3, orf4, and orf5 (black). The locations of primers (TMP3 and TMP4) used for amplification of this area are also shown.

dfrA1 mediates Tm resistance in V. cholerae O1 constin.

Although the constin in strain C10488 lacked dfr18, we strongly suspected that the determinant of Tm resistance in this strain would be part of SXT^ET, since the Su, Sm, Cm, and Tm resistance determinants were cotransferred by C10488 (data not shown). We constructed a plasmid library with insert DNA derived from C10488 chromosomal DNA to isolate the Tm resistance determinant(s) from this strain. We identified two recombinant plasmids (pYL1 and pYL8) that allowed their host cells to grow on media containing Tm. Determination of their respective insert DNA sequences revealed that they contained overlapping inserts and that the overlap included an ORF with nucleotide sequence identity to the previously described gene dfrA1 (Fig. 4) (14). dfrA1 encodes a trimethoprim resistance dihydrofolate reductase which until now has been found exclusively as a cassette within class 1 and 2 integrons (11, 32). Instead, dfrA1 from C10488 appears to be part of a novel (class 4) type of integron; 271 bp upstream of the dfrA1 cassette was a gene of 320 codons whose deduced amino acid sequence showed similarity to the site-specific recombinases found in integrons and which has been named intI9. Its predicted product, IntI9, shows 53% identity to IntI2∗ (a 325-amino-acid protein obtained through readthrough of the stop codon at position 178 in intI2 [accession no. NP_065308]), a putative integrase of the class 2 resistance integrons. The paradigm of class 2 integrons is found on Tn7. The second closest relative of IntI9 is SpuIntIA, the Shewanella putrefaciens chromosomal integron integrase (47% identity) (34). dfrA1 and intI9 are oriented in opposite directions, an arrangement characteristic of integrons. Furthermore, the DNA sequence of the dfrA1 cassette is 99.8% identical to the dfrA1 cassette of class 1 and 2 resistance integrons.

FIG. 4.

Organization of the integron in SXT^ET constin. The five cassettes found in the SXT^ET integron are shown. The attC sites are represented by triangles, and ORFs are represented below the cassettes as arrows. Also shown are the genes traF and orf73. DNA sequences identical to SXT^MO10 are shown as black lines. The insert DNA in pYL1 and pYL8 is shown below. The positions of primers YL6 and YL3, used to amplify the upstream boundary of the integron insertion in SXT^ET, are also indicated.

The sequence downstream of the dfrA1 cassette did not show similarity to any known genes. However, analysis of this sequence revealed the presence of four putative consecutive integron cassettes (Fig. 4). Cassettes 2, 3, and 4 each carry a single ORF, while cassette 5 contains two ORFs in opposite orientation. The putative product of orfC2 (142 aa) is predicted to be located in the cytoplasmatic membrane. The deduced amino acid sequence of orfC3 (136 aa) contains a region with similarity to an Xre-type HTH motif and is predicted to the membrane associated. Finally, the product of orfC5A (233 aa) has an AraC-type HTH motif, while the putative product of orfC5B (82 aa) contains a domain conserved among bleomycin resistance proteins. Although integrons were originally described as systems to capture antibiotic resistance genes, analysis of superintegron cassettes has revealed that many of the genes contained therein are of unknown function (32, 34).

As seen for the cassettes carried in the multiresistance integrons, the attC sites carried by the SXT^ET integron cassettes are extremely different in length (58 to 99 bp) and sequence. Interestingly, the attC site of cassette 2 is almost identical to the attC site of the first cassette of the S. putrefaciens CIP 69.34 superintegron (accession no. AF324211) (34), while the genes carried in both cassettes are unrelated. In contrast, the attC sites of the other three cassettes do not show any significant homology (<50% identity) with the attC sites found either in previously described resistance cassettes or in any of the superintegron cassettes, including those of the V. cholerae superintegron.

In our ongoing research, we are determining the complete nucleotide sequence of SXT^MO10. We took advantage of the partially completed sequence to determine where the dfrA1-containing integron is located in the C10488 constin. Downstream of intI9 in the insert of pYL1 was a region with near nucleotide sequence identity to SXT^MO10 (Fig. 4). In SXT^MO10, this region encodes a putative gene (traF) thought to be required for pilus assembly; it is located about 70 kbp away from the resistance gene cluster. Unlike the insert in pYL1, the sequence of the pYL8 insert did not show any similarity to the sequence of SXT^MO10 (Fig. 4).

To identify the upstream boundary of the apparent insertion of the dfrA1-containing integron, we designed PCR primers to amplify the junction between this integron-like element and predicted upstream sequences in SXT^MO10 (Fig. 4). With the primer pair YL6/YL3, we amplified a product of ∼1 kbp with C10488 chromosomal DNA as a template. Sequence analysis of this PCR product, combined with the sequences of the pYL1 and pYL8 inserts, revealed that relative to SXT^MO10, an insert of 4.77 kbp is present in SXT^ET between traF and an ORF of unknown function, orf73 (Fig. 4). Examination of the borders of the integron in SXT^ET did not reveal sequences such as inverted or direct repeats that might suggest the mechanism by which this integron was acquired. However, we noticed that the divergence between the sequences of SXT^MO10 and SXT^ET downstream of orf73 coincided exactly with the core site sequence in attC of cassette 5. In this region the MO10 sequence does not show any of the attC site structural characteristics apart from a conserved CGTT sequence, which is precisely located at the beginning of the identity with the SXT^ET sequence. Integrase-mediated recombination between attC sites and noncanonical sites, known as secondary sites of consensus GWTMW (15), has been reported several times (15, 16, 33). This suggests that this boundary of the integron likely corresponds to a recombination between the attC site of cassette 5 and the sequence AACGTTCTGC (bases corresponding to bases fitting the secondary site consensus shown above are underlined) of the SXT backbone. To our knowledge, this is the first evidence of such an event to explain the 3′ end of a cassette array in an integron. The only natural case of likely recombination between an attC site and a secondary site described so far was the integration of a single aadB cassette, not an integron, into an RSF1010 plasmid (33).

Like C10488, the other two El Tor strains we studied, CO943 and 1811, also contained a 4.77-kbp sequence inserted between orf73 and traF. Insertion of a dfrA1-containing integron into this locus was not limited to the constins found in El Tor strains. We identified a nontoxigenic O139 isolate, E712, that also contained this insertion (Table 4). In fact, like SXT^ET, the E712 constin also lacked the 3.34-kbp region containing dfr18, present in SXT^MO10 (Table 4), suggesting that the E712 constin was very similar (or identical) to the El Tor constin.

aphAI in R391 is part of a transposon.

Although SXT^MO10 and R391 are closely related constins, they encode different sets of antibiotic resistances. Cells carrying R391 are sensitive to Cm, Tm, Su, and Sm but resistant to Kn. As expected, PCR assays and Southern analyses revealed that R391 does not carry any of the resistance genes or putative transposase genes encoded in SXT^MO10 (20; data not shown). Also, R391 contains an intact rumAB operon (19, 24). This operon encodes proteins that are phylogenetically related to a superfamily of novel error-prone DNA polymerases found in all three kingdoms of life (46). While R391 complements the DNA repair functions encoded by the umuDC operon in E. coli strains missing these genes (19), SXT^MO10 failed to complement a ΔumuDC strain (data not shown), confirming the inactivation of rumB.

DNA sequence analysis of the ∼11-kbp EcoRI fragment carrying the R391 Kn resistance determinant revealed on ORF some ∼4 kbp from the rumAB_R391 locus that is identical to the previously described aphAI gene, which encodes an aminoglycoside phosphotransferase (29). Immediately 5′ and 3′ of aphAI, we identified two copies of IS26 in opposite orientations (data not shown), indicating that aphAI is part of a novel transposon. Interestingly, linkage of aphAI with IS26 is also found in the multiresistance plasmid pSP9351 from P. damselae (23); however, the organization of IS26 relative to aphAI differs between pSP9351 and R391. Taken together, our data indicate that antibiotic resistance genes can be added to SXT-like constins at several locations and via different mechanisms.

Conclusions.

SXT-related constins constitute an important group of transmissible genetic elements that have contributed to the spread of resistance to antimicrobial agents in clinical isolates of V. cholerae from Asia. Our surveys of V. cholerae O139 and O1 clinical isolates from this region indicate that the great majority of post-1993 isolates contain an SXT-related element integrated on the large V. cholerae chromosome at prfC. Thus far, all of the elements tested are self-transmissible and encode Int_SXT, the defining features of SXT-related constins. Although the genetic determinants of the transfer and integration functions of these related elements appear to be nearly identical, in the current study we found that the antibiotic resistance genes in these elements differed. In the SXT constin found in the original 1992 O139 outbreak strains (SXT^MO10), as exemplified by MO10 (and found in other isolates as well), the antibiotic resistance genes were all clustered together within a ∼17-kbp composite transposon-like structure. In contrast, in the SXT^ET constin found in the reemerged (post-1993) El Tor O1 strains, this cluster is missing a 3.3-kbp segment that includes the novel dfr18 found in SXT^MO10. Instead, SXT^ET contains a novel integron-like structure that includes dfrA1, located ∼70 kbp away from the other antibiotic resistance genes in this constin. Finally, R391 contains a transposon-associated Kn resistance gene located ∼3.5 kbp away from the site where the composite transposon-like element apparently inserted in SXT^MO10. The differences in the antibiotic resistance genes in SXT-related constins suggest that these genes are not intrinsic features of this family of constins; they appear to have inserted themselves on these elements as a way to become transmissible through bacterial populations. Selective pressure to become and remain resistant to antibiotics does not seem to be the only explanation for the dissemination and persistence of SXT-related constins in Asian V. cholerae. This is clear from the absence of antibiotic resistance genes from the SXT-like constins found in many recent O139 isolates, such as strain 2055 analyzed in this study. The advantage(s) conferred by constins lacking resistance genes remains to be elucidated.

A plausible scheme outlining the steps in the acquisition and loss of antibiotic resistance genes in the V. cholerae derived SXT family of constins is shown in Fig. 5. First, in one or several steps, a transposon(s) that included sulII, strAB, and floR inserted into rumB, a gene that is intact in R391, an SXT-related constin. Then, the resulting Su^r, Sm^r, and Cm^r constin (such as was found in O139 strain AS207) could have become Tm^r by acquiring either the novel integron containing dfrA1, to give rise to SXT^ET, or dfr18, orf3, orf4, and orf5, to give rise to SXT^MO10. This latter event likely depended on orfA (by an unknown mechanism), since orfA is associated with antibiotic resistance genes in several instances. Subsequently, the Su^r, Sm^r, and Cm^r constin could have undergone a deletion event, likely mediated by homologous recombination, to give rise to constins that lack antibiotic resistance genes such as those found in O139 strains 2055 and HKO139-SXT^S. Even though SXT^MO10 was the first SXT-family constin that we identified (from a 1992 O139 isolate) and we did not detect SXT^ET in O1 strains until 1994, given the differences in the antibiotic resistance genes between these two constins, it seems unlikely that SXT^MO10 is an immediate precursor of SXT^ET. Rather, SXT^ET, the constin found in most recent O1 isolates, seems to have arisen independently of SXT^MO10. We detected an SXT^ET-like element in an O139 isolate (E712), indicating that SXT^ET is not limited to the V. cholerae O1 serogroup. Additionally, we found SXT^ET (or at least very similar elements) in Providencia alcalifaciens isolates from patients in Bangladesh (data not shown). This suggests a recent gene transfer between V. cholerae and P. alcalifaciens. Finally, although SXT family constins are present in virtually all clinical V. cholerae isolates from Asia, these elements are a relatively recent addition to the V. cholerae genome. They are not present in seventh-pandemic V. cholerae isolates, as exemplified by their absence from the genome of N16961, the type strain used for determination of the complete nucleotide sequence of the V. cholerae chromosomes by the Institute for Genome Research. The bacterial species that donated SXT family constins to Asian V. cholerae remains to be determined.

FIG. 5.

Model for antibiotic resistance gene flux in the SXT family of constins.