Integron-Associated DfrB4, a Previously Uncharacterized Member of the Trimethoprim-Resistant Dihydrofolate Reductase B Family, Is a Clinically Identified Emergent Source of Antibiotic Resistance

ABSTRACT

Whole-genome sequencing of trimethoprim-resistant Escherichia coli clinical isolates identified a member of the trimethoprim-resistant type II dihydrofolate reductase gene family (dfrB). The dfrB4 gene was located within a class I integron flanked by multiple resistance genes. This arrangement was previously reported in a 130.6-kb multiresistance plasmid. The DfrB4 protein conferred a >2,000-fold increased trimethoprim resistance on overexpression in E. coli. Our results are consistent with the finding that dfrB4 contributes to clinical trimethoprim resistance.

TEXT

Public health agencies worldwide rank trimethoprim (TMP) a broad-spectrum antibiotic of importance in human medicine (1). Widely used as a result of its low cost and effectiveness, TMP inhibits the activity of many microbial chromosomal dihydrofolate reductases (DHFRs); thus, DHFRs have long served as prioritized targets of antiproliferative drugs (2). Although the majority of living cells harbor a chromosomal member of the type I DHFR family, encoded by a dfrA homolog, the dfrB genes encode a family of plasmid-borne type II DHFRs that are evolutionarily unrelated to type I DHFRs. The dfrB genes have been found in pathogenic bacteria recovered from many food sources, including fish (3), pigs (4, 5), and cows (6), where they confer TMP resistance. Bacteria carrying dfrB genes have also been identified in wastewater samples (7). Over the past decade, dfrB genes have been tracked indirectly in antibiotic resistance studies through identification of integron-related elements (8–10). Therefore, the importance of dfrB genes in TMP resistance in human pathogens may be underappreciated (11, 12).

To date, only seven members of the dfrB gene family are known, and they are highly homologous (77% to 94% genetic identity, 77% to 99% amino acid identity) (Table 1). Among these, the DfrB1 protein (also known as R67 DHFR) is the best-studied type II DHFR (13–17). It is proposed to be recently evolved, and it confers a significant survival advantage under TMP exposure to microbes that harbor it (18). To date, the family of dfrB genes has consistently been reported to be contained within the following mobile genetic elements: a highly variable 57- to 141-bp recombination binding site (attC) and the 7-base canonical sequence (7-be), which is identical in all dfrB genes (5′-GTTRRRY) except dfrB2 (GTTAGGC) and includes a 7-be at the 3′ end, which is the reverse complement of the 7-be found upstream of the coding sequence (r7-be).

TABLE 1.

DNA sequence query coverage and identity of the main type II DHFRs, with the expected value

DHFR	Query coverage (% identity) and expected valuea for:
	dfrB1	dfrB2	dfrB3	dfrB4	dfrB5	dfrB6	dfrB7
Truncated dfrB1	100	95 (77) 2e−47	90 (88) 3e−77	89 (83) 5e−61	100 (89) 9e−90	100 (92) 7e−98	100 (92) 7e−98
dfrB2		100	85 (86) 2e−66	72 (85) 2e−54	83 (79) 3e−45	97 (78) 4e−50	95 (77) 2e−47
dfrB3			100	94 (85) 2e−73	96 (86) 4e−75	97 (87) 7e−79	96 (86) 8e−78
dfrB4				100	92 (81) 8e−59	94 (80) 1e−57	92 (80) 4e−56
dfrB5					100	100 (91) 3e−96	100 (92) 1e−100
dfrB6						100	93 (94) 7e−111

^aExpected values indicated by boldface (coding sequence only). The expected value (e) represents the probability of randomly matching two different sequences. The lower the e value, the more significant the match.

Clinical sample library.

We examined whole-genome shotgun sequencing data for 593 Escherichia coli isolates, including 380 E. coli isolates recovered from 324 individuals with hospital-associated human extraintestinal infections (19) and 189 E. coli isolates recovered from 189 women diagnosed with a community-acquired urinary tract infection (UTI) (A. R. Manges, unpublished data). Genomic E. coli DNA from women with UTIs (Manges collection) was extracted using the PureLink genomic DNA minikit (Thermo Fisher Scientific). Purified DNA was sheared in water using the Biorupter Pico (Diagenode), and sequencing libraries were prepared using the TruSeq DNA PCR-free library preparation kit (Illumina), according to the manufacturer's instructions. All E. coli isolates were sequenced on the Illumina HiSeq 2500 at the University of British Columbia's Pharmaceutical Sciences Sequencing Centre and British Columbia Genome Sciences Centre (Vancouver, BC).

In silico screening.

Seven members of the dfrB gene family were used as templates for in silico screening, including two variants of the dfrB1 gene that differ by the absence or presence of 19 additional N-terminal amino acids not essential for the reductase activity (20; see http://www.esi.umontreal.ca/~pelletjo/ToulouseSupplemental-material.pdf). The in silico dfrB screening was performed by aligning paired-end reads from 569 whole-genome-sequencing data sets to the set of dfrB genes in FASTA format with the Burrows-Wheeler aligner (BWA), using standard alignment parameters. Sequencing reads from E. coli isolates recovered from blood samples of an individual taken 2 days apart aligned exactly over the entire dfrB4 gene sequence (19 [https://www.ncbi.nlm.nih.gov/sra/SRX560289 and https://www.ncbi.nlm.nih.gov/sra/SRX560290]). The DNA consensus sequences were identical to the previously deposited 237-bp dfrB4 sequence (GenBank AY968808 and KP314737.1). Three further samples held one read that overlapped with the dfrB4 integron with 0 to 2 mismatches. Because no contig within assemblies of these three data sets aligned with dfrB4, they were not considered further.

Reconstruction of dfrB4 mobile genetic element and contiguous DNA segments.

Read sets with significant alignment to any reference dfrB gene were assembled into contigs, and the dfrB genes were aligned with BWA to these assembled contigs. The contigs from both samples have an r7-be (GTTGGGC) 61 bp upstream of the dfrB4 coding sequence and an attC sequence downstream (Fig. 1). The 7-be found in the attC (GCCCAAC) is 53 bp downstream from the coding sequence. These flanking sequences are identical to those of a previously reported dfrB4 mobile genetic element (GenBank accession no. AY970968.1) from Klebsiella pneumoniae (21). The attC sequence differs by one nucleotide from another dfrB4 mobile genetic element from E. coli (GenBank accession no. KP314737.1) (22; http://www.esi.umontreal.ca/~pelletjo/ToulouseSupplemental-material.pdf).

FIG 1.

Scheme of the 9,029-bp DNA segment derived from sample SRX560289. Purple, indispensable genes for integration of mobile genetic elements. Yellow, intI (integrase/recombinase), tnpA (transposase), and the integron-associated recombination site (attI). Red, resistance genes (qacEΔ1, biocide resistance; sulI, sulfonamide resistance; the mphR(A)/mphA/mrx gene cluster, erythromycin resistance; and chrA, chromate resistance), except dfrB4 (dihydrofolate reductase, TMP resistance), which is blue. Orange, recombination binding site (attC). Dashed lines, 7-be sequences. Green, DNA sequences sharing similarity with known genes: GNAT/sulI3 (acetyltransferase/sulfonamide resistance) and padR (transcriptional regulator). Gray, the 62- and 4-bp sequence differences between DNA contigs.

Contiguous to the mobile genetic element, we identified a 8,963-bp segment derived from sample SRX560290 and a 9,029-bp segment derived from sample SRX560289 (Fig. 1) (19). The 66-bp difference did not belong to an open reading frame (ORF) and lay outside the dfrB4 mobile genetic element. The dfrB4 gene was carried with intI1 (class 1 integron integrase/recombinase) and tnpA (transposase), which are indispensable for integration of mobile genetic elements (23). Moreover, multiple antibiotic, biocide, and metal resistance genes were identified: qacEΔ1 (biocide resistance [24, 25]), sulI (sulfonamide resistance [26]), the mphR(A)/mphA/mrx gene cluster (macrolide/erythromycin resistance [27]), and chrA (chromate resistance [28, 29]). The qacEΔ1 gene encodes a truncated and less-efficient cation efflux pump version of qacE (24).

Additionally, both segments included similarity to the GNAT (acetyltransferase) and sulI3 genes (30). The DNA sequence was 96% identical to GNAT from Aeromonas hydrophila with 72% query coverage (GenBank accession no. JX141473.1). The sulI3 gene was 100% identical with 55% query coverage (GenBank accession no. EF382672.1) yet included a 140-nucleotide 3′ deletion. The segments included the little-known padR transcriptional regulator. Nine additional ORFs were identified as hypothetical proteins.

A BLAST alignment of the contig derived from SRX560289 against all organisms resulted in 957 hits with an E value of ≤1 × e⁻²⁴ (31, 32). Among these, a single hit (GenBank accession no. CP014320.1) included the dfrB4 mobile genetic element. It was a 130.6-kb plasmid from a clinical isolate of a patient with recurrent E. coli sequence type 131 (ST131) cystitis, the most prevalent E. coli lineage found in recurrent UTIs (33). Other hits included mismatches (≤15) and gaps (≤38) and excluded the dfrB4 gene sequence. This suggests that the dfrB4 mobile genetic element has recently been incorporated into the CP014320.1 multiresistance plasmid.

The DfrB4 protein confers TMP resistance.

The high sequence identity of the DfrB4 protein to the well-characterized TMP-resistant DfrB1 protein (Table 1) and the 6-mm TMP zone of inhibition for these clinical samples (S. Salipante, personal communication) corresponding to an MIC of >16 μg/ml strongly argue that DfrB4 should confer TMP resistance to E. coli (34). To corroborate this hypothesis, the consensus dfrB4 coding sequence was obtained (GeneArt gene synthesis; Thermo Fisher Scientific) with an N-terminal hexahistidine affinity tag for eventual purification. It was cloned into pET24-cTEM19m between NdeI and HindIII, under IPTG (isopropyl-β-d-thiogalactopyranoside)-inducible overexpression in E. coli BL21(DE3). The MIC value was determined at least in duplicate through broth microdilution after incubation at 37°C for 14 to 16 h. We observed high resistance to TMP when expressing DfrB4 (>600 μg/ml, which is the maximal concentration of TMP that is soluble in 5% methanol) (35). The MICs for all negative controls were at least 2,000-fold lower, specifically E. coli BL21(DE3) (0.30 μg/ml) and E. coli BL21(DE3)/pET24-cTEM19m expressing an IPTG-inducible β-lactamase (0.075 μg/ml) (36). Thus, the MIC value of DrfB4 greatly surpasses the E. coli TMP resistance threshold (2 μg/ml) and is associated with 2.5% of all E. coli strains considered to exhibit dangerously high TMP resistance (≥512 μg/ml) (35, 37). In these experiments, DfrB4 was overexpressed; the expression level of DfrB4 in its native form is unknown.

To our knowledge, this work constitutes the first report of the type II TMP-resistant dfrB4 gene identified in a clinical sample by whole-genome sequencing. The 2,000-fold increase in MIC for TMP that we observed in a transformed E. coli isolate points to TMP resistance in the clinical isolate identified carrying the dfrB4 gene. The observation of this highly TMP-resistant DfrB4 within a known class I integron indicates that it is being disseminated, although the extent and rate of its propagation are currently unknown as it has not been systematically tracked. Our results highlight the importance of tracking the presence of this gene family in TMP-resistant clinical samples.