High Prevalence of qnr Genes in Commensal Enterobacteria from Healthy Children in Peru and Bolivia

Abstract

A remarkable prevalence of qnrB (54%) and, at a lower level, of qnrS (14%) was discovered in pools of commensal enterobacteria from 310 healthy children living in Peru and Bolivia, using a metagenomic approach. Analysis of randomly selected enterobacterial pools revealed that qnrB was mainly carried by Escherichia coli and qnrS by Klebsiella pneumoniae. Investigation of 11 qnrB-positive isolates and 9 qnrS-positive isolates revealed the presence of plasmid-borne qnrB19 (n = 8), qnrB2 (n = 2), qnrB10 (n = 1), and qnrS1 (n = 9) genes.

Several plasmid-mediated quinolone resistance (PMQR) mechanisms have been discovered during the past decade, including Qnr proteins, QepA transporters, and the acetyltransferase AAC(6′)-Ib-cr (11, 13). Qnr proteins, the first discovered PMQR mechanism, are small pentapeptide repeat proteins that bind and protect type II DNA topoisomerases from inhibition by fluoroquinolones (17-19). Different lineages of Qnr proteins have been described (QnrA, QnrB, QnrS, and, more recently, QnrC and QnrD), with several allelic variants known for some of them (3, 9, 11, 20). qnr-like genes have also been detected on chromosomes from both gram-negative and gram-positive bacteria (9), and, recently, as class 1 integron gene cassettes in the chromosome of Vibrio cholerae (7). Although Qnr proteins only determine a moderate reduction of quinolone susceptibility, this may favor the selection of additional resistance mechanisms leading to higher-level quinolone resistance of clinical significance, and the dissemination of qnr genes and other PMQR determinants is believed to be an important promoter for evolution of quinolone resistance (11-13).

qnr genes have been reported worldwide, especially in enterobacteria. However, they have mostly been investigated in clinical isolates with specific resistance traits (e.g., quinolone resistance or extended-spectrum β-lactamase phenotype) (11 and references therein), while their prevalence in commensal bacteria remains largely unknown. In this study, we investigated the prevalence of qnr genes in commensal enterobacteria from healthy children by a PCR-based metagenomic approach. We also tested a simplified dot blot DNA hybridization method as a less-labor-intensive and expensive tool to perform the metagenomic analysis.

The analysis was carried out on commensal enterobacterial pools from 310 healthy children, ages 6 to 72 months, living in four urban areas of Latin America: two in Peru (Moyobamba, San Martin Department; and Yurimaguas, Loreto Department) and two in Bolivia (Camiri, Santa Cruz Department; and Villa Montes, Tarija Department). The enterobacterial pools, consisting of the bacterial growth obtained by plating fecal samples (one sample per child) onto MacConkey agar (MCA) plates (Oxoid, Milan, Italy), were randomly selected among samples (n = 3,193) obtained during a survey performed in 2005 in the same areas (1) and stored at −70°C. A loopful of each pool was plated on an MCA plate supplemented with 0.12 μg/ml ciprofloxacin (MCA-CIP). This ciprofloxacin concentration was used for screening purposes since (i) it was lower than MICs usually exhibited by enterobacterial strains harboring qnr genes as the sole quinolone resistance mechanism (reference 11 and references therein), while being higher than the wild-type MIC distribution for Escherichia coli and Klebsiella pneumoniae (http://www.escmid.org/research_projects/eucast/); (ii) a similar ciprofloxacin MIC threshold has previously been used for screening of qnr-positive bacteria (4, 8, 14, 15, 21). In case of growth onto MCA-CIP, a loopful of bacteria was directly used for total DNA extraction (10), and about 100 ng of metagenomic DNA was used as template in PCRs (50 μl) to detect the presence of qnr genes. Primers and conditions for PCR amplification of qnr genes were described previously in references 14 (qnrA and qnrS) and 2 (qnrB). Controls for qnr genes were kindly provided by Patrice Nordmann and Laurent Poirel (Université Paris-Sud, K.-Bicêtre, France). The specificity of the PCR products was confirmed by partial sequence analysis of randomly selected amplicons (Macrogen, Seoul, Korea).

The presence of qnr genes in the enterobacterial pools was also investigated by dot blot DNA hybridization using a rapid method, essentially as described by Srinivasan et al. (16). Briefly, a loopful of the bacterial growth on MCA-CIP was transferred to 150 μl of lysis solution (0.4 N NaOH, 10 mM EDTA) and incubated at 70°C for 2 h. The bacterial lysate (100 μl) was directly blotted onto Hybond-N+ nylon membranes (Amersham Bioscience, Buckinghamshire, United Kingdom) using a BIO-dot microfiltration apparatus (Bio-Rad Laboratories, Milan, Italy). Nylon membranes were washed twice with 2× SSC (1× SSC is 0.15 M NaCl plus 0.015 M sodium citrate), fixed by UV light, and hybridized with digoxigenin-dUTP (DIG)-labeled qnr probes (generated by PCR using the positive controls, as described above) using the DIG system according to the manufacturer's instructions (Roche Diagnostics SpA, Milan, Italy).

Growth of enterobacterial pools on MCA-CIP was observed with 275 of 310 samples (89%). Analysis of the metagenomes prepared from this bacterial growth by PCR revealed overall positivities of 54% for qnrB and 14% for qnrS, while qnrA was not detected (Table 1). Dot blotting showed signals of variable intensity (Fig. 1) and an overall lower detection sensitivity, but results were fully consistent with those of PCR (i.e., all positive samples in the dot blot, including weak signals, were also positive by PCR) (Table 1). This evidently reflected a variable copy number of target genes in the metagenomic DNA from various samples, with a number of cases in which the amount of target genes could only be detected by the more sensitive gene amplification approach.

TABLE 1.

Prevalence of qnr genes in commensal enterobacteria from 310 healthy children living in Peru and Bolivia^a

Study area	No. of samples	No. (%) of samples grown on MCA-CIPb	No. of samples (% [95% CI]) forc:
			Any qnr		qnrB		qnrS
			PCR	Dot blot	PCR	Dot blot	PCR	Dot blot
Peru	164	154 (94)	113 (69 [61-76])	93 (57 [49-64])	107 (65 [57-72])	87 (53 [45-61])	37 (23 [16-30])	11 (7 [3-12])
Bolivia	146	121 (83)	63 (43 [35-52])	51 (35 [27-43])	60 (41 [33-50])	49 (34 [26-42])	7 (5 [2-10])	2 (1 [0-5])
Total	310	275 (89)	176 (57 [51-62])	144 (46 [40-52])	167 (54 [48-60])	136 (44 [38-50])	44 (14 [11-19])	13 (4 [2-7])

^aStatistical differences were determined by the chi-square test with the EpiInfo software package, version 6 (Centers for Disease Control and Prevention, Atlanta, GA). The binomial exact 95% confidence interval (95% CI) was calculated by Stata Software release 8.0 (2003; StataCorp LP, College Station, TX).

^bRates of growth on MCA-CIP plates were not significantly different (P > 0.5).

^cThe percentage of samples was calculated for the total samples. qnr genes were significantly more prevalent in Peru than in Bolivia: P = 0.01 for any qnr and qnrB, and P < 0.001 for qnrS. All positive samples in the dot blot were also positive in the PCR.

FIG. 1.

Nylon membrane prepared with bacterial lysates, hybridized with DIG-labeled qnrB probe. All positive signals, including the weakest ones, were positive in PCR experiments. Positive and negative controls (C+ and C−, respectively) are indicated by arrows.

Concerning geographical differences, even though the rates of growth of enterobacterial pools on MCA-CIP were not significantly different, qnr genes were significantly more prevalent in Peru than in Bolivia (Table 1). Interestingly, these differences mirrored the higher quinolone resistance rates found in the commensal E. coli microbiota of the same population of children from Peru (62% for nalidixic acid and 39% for ciprofloxacin) as compared to Bolivia (51% for nalidixic acid and 26% for ciprofloxacin) (1), which supports an association between dissemination of PMQR determinants and quinolone resistance.

To investigate the nature of the bacterial hosts of qnr genes in the commensal microbiota, enterobacterial pools yielding metagenomes positive for qnrB (n = 42) or qnrS (n = 22), selected at random, were plated on MCA-CIP to yield isolated colonies. All colonies with a different morphological appearance were collected and subjected to molecular analysis to investigate the presence of qnrB and qnrS genes by PCR (up to four colonies were analyzed per sample). All qnr-positive isolates were identified by the API 20E system (BioMérieux, Marcy l'Etoile, France). When two or more isolates of the same species and carrying the same qnr gene were observed in a sample, only one isolate was included in the data analysis.

Isolates carrying qnrB or qnrS were detected in 36 of 42 (86%) and 14 of 22 (64%) of the selected qnrB- and qnrS-positive enterobacterial pools, respectively (Table 2). The lack of recovery of qnr-positive isolates from some samples was likely due to a lower frequency of such isolates in those enterobacterial pools. In fact, the success in isolating qnr-positive isolates from the enterobacterial pools was consistently higher when the corresponding metagenome had been found qnr positive also by dot blotting (89% versus 47%) (Table 2), in agreement with the hypothesis that, in those cases, qnr genes were carried by a dominant bacterial population.

TABLE 2.

Detection and identification of qnr-carrying bacterial hosts from enterobacterial pools yielding qnr-positive metagenomes

Enterobacterial pool (no. of samples)	No. (%) of samples in which qnr hosts were identified	No (%) of qnr bacterial hosts
		E. coli	K. pneumoniae	Othersa
qnrB positive
PCR and dot blot positive (35)	32 (91)	22 (69)	5 (16)	5 (16)
PCR positive only (7)	4 (57)	4 (100)
Total (42)	36 (86)	26 (72)	5 (14)	5 (14)
qnrS-positive
PCR and dot blot positive (12)	10 (83)	3 (30)	7 (70)
PCR positive only (10)	4 (40)	1 (25)	1 (25)	2 (50)
Total (22)	14 (64)	4 (29)	8 (57)	2 (14)

^aqnrB was detected in E. cloacae (n = 2), K. oxytoca (n = 1), C. freundii (n = 1), and E. hermannii (n = 1). qnrS was detected in K. oxytoca (n = 2).

Identification of the qnr-positive isolates showed that qnrB was mainly carried by E. coli and, more rarely, by K. pneumoniae or other enterobacterial species, including Enterobacter cloacae, Klebsiella oxytoca, Citrobacter freundii, and Escherichia hermannii (Table 2). On the other hand, qnrS was mainly found in K. pneumoniae and, more rarely, in E. coli and K. oxytoca (Table 2). In two cases, both qnr genes were found to be carried by the same E. coli isolate.

MICs of nalidixic acid, ciprofloxacin, and levofloxacin were determined by agar dilution and interpreted according to CLSI (5, 6) for the 48 qnr-harboring isolates. Resistance to nalidixic acid was common (77%), while 32% and 17% of isolates were nonsusceptible to ciprofloxacin and levofloxacin, respectively (analytical data are reported in Table S1 in the supplemental material).

The location and nature of qnr genes were investigated in 20 selected isolates representative of different species (11 qnrB and 9 qnrS) by Southern blotting on nylon membranes, as described for dot blot hybridization, and by sequencing of PCR amplicons generated with primers designed on flanking sequences (EU624315 and EU715254 for qnrB and EU939771 for qnrS). qnrB genes were located on either low- or high-molecular-weight plasmids and included mostly qnrB19 but also qnrB2 and a new allele of qnrB10 showing 7 nucleotide differences compared to qnrB10 DQ631414. All qnrS genes were qnrS1 and were located on high-molecular-weight plasmids (Table 3; and data not shown).

TABLE 3.

qnr genes in commensal enterobacteria from healthy children in Peru and Bolivia

Isolatea	Origin	qnr gene	MIC (μg/ml)b
			NAL	CIP	LEV
E. coli Y1	Yurimaguas, Peru	qnrB19	32	0.5	1
E. coli M1	Moyobamba, Peru	qnrB19	16	0.25	0.5
E. coli C1	Camiri, Bolivia	qnrB19	32	0.5	1
E. coli V1	Villa Montes, Bolivia	qnrB19	32	0.25	1
K. pneumoniae Y1	Yurimaguas, Peru	qnrB19	64	2	1
K. pneumoniae M1	Moyobamba, Peru	qnrB19	32	2	2
K. pneumoniae V1	Villa Montes, Bolivia	qnrB2	32	1	1
E. hermannii C1	Camiri, Bolivia	qnrB19	32	0.25	0.25
E. cloacae V1	Villa Montes, Bolivia	qnrB2	32	2	2
K. oxytoca M1	Moyobamba, Peru	qnrB19	16	0.25	0.5
C. freundii V1	Villa Montes, Bolivia	qnrB10	16	0.25	0.5
E. coli Y2	Yurimaguas, Peru	qnrS1	16	0.5	0.5
E. coli Y3	Yurimaguas, Peru	qnrS1	16	0.5	0.5
E. coli M2	Moyobamba, Peru	qnrS1	16	0.5	0.5
E. coli M3	Moyobamba, Peru	qnrS1	32	0.5	2
K. pneumoniae Y2	Yurimaguas, Peru	qnrS1	32	1	2
K. pneumoniae M2	Moyobamba, Peru	qnrS1	32	4	1
K. pneumoniae C2	Camiri, Bolivia	qnrS1	32	2	2
K. pneumoniae V2	Villa Montes, Bolivia	qnrS1	16	1	2
K. oxytoca M2	Moyobamba, Peru	qnrS1	>128	4	4

^aThe 11 qnrB-harboring isolates were selected as follows: four E. coli isolates, one from each study area; three K. pneumoniae isolates, each from different study areas (no qnrB-harboring K. pneumoniae isolate detected in Camiri, Bolivia); and four isolates representative of species other than E. coli or K. pneumoniae. The nine qnrS-harboring isolates were selected as follows: all of the qnrS-harboring E. coli isolates (n = 4); four K. pneumoniae isolates, one from each study area; and one of the two qnrS-harboring K. oxytoca isolates.

^bNAL, nalidixic acid; CIP, ciprofloxacin; LEV, levofloxacin.

Concluding remarks.

This study demonstrated a remarkable dissemination of qnrB and, at a lower level, qnrS determinants in commensal enterobacteria from healthy children living in urban areas of Peru and Bolivia. The few studies that have investigated the prevalence of qnr genes in clinical isolates from South American countries showed that alleles belonging to the qnrB are the most frequently reported (11 and references therein), in accordance with the high prevalence of qnrB observed in this study. To our best knowledge, this is the first study on the prevalence of qnr genes in human commensal bacteria, and present findings suggest that the commensal enterobacterial microbiota could be an important reservoir of similar genes. It will thus be interesting to investigate if a similar situation is also found in the adult population and in other geographical settings.

Since data collected about household use of antibiotics excluded previous use of fluoroquinolones in the children included in this study (1), selection of qnr genes could be related to linkage with other resistance genes carried on the same plasmids. Further studies are under way to investigate this issue.

Nucleotide sequence accession number.

The nucleotide sequence of the new allele of qnrB10 has been submitted to GenBank and assigned accession no. FJ769283.