During October 19–27, 2012, we collected environmental water/sewage samples from 7 regions (58 sites) in Dhaka, Bangladesh (Figure 1). Control samples were from the United Kingdom. Each sample was investigated for bacterial growth on UTI brilliance agar plates (Thermo Fischer Scientific, Basingstoke, UK) containing vancomycin (30 mg/L) plus meropenem (0.5 mg/L). The species of individual colonies of different colors and morphologies were determined by using matrix-assisted laser desorption/ionization time-of-flight mass spectrometry. Bacteria were genetically characterized by bla_NDM-specific PCR. Genetic location of the bla_NDM-1 gene was determined by probing S1 nuclease pulsed-field gels. A subset of isolates of each species was further investigated for MICs of relevant antimicrobial drugs. All E. coli isolates were genotyped to determine multilocus sequence typing group; examples of each group were characterized for additional relevant resistance mechanisms. Details are provided in the Technical Appendix.

The carbapenemase and extended-spectrum β-lactamase genes bla_NDM-1 and bla_CTX-M-15 were detected by PCR in 36 (62%) and 41 (71%), respectively, of the 58 water samples. Both genes were found at all 7 sample region sites in Dhaka. Gene bla_CTX-M-15, but not bla_NDM-1, was detected in sewage samples from the United Kingdom; neither was detected in UK water samples from the River Thames.

We identified 226 gram-negative NDM-1–producing isolates to the species level (Figure 1; Technical Appendix Table 1); 15 isolates harboring bla_NDM-1 could not be identified and were not investigated further. The most widely disseminated bacteria in samples from Dhaka were pseudomonads (6/7 regions) and Klebsiella pneumoniae (4/7 regions). Nine different species of Pseudomonas spp. and 5 Acinetobacter spp., mostly belonging to nonpathogenic strains, were among the nonfermentative bacteria (Technical Appendix Table 1). Carbapenem resistance in the Pseudomonas spp. isolates was unstable; all strains lost the bla_NDM-1 gene after 2 days’ growth or when frozen for storage.

With the exception of 4 isolates, all bacterial isolates contained the original bla_NDM-1 allele; 3 E. coli sequence type (ST) 101 isolates carried the bla_NDM-3 variant, and 1 ST648 isolate carried the bla_NDM-4 variant (Technical Appendix Table 2). S1 nuclease pulsed-field gel electrophoresis combined with bla_NDM-1 probes detected bla_NDM-1 on plasmids of limited size diversity in E. coli (ST101, 160 kb; ST405, 100 kb; ST648, 150 kb); however, other species included bla_NDM-1–positive plasmids in a wide diversity of sizes (30 kb–450 kb); some of these species had multiple positive plasmids, and bla_NDM-1 was also found on the chromosome (Technical Appendix Table 1 and Figure 1).

The E. coli isolates were further analyzed by PCR to identify additional resistance mechanisms often associated with bla_NDM-1 (Table). bla_CTX-M-15 and 16s ribosomal methylase genes (armA or rmtB) were associated with most E. coli strains, which explains the extensively drug-resistant phenotype of the E. coli isolates (Technical Appendix Table 3). Plasmids of plasmid incompatibility groups incFII (ST101, ST405, ST648) and incX (ST405, ST648) were also closely associated with E. coli strains (Table). E. coli harboring bla_NDM-1 were isolated from 10 sampling sites (Figure 1; Technical Appendix Table 1). The E. coli isolates belonged to 3 different multilocus sequence typing groups: ST101 (phylogroup B1, 20/53 samples); ST405 (phylogroup D, 5/53 samples); and ST648 (phylogroup D, 28/53 samples) (Technical Appendix Table 2). ST101, which was found in samples from 6 (10.3%) of the 58 sites, was the most prevalent NDM-1–encoding E. coli genotype. ST648 represented an intermediate prevalence (5/58 [8.6%] sites), and ST405 was the least prevalent (1/58 [1.7%] sites) (Technical Appendix Tables 1, 2).

Our findings indicate that NDM-1 is widespread in the Dhaka environment. We detected 241 NDM-1–encoding bacterial isolates; they were found in all 7 sampled regions and at 36 (62%) of the 58 sampling sites. This high level of environmental bla_NDM-1 contamination is of concern, especially because drinking water in Bangladesh usually carries high levels of sewage-derived bacteria (11). It is therefore likely that bla_NDM-1 carriage rates will rise rapidly. Future environmental studies could provide indicators of epidemics of emerging resistant bacteria before they are realized in hospitals.

Despite the widespread presence of NDM-1 in Dhaka, it appears that this carbapenemase has recently emerged in the Bangladesh environment. Studies in northern Bangladesh did not find NDM-1 in wild ducks and poultry in 2009 (9) or in crow and gull feces in 2010 (10). Similarly, NDM-1 was not detected in drinking water in Dhaka during 2008–2009 (11) even though all samples had high levels of fecal and bla_CTX-M-15 contamination. Furthermore, a study of 1,879 clinical E. coli and Shigella spp. isolates collected during 2009–2010 in Bangladesh did not detect bla_NDM-1 (12). The first known clinical isolates date from 2008 (12), and the first evidence of human gut carriage of bla_NDM-1 was found in samples collected in Dhaka (13) a month before our study.

Because E. coli is the leading cause of human urinary tract infections, bloodstream infections, and neonatal meningitis, the ability of NDM-1 to give this bacterium clinical resistance to carbapenems is of concern (14). E. coli is also universally carried in the human gut. Therefore, we focused on this species because it is likely to be the greatest threat to human health. E. coli encoding NDM-1 were found in 3 of the 7 sampled regions, and genotyping showed they belonged to only 3 STs: ST648, ST101, and ST405. These same 3 E. coli genotypes are responsible for 80% of clinical NDM-1–encoding E. coli isolates in the United Kingdom (15). Furthermore, ST101 is the most common E. coli genotype in the Bangladesh environment (10.3% prevalence) and in clinical isolates from the United Kingdom (50%). Results of a literature search for NDM-1–encoding E. coli belonging to ST101 showed that this genotype has been detected in 15 nations (Figure 2). Thus, E. coli ST101 appears to be a successful global genotype that is often associated with NDM-1. This association with a single global genotype is analogous to the association between E. coli ST131 and the cephalosporinase CTX-M-15. Because of the critical nature of extensively drug-resistant bacteria, we are investigating the underlying factors responsible for the success of these particular antimicrobial drug–resistant strains.