NDM Metallo-β-Lactamases and Their Bacterial Producers in Health Care Settings

Distribution and Prevalence of NDM-Positive Strains in Health Care SettingsAfter the initial discovery of NDM-1, a follow-up study revealed the widespread existence of bla_NDM-1 in the Indian subcontinent, including India, Pakistan, and Bangladesh (33). Since then, NDM-positive strains have been shown to be globally distributed, with virtually all countries conducting epidemiological searches detecting NDM-positive strains (Fig. 2; a complete list of countries with documented NDM-positive strains is available in Appendix S1 in the supplemental material).

• Open in new tab
• Download powerpoint

FIG 2

Worldwide distribution of NDM-positive strains of the Enterobacteriaceae. Countries (Egypt, India, Pakistan, Serbia, and UAE) with evidence showing a prevalence of NDM-positive strains among the Enterobacteriaceae of ≥5% are indicated in red, while countries with reports of NDM-positive strains but without evidence of a ≥5% prevalence are shown in light brown. Countries without reports or data on NDM-positive strains are indicated in white.

The worldwide distribution of NDM-positive strains appears to be heterogeneous with regard to prevalence. The SMART global surveillance program collected 103,960 isolates of Enterobacteriaceae in 55 countries from 2008 to 2014 and demonstrated that 290 strains (0.28% of all strains) were NDM positive, suggesting a relatively low prevalence (34). In the SMART program, the prevalence of NDM-positive strains varied significantly across countries: up to 5.01% in the United Arab Emirates (UAE), 6.15% in Egypt, 6.22% in India, and 6.26% in Serbia (34). This supports the observation that NDM-positive strains are more highly prevalent in South Asia, the Balkans, North Africa, and the Middle East. The high prevalence of NDM-positive strains in the Middle East has been proposed to be a result of population exchange with the Indian subcontinent (8). INFORM is another large-scale multinational study, which collected 38,266 Enterobacteriaceae isolates and 8,010 P. aeruginosa strains from 40 countries between 2012 and 2014 (35). The proportions of NDM-positive strains were 0.19% (72/38,266) in Enterobacteriaceae and 0.04% (3/8,010) in P. aeruginosa strains (35), consistent with the relatively low prevalence revealed by the SMART project. Unfortunately, the prevalence of NDM-positive strains in individual countries was not provided in that study (35). Other than the SMART and INFORM global surveillance programs, there are very few appropriately designed large-scale studies to determine the true prevalence of NDM carriage in given species or genera at a national level. In Pakistan, 18.5% of hospitalized patients at two military hospitals were found to carry NDM-positive Enterobacteriaceae (36). In China, a study of 1,162 clinical isolates of Enterobacteriaceae and Acinetobacter spp. collected at multiple sites reported that 3.9% were NDM positive, but the proportion of NDM-positive strains among the Enterobacteriaceae remains unclear, as the exact number of isolates of Enterobacteriaceae was not given (37). More large-scale surveillance studies, in particular on strains collected after 2014, are required to reveal changing trends and changes in the prevalence of NDM-positive strains.

In the SMART global surveillance program, bla_NDM was the third most common carbapenemase-encoding gene and accounted for 19.42% of carbapenemase positivity, after bla_KPC (53.18%) and the bla_OXA-48-like gene (20.09%) (34). In China, which was not included in the SMART program, 31% of 1,105 carbapenem-resistant Enterobacteriaceae (CRE) strains were NDM positive (38). The EuSCAPE survey across Europe revealed that 7.7% of carbapenem-resistant K. pneumoniae (n = 1,203) and 10.3% of carbapenem-resistant E. coli (n = 194) isolates were NDM positive (39).

In addition to clinical samples, bla_NDM genes have also been detected in hospital sewage in several countries, including China (40, 41), India (42), and Lebanon (43). Some NDM-positive strains recovered from hospital sewage belong to the Enterobacteriaceae (42, 43), which may reflect intestinal carriage of NDM-positive strains among the population within health care settings. However, NDM-positive strains of various Acinetobacter species have also been recovered from hospital sewage (40, 41). This suggests that sewage may be a reservoir of bla_NDM- and NDM-positive strains. The links between hospital sewage and the spread of bla_NDM are yet to be established and require more studies. Nonetheless, hospital sewage should be properly treated according to existing guidelines and regulations (44).

Spread of NDM-Positive Strains and International TravelThe rapid spread of NDM from its initial emergence in India to all continents is significantly associated with global travel (33). The initial discovery of bla_NDM-1 in India, Pakistan, and the United Kingdom showed that almost all United Kingdom cases were associated with travel to the Indian subcontinent (33). Following the first report, the incidence of NDM cases rose sharply, with countries in the Mediterranean region of Europe showing the largest increase (45). A detailed study of the first reported cases of NDM in Europe showed that 57% of all cases were associated with previous hospitalization in the Indian subcontinent or Balkans region (46). The first reported outbreak in Europe was in Italy in 2011. The bla_NDM-1 gene was detected in both Klebsiella and E. coli strains isolated from clinical infections in a hospital in Bologna, Italy, with the index case being a patient initially treated for an infection due to NDM-positive bacteria in New Delhi, India, before traveling to Bologna, where further treatment was required (47). This initial seeding led to Italy having some of the highest rates of NDM cases in Europe by 2017 (39). By 2014, Greece was reporting sustained cases of hospital-associated infections caused by K. pneumoniae sequence type 11 (ST11) strains carrying bla_NDM-1 (48), believed to be first introduced via travel from the eastern Balkans region. International travel has also been associated with the movement of NDM into North America, with direct patient transfer from India to Canada resulting in the introduction of bla_NDM-1-carrying P. aeruginosa ST654 (49) as well as cases of bla_NDM-1-carrying E. coli and K. pneumoniae (50). Direct travel from India and then Iran was also implicated in the first cases of NDM isolation in the United States (51–53).

While the role of travel in observed clinical cases of infection with bla_NDM-1-carrying bacteria is clear, less is known of the role that travel may play in introducing NDM into the wider community. A number of high-quality studies have been performed, showing that travel to regions of endemicity, such as India and Southeast Asia, leads to significant levels of intestinal colonization by bacteria carrying extended-spectrum β-lactamase (ESBL) genes (54–56). However, to date, there has been only one study examining the risk of asymptomatic colonization by bla_NDM-1-carrying bacteria during travel (57). This small study of French travelers to India identified intestinal colonization by a bla_NDM-1-carrying E. coli strain and reported that colonization lasted for only 1 month, compared to as long as 8 months for ESBL-carrying E. coli (55). This suggests that there may be intrinsic differences in the abilities of bla_NDM-carrying bacteria to successfully outcompete intestinal microbiota and colonize the human intestinal tract, an area that merits full and intensive study.

Host Species of NDMTo obtain a comprehensive picture of the distribution of NDM across bacteria, we retrieved all bacterial genome sequences containing bla_NDM from GenBank (n = 766; accessed on 8 January 2018) (see Data Set S1 in the supplemental material), in addition to reviewing the available literature. bla_NDM genes have been found in species belonging to 11 bacterial families (Aeromonadaceae, Alcaligenaceae, Cardiobacteriaceae, Enterobacteriaceae, Moraxellaceae, Morganellaceae, Neisseriaceae, Pseudomonadaceae, Shewanellaceae, Vibrionaceae, and Xanthomonadaceae) of the class Gammaproteobacteria (Table 1). A bla_NDM gene has also been identified in the genome sequence (GenBank accession no. JPNZ00000000) of a strain of Bacillus subtilis, which is a Gram-positive bacterium of the family Bacillaceae. This is very unusual, and resequencing of this strain would be advised to exclude any possibility of sequence read contamination. According to the literature, the Enterobacteriaceae are the major hosts of bla_NDM, among which K. pneumoniae is the most common species, accounting for just over half of all isolates, followed by E. coli and the Enterobacter cloacae complex (Table 2). The Enterobacteriaceae are able to cause a variety of community-onset or hospital-acquired infections, such as abscesses, bloodstream infection, intra-abdominal infection, meningitis, pneumonia, and urinary tract infection (58).

Acinetobacter spp. are also frequently identified as hosts of bla_NDM (41). Acinetobacter strains carrying bla_NDM have been found in at least 25 countries in Africa (Algeria, Libya, Morocco, Egypt, Ethiopia, Kenya, and Tunisia), the Americas (Argentina, Brazil, Cuba, Honduras, and Paraguay), Asia (Bangladesh, China, India, Israel, South Korea, Lebanon, Malaysia, Palestine, and Thailand), and Europe (Croatia, Denmark, Greece, and Slovenia). Acinetobacter baumannii is a notorious opportunistic pathogen associated with hospital-acquired infections and pneumonia in particular (59). Surprisingly, bla_NDM is also found in at least 18 other Acinetobacter species, most of which have been recovered from sewage and are rarely associated with human diseases (Table 1). This suggests that Acinetobacter may play a vital role in the dissemination of NDM-encoding genes and raises the question of whether Acinetobacter species could be the origin of bla_NDM. This wide species distribution may also be due to multiple mechanisms facilitating the transfer of bla_NDM across bacterial populations. In addition to conjugation, which is the major mode of horizontal transfer of bla_NDM in the Enterobacteriaceae, it has been found that OMVs of A. baumannii are able to mediate the intra- and interspecies transfer of bla_NDM plasmids at high transformation frequencies (10⁻⁵ to 10⁻⁶ transformants in the total cell count [CFU per milliliter]) (23, 60). General transduction facilitated by prophages present in the chromosome can also mediate the horizontal transfer of bla_NDM between A. baumannii strains (61). More studies are required on this remarkable diversity of Acinetobacter strains carrying bla_NDM and its relevance to the successful emergence of NDM across Gram-negative bacteria.

Among the 24 NDM variants, NDM-1 has the widest host spectrum identified so far and has been found in a number of species belonging to 11 bacterial families. Publicly available genome sequences also reveal that most NDM-positive Acinetobacter species, Enterobacter species, and K. pneumoniae isolates have NDM-1, while NDM-5 is most common in E. coli. NDM is rare in P. aeruginosa, with VIM being the most common MBL in this species (35).

Clonal Background of NDM-Positive StrainsThere is a limited number of large-scale studies that have examined the clonal background of NDM-producing strains. These include one study involving multiple nations (62), one study across three countries (63), three studies conducted at a national level (38, 64, 65), and two municipal-level studies (66, 67) (see Table S2 in the supplemental material). While studying genomes deposited in GenBank can result in inherent bias, it is also the most comprehensive approach available to provide additional insights into the clonal background of NDM-positive strains. Among the 766 NDM-positive bacterial genomes available in GenBank, E. coli (n = 305), K. pneumoniae (n = 214), Acinetobacter spp. (n = 84), and Enterobacter spp. (n = 67) are the most common (Data Set S1). NDM-positive E. coli and K. pneumoniae strains are distributed across more than 40 STs for each species. NDM-positive Enterobacter and Acinetobacter strains are also distributed across various species and multiple STs (Data Set S1). This suggests heterogeneous clonal backgrounds of NDM-positive strains and multiple acquisitions of bla_NDM genes across bacterial species. Despite this heterogeneity, a small number of STs of the Enterobacteriaceae and A. baumannii have been identified to be the most common carriers of bla_NDM (see below). These STs warrant further investigation to identify high-risk clones mediating the international spread of bla_NDM genes as well as determine underpinning factors that may make a clone more likely to emerge as a successful multidrug-resistant (MDR) pathogen.

NDM-positive E. coli strains belong to a variety of STs, and there are no predominant STs. NDM has been found in strains of E. coli ST131 (68–70), the pandemic clone mediating the global spread of the ESBL gene bla_CTX-M-15 (71, 72). However, NDM-positive ST131 strains remain uncommon (Table S2 and Data Set S1). In contrast, ST167 is relatively common among NDM-positive E. coli strains (Table S2 and Data Set S1) and accounted for 14.4% (44/305) of the 305 NDM-positive E. coli genomes available (Data Set S1). ST167 has been detected in multiple countries (India, Niger, South Africa, South Korea, Switzerland, and the United States) and throughout China and has been predominantly recovered from humans (Data Set S1). ST617 and ST410 are two other types of E. coli strains seen in multiple countries, although they are less common than ST167 (Data Set S1), and both types have sequenced isolates recovered from animals in addition to humans. In a multiple-site study in China, ST167 was the most common type of NDM-positive E. coli strain, followed by ST410 (38). Our previous study also suggested that E. coli ST167 and ST617 appear to be two major types of globally disseminated, NDM-positive E. coli strains (73). More studies are required to investigate whether ST167, ST410, and ST617 are international epidemic clones of NDM-positive E. coli.

NDM-positive K. pneumoniae strains are also distributed across a large number of STs, with no predominant lineages, suggesting that there are no obvious high-risk clones of NDM-positive K. pneumoniae. This is in contrast to KPC-positive K. pneumoniae, whose global spread is largely due to clonal complex 258 comprising ST258 and ST11 (74, 75). ST11, ST14, ST15, and ST147 strains are relatively common NDM-positive K. pneumoniae lineages and have been found in multiple countries across several continents, almost all of which were isolated from humans (Data Set S1). In the literature, ST14 is repeatedly reported as one of the most common types of NDM-positive K. pneumoniae strains (63–65). ST11 is another common type in multiple studies (38, 63–65). Of note, ST11 is the predominant ST of carbapenem-resistant K. pneumoniae in China but mainly carries KPC-2 rather than NDM (76). Although the currently available evidence is insufficient to demonstrate that ST11, ST14, ST15, and ST147 are truly epidemic clones mediating the international spread of bla_NDM, their distribution in multiple countries warrants further study. A multiple-site study reported that ST23 is the most common type of NDM-positive K. pneumoniae strain in China (38). However, ST23 is rarely seen in other countries, and there is only one ST23 genome in GenBank, isolated in China. This suggests that ST23 may be largely restricted to China (38). The well-known international epidemic carbapenem-resistant K. pneumoniae ST258, which carries bla_KPC-2 or bla_KPC-3, has not yet been found to carry bla_NDM (77).

Most NDM-positive strains of Enterobacter spp. are either Enterobacter xiangfangensis or Enterobacter hormaechei strains (Data Set S1). ST78 and ST171 have been reported as two emerging lineages of carbapenem-resistant Enterobacter spp., but strains of these lineages usually produce KPC rather than NDM (78). There are only three NDM-positive ST78 (belonging to E. hormaechei) and three NDM-positive ST171 (belonging to E. xiangfangensis) genomes in GenBank (Data Set S1). ST114 (belonging to E. hormaechei) strains have been found in multiple countries (Data Set S1). A study on an international collection of carbapenem-resistant Enterobacter strains demonstrated that ST114 is the most common type of NDM-positive Enterobacter strain, although it accounts for only a minority of all strains due to the very diverse clonal background of NDM-positive Enterobacter strains (62). Therefore, no obvious international epidemic clones of NDM-positive Enterobacter strains have been identified at present.

NDM-positive A. baumannii strains are very genetically diverse with respect to ST lineage (67), with ST85 being the most commonly isolated (11 genomes in GenBank) (Data Set S1) across several countries. NDM has also been found in P. aeruginosa in multiple countries (35, 49, 79, 80). Whole-genome sequences exist for a small number of P. aeruginosa strains from lineage ST308, all isolated in Singapore (Data Set S1). NDM-positive ST308 P. aeruginosa has also been detected in neighboring Malaysia (81), suggesting that this type of NDM-positive P. aeruginosa strain may be circulating in the region.

Plasmids Carrying bla_NDMAlthough bla_NDM has been found on bacterial chromosomes (82–84), the vast majority of carriage occurs on plasmids, which play a vital role in dissemination. bla_NDM has been reported to be carried on plasmids with a variety of replicon types (33, 82, 85–104). There are a total of 355 bla_NDM-carrying plasmids with complete sequences available in GenBank (accessed on 8 January 2018) (see Data Set S2 in the supplemental material). We determined their replicon types using PlasmidFinder (https://cge.cbs.dtu.dk/services/PlasmidFinder/). There are 20 replicon types of bla_NDM-carrying plasmids in the Enterobacteriaceae, including IncC, IncB/O/K/Z, IncFIA, IncFIB, IncFIC, IncFIII, IncHI1, IncHI2, IncHI3, IncN, IncN2, IncL/M, IncP, IncR, IncT, IncX1, IncX3, IncX4, IncY, and ColE10 (Table 3) (e.g., see references 33, 82, and 85–104). This suggests multiple acquisitions of bla_NDM by various plasmids and also highlights that the horizontal transfer of bla_NDM is mediated by multiple plasmids. The global distribution of the replicon types of bla_NDM-carrying plasmids is shown in Fig. 3.

• Open in new tab
• Download powerpoint

FIG 3

Worldwide distribution of the replicon types of bla_NDM-carrying plasmids in Enterobacteriaceae. Detailed information about the distribution of the replicon types of bla_NDM-carrying plasmids is available in Table 3 and Data Set S2 in the supplemental material.

IncX3 appears to be the most common type of plasmid carrying bla_NDM. Among the 355 bla_NDM-carrying plasmids available in GenBank, 117 (about one-third) had the IncX3 replicon, including 112 plasmids with IncX3 alone and 5 with IncX3 plus other replicons. IncX3 plasmids are narrow-host-range plasmids and have so far been seen only in the Enterobacteriaceae. Most of the IncX3 plasmids in GenBank (67/117; 57.3%) were present in E. coli, followed by K. pneumoniae (20/117; 17.1%). Although bla_NDM-carrying IncX3 plasmids have been found in Europe and North America, most of the plasmids deposited in GenBank (80/117; 68.4%) have been recovered in China and neighboring countries, such as South Korea (n = 7), Vietnam (n = 2), and Myanmar (n = 3). This suggests that IncX3 plasmids may be a major vehicle in mediating the dissemination of bla_NDM in East Asia, particularly in China. Variants of bla_NDM, including bla_NDM-1, bla_NDM-4, bla_NDM-5, bla_NDM-6, bla_NDM-7, bla_NDM-13, bla_NDM-17, bla_NDM-19, bla_NDM-20, and bla_NDM-21, have also been found on IncX3 plasmids (Table 3). This highlights that IncX3 plasmids may serve as one of the major platforms on which bla_NDM genes are evolving with the generation of new NDM variants.

There are 99 bla_NDM-carrying plasmids containing an IncFII replicon, alone or in combination with other types of replicons, commonly IncFIB in GenBank. A replicon sequence typing (RST) scheme is available for IncF plasmids (105). We performed RST for these plasmids using the pMLST tool (https://cge.cbs.dtu.dk/services/pMLST/). It is evident that IncFII plasmids of the FIA⁻:FIB36:FIIY4 allele type (n = 27) or the FIA⁻:FIB⁻:FII2 type (n = 21) are particularly common and have been found in various species of the Enterobacteriaceae from multiple countries in several continents (Data Set S2). These two IncFII plasmids have mainly been found in strains from human samples (Data Set S2).

Another common type of bla_NDM-carrying plasmid (53/355; 14.9%) is IncC (also incorrectly known as IncA/C2) (106). IncC plasmids carrying bla_NDM have a worldwide distribution and are found on all continents except Antarctica (Data Set S2). IncA/C has a broad host range, and IncC plasmids carrying bla_NDM have been found in the Morganellaceae and the Vibrionaceae in addition to the Enterobacteriaceae. A plasmid multilocus sequence typing (pMLST) scheme exists for IncA/C plasmids (https://pubmlst.org/plasmid/). Almost all IncC plasmids carrying bla_NDM belong to either ST1 (39/53) or ST3 (12/53) (Data Set S2). However, there are only 13 STs of IncA/C plasmids in the database, suggesting low diversity of these plasmids or low resolution of the scheme for typing such plasmids. A core genome pMLST (cgPMLST) scheme also exists for IncA/C plasmids (https://pubmlst.org/plasmid/). IncC plasmids of cgST1.2 are particularly common (n = 17) (Data Set S2) and have been found in several species of the Enterobacteriaceae and Morganellaceae from all continents except Antarctica. Plasmids of the above-mentioned IncX3 type, FIA⁻:FIB36:FIIY4 and FIA⁻:FIB⁻:FII2 types, and cgST1.2 of IncC warrant further investigation to understand their epidemiology, their true contribution to the NDM problem, and the mechanisms mediating their wide spread.

Plasmids carrying bla_NDM have been well documented in Acinetobacter (e.g., see references 107–110). A replicon typing scheme for plasmids of A. baumannii has been developed (111), but the replicon types of bla_NDM-carrying plasmids in Acinetobacter have rarely been reported, and the scheme has not been incorporated into PlasmidFinder or any other commonly used plasmid typing tools. Therefore, the replicon types of bla_NDM-carrying plasmids in Acinetobacter remain largely unknown. These plasmids vary significantly in size from 1,634 bp to 354,308 bp (Data Set S2), suggesting that multiple plasmids are involved in the spread of bla_NDM in Acinetobacter.

Genetic Contexts of bla_NDMMobile genetic elements, such as insertion sequences, transposons, and integrons, can mobilize antimicrobial resistance genes. This mobilization can be between different plasmids as well as between plasmids and the chromosome. bla_NDM has been found in a variety of genetic contexts, which suggests that multiple mechanisms have been involved in the mobilization of bla_NDM. The genetic contexts of bla_NDM share two common features. The insertion sequence ISAba125 (intact or truncated) is always upstream of bla_NDM, while a bleomycin resistance gene, ble_MBL, is always downstream. Further downstream of ble_MBL, there is usually a complete or remnant form of a set of several genes, including trpF (encoding a phosphoribosylanthranilate isomerase), dsbC (also called tat, encoding a twin-arginine translocation pathway signal sequence domain protein), cutA1 (also called dct, encoding a periplasmic divalent cation tolerance protein), and groES-groEL (encoding chaperonin), and the insertion sequence ISCR27. ISAba125 provides the −35 region of a promoter for the expression of bla_NDM-1 (112, 113). Another ISAba125 element has been found downstream of ISCR27 in Acinetobacter, and the two ISAba125 elements form a composite transposon carrying bla_NDM-1, termed Tn125 (Fig. 4A). It appears that the genetic components within Tn125 have different origins, as the groES-groEL-ISCR27 section may originate from Xanthomonas spp. (114, 115). The exact origin of bla_NDM-1 remains unknown. Careful examination of the genetic context and sequence of bla_NDM-1 reveals that bla_NDM-1 is a chimeric gene. The first 19 bp of nucleotide sequence (encoding the first 6 amino acids of NDM-1) originate from an aminoglycoside resistance gene, aphA6. The remaining nucleotide sequence originates from a yet-to-be-identified preexisting MBL gene (116). As ISAba125 and aphA6 are widespread in Acinetobacter spp. (117, 118), it is very likely that the fusion of genes to form bla_NDM-1 occurred in Acinetobacter (116). ISCR elements are known to acquire and accumulate genetic components via a rolling-circle mechanism (119, 120). As an ISCR element, ISCR27 may have initially acquired the progenitor gene of bla_NDM-1 (115, 116, 121) and mobilized the gene together with ble_MBL, trpF, dsbC, cutA1, and groES-groEL into aphA6 (downstream of a copy of ISAba125), allowing fusion and the formation of bla_NDM-1 (115, 116). The second copy of ISAba125 then inserted downstream of ISCR27 to form the ISAba125-based composite transposon Tn125 (116).

• Open in new tab
• Download powerpoint

FIG 4

Examples of genetic contexts and mobilization mechanisms of bla_NDM. (A) Tn125 formed by two copies of ISAba125. (B) Composite transposon formed by two copies of IS26. (C) Composite transposon formed by two copies of IS903. orf, open reading frame. (D) Element formed by two copies of the TIME. (E) Composite transposon formed by two copies of IS3000. (F) Genetic contexts containing ISCR1. This element is also flanked by two copies of IS26. The plasmid names and GenBank accession numbers are shown. Δ represents truncated genes or elements.

Acinetobacter spp. serve as the intermediate source for the mobilization of bla_NDM into the Enterobacteriaceae (116, 122). The bla_NDM-1-carrying element Tn125 has been interrupted or truncated in Enterobacteriaceae, to generate a variety of complex genetic contexts for bla_NDM. This interruption is largely due to the insertion of many other mobile genetic elements (e.g., IS1, IS5, IS26, IS903, ISEc33, and ISKpn14, etc.) and recombination. The flanking sequences of various remnants of Tn125 have also formed mechanisms involved in the mobilization of bla_NDM-1. These mechanisms included a number of composite transposons formed by two copies of the same insertion sequence, such as IS26 (123) (Fig. 4B), IS903 (124) (Fig. 4C), and IS3000 (termed Tn3000) (Fig. 4F) (125). The duplication of bla_NDM on the same plasmid is due to the action of an IS26-formed composite transposon (94). The mobilization of bla_NDM may be associated with another ISCR element, ISCR1 (Fig. 4E) (95, 126). Two tandem copies of bla_NDM-1 genes have been found on the chromosomes of both an ST167 E. coli strain in China (127) and a P. aeruginosa strain in Serbia (128). In both cases, the tandem copies of bla_NDM-1 are associated with ISCR1, which uses a rolling-circle mechanism of transposition and may generate a tandem duplication of its mobilized sequence via homologous recombination (119). Tn3-derived inverted-repeat transposable elements (TIMEs), which were previously described as miniature inverted-repeat transposable elements (MITEs) (94, 129), have also been found to mobilize bla_NDM (Fig. 4D) (95). TIMEs are a type of mobile genetic element bounded by 38-bp inverted repeats characteristic of the Tn3 family but lacking the transposase gene tnpA and, usually, the resolvase gene tnpR (130). Two copies of the TIME can form a composite transposon-like element, which is able to mobilize the intervening genetic components in the presence of the external Tn3-like transposase (94, 129).

Genetic contexts similar to those of bla_NDM-1 are shared in other bla_NDM variants, which are commonly associated with ISAba125 (intact or truncated) upstream and ble_MBL downstream (29, 95, 123, 131–135). This suggests that other bla_NDM variants emerged from bla_NDM-1 via nucleotide mutations. Due to the highly mobile nature of bla_NDM, the gene can also be lost by bacterial cells. Loss can be due to the deletion of DNA fragments as a result of insertion sequence and transposon activity, such as recombination (136, 137), or the complete loss of the bla_NDM-carrying plasmids (138).

Detection of Carbapenemase ActivityNew methods and tools are continuously being introduced for the detection of carbapenemase activity. A number of phenotypic methods to detect carbapenemase activity have been developed. These include the modified Hodge test (MHT) (139), the Carba NP (CNP) test (140) and its variants, the β-Carba test (141, 142), the carbapenem inactivation method (CIM) (143), the modified carbapenem inactivation method (mCIM) (144), matrix-assisted laser desorption ionization–time of flight mass spectrometry (MALDI-TOF MS) (145), isothermal titration calorimetry (ITC) (146), and UV spectrophotometry (147). The MHT, CNP, and mCIM have been extensively evaluated, and the latter two are currently recommended by the CLSI for detecting carbapenemases in carbapenemase-producing Enterobacteriaceae (CPE) and carbapenemase-producing P. aeruginosa (148). The MHT had been recommended by the CLSI for epidemiological or infection control purposes since 2009 (149), but the recommendation was removed in 2018 due to the availability of newer tests (CNP and mCIM) with higher accuracy (148). None of these phenotypic tests are specific for NDM, as they are designed to detect all carbapenemases, including class A (e.g., KPC) and class D (e.g., OXA-48) carbapenemases and other MBLs (e.g., IMP and VIM). We do not review these methods in detail here but refer the reader to the latest review on phenotypic methods for detecting carbapenemases (150).

Detection of NDM-Encoding GenesMolecular techniques are the reference methods for detecting carbapenemase genes, including bla_NDM, due to their excellent sensitivity and specificity and robust performance (187, 205). Molecular techniques are mostly based on PCR, but whole-genome sequencing (WGS) is being increasingly used. The main limitations of molecular technologies are the high costs and the requirement for trained technicians (184, 206).

Aztreonam-Avibactam or Aztreonam Combined with Ceftazidime-AvibactamAlthough aztreonam is stable against MBLs, NDM-producing strains usually have ESBLs and/or AmpC enzymes that are able to hydrolyze aztreonam. Aztreonam alone therefore has limited clinical utility against NDM-producing strains. Avibactam is a non-β-lactam β-lactamase inhibitor with the ability to inhibit most serine β-lactamases, such as class A (e.g., KPC, CTX-M, TEM, and SHV), class C (AmpC), and some class D (e.g., OXA-48) enzymes. However, avibactam has no ability to inhibit MBLs, including NDM, and cannot effectively protect β-lactams from the hydrolysis of MBLs (245). The combination of avibactam with aztreonam can protect the latter from the hydrolysis of ESBLs and AmpC and therefore can expand the spectrum of aztreonam. The combination of aztreonam and avibactam has demonstrated potent in vitro activity against NDM-positive ESBL-producing Enterobacteriaceae (246). Aztreonam-avibactam is currently in clinical development and is not available for clinical use. Ceftazidime-avibactam has been approved for clinical use and is highly effective against KPC-positive CRE but has no activity against NDM-positive strains. In vitro studies have demonstrated synergistic activity and a bactericidal effect of the combination of ceftazidime-avibactam and aztreonam against CRE (247, 248). A small case series of 10 patients revealed that 7 of the patients survived after receiving the combination of aztreonam and ceftazidime-avibactam. The other three patients died as a result of other underlying conditions and therefore should not be considered a failure of the combination treatment (249). Several case reports have also demonstrated that the combination of aztreonam and ceftazidime-avibactam can successfully treat infections with NDM-positive CRE strains (247, 250). This combination could be considered if there are no alternative therapeutic options, and its efficacy warrants further, large-scale studies.

Polymyxins (Colistin and Polymyxin B) Alone and in CombinationPolymyxins are the current mainstay choice of antimicrobial agents against CRE and carbapenem-resistant A. baumannii, including NDM-positive strains. The treatment of infections due to multidrug-resistant organisms (MDRO) with polymyxins has been reviewed previously (251–253), and there are many studies on the efficacy of colistin against CRE, most of which have KPC rather than NDM (254, 255). In this review, we focus on treatment against NDM-positive strains only.

Polymyxin E (colistin) and polymyxin B have been evaluated for the treatment of infections caused by NDM-positive strains. NDM-positive strains are usually susceptible to polymyxins. There is a case report that colistin alone has been used successfully for treatment of a patient with a polymicrobial infection, including an NDM-positive Vibrio cholerae strain, with an increased dose (from 1 million units [MU] to 2 MU three times a day) and monitoring serum concentrations (256). However, an in vitro time-kill assay revealed that although initial killing of bacterial cells could be achieved by colistin alone, considerable regrowth occurs at 24 h, and resistant subpopulations are frequently detected after exposure to colistin alone (257). A murine infection model demonstrated that unlike in combination with amikacin, colistin alone is unable to achieve 1.5- to 2.8-log₁₀ killing after 24 h of therapy (258). Another study using murine models revealed that colistin alone was inappropriate for treating pneumonia due to NDM-positive K. pneumoniae, although the strain was susceptible to colistin in vitro (259). The use of colistin has also been hampered by the occurrence of renal toxicity (260, 261) and, to a lesser extent, neurological adverse effects (262). Many NDM-positive Enterobacteriaceae strains have become resistant to colistin by acquiring plasmid-borne mcr genes or have mutations/interruptions of chromosomal genes such as phoP-phoQ (encoding a two-component system) and mgrB (a regulator of phoP-phoQ) (263–265).

Evidence suggests that treatment with colistin-based combinations may offer a benefit compared to colistin alone. Colistin is usually recommended in combination with other agents for treatment (266). Although various colistin-based combinations have been examined in vitro and occasionally in vivo, it remains unclear what the best combination is, as studies are usually done with small sample sizes or single case reports, are heterogeneous in methodology, and have generated inconsistent findings.

A patient with complicated health care-associated cystitis due to an NDM-positive E. coli strain recovered after receiving colistin and rifampin (267). The combination of colistin, rifampin, and meropenem was successfully used to treat pyoderma caused by multiple microorganisms, including an NDM-positive E. coli strain, in a 49-year-old male patient (268). A patient with acute pyelonephritis due to NDM-positive P. aeruginosa recovered when treated with colistin combined with aztreonam (269). Polymyxin B in combination with aztreonam and meropenem rescued a 65-year-old patient with acute myeloid leukemia from bloodstream infection caused by an NDM-positive K. pneumoniae strain (270).

The colistin-tigecycline combination was not bactericidal against two NDM-positive colistin-susceptible K. pneumoniae strains in a 24-h time-kill assay (271). Another study revealed that the addition of tigecycline to colistin did not produce increased bacterial killing. Instead, it may cause antagonism at lower concentrations (263). In a study including 28 NDM-positive Enterobacteriaceae strains, in vitro synergistic activity was observed with colistin plus tigecycline in very rare cases (272).

The combination of colistin and fosfomycin achieved increased bacterial killing and decreased the chance of emergence of resistance compared to either agent alone in 6 NDM-positive Enterobacteriaceae strains (273). Synergistic or bactericidal activity is present for fosfomycin and colistin in a 24-h time-kill assay (271). However, in another study including 28 NDM-positive Enterobacteriaceae strains, in vitro synergistic activity was observed for colistin plus fosfomycin only in very rare cases (272).

A time-kill assay revealed that the addition of amikacin is able to restore the susceptibility of four NDM-positive and mcr-positive E. coli strains to colistin (258). However, another time-kill assay demonstrated that the combination of amikacin and polymyxin B failed to eradicate NDM-positive and mcr-positive E. coli strains but that the addition of aztreonam with amikacin and polymyxin achieved eradication (274).

TigecyclineThe susceptibility of NDM-positive Enterobacteriaceae strains to tigecycline varies significantly. Some studies report that these strains are mostly susceptible to tigecycline in vitro (275, 276), while a multicenter study in China found that most NDM-positive strains were nonsusceptible to tigecycline (38). Nonetheless, tigecycline does not have desirable pharmacokinetic properties and is a bacteriostatic agent. There are concerns that in vitro susceptibility to tigecycline may not be translated to in vivo efficacy (277). A previous study demonstrated that treatment with tigecycline was associated with high mortality rates (40.1%) when used for treatment of infections due to CRE (most of which are likely to have KPC rather than NDM), whereas the mortality rate for inactive therapy was 46.1% (278). Both in vivo and clinical studies on tigecycline against NDM-positive strains are scarce. In an in vivo study using a murine infection model, a high-dose tigecycline scheme was effective for treating pneumonia due to NDM-positive E. coli and K. pneumoniae and was more active than colistin (259). The combination of tigecycline and levofloxacin was used to successfully treat a patient with hospital-acquired pneumonia caused by NDM-positive Raoultella planticola (279). However, during treatment against NDM-positive strains, high-level tigecycline resistance can emerge rapidly (280). Nonetheless, these limited data fail to provide a convincing argument for or against the use of tigecycline for treating infections due to NDM-positive strains, and further studies are therefore required.

EravacyclineEravacycline is a novel synthetic fluorotetracycline with potency that is 2- to 4-fold higher than that of tigecycline in NDM-positive strains. In an in vitro analysis of 2,644 Gram-negative pathogens, eravacycline demonstrated excellent activity against 18 E. coli strains that had carbapenem resistance associated with OXA and NDM (281). Eravacycline was approved for treating adults with complicated intra-abdominal infections by the FDA in August 2018. Eravacycline is also currently being tested against complicated urinary tract infections in a clinical trial (ClinicalTrials.gov registration no. NCT01978938).

Dual CarbapenemsThe rationale for using two carbapenems together is to provide a competitive substrate for the β-lactamase. These studies are mainly aimed at CRE as a whole, most of which have serine-based β-lactamases, such as KPC and OXA-48, rather than NDM.

Carbapenem MICs for carbapenemase-producing K. pneumoniae isolates may vary within a broad range of values, from 0.12 to >256 mg/liter. This variation depends on the clonal background of the bacterial isolates and the type of carbapenemase produced. Isolates producing NDM usually have high carbapenem MICs (≥32 mg/liter) (33, 282). This has made dual-carbapenem-based treatment more challenging for NDM-positive strains.

Twenty carbapenem-resistant K. pneumoniae (CRKP) clinical strains, 6 of which were NDM positive, were tested with dual-carbapenem (any two of doripenem, ertapenem, imipenem, and meropenem) combinations. The data strongly support the hypothesis that dual-carbapenem combinations might be effective against KPC and OXA-48, but no synergy was observed for any of the NDM-positive strains (283), which might be due to the different action mechanism of MBLs compared to those of serine-based carbapenemases (15). NDM-positive Enterobacteriaceae were tested in immunocompetent and neutropenic murine thigh infection models using humanized regimens of standard (500 mg given every 8 h) and high-dose, prolonged infusion (2 g given every 8 h; 4-h infusion) of doripenem and 1 g of ertapenem given intravenously every 24 h. Doripenem and ertapenem demonstrated efficacy against several NDM-positive strains, especially using high-dose and prolonged infusion (284). The findings in this in vivo study are inconsistent with those of the in vitro study (283). Dual carbapenems (meropenem and ertapenem) plus fosfomycin were used to successfully treat urinary tract infections caused by NDM-positive Enterobacteriaceae in two patients (285). In general, there are very few studies of dual carbapenems against NDM-positive strains, and the current evidence is contradictory. More studies are therefore required.

AminoglycosidesSusceptibility to aminoglycosides may be unpredictable and can vary according to the strain type. Aminoglycosides may be considered a viable option for combination therapy against NDM-positive strains. Plazomicin is a synthetic derivative of sisomicin that evades many aminoglycoside-modifying enzymes but is not active against bacterial strains having 16S rRNA methyltransferases (286). However, NDM-positive strains are usually resistant to plazomicin (MIC ≥ 64 mg/liter) (287). Apramycin is of the 4-monosubstituted deoxystreptamine (DOS) subclass and is active against ribosomes modified by all 16S rRNA methyltransferases except NpmA (288), which is not common in the Enterobacteriaceae. An in vitro study demonstrated that almost 90% of NDM-positive strains are susceptible to apramycin (289). However, apramycin is a veterinary agent and has not been approved for clinical use, likely due to its narrow therapeutic index (287). Apramycin therefore warrants further investigations as a repurposed agent against CRE, including NDM-positive strains.

FosfomycinFosfomycin is available as an oral agent in the United States and is also available for intravenous use in Europe and China. Few large-scale studies have addressed the susceptibility of NDM-positive Enterobacteriaceae to fosfomycin (267). Sufficient data are lacking to support the use of fosfomycin alone, but fosfomycin-containing combination therapy has demonstrated promising results, as mentioned above (273).

Other Non-β-Lactam Serine β-Lactamase Inhibitors in DevelopmentRelebactam (MK7655) (300) and nacubactam (RG6080 or OP0595) (301) are two non-β-lactam β-lactamase inhibitors (see Fig. S1 in the supplemental material), like avibactam, and can inhibit serine β-lactamases but cannot inhibit MBLs. Relebactam and nacubactam are being developed in combination with imipenem and meropenem, respectively, and imipenem-relebactam has successfully completed a phase 3 trial. Although relebactam and nacubactam have no activity against NDM-positive strains, their combination with aztreonam may have potential, like aztreonam-avibactam (see above), against NDM-positive strains.

MBL InhibitorsAn attractive strategy to combat NDM-positive bacteria is to develop MBL (including NDM) inhibitors. A number of NDM inhibitors of various classes have been identified and characterized, but none of the compounds have been approved for clinical use thus far. The inhibitors have been summarized in several excellent reviews (19, 302, 303). In this review, we summarize these inhibitors, including those that have been developed in the past 3 years, in Table 4.

Many of the NDM inhibitors have been tested only in vitro. However, several, including ANT431 (304), aspergillomarasmine A (305), calcium disodium EDTA (306), and colloidal bismuth subcitrate (CBS) (an anti-Helicobacter pylori drug) (307), have also been studied using in vivo models and have been shown to be efficacious. None of the NDM inhibitors have been approved for clinical use yet.

Antimicrobial Adjuvants in DevelopmentWith limited treatment options available, we are in urgent need of new therapeutic options. One approach to combat this growing problem is the use of combinations containing adjuvants. Antimicrobial adjuvants include not only antimicrobial agents but also compounds from plants.

Immunotherapeutic AgentsNonantimicrobial immunotherapeutic agents represent a new, alternative way to treat infections due to NDM-positive strains. Two recombinant human cysteine proteinase inhibitors, cystatin 9 (rCST9) and cystatin C (rCSTC), have been tested using murine pneumonia models of infection with NDM-positive K. pneumoniae (313). That study demonstrated that the combination of rCST9 and rCSTC led to significantly improved survival compared to that of infected mice treated with one of the inhibitors or without treatment (313). This suggests that rCST9-rCSTC is a promising therapeutic candidate for treating bacterial pneumonia (313).

OutbreaksMany outbreaks caused by NDM-producing strains of Enterobacteriaceae and A. baumannii have been reported in the literature (Table 5). Many reports describe only the microbiological aspects of the outbreak and fail to provide full information regarding outbreak investigations. The source of outbreaks has often been identified as a contaminated near-patient environment or communal medical device (Table 5). Outbreaks have also been seen in patients sharing rooms with those infected or colonized by NDM-positive strains (314, 315).

Outbreaks are generally controlled by strengthening commonly used infection control measures, such as contact precautions, hand hygiene, environmental cleaning, isolation, and active surveillance, with or without implementing additional measures (314–325). It has been reported that outbreaks due to CRE have been effectively controlled by enhanced environmental disinfection using hydrogen peroxide (326). Particular efforts should be taken to investigate possible environmental reservoirs during outbreaks (327), and if the source is identified, targeted measures are required to control the outbreak. For instance, duodenoscopes have been identified as the contamination source for several outbreaks, which were subsequently controlled by enhanced cleaning and a change from high-level disinfection to gas sterilization (328, 329). A “search-and-destroy” approach has been successfully used to control outbreaks but can be costly (330). The sources of outbreaks may not be identified in many cases. In such cases, the entire environment of affected wards may become heavily contaminated, and temporary ward closures have been recommended for enhanced cleaning and limiting the further spread of the pathogen (327, 331–333). External audits of infection control measures and extensive health service-wide education have also been recommended (334). If outbreaks persist despite the implementation of measures, screening of health care workers for carriage of MDRO is advised (332, 334), but the true relative value of this exercise is yet to be proven.

Infection Prevention and ControlThe reservoirs and the mode of transmission of MDRO, including NDM-positive strains, are largely underinvestigated (335). Based on outbreak investigations, the source of NDM-positive strains in health care settings has been suggested to be patients infected or colonized with the strains or contaminated environment or devices (Table 5) (336). Unlike bla_KPC-2, the spread of bla_NDM is largely due to horizontal transfer, which is mainly mediated by plasmids but can also be mediated by OMVs, at least in Acinetobacter (60). The transmission of NDM-positive strains in health care settings is believed to be due to close contact (331). Hospitalized patients are the most vulnerable populations for infection or colonization by NDM-positive strains. Healthy persons can also acquire NDM-positive strains, although colonization is usually for a short period of time (less than 1 month, shorter than the period seen for hospitalized patients [see below]) (57).

Several guidance documents are available to address the prevention and control of carbapenem-resistant Gram-negative bacteria in health care settings. The CDC’s guidance for the control of CRE in health care facilities was published in 2012 and updated in 2015 (336). The European Centre for Disease Prevention and Control (ECDC) issued its guidance on prevention of the entry of CRE into health care settings in 2017 (337). The World Health Organization (WHO) also released global guidance on the prevention and control of carbapenem-resistant Enterobacteriaceae, A. baumannii, and P. aeruginosa in 2017 (331). The WHO guidance also identified research gaps and the need for further research. Measures (summarized in Appendix S2 in the supplemental material) recommended by these guidance documents are evidence based and are overall highly similar (Table 6), although the emphasized points and the terminology are slightly different. In addition to the guidance documents, the Society for Healthcare Epidemiology of America (SHEA) established a step-by-step strategic roadmap for CRE infection control based on evidence and expert opinion (327). This roadmap has six successional steps, including determining whether CRE have been isolated, determining affected wards and the occurrence of intrafacility transmission, implementing early CRE detection and CRE containment measures, enhancing existing infection control requirements, regional strategy, and investigating for community spread of CRE, which should be applied in different situations, such as sporadic cases, single-hospital outbreaks, and settings of endemicity (327). However, as the WHO guidance has identified, the quality of the evidence for the recommended measures in guidance is usually very low or low (331). The efficacy of the recommended measures within the guidelines remains to be verified in health care settings. In resource-limited settings, such as health care institutions in the developing world, overcrowding, shortage of staffing, and poor hygiene may be the critical factors contributing to the spread of NDM-positive strains. Such factors may not be addressed by studies that have been performed in developed countries and have been included as evidence during guideline development.

There are several other major challenges to the prevention and control of colonization and infection due to NDM-positive strains in health care settings. One such challenge is long-term carriage (colonization) in human hosts. It is well known that CRE can colonize the human gut for extended periods of time (338, 339), with reports of colonization for more than 3 years (340). Although studies on the duration of colonization usually do not specifically target NDM-positive strains, prolonged colonization by an NDM-positive strain for months was previously reported (341). A study in South Korea reveals that NDM-positive strains are difficult to clear, as only 11.3% (12/106) of colonized patients had no such strains upon discharge from the hospital (342). These colonized patients could serve as cryptic sources of further transmission of NDM-positive strains. Another particular challenge for infection control is the implementation of recommended measures. Even within a single geographic area in developed countries, the infection control strategies and practices against CRE vary remarkably (343). As stated in the WHO guidance (331), compliance should be monitored and analyzed to identify barriers to the implementation of countermeasures. A strategic roadmap for the control of CRE based on the best available evidence and expert opinion has been developed to facilitate implementation (327), and the WHO guidance has also detailed ways to help implementation (331).

The WHO guidance states that the same infection control measures should apply to carbapenem-resistant organisms regardless of the resistance mechanisms (331). NDM- and KPC-positive strains have some important differences, however. Unlike the well-known international spread of KPC-positive strains of K. pneumoniae clonal complex 258 (including ST258, ST11, and a few closely related STs), the spread of NDM-positive strains is less clonal, as there is no sustained global spread of certain high-risk clones. The spread of bla_NDM is primarily mediated by plasmids. bla_NDM is mainly seen in the Enterobacteriaceae, which usually colonize the gut of warm-blooded animals, including humans. bla_NDM has been found in various species of the Enterobacteriaceae. This suggests that the human gut may serve as a hot spot for the spread of bla_NDM in health care settings. However, the predisposing factors of colonization by NDM-positive strains remain largely underinvestigated. The avoidance of disturbance of the gut microflora and selective decontamination may be effective measures against NDM-positive strains and warrant further investigation.