Novel Enterobacter Lineage as Leading Cause of Nosocomial Outbreak Involving Carbapenemase-Producing Strains

Racha Beyrouthy; Marion Barets; Elodie Marion; Cédric Dananché; Olivier Dauwalder; Frédéric Robin; Lauraine Gauthier; Agnès Jousset; Laurent Dortet, François Guérin; Thomas Bénet; Pierre Cassier; Philippe Vanhems; Richard Bonnet


Emerging Infectious Diseases. 2018;24(8):1505-1515. 

In This Article


Emergence of Carbapenemase-producing ECC strains

During January 12, 2014–December 31, 2015, a total of 320 positive cultures for ECC with antibiogram were identified in the Edouard Herriot Hospital in Lyon. Each of 7 (2.2%) ECCs recovered from 7 patients (designated P1–7) produced a carbapenemase (Table 1). These isolates were resistant to penicillins and combinations of penicillins, reacting by releasing β-lactamase inhibitors, oxyimino cephalosporins, and ertapenem (Table 2). PCR and sequencing showed the presence of the carbapenemase-encoding gene bla VIM-4 in the 7 isolates. During January 11, 2013–November 30, 2014, no positive cultures for ECC with antibiogram showed bla VIM-4. These results suggest an epidemic spread of VIM-4–producing ECC strains in the hospital.

Temporal and Spatial Links of the Carbapenemase-encoding ECC Cases

The attack rate of CPE was 0.7/10,000 hospital stays during January 12, 2014–December 31, 2015, compared with 0.0/10,000 hospital stays during January 11, 2013–November 30, 2014 (p = 0.008; Figure 1). The all-causes crude mortality rate among patients with ECC isolates was 43% (n = 3). We provide additional clinical data in the online Technical Appendix. None of the patients had a recent history of travel or hospitalization in foreign countries. Only 2 patients (P6 and P7) were hospitalized in the same unit at the same time. Five patients (P1, P2, P3, P4, and P7) had undergone surgery in the same operating room but at different dates and with different operating teams. We identified CPE in urine samples of 5 patients (P1–P5) that were drawn from the patients' urinary catheters. CPE was isolated from a skin sample that we excised from a necrotic ulcer from P6 and in a sample of operative peritoneal fluid from P7. Patient P1 was simply colonized, whereas patients P2–P7 were infected. All patients had >1 negative samples from the same clinical site during hospitalization before colonization/infection by ECC (Table 1). These findings showed no clear-cut epidemiologic or temporal links between the VIM-4 ECC case-patients, except that surgical procedures were performed in the same operating room on 5 of the 7 patients.

Figure 1.

Epidemic curve of VIM-4–producing Enterobacter cloacae complex isolates (n = 7) in nosocomial outbreak involving carbapenamase-producing Enterobacter strains, Lyon, France, January 12, 2014–December 31, 2015. The attack rate was 0.7/10,000 hospital stays during the study period versus 0.0/10,000 hospital stays during January 11, 2013–November 30, 2014 (p = 0.008). The patients (P1–7) are labeled according to the ST of isolate with which they were infected or colonized. ST, sequence type.

Environmental Investigations

Putative sources previously described in other settings, such as handwashing sinks[23] and endoscopes,[24] were not assumed to be a source of bla VIM-4 ECC because surveillance samples were negative during the study period. In addition to the 102 contact patients, we screened 65 persons during the first 5 episodes (illnesses in P1–P5) without identifying any secondary cases. For the last episode, which involved P6 and P7, we screened 125 of 160 contact patients; all were negative. These findings support a key role for 1 or multiple environmental reservoirs in the nosocomial transmission of bla VIM-4 ECC to patients. However, transmission by healthcare workers (HCWs) cannot be completely ruled out. We did not screen HCWs for possible CPE carriage, but no particular HCW was involved in care of all CPE cases.

The data, including the temporal distribution of the cases over 13 months and the molecular characterization of the isolates, suggested intermittent transmission of human or environmental origin. We therefore implemented environmental screening in 3 rooms occupied by patients P3 and P7, including the beds, mattress covers, and shared equipment, in June and August 2016. These rooms were investigated because they had been occupied >1 time by patients with VIM-4–producing ECC (data not shown). Analysis of the environmental samples after the discharge of patients with VIM-4–producing ECC showed that those collected from a radiator and the mattress cover in 1 patient's room (P7) were contaminated by VIM-4–producing ECC isolates (E14 and E16). The antibedsore mattresses were used in different rooms for several patients. The incidence of CPE was reduced by discarding the mattress covers after the patients known to be VIM-4–producing ECC carriers were discharged from the hospital (data not shown). These data suggest that the transmission of ECCs by contact with mattress covers could be 1 of the key causative factors, especially for the last episode (P6–P7).

Multiclonal Spread of the VIM-4 Carbapenemase

We sequenced the genomes of isolates to assess the molecular links between the VIM-4 ECC cases (Table 2). Six isolates (C45, C46, C48, C310, E14, and E16) formed a clonal cluster designated clone A, in which core genome SNP analysis showed a diversion of <10 SNPs and wgMLST. (Figure 2).

Figure 2.

Whole-genome typing of Enterobacter cloacae complex isolates from nosocomial outbreak involving carbapenamase-producing Enterobacter strains, Lyon, France, January 12, 2014–December 31, 2015. A) Dendrogram inferred by the maximum-likelihood method on the basis of core genome SNPs. The node sizes are proportional to the bootstrap values; values >80 are indicated. Scale bar indicates SNPs. The relatedness of the strains was determined by using <15 variant sites as clonality criteria. B) Minimum-spanning tree based on a whole-genome multilocus sequence typing approach, combining the analysis of core genome loci and the presence or absence of accessory genes. Labels on branches indicate the absolute number of variant loci (clonality threshold ≤10 variant loci). SNP, single-nucleotide polymorphism; ST, sequence type.

Isolate C309 belonged to ST110, isolates C47 and C308 to ST118, and the 6 isolates of clone A to ST873 (dnaA:85/fusA:63/gyrB:101/leuS:103/pyrG:96/rplB:6/rpoB:53). By screening a collection of 30 ECCs isolated in France during the same period, we identified 2 ST873 isolates (CNR1568 and CNR1569) containing the extended-spectrum β-lactamase–encoding gene bla CTX-M-15 in the teaching hospital of Caen. These related isolates differed by 28 core genome SNPs and 27 loci (Figure 2). The clustering of the ECC isolates was also apparent in the analysis of the antimicrobial resistance gene contents (Figure 3). Overall, these data show a multiclonal spread of bla VIM-4 ECCs, the predominance of ECC genotype ST873 among bla VIM-4 ECC, and the interregional spread of this ECC lineage.

Figure 3.

Genetic resistance determinants in Enterobacter cloacae complex isolates from nosocomial outbreak involving carbapenamase-producing Enterobacter strains, Lyon, France, January 12, 2014–December 31, 2015. Black cells indicate presence and white cells absence of resistance determinants. The isolates were classified according to the content in resistance determinants by using a binary distance matrix and UPGMA clustering method. Scale bar indicates the dissimilarity in resistance gene content.

VIM-4–encoding Plasmids

To explore possible links between clone A and the other VIM-4–encoding isolates, we investigated the plasmid contents and the transferability of bla VIM-4 by conjugation. The transfer of ertapenem resistance into Escherichia coli C600 was successful; all of the 9 isolates were at room temperature. Hybridization of plasmids with a specific probe revealed the location of bla VIM-4 on conjugative plasmids of ≈300 kb (n = 6) in the clone A isolates, C47, and C308 and of ≈245 kb (n = 1) in isolate C309. In all isolates, analysis of the assembled genomes showed the presence of an ST1-IncHI2 replicon encoding bla VIM-4.

The plasmid from isolate C45 designated pC45-VIM4 formed a circular 299,117-bp sequence (Figure 4, panel A). The antimicrobial drug resistance genes were located in a ≈50-kb region (bases 97,253–154,784); bla VIM-4 gene was included in a 16kb Tn21-like transposon designated Tn6540 (bases 97,253–113,368). Tn6540 comprised a class 1 integron including bla VIM-4 as the first gene cassette, followed by aac(6′)-Il, dfrA1b, Δant(3″), and smr2. Seven heavy metal resistance loci were also encoded by pC45-VIM4: the tellurite resistance genes terY3Y2XY1W (bases 65,568–69,439) and terZABCDEF (bases 76,028–82,281); the cobalt-zinc-cadmium-resistance protein (bases 118,576 to 119,544); operon copS/copE (bases 165,340 to 167,337); the efflux system rcnR/rcnA (bases 167,595 to 169,105); the arsenic resistance genes arsCBRH (bases 181,666–184,550 bp); and 1 complete mercury resistance operon, merRTCADE (base 134,519–138,533).

Figure 4.

Analysis of bla VIM-4–encoding plasmids from study of nosocomial outbreak involving carbapenamase-producing Enterobacter strains, Lyon, France, January 12, 2014–December 31, 2015. A) Schematic representation of ST1-IncHI2 plasmid pC45-VIM4. The first ring indicates the coordinates of the complete plasmid circle. The 2 outer rings represent the forward and reverse open reading frames, respectively. B) Comparative sequence analysis of ST1-IncHI2 bla VIM-4–encoding plasmids from this study. The plasmids of isolates C45, C46, C47, C48, C308, C309, C310, E14, and E16 are designated pC45-VIM4, pC46-VIM4, pC47-VIM4, pC48-VIM4, pC308-VIM4, pC310-VIM4, pE14-VIM4, pE16-VIM4, and pC309-VIM4, respectively. C) Comparative sequence analysis of bla VIM-4–encoding plasmid pC45-VIM4 to the related bla VIM-1–encoding IncHI2 plasmids pMRVIM0813 (GenBank accession no. KP975077), pRH-R27 (GenBank accession no. LN555650), and pRH-R178 (GenBank accession no. HG530658). Vertical blocks between sequences indicate regions of shared similarity shaded according to blastn ( Blue indicates matches in the same direction; red indicates inverted matches

We identified a similar organization in the other bla VIM-4 isolates (Figure 4, panel B). However, in C309, the VIM-4–encoding plasmid designated pC309-VIM4 (254,277-bp) differed by 2 deletion sites (≈24 kb and ≈21 kb) flanked by mobile elements (IS26 and ISPa21). These deletions resulted in the loss of resistance genes aac(6′)-Ib7, tetA, bla TEM-1, ant(3″), and ant(2″)-Ia. These results suggest the horizontal transfer of the same bla VIM-4–encoding plasmid in several lineages of ECC.

Three related IncHI2 plasmids encoding bla VIM-1 were identified in GenBank (Figure 4, panel C). Except in the region encoding resistance genes, pC45-VIM4 shared 94% of its sequence and most of gene synteny with pMRVIM0813; pRH-R27 and pRH-R178 reported from Germany are more distantly related (85% and 60% of overlap).

A new Cluster in the Phylogenomic Tree of the ECC

Because specific subgroups within the ECC are more prone to cause nosocomial infections or outbreaks, we characterized the isolates at the species and subspecies levels as described by Hoffmann and Roggenkamp[14] and Chavda et al..[15] In the hsp60-based neighbor-joining tree comprising 52 representative reference and type strains (Technical Appendix Figure 1), the sequences of C47 and C308 co-localized with hsp60 cluster III and that of C309 localized with hsp60 cluster VIII; both had >99.3% identity within the clusters. The hsp60 sequences of the ST873 isolates formed a new cluster that shared only 96.7%–97.1% identity with the closest related sequences.

To confirm that the ST873 isolates formed a new subgroup, we performed a phylogenomic analysis with 398 ECC genomes downloaded from GenBank. In the resulting phylogenomic tree (Figure 5), the genomes were distributed in 2 major branches corresponding to the E. hormaechei and E. cloacae metaclusters as previously reported.[14,15] The E. hormaechei metacluster comprised 6 branches corresponding to Chavda's phylogenomic groups A–E, and a new phylogenomic group, designated S, comprising solely the ST873 isolates. As expected, the C47, C308, and C309 isolates clustered in Chavda's phylogenomic groups D and B, which correspond to Hoffmann's hsp60 clusters III and VIII (Figure 5). These findings fortify the hypothesis that the ST873 isolates could be a new species or subspecies in the E. hormaechei metacluster.

Figure 5.

Approximately maximum-likelihood phylogenetic trees based on recombination free core single-nucleotide polymorphisms (SNPs) inferred from ST873, ST110 and ST118 genomes and 398 representative genomes of Enterobacter cloacae complex strains in study of nosocomial outbreak involving carbapenamase-producing Enterobacter strains, Lyon, France, January 12, 2014–December 31, 2015. All nodes are supported by Shimodaira-Hasegawa test values ≥97%. Scale bar indicates SNPs. NA, nonattributed; ST, sequence type.

A new Species in the E. Hormaechei Metacluster

Average nucleotide identity (ANI) and percentage of conserved DNA (PCD) can accurately replace DNA–DNA hybridization values for species delineation by using 0.95 and 0.69 as ANI and PCD thresholds, respectively.[25,26] We therefore calculated ANIs and PCDs for the ST873 isolates against 398 ECC genomes by using BLAST ( The PCD values were high enough (>0.69) within the E. hormaechei and E. cloacae metaclusters for the delineation of species and subspecies by ANI calculations (Technical Appendix Figure 2). Genomes within the same phylogenomic group shared ANI mean values >98% (Figure 6). In the E. cloacae metacluster, the ANI values supported the designation of the phylogenomic groups as different species (ANI, 0.87–0.94), except for E. cloacae subsp. cloacae and E. cloacae subsp. dissolvens (ANI, 0.95). In the E. hormaechei group, most phylogenomic groups shared ANI values 0.96–0.98, supporting the split of the E. hormaechei metacluster into different subspecies. As expected, isolates C47 and C308 clustered in the E. hormaechei phylogenomic group D and isolate C309 in the phylogenomic group B (also designated E. hormaechei subsp. steigerwaltii). The new phylogenomic group S, comprising the ST873 isolates, had ANI values below the species cutoff (<0.95) against all groups, including those in the E. hormaechei metacluster. We obtained similar results by using the MUMmer-based approach (Technical Appendix Figures 3, 4), suggesting that the ST-873 isolates are a new species of the E. hormaechei metacluster.

Figure 6.

Average nucleotide identity calculated from BLAST ( pairwise comparisons of ST873 genomes and 398 Enterobacter cloacae complex genomes in study of nosocomial outbreak involving carbapenamase-producing Enterobacter strains, Lyon, France, January 12, 2014–December 31, 2015. NA, nonattributed.

The 3 subspecies E. hormaechei subsp. hormaechei, E. hormaechei subsp. oharae, and E. hormaechei subsp. steigerwaltii can be differentiated by using D-adonitol, D-arabitol, D-sorbitol, and D-melibiose fermentation tests (bioMérieux). By using the API 50CH system, we found that the biochemical characterization of our isolates yielded results compatible with E. hormaechei subsp. steigerwaltii (hsp60 cluster VIII and phylogenomic group B), as expected (Technical Appendix Tables 1, 2). Of interest, the ST873 isolates produced a distinguishable biochemical phenotype in the E. hormaechei metacluster by growing on only D-melibiose as the sole carbon source (Technical Appendix Table 2). Overall, our molecular and biochemical data agree with those of previous studies in the field and suggest that the ST873 isolates are a new species of the E. hormaechei metacluster.

Functional Genomics in the E. hormaechei Metacluster

To investigate the functional features of the ST873 isolates and other phylogenomic subgroups in the E. hormaechei metacluster, we constructed a pangenome by using 245 strains including our isolates. The pangenome was divided into 3 sections: 1) the core genome (the set of genes shared by 99% of strains), 2) the accessory genome (the set of genes present in some but not all representatives), and 3) the unique genome (genes unique to individual strains). The 245 strains examined yielded a pangenome of 25,221 genes. On the basis of this dataset, the core genome is composed of 2,575 genes, the accessory genome of 14,849 genes, and the unique genome of 7,797 genes (Figure 7, panel A). Functional annotation of the pangenome on the COG database showed an overrepresentation of genes belonging to the groups with housekeeping functions (COG categories C, G, E, F, I, P, H, J, O, D, and T) in the core genome (Figure 7, panel B). The accessory genome and the unique genes had a similar distribution of functional annotations with an overrepresentation of genes involved in DNA recombination (genes encoding integrases, transposases, or resolvases) and defense mechanisms, as well as those belonging to the mobilome (COG categories L, U, V, and X), as expected.

Figure 7.

Pangenome analysis of metacluster Enterobacter hormaechei in study of nosocomial outbreak involving carbapenamase-producing Enterobacter strains, Lyon, France, January 12, 2014–December 31, 2015. A) Distribution of COGs); B) functional annotations in the pangenome; C) functional annotations in the variable genome (accessory genome + unique genes); and D) functional annotations for specific genes. Bar charts show the enrichment of COG categories as odds ratios; error bars indicate 95% CIs. Asterisks indicate certain COG categories that are significantly enriched: *p<0.05; **p<0.01;***p<0.001, all by Fisher exact test. Each COG category is identified by a 1-letter abbreviation: C, energy production and conversion; D, cell cycle control and mitosis; E, amino acid metabolism and transport; F, nucleotide metabolism and transport; G, carbohydrate metabolism and transport; H, coenzyme metabolism; I, lipid metabolism; J, translation; K, transcription; L, replication, recombination and repair; M, cell wall/membrane/envelope biogenesis; N, cell motility; O, post-translational modification, protein turnover, and chaperone functions; P, inorganic ion transport and metabolism; Q, secondary metabolism; T, signal transduction; U, intracellular trafficking and secretion; V, defense mechanisms; and X, mobilome. COG, clusters of orthologous groups.

The distribution of variable genome annotations among the phylogenomic groups showed closely related distributions of functional annotations (Figure 7, panel C), except for cell mobility annotations (COG category N, comprising the flagellar- and adhesion-related functions), which were overrepresented in phylogenomic group B and the ST873 isolates (phylogenomic group S), and mobilome annotations (COG category X), which were overrepresented in phylogenomic group A. The differences in functional distribution were enhanced by the analysis of variable genes specific to phylogenomic groups (Figure 7, panel D). Phylogenomic group B (hsp60 group VIII), the most prevalent subgroup in human infections, and the new epidemic phylogenomic group S (ST873) had an overrepresentation of genes involved in cell motility (COG category N), as previously observed among variable genes, showing that these functions are overrepresented and based on specific genes in these ECC lineages. The ST873 group also exhibited enrichment in specific genes related to signal transduction (COG category T), and phylogenomic group A exhibited enrichment in genes related to recombination and the mobilome (COG categories L and X). We also observed other differences in genes related to metabolic functions. Phylogenomic group A demonstrated an enrichment in specific genes involved in lipid metabolism (COG category I), and phylogenomic group D has specific genes involved in coenzyme, nucleotide, and secondary metabolism (COG categories F, H and Q); phylogenomic group B accumulated specific genes involved in energy production and carbohydrate, amino acid, and ion metabolism (COG categories C, G, E and P). Overall, these data suggest that there exist distinct lifestyles in ECCs explaining varied abilities to colonize the hospital environment and to induce nosocomial infections.

Adhesion to Abiotic Surface and Epithelial Cells

To investigate the overrepresentation of mobility/adhesion functions in the ST873 isolates in the context of their prolonged persistence in hospital, we compared the ability of our isolates to initiate biofilm formation on an abiotic surface and their adhesion to intestinal epithelial cells. The ST873 isolates had greater ability to initiate biofilm on PVC than did isolates of ST118 and ST110 (Technical Appendix Figure 5, panel A). We also observed slight differences of adhesion to HT29 intestinal epithelial cells (Technical Appendix Figure 5, panel B). These results suggest that ST873 isolates have original adhesion features, as suggested by the pangenomic analysis.