Comparative genomics of global optrA-carrying Enterococcus faecalis uncovers a common chromosomal hotspot for optrA acquisition within a diversity of core and accessory genomes.

Linezolid-resistant Enterococcus faecalis (LREfs) carrying optrA are increasingly reported globally from multiple sources, but we lack a comprehensive analysis of human and animal optrA-LREfs strains. To assess if optrA is dispersed in isolates with varied genetic backgrounds or with common genetic features, we investigated the phylogenetic structure, genetic content [antimicrobial resistance (AMR), virulence, prophages, plasmidome] and optrA-containing platforms of 27 publicly available optrA-positive E. faecalis genomes from different hosts in seven countries. At the genome-level analysis, an in-house database with 64 virulence genes was tested for the first time. Our analysis showed a diversity of clones and adaptive gene sequences related to a wide range of genera from Firmicutes. Phylogenies of core and accessory genomes were not congruent, and at least PAI-associated and prophage genes contribute to such differences. Epidemiologically unrelated clones (ST21, ST476-like and ST489) obtained from human clinical and animal hosts in different continents over eight years (2010-2017) could be phylogenetically related (3-126 SNPs difference). optrA was located on the chromosome within a Tn6674-like element (n=10) or on medium-size plasmids (30-60 kb; n=14) belonging to main plasmid families (RepA_N/Inc18/Rep_3). In most cases, the immediate gene vicinity of optrA was generally identical in chromosomal (Tn6674) or plasmid (impB-fexA-optrA) backbones. Tn6674 was always inserted into the same ∆radC integration site and embedded in a 32 kb chromosomal platform common to strains from different origins (patients, healthy humans, and animals) in Europe, Africa, and Asia during 2012-2017. This platform is conserved among hundreds of E. faecalis genomes and proposed as a chromosomal hotspot for optrA integration. The finding of optrA in strains sharing common adaptive features and genetic backgrounds across different hosts and countries suggests the occurrence of common and independent genetic events occurring in distant regions and might explain the easy de novo generation of optrA-positive strains. It also anticipates a dramatic increase of optrA carriage and spread with a serious impact on the efficacy of linezolid for the treatment of Gram-positive infections.

Data Summary

The whole genome of all strains was downloaded from NCBI (https://www.ncbi.nlm.nih.gov/Traces/wgs/) and EMBL (https://www.ebi.ac.uk/ena). A list of the accession numbers, related epidemiological data, and references (if available) can be found in Table 1. The authors confirm all supporting data, code and protocols have been provided within the article or through supplementary data files. Seven supplementary tables and four supplementary figures are available with the online version of this article.

Table 1.

Epidemiological background of the 28 genomes analysed in this study

Isolate	Source	Sample	Country	Year	ST	No Contigs	Genome size (bp)	GenBank/ENA	Ref
12E	RCk	Meat	COL	2010–2011	59	234	2 834 796	ERX1670553	[33]
34E	RCk	Meat	COL	2010–2011	59	220	2 776 507	ERX1670554	[33]
745E	RCk	Meat	COL	2010–2011	489	277	2 979 622	ERX1670555	[33]
612T	RCk	Meat	TUN	2017	21	55	2 944 588	SEWV01000000	[64]
728T	RCk	Meat	TUN	2017	859	62	3 278 897	SEWT01000000	[64]
697T	RCk	Meat	TUN	2017	476	28	2 973 239	VDMY00000000	[64]
N60443F	RCt	Meat	USA	2014	489	3	2 853 753	CP028724-6	[5]
N48037F	RP	Meat	USA	2013	141	4	3 151 142	CP028720-3	[5]
712T	Chicken	Faeces	TUN	2017	476	55	2 982 138	SEWU01000000	[64]
P.En090	Pig	Faeces	MYS	2012	756	60	3 071 086	MJBX01000000	[23]
Enfs51	Pig	Faeces	MYS	2012	754	58	3 024 957	MJEK01000000	[23]
P.En250	Pig	Faeces	MYS	2012	476	42	2 834 146	MJBZ01000000	[23]
Enfs85	Pig	Faeces	MYS	2012	755	62	2 834 811	MJBV01000000	[23]
IBUN9046YE	Pig	Faeces	COL	2014	16	113	2 945 287	NCAD01000000	–
12T	WW	ww	TUN	2014	86	27	3 009 161	NHNF01000000	[37]
P.En218	H (PF)	Faecal	MYS	2012	765	61	2 862 267	MJBY01000000	[23]
Enfs94*	H (PF)	Faecal	MYS	2012	227	53	2 625 583	MJBW01000000	[23]
Efs 599	HH	Unknown	USA	Unknown	16	130	3 053 367	ALZI01000000	–
11 340	HP	Ascites	CHN	2013	476	119	2 920 750	QNGJ01000000	[30]
13 484†	HP	Fester	CHN	2013	21	122	2 918 217	QNGN01000000	[30]
18 026	HP	Ascites	CHN	2014	476	231	2 959 223	QNGT01000000	[30]
27 149†	HP	Ascites	CHN	2014	858	97	2 836 434	QNHA01000000	[30]
Iso1Bar	HP	Urine	ESP	2016	585	56	2 878 959	ERX2067873	[12]
Iso2Bar	HP	Urine	ESP	2016	585	72	2 902 425	ERX2067874	[12]
Iso3Bar	HP	Urine	ESP	2016	585	79	2 872 904	ERX2067875	[12]
Iso4Bar	HP	Urine	ESP	2016	474	128	2 943 732	ERX2067876	[12]
Iso5Bar	HP	Urine	ESP	2016	474	204	2 949 268	ERX2067877	[12]
16–372	HP	Urine	FRA	2016	480	6	2 779 791	PHLF01000000	[10]

*Enfs94 strain does not carry optrA.

†These isolates were obtained from hospitalized patients that work as farmers.

‡These isolates were obtained from patients that work as farmers.

CHN, China; COL, Colombia; ESP, Spain; FRA, France; H, Human; HH, Healthy Human; HP, Hospitalized Patient; MYS, Malaysia; PF, Pig Farmer; RCk, Retail Chicken; RCt, Retail Cattle; RP, Retail Pig; ST, sequence type; TUN, Tunisia; USA, United States of America; WW, Wastewater.

–

–

rank as one of the most common causes of nosocomial infections and they are well known for their ability to acquire and transfer different gene adaptive traits including antimicrobial resistance and virulence genes, their genomic plasticity and high numbers of extra-chromosomal elements. In recent years, transferable genes conferring resistance to last-resort antibiotics as linezolid, especially optrA, emerged in different regions and, in particular, in strains obtained from different hosts. Nevertheless, little is known about the similarity of populations and adaptive genes circulating globally. Here, we compared the genomes of optrA-carrying strains from disparate origins and regions and identified some epidemiologically unrelated strains from human clinical and animal hosts in different continents as being phylogenetically related. optrA was commonly associated with a Tn6674-like element (besides diverse plasmids) inserted into the same chromosomal integration site in different strains, here proposed as a genetic hotspot for optrA integration. Furthermore, optrA was found to be embedded in a common 32 kb chromosomal platform that is conserved among hundreds of genomes. The data provide novel insights about optrA mobilization across different settings that might help to explain the easy de novo generation of optrA-positive strains. It also highlights the relevance of further genomic studies exploring the dissemination and reservoirs of optrA at a global and One Health scales.

Introduction

Oxazolidinones (linezolid and tedizolid) are increasingly used for the treatment of human infections caused by relevant pathogens such as methicillin-resistant and vancomycin-resistant enterococci [1, 2]. Linezolid resistance rates remain generally low in enterococci causing infections worldwide (<1 %) [1, 3], however, acquired linezolid resistance genes (cfr, optrA and/or poxtA) are being increasingly reported in different enterococcal species and across different settings [2-7]. Among the three transferable linezolid resistance genes, optrA has been the main one responsible for the recent increase in linezolid-resistant enterococci (LRE) in human isolates [8-13]. According to available studies, optrA-carrying LRE are globally circulating in hospitals since at least 2005 [14], while the first description in food-producing animals dates from 2008 [15].

optrA codifies an ABC-F protein targeting the ribosome of Gram-positive bacteria and mediates resistance to both phenicols and oxazolidinones through ribosomal protection [16]. As phenicols are allowed in veterinary medicine to control respiratory infections, the use of florfenicol in food-producing animals might co-select resistance for oxazolidinones and promote the occurrence of an optrA animal reservoir [7]. optrA-carrying LRE strains were already described in a variety of hosts including hospitalized humans, healthy humans and different food-producing animals (poultry, pigs, cattle, and ducks) worldwide, and embedded in a high diversity of genetic backgrounds in association with different chromosomal or plasmid genetic platforms [1–3, 14, 15, 17–19]. Moreover, optrA has been more commonly described in than in [2, 20].

Even though optrA-carrying LRE strains pose a growing threat to public health, little has been done regarding a global comparative genomics analysis and detailed optrA genetic context characterization [5, 9, 21, 22]. Indeed, most optrA-positive strains are only analysed with respect to their optrA-carrying platform neglecting the deep characterization of their genetic makeup. Although no clear distinction between clinical and non-clinical isolates exists in terms of gene repertoire, we aimed to address if optrA-carrying strains from different hosts carry similar genetic markers by performing a comprehensive comparison of publicly available optrA-carrying genomes. Given the exponential increase of optrA among from different settings and areas, we also aimed to investigate deeper about optrA genetic platforms and their mobilizable nature.

Methods

Bacterial collection and comparative genomics

Publicly available optrA-carrying genomes (n=27) at the NCBI and EMBL databases (last search March 2019) were retrieved together with all available epidemiological information. An additional genome not carrying optrA was also considered since it is included in the same research project (“Gastrointestinal health of animal handlers with special emphasis on potential zoonotic transmission of bacterial infection”) of other related optrA-carrying isolated from pigs [23], accounting for a total of 28 genomes. Only complete, assembled and identifiable genomes were included in this analysis (Table 1).

All 28 genomes were mapped against the V583 reference strain to infer a phylogeny based on the concatenated alignment of high-quality single nucleotide polymorphisms (SNPs) using CSI Phylogeny standard settings (CGE, https://cge.cbs.dtu.dk/services/CSIPhylogeny). Recombination was removed from the lineage-specific analyses using Gubbins [24]. Accessory genome was determined using PATO (https://github.com/irycisBioinfo/PATO), genes were considered homologous if they had 80 % identity and 80 % coverage. Both trees were obtained with IQ-tree [25]; in the case of the accessory genome tree a binary model was made by using a presence/absence matrix. Comparison between SNPs-based and accessory genome trees was performed using R package phytools [26].

Genomes were screened for genes encoding antibiotic resistance (ABR), plasmids, virulence and MLST using in silico genomic tools (ResFinder 3.0, PlasmidFinder 2.1, VirulenceFinder 1.5 and MLST 1.8 tools, respectively) available at the Center for Genomic Epidemiology (CGE; http://www.genomicepidemiology.org). As the VirulenceFinder database is not complete for the species (n=22 virulence genes typical of this species), we created an in-house database with 42 additional virulence genes and tested it using the MyDbFinder (blast) tool available at CGE, making a total of 64 virulence genes (Table S1, available in the online version of this article). The 28 assembled genomes were also analysed by PHASTER (PHAge Search Tool Enhanced Release) to identify the presence of prophage sequences [27]. Genomic data for 23S rDNA mutations and for ribosomal protein genes rplC, rplD and rpIV were compared to reference gene loci of the linezolid-susceptible V583 (GenBank acc. no. AE016830) using Geneious Prime blastn. The presence of cfr variants, poxtA and plasmid replicons was also performed using Geneious Prime blastn and reference cfr/poxtA sequences [GenBank accession numbers AM408573 (cfr1), AJ879565 (cfr2), NG_055643 (cfrC), PHLC01000011 (cfrD) and MF095097 (poxtA)].

PATO was used for the analysis of the accessory genome allowing us to infer the accessory genome from the proteomes and cluster them based on 80 % protein similarity and 80 % of coverage. The set of representative proteins was used to build a binary matrix (presence/absence of proteins in the accessory genome) in the R-environment (http://www.r-project.org) and a tree to classify the strains according to the accessory genomes. The tree was build using IQTree and binary data with model GTR2+FO+ASC+R2 (selected by ModelFinder).

Characterization of optrA genetic platforms

Prokka annotation of the 612T optrA-containing contig was confirmed using blastn, blastp and ISFinder (https://isfinder.biotoul.fr/). The same strains used for comparative genomics analysis were mapped against the optrA-containing contig of 612T using Geneious Prime in order to determine the presence of the same optrA platform and the genome region where it was inserted. Vector NTI advance v11 and Easyfig v2.2.2 [28] were used for drawing optrA platform schemes and contig comparisons. Gene functions of the 612T optrA-containing contig and the different optrA platforms were classified according to the database of Clusters of Orthologous Groups of proteins (COGs) using eggNOG 4.5.1 (http://eggnogdb.embl.de/#/app/home). The identification of contigs belonging to the optrA-containing plasmid in 728T strain and the corresponding annotation was determined by comparison with the GenBank database using blastn, blastp and ISFinder. Vector NTI advance v11 and EggNOG 4.5.1 were used to determine gene function and represent the different plasmid contigs.

Results and discussion

Phylogenetic relatedness of strains

Fig. 1 represents the phylogenetic tree obtained with SNPs, not afected by recombination, of the 28 genomes using as reference the genome of V583. The healthy humans included correspond to a human volunteer (Efs 599) of the NIH Human Microbiome Project [29] and two swine farmers (Enfs94/P.En218) working in pig farms from Malaysia [23]. In the four genomes from hospitalized patients in China, linezolid resistance was not associated with the consumption of linezolid and an asymptomatic gut colonization was speculated by the authors [30]. Moreover, two of them (13484/27149) were recovered from Chinese farmers [30].

Fig. 1.

Phylogenetic tree of the strains (n=28) analysed in this study. SNP alignment was obtained using standard conditions of CSI Phylogeny and strains V583 as reference strain. Recombination regions were eliminated using Gubbins and maximum likelihood tree was obtained using IQ-tree. Stars indicate the presence of the optrA gene. The colour of the star indicates the different optrA variants as indicated in the key. Triangle filled or not filled with dark-green respectively indicates the presence or absence of the chromosomal 612T contig containing optrA used as a reference for comparison of optrA chromosomal platforms (please refer to this section below). The different optrA genetic platforms are represented with circles filled with red (group 1), green (group 2) and blue (group 3) as in Fig. 6. Group 1 includes isolates carrying optrA in the chromosomal Tn6674, group 2 includes isolates carrying optrA-fexA on medium-size plasmids, and group 3 includes isolates carrying optrA-araC on plasmids/chromosome. Coloured cells represent the presence of acquired AMR genes with each colour indicating the correspondent family as indicated in the key (oxazolidinones in red; phenicols in yellow; aminoglycosides in green; macrolides, lincosamides or streptogramins in pink; tetracyclines in purple; and trimethoprim in blue). AMR genes present in only one isolate are not represented in the figure: cfr and poxtA (Efs599), tet(O) (728T) and tet(K) (12E). Abbreviations: ST, sequence type. White cells represent the absence of genes.

Fig. 6.

Different platforms carrying the optrA gene. Panel (a) group 1; (b) group 2; (c) group 3. Group 1 includes isolates carrying optrA in the chromosomal Tn6674, group 2 includes isolates carrying optrA-fexA on medium-size plasmids, and group 3 includes isolates carrying optrA-araC on plasmids/chromosome. Coloured arrows represent genes and the arrows indicate the orientation. Grey arrows represent hypothetical proteins of unknown function; the remaining colours represent the gene function according to the COGs database classification indicated in the key. Abbreviations: ST, sequence type; ENA, European Nucleotide Archive.

Based on SNP analysis, we observed an overall diverse population of optrA-positive (range 3–29 258 SNPs) with a correspondence between the sequence types identified (17 STs, 15 nodes) (Fig. 1, Table S2). Strains originating from the same source or country were dispersed across the different tree nodes, suggesting no significant correlation between strain origin and phylogeny as commonly observed in [31]. These results further confirm the easy promiscuity of optrA that has been identified in enterococci from highly variable epidemiological and genetic backgrounds [2, 32]. The exceptions corresponded to specific strains originating both temporally and locally from the same scenario namely the clinical isolates Iso1Bar/Iso2Bar/Iso3Bar (15–35 SNPs) or Iso4Bar/Iso5Bar (69 SNPs) from an hospital in Barcelona, and the chicken meat isolates 12E/34E (39 SNPs) from Colombia, with close clonal relationships as described [12, 33]. Still, strains from different origins and lacking evident epidemiological relationships were identified as possibly related (<126 SNPs) given their distant geographical and temporal relationships: (i) five ST476 and one ST858 (single locus variant of ST476) isolates from chicken faeces and chicken meat (712T and 697T, respectively) in Tunisia, pigs in Malaysia (P.En250) and hospitalized patients in China (11340, 18026, 27149) differing between 3–121 SNPs; (ii) two ST21 isolates from chicken meat in Tunisia (612T) and a hospitalized patient in China (13484) differing in 113 SNPs; and (iii) two ST489 isolates from retail chicken in Colombia (745E) and cattle in USA (N60443F) differing in 126 SNPs (Table S2).

The finding of the related optrA E. faecalis strains (ST21 and ST476) in human clinical and animal hosts from different continents over >5 years (2012–2017) was remarkable and demonstrates that they might originate from an animal reservoir and are able to easily colonize humans. An inter-hospital spread of ST476 optrA-carrying in different Chinese cities has been previously reported [22]. Noteworthy, most of the food animal strains included in this phylogenetic analysis (ST21, ST59, ST86, ST141, ST227, ST474, ST476, ST489, ST585) have been previously associated with from human infections in different continents, linked to linezolid resistance or not [www.pubmlst.org] [9, 12, 13, 34].

Comparative pan-genomic analysis of optrA-carrying : antibiotic resistance, virulence, prophages and plasmid replicons

We constructed an additional tree using the accessory genome to segregate the strains based on their unique accessory gene repertoire. If we consider two main clades in the core genome tree, the lower clade including the bottom cluster with two main nodes, in addition to strains Enfs51 and Enfs94, corresponded to the strains that remained stable in both trees (Fig. 2). The node placement in both trees was not consistent with the two methods used with 19 out of the 28 genomes changing their tree positions and, in some cases, also their cluster’s relatives. This confirms that even though some strains are quite close to one another from a phylogenetic (SNPs) point-of-view, they were not necessarily close in their overall genetic repertoires. Isolates changing their nodes within a given cluster were: 16–372 (ST480) and P.En218 (ST765) co-clustering with ST476-like strains in the accessory genome tree; the phylogenetically distant P.En90 (ST756) and N48037F (ST141) joining and co-clustering into the V583 strain group that moved to the upper clade of the accessory tree; and the ST21 612T and 13 484 strains moving into the lower clade of the accessory tree (Fig. 2).

Fig. 2.

Comparison of core and accessory genome trees. SNP alignment was obtained using standard conditions of CSI Phylogeny and strains V583 as reference strain. Recombination regions were eliminated using Gubbins. Analysis of the accessory genome was performed using PATO and the resulting presence/absence matrix was used to obtain a maximum likelihood tree. Both trees were obtained using IQ-tree. Tree comparison was performed using R package phytools. Abbreviations: ST, sequence type.

The analysis of the 28 genomes using the AcCNET tool [35] yielded a set of 4792 protein families representing the accessory genome (Fig. 3). Again, the accessory genome network did not reveal a trend for host-specific proteins (Fig. 3a) but strains grouped according to their ST (Fig. 3b) with protein sharing in specific cases. A total of 1570 unique proteins were identified across all strains when grouping them by ST, with more than half (n=921 proteins) corresponding to hypothetical proteins. The greatest number of unique proteins by ST (n=388 genes) belonged to a single strain from a retail chicken in Tunisia (ST859). When analysing the unique proteins by origin, all of them (n=139) were identified from wastewaters (one ST86, Tunisia), with all other origins sharing protein content. Unique proteins were diversely annotated mainly as surface proteins, ABC transporters, conjugative and mobile proteins, phages, carbohydrate metabolism and transporters (Table S3).

Fig. 3.

AcCNET analysis of the 28 strains analysed in this study. (a) Representation of the accessory genome network by the origin of isolates. Each colour indicates a different origin as indicated in the figure key. (b) Representation of the accessory genome network by the ST of each isolate. Each colour indicates a ST origin as indicated in the figure key. Abbreviations: ST, Sequence type. Clusters of isolates (larger nodes) as well as proteins based on 80 % similarity can be visualized according the colours of the nodes.

When we compared the isolation source and the genetic content of the 28 strains by clustering them based on their accessory genome (1Fig. S1 and Fig. S2), no clear patterns emerged among related strains. The only exceptions corresponded to those strains carrying esp and cyl genes as they clustered in the upper part of the accessory genome tree while being more dispersed in the core genome-based tree. Esp (enterococcal surface protein) is located within a pathogenicity island (PAI) and the cyl operon encodes a cytolysin with haemolytic activity located on pheromone-responsive plasmids or within a PAI [36], which explains their contribution to the accessory genome of the strains carrying them. Most of these esp- and/or cyl-positive strains (N48037F, P.En090, Iso1Bar, Iso2Bar, Iso3Bar, Efs599, IBUN9046YE) clustered together in a pan-matrix showing the sharing of accessory proteins, but other pan-proteins may contribute to their close relationship (Fig. S3). This heatmap (Fig. S3) shows the highest number of shared proteins for related isolates (same ST) and according to the accessory strain clusters.

Specific genes codifying for antibiotic resistance, virulence, phages and plasmids are analysed in detail below (Figs 1 and 4).

Antibiotic Resistance

A high amount of antimicrobial resistance genes was identified in most of the genomes, with no clear distinct patterns being observed among related strains. The overall acquired ABR gene profiles identified are the same as that commonly identified in optrA-carrying strains from non-clinical sources in different regions [15, 18], with no significant differences between clinical (8–15 genes, 11 in average) and non-clinical (1–15 genes, 10 in average) isolates (Fig. 1). They included acquired genes putatively conferring resistance to oxazolidinones, phenicols, aminoglycosides, macrolides, lincosamides or streptogramins, tetracyclines, and trimethoprim, and varying in their homology against reference strains (Table S4). Chromosomal point mutations putatively conferring resistance to linezolid (in 23S rRNA and L3/L4/L22 ribosomal proteins) and ciprofloxacin [gyrA (S83I, E87G) and/or parC (S80I)] were also identified in most strains. Regarding linezolid resistance mutations, only synonymous ones were identified in rplC/L3 and rplV/L22 ribosomal proteins while in the case of the 23S rRNA some strains (n=10/28) exhibited mutations not previously described (Table S4). In common to previous studies, all strains exhibited an amino acid substitution at position 101 of the rplD/L4 protein, which has been linked to both linezolid-resistant and susceptible isolates [7, 9, 37]. Interestingly, the optrA-negative Enfs94 is described as linezolid-resistant (MIC not available) according to Tan et al. [23], and so its 23S rRNA (G388A/D130N) and rplD/L4 protein (T301C/F101L) mutations might contribute to linezolid resistance expression as this strain does not have any other known linezolid resistance mechanism. Curiously, strains from healthy humans presented very different ABR profiles even from the same local/country: ST227 (Enfs94, strain without optrA) and ST765 (P.En218) obtained from pig farmers in Malaysia, respectively, carried lsa(A) only and 15 antibiotic resistance different genes. The two related clusters of ST476- and ST21-like strains from foodborne animals and hospitalized patients exhibited similar profiles of both acquired ABR genes and point mutations, with genes conferring resistance to aminoglycosides being the more variably present among the different strains. Enterococci displaying the referred ciprofloxacin point mutations in gyrA/parC have been commonly identified in samples obtained from animals and foodstuffs, but also from human infections [38, 39]. Interestingly, some of the strains clustering into the same node in the phylogenetic tree shared similar combinations of gyrA/parC mutations but the same was not observed for linezolid chromosomal mutations or optrA variants which greatly varied even between closely related strains (Fig. 1, Table S4).

Virulence gene profiles

We herein firstly provide an in silico analysis of an extended number of putative virulence genes (n=64), as the few available studies analysing virulence generally include a small set of 22 genes available in the VirulenceFinder database [12, 33, 37]. The genomes here analysed contained a quite high number of virulence genes (41–61, 49 in average), though with variable gene identity in comparison to reference strains and some in truncated or increased size versions (Table S5). Virulence profiles were generally identical beweteen clinical (41–59 genes) and non-clinical (41–61 genes) isolates (Fig. 4). The related clusters comprising ST476-like, ST21 and ST489 strains exhibited identical virulence gene profiles including 51–55, 44 and 42–44 genes, respectively.

Fig. 4.

Profiles of virulence, prophages and replication initiation genes present in the (n=28) strains analysed in this study. SNP alignment was obtained using standard conditions of CSI Phylogeny and strains V583 as reference strain. Recombination regions were eliminated using Gubbins and maximum likelihood tree was obtained using IQ-tree. The presence of the virulence genes, prophages and plasmid replication initiation genes is indicated by coloured squares as indicated in the key (virulence genes in green; prophages in blue; replication initiation genes in pink) and white cells indicate the absence of genes. Group 1 includes isolates carrying optrA in the chromosomal Tn6674, group 2 includes isolates carrying optrA-fexA on medium-size plasmids, and group 3 includes isolates carrying optrA-araC on plasmids/chromosome. Abbreviations: ST, sequence type.

Thirty-six out of the 61 virulence genes detected were present in all strains and they were related with adhesion, survival, general stress, cell surface behaviour, capsule production and sex pheromones, some of which associated with PAI. Interestingly, all strains carried four major virulence genes or operons relevant for endocarditis, including: bgsA/bgsB (glucosyltransferases with a key role in the production of the major cell wall glycolipids, biofilm production and presumably in endocarditis) [40], gls24/glsB (stress genes important for survival in vivo, and gls24 in particular to bile salts resistance and hypothetically for virulence in peritonitis and endocarditis) [41], efbA (a PavA-like fibronectin adhesin previously shown to be important in experimental urinary tract infection and endocarditis) [42], and eep (a membrane metalloprotease crucial in biofilm formation and required for endocarditis) [43], all exhibiting nucleotide identities varying between 97 and 100 % (Table S5). Except for epaM missing in three strains, all genes comprising the epa locus (epaA-epaR) were found in all strains. The epa locus is required for the biosynthesis of an antigenic rhamnopolysaccharide that plays a role in urinary tract infections and peritonitis, in intestinal colonization and in protecting against human host defenses [44]. Teng et al. have also described that epa genes are widespread among despite showing differences in sequence and cluster composition [45], but the impact of such differences on their expression remains unknown. Survival [elrA (leucine-rich protein A associated with macrophage persistence), tpx (thiol peroxidase for oxidative stress resistance)] and sex pheromone-associated (cad, camE, cCF10, coB1) genes were also common to all strains.

Among the six virulence gene operons identified (hylA/hylB, ebpA/ebpB/ebpC, gls24/glsB, cpsC-cpsK, epaA-epaR and bgsA/bgsB), only the bgsA/bgsB was complete in all strains. The six cps genes essential for capsule biosynthesis (cpsC, cpsD, cpsE, cpsG, cpsI, and cpsK) were identified in 14 of the 28 strains obtained from human, environmental and animal (including the ST476 cluster) sources. The identity of these genes varied from those carried by the V583 strain (97–100 % in nucleotides), and some corresponded to pseudogenes (cpsF and cpsG; Table S5). Isolates were classified based on capsule operon polymorphisms in type 1, 2 or 5 as described: (i) CPS type 1 do not express the CPS as they possess only cpsA and cpsB lacking all remaining genes (cpsC-cpsK) essential for capsule production; (ii) CPS type 2 possess all genes (cpsA-cpsK) of the cps operon (e.g. V583); (iii) and the cpsF gene is absent in the CPS 5 serotype [46]. Within each cluster of strains, of the core genome tree, the CPS type was invariant. In common to other studies, the most common CPS type was type 1 (n=14), followed by type 2 (n=12) and type 5 (n=2) [46]. All isolates with the CPS type 2 carried cpsF with different mutations in comparison to V583 so the expression of CPS cannot be ruled out only by capsule genotyping. Although isolates with CPS type 1 should not express a capsule, non-encapsulated strains have been already linked to endocarditis [47].

Even though we cannot infer by genomics the true opportunistic infection potential of the 28 strains, several of them were enriched in different virulence genes that are known to play key roles in pathogenicity. Strains recovered from human and animal sources carried most virulence genes tested including some (ace-89%; gelE-fsrB-71%; agg-50%; elrA-100%) that were previously described in hospital/epidemic strains [48, 49]. Four genes from the module A of PAI (EF0500, EF0501, EF0502 and EF0503) have been considered the most discriminatory within the PAI to identify putative pathogenic strains [50]. In that study, those genes were present in all strains characterized as potential pathogens and not in commensals. In our 28 genomes, they were identified in eleven strains dispersed in the phylogenetic tree, seven of which were obtained from hospitalized patients (ST474, ST476, ST585, ST858), two were obtained from pig faeces (ST16) and the remaining two were obtained from a wastewater isolate (ST86) and from a healthy human (ST16). In general, no obvious differences arose from the profiles and sequences of virulence genes between strains obtained from foodstuffs/livestock and those causing human infections and only the occurrence of cyl and cpsC-K seemed lineage (ST)-specific (Table S5). These data support previous works indicating the generalist lifestyle of and the opportunist nature of the human infections they cause [31, 51].

Prophages

Prophages-associated sequences are commonly found in clinical strains such as the well-known V583 harbouring seven prophage-like elements, of which only two correspond to intact ones [52]. They are also commonly found in human and animal sources but the identification and distribution of phages is scarcely known in well-characterized populations [53].

Even though its difficult to assess the precise content of phages for each genome given the similarity of different phage proteins in several phages, a remarkable variety of putative prophages and number of intact phages were identified in most strains of this study (Table S6, Fig. 4). PHASTER identified the presence of intact phages in 16 out of 28 isolates from both clinical and non-clinical sources. Four isolates from varied sources (farm pig, chicken meat, clinical, wastewaters) carried two intact phages each. Isolates carrying optrA::Tn6674 contained the highest number of prophages (two carrying 5–6 per genome). We did not identify any obvious relationship between the presence of prophages and isolation source or phylogeny. Nevertheless, and in common to the aforementioned PAI-related genes, the lower clade of the pangenome tree included genomes without intact phages and with a generally poorer number of prophages. Intact or relics of prophages highly identical to sequences from common enterococcal prophages such as phiL1A and vB_IME197 were identified in most strains, but the exact phage(s) present in each genome cannot be ruled out with accuracy. The phiL1A and vB_IME197 phages are widespread in from variable sources [31] and were originally described in clinical settings in UK and China, respectively [54, 55]. Putative prophages associated with other bacterial genera (Listeria, Bacillus, Lactobacillus) from Firmicutes, all classified as temperate Siphoviridae bacteriophages and originally obtained from variable sources (humans, farms, hospital sewage) [55-57], were also identified. A comparative genomic approach performed by La Rosa et al. demonstrated that phage03-like elements (phage03 from V583 similar to the phage B054; GenBank acc. no. NC_009813) have an increased in vivo virulence and pathogenicity and are particularly enriched in clinical isolates [58]. In this study, this phage was putatively identified in three related ST476 strains from hospitalized patients, farm and retail chicken in China and Tunisia (Table S6), once more enphasizing the lack of a demarcation between clinical and non-clinical genomes in relevant mobile genetic elements (MGEs).

We cannot discard the possibility of an erroneous output of the ‘defective’ prophages if these are distributed in different contigs, but the finding of several intact prophages might indicate recent integrations. In any case, the number of prophage sequences found in such a small set of genomes evidences the contribution of prophages to the diversity of populations by acting as vectors of different gene traits during intra- and inter-species genetic exchange, and that their spread is independent from the isolation source [53].

Plasmidome

The extraction and analysis of plasmid replicon sequences was performed using the PlasmidFinder 2.1 (CGE) which in turn is based on the recent and comprehensive review of Lanza et al. [59]. All but the ST227 optrA-negative strain from the healthy farmer in Malaysia (with no rep) carried at least one rep gene (1–5 rep genes) (Fig. 4, Table S7). The strains analysed carried plasmids belonging to all the six families known to date among enterococci, either replicating by a theta mechanism (Rep_3, Inc18, RepA_N) or by rolling circle replication (RCR: Rep_trans, Rep_1 and Rep_2) [60]. When it was possible to associate optrA to a specific plasmid, different plasmid types were identified (Rep_3, Inc18, RepA_N) (Table S7).

Within rolling-circle replicating (RCR) plasmids, most strains (21 out of 28) carried one or two replicons of the Rep_trans family, namely reppDO1 from (n=21), and additionally reppGB354 from (n=5) linked to chloramphenicol resistance (catA7) or repSAP093A from (n=1). Rep_trans plasmids are of small size and widespread among diverse enterococcal populations, being often mobilized by other ABR conjugative theta-replicating plasmids or integrated in the chromosome [59, 61].

The content in the theta-replicating plasmids was highly variable and included replicons from pheromone-responsive RepA_N plasmids, typical of , the broad-host range Inc18, and Rep_3 which includes small non-conjugative and cryptic plasmids obtained from a variety of environments but often associated with bacteriocin production. Only the pHTβ-like plasmids were not identified as they are usually exclusive of species [61]. Notably, all but two strains carried at least one rep from Inc18 or RepA_N plasmids that correspond to the conjugative plasmids usually involved in the acquisition and spread of ABR among different genera. Within RepA_N, pheromone-responsive plasmids were present in most strains. They are unique and highly important in the diversification and evolution of species and the finding of truncated pTEF1 rep genes in most strains might indicate events of chromosomal integration across similar IS elements with the transfer of plasmid and large genomic DNA portions as proposed by Manson et al. [62]. In fact, also several PAI-related genes were identified as truncated (Table S7), which can be related to the movement of the PAI by pheromone-like plasmid integration, as described [62].

The high variability of plasmid content found in optrA-carrying (Table S7) illustrates the diverse nature of MGE that can be found in in general and in particular in optrA-carrying strains.

optrA variants and genetic context

The optrA nucleotide sequences of the 28 genomes were compared against the original optrA gene (optrA_1 or optrA WT; GenBank acc. no. KP399637). Fourteen different variants (seven new), with point mutations not including the original one, were identified (Fig. 1). As the nomenclature of optrA variants is not uniformed across different studies [9, 63] and available databases (CGE), we compiled the different designations and gene mutations in Fig. 5 to facilitate results' interpretation. Three different optrA variants were identified among the four epidemiological related chicken strains from Tunisia (optrA_5, optrA_7 and optrA_8) which have been also identified in different clones (e.g. ST476) from humans and pigs in China [21]. optrA location was inferred, when possible, from the genomic sequences. This gene was either located on the chromosomally-borne Tn6674 (n=10) or on medium-size plasmids (30–60 Kb; n=14) belonging to the plasmid families (RepA_N, Inc18 and Rep_3) usually involved in the acquisition and spread of AMR among different genera of [59]. Independently of optrA localization, the analysis of the optrA genetic platforms carried out in the 27 strains enabled their grouping into three main groups (Fig. 6). These groups, dispersed through the nodes of the phylogenetic tree, showed no obvious correlation with their strain genetic background, neither with the respective optrA gene variant (Fig. 1). They corresponded to:

Fig. 5.

Representation of the alignment of relevant SNPs in the different optrA variants. Different colours indicate the different nucleotides: green – adenine; blue – thymine; red – cytosine; purple – guanine. Squares in grey indicate non-synonymous amino acid mutations. The variant names used are the same as those included in the CGE database until optrA_19 and new variants were attributed from optrA_20 to optrA_27. Variant optrA_4 is also labelled as optrA_6 in CGE database; variant optrA_8 is also labelled as optrA_9 and optrA_10 in CGE database; variant optrA_11 is also labelled as optrA_12 in CGE database. areference [9]. breference [63] . Abbreviations: NA, European Nucleotide Archive.

Group I (n=10) included three chicken strains from Tunisia (one ST21 and two ST476), all the pig strains from Malaysia (ST476, ST755, ST756 and ST765) and three hospital strains from France (ST480) and China (ST21 and ST858) carrying five different optrA variants (Fig. 6). Three additional genomic regions were included for comparison (Tn6674, TZ2 and A101). Whole genome sequences of these strains are not available so they were not included in the genomic analysis. These strains shared a chromosomal segment containing fexA-optrA adjacent to the entire Tn554 from , varying between 11.7–12.9 kb, which is identical to the recently identified Tn6674 [64]. Tn6674 is a chromosomal 12.9 kb Tn554-related transposon obtained from the strain E1731 (GenBank acc. no. MK737778.1) that contains 7 ORFs: tnpABC and resistance genes [spc-erm(A)-fexA-optrA] conferring resistance to spectinomycin, macrolides, phenicols and oxazolidinones, respectively; so fexA and optrA are additionally carried in Tn6674 in comparison to Tn554. As illustrated in Fig. 6, that includes not only some of the genomes of this study but also optrA platforms available in GenBank, Tn6674 has been driving optrA spread amongst hospitalized patients (France, China), healthy humans (China, Malaysia) and food-producing animals (China, Malaysia, Tunisia) isolates across different continents during 2012–2018. A similar structure was also recently described in an optrA-carrying ST16 from a patient in Greece (sequence not available) [11]. For the isolates and corresponding studies possible to access, this genetic platform could be positively transferred by filter mating [25].

Group II (n=11) included chicken strains from Colombia (two ST59 and one ST489) and Tunisia (ST859), cattle and pig samples from USA (ST141 and ST489), and the clinical strains from Spain (three ST585 and one ST474) and China (two ST476) carrying eight optrA variants. optrA-carrying regions of pE349 and p10-2-2 plasmids were also included. Whole genome sequences for these strains are not available so they were not included in the genomic analysis. Most of them (n=7) contained a common impB-fexA-optrA segment that showed high similarity to the originally described optrA-positive pE349 plasmid (human, ST116; GenBank acc. no. KP399637) as well as other plasmids obtained from human and pig strains in China (ST27, ST59, ST480, ST585) [21]. This plasmid backbone has been also recently described in isolates from Sweden and Ireland [3]. impB showed the highest homology with a type VI secretion system from (GenBank acc. no. ERK33158) and has been suggested to be a plasmid-associated DNA repair gene eventually playing a role in the survival under stressful environments [65]. The ST859 chicken strain (728T) additionally carried erm(A) in the same contig of 12.9 kb in common to pig ST59 (p10-2-2 plasmid) and hospitalized patient ST476 strains from China [21]. optrA was located on plasmids ranging from ca. 30 to 60 kb in all cases [strains from Group 2 of this study (Fig. 6) and those referred from China, Colombia and the USA] [5, 12, 14, 21] belonging to RepA_N, Inc18 or Rep_3 families whenever known. Some optrA-carrying plasmid segments (n=5) of this group were flanked by a transposase of the ISL3 family from (GenBank acc. no. WP_029876135.1), a species previously recognized as a reservoir of optrA [32].

Group III comprised four miscellaneous strains from pigs (ST16 and ST754), humans (ST16) and wastewaters (ST86). ST16 from a healthy human in USA (Efs 599) and ST86 from wastewaters (12T) in Tunisia commonly carried optrA adjacent to an araC-NADH-spoVG-helicase region highly similar to that of a blood E016 strain from China, despite the fact that in these cases optrA was variably located on a plasmid (ST86) or on chromosome (E016) [21, 37]. The two remaining strains (IBUN9046YE and Enfs51) only shared the araC-optrA segment with those aforementioned, which has been commonly described flanked by transposases and Tn558 relics in different enterococcal species [12, 21].

Features of a novel chimeric pheromone-responsive-like optrA-carrying plasmid

We were able to presumptively identify the contigs belonging to the optrA-carrying plasmid of the Tunisian 728T strain from this group (Fig. S4). The sum of such contigs was 39 173 bp, which is compatible with the size inferred from the S1-PFGE gel (~40 kb) [64]. Using EggNOG it was possible to identify the function of some of the proteins codified by the putative ORFs encoded in the plasmid, including ABR [optrA, fexA, erm(A), tet(L)], plasmid replication and stability (repA, parA, prgN), but not of conjugative transfer (compatible with the negative conjugation results), a toxin-antitoxin (TA) system (mazE, mazF), class II bacteriocin (bacA, bacB), cell fillamentation (fic), and a putative ABC-type multidrug transporter (CDS_2 and CDS_3 in the figure). The entire contig 23 of 12 971 bp (ermA-optrA-fexA-impB; Figs 6 and 7) was identical to that of p10-2-2 plasmid from a pig ST59 in China [21]. RepA had the highest homology (92 % in nucleotides) with that from the pheromone-responsive plasmids pTW9 (GenBank acc. no. NC_014726), pPD1 (GenBank acc. no. D78016) and pCF10 (GenBank acc. no. AY855841.2) previously identified in MDR strains obtained from different hosts and areas. Upstream of repA is parA and prgN, previously described in the pheromone-responsive pTEF2 (GenBank acc. no. AE016831) and pPD1 plasmids as putatively involved in plasmid partitioning and replication control, respectively. The region spanning from repA to prgN had the highest homology with the same repA-parA-prgN region of the VanB-type pMG2200 plasmid (GenBank acc. no. AB374546), still of only 73 % in nucleotides, meaning this corresponds to a putative new plasmid replication region. Bacteriocin and TA gene clusters were similar to that from the pheromone-responsive pAMS1 plasmid (GenBank acc. no. EU047916) and that described in an optrA-positive pEF123 plasmid from chicken in China (GenBank acc. no. KX579977), respectively. The ABC transport system was described in pVEF3 (GenBank acc. no. NC_010980) and pVEF4 (GenBank acc. no. FN424376) plasmids from chicken strains, though it was not associated with a reduced susceptibility to several substances tested at that time by Sletvold et al. [66]. Taking all together, our data suggests that this optrA-carrying plasmid from chicken meat corresponds to a putative novel pheromone-responsive-like plasmid containing genes of different origins including the first optrA pE349 plasmid described, as has been commonly observed worldwide and particularly in the non-clinical setting [2, 14, 21, 37].

Comparison of chromosomal optrA-carrying contigs

We further compared the entire chromosomal optrA-carrying contig (~44 kb) of a strain from group 1 (612T) by mapping it against the remaining genomes (Fig. 7).

Fig. 7.

Comparison of the chromosomal 612T optrA-containing contig with the remaining strains. Coloured arrows represent genes and the arrows indicate the orientation. Grey arrows represent hypothetical proteins of unknown function; the remaining colours represent the gene function according to the COGs database classification indicated in the key. Abbreviations: ST, sequence type; ENA, European Nucleotide Archive.

In addition to the Tn6674-like element, this contig included genes involved in different cellular functions like oxidative stress response (tpx), redox-sensing transcriptional regulation (rex), carbohydrate transport and metabolism [rbsUDKR operon (ribose transport and regulation), ulaA (PTS system ascorbate-specific transporter)], coenzyme transport and metabolism (folC and thiI – folate and thyamine biosynthesis respectively), protein translation (valS, valyl-tRNA synthetase), phosphate transport and regulation (phoB), and DNA mismatch repair (mutS2). It was interesting to identify the same conserved chromosomal region (32 199 bp from ∆radC to mutS2) in all but the animal/food strains from Colombia and one from the USA (n=24/28), independently if strains carried optrA on this or other chromosomal contigs, on plasmids or even lacked it. optrA adjacent to Tn554 or Tn554 relics has been described either embedded on chromosomal platforms or on plasmids, in a plethora of enterococci strains from different host and regions [19, 21, 63].

As the transposons of Tn554 family, Tn6674 does not fall into any standard transposon category since it lacks inverted or direct terminal repeats, and it does not generate a duplication of a target sequence upon transposition. It has been suggested that Tn554 is a prophage-like element characterized by a highly efficient site-specific integration-excision mechanism [67]. Transfer of Tn554 between Gram-positive species, especially known between , typically involves integration into another mobile element and hitch-hiking across via conjugation, transformation or transduction, so it can be found in multiple copies in the cell. Li et al. [68] have experimentally confirmed the ability of Tn6674 to form circular forms and suggested its functional activity. Tn554 has an exceptional preference for integration into a chromosomal insertion site called att554, at a lower frequency into secondary insertion sites (mostly in the chromosome) or more rarely on plasmids if that site is occupied or deleted, following transfer [67, 69]. The transposition of Tn554 generally interrupts the radC gene encoding a putative DNA repair protein and we herein confirm the same with Tn6674 (∆radC in Fig. 7) [70]. The identification of the optrA-carrying Tn6674 adjacent to a common chromosomal 32 kb platform in most of the strains analysed suggests that this specific chromosomal region may be a hotspot for Tn6674 integration and mobilization as observed for other bacterial species in which hotspots tend to be flanked by recombination and repair core genes [71]. A GenBank search of this 32 kb nucleotide sequence revealed its presence in >500 other genomes (562 out of 2701 subject genomes) from different sources (data not shown), thus demonstrating its conserved nature among this species. Remarkably, the Enfs94 strain from the healthy farmer lacking optrA, whom was in contact with pigs colonized with optrA-positive , also contained this conserved chromosomal region and an optrA acquisition might occur.

According to available studies, Tn554 seems to transfer primarily by transducing phages independently of flanking regions [69]. Although the extent of horizontal gene transfer due to transduction with bacteriophages among enterococci remains largely unknown, the transfer of ABR genes between different enterococcal species mediated by phages has been already demonstrated [54, 60], as well as the effect of specific antibiotics (e.g. fluoroquinolones [52]) in prophage release and gene spread within isolates. optrA was not identified as integrated in a prophage in this work, but a diversity of prophages was found integrated in the chromosome of several strains, with those carrying optrA in Tn6674 being generally the most enriched in prophage sequences including intact phages. Strikingly, phage-mediated lateral transduction was recently described as a putative new universal mechanism of transfer promoting the efficient exchange of large portions of bacterial genomes at high frequencies [72]. optrA has been identified in many and diverse genetic platforms (plasmids, ICEs, phages) amongst different Gram-positive genera (enterococci, staphylococci, streptococci). Among enterococci, mostly , optrA has been commonly described as adjacent to fexA (conferring resistant to phenicols), relics of Tn554 and Tn558 [optrA-erm(A) clusters] and surrounded by ISs (e.g. IS1216 elements) in a great variety of several and unique combinations of these and other sequences, on plasmids or chromosomal platforms [21]. In pig isolates, optrA has been recently described within a novel composite IS1216E-araC-optrA-hp-catpC194-IS1216E transposon integrated on a phage similar to Φm46.1 [73]. optrA was identified in prophages and ICEs only in [32, 73, 74], but the mobilization of optrA through phage-mediated transduction cannot be discarded also in enterococci. Wang et al. [75] also recently demonstrated by a metagenomic approach in swine feedlot wastewater samples that phages were more likely to harbor ABC transporter family and ribosomal protection genes including optrA. Adding to the fact that transfer frequency of Tn554 via transduction is higher than most plasmids [69], we could expect a tremendously fast dissemination of optrA among different Gram-positive species and across different hosts and settings, as seems to be occurring more expressively in the animal production setting. It has been recently hypothesized that the virome may play a larger role in ABR transfer in non-clinical environments where phages are abundant and frequently carry ABR genes [76].

Conclusions

This work constitutes one of the most comprehensive studies in comparing optrA-positive genomes from disparate origins and the first to depict extended profiles in key adaptive features including antibiotic resistance, virulence, prophages and the plasmidome. Our study evidenced the amazing variability of clones and optrA-genetic platforms or plasmids that can be found in a single species, and on the other hand the easy possibility of finding identical strains or optrA genetic platforms in samples from different sources and distant countries. The analysis of publicly available strains carrying optrA revealed that its immediate gene vicinity was generally identical in chromosomal (Tn6674) and plasmid (impB-fexA-optrA) backbones. We herein identified a common Tn554::fexA-optrA chromosomal platform, the recently identified novel Tn6674, in different linezolid-resistant clones from livestock, meat, animal farmers and hospitalized patients across different continents, and proposed a conserved chromosomal hotspot for optrA-fexA integration. optrA-carrying strains from livestock/foodstuffs and human infections could be closely related and harbour indistinguishable key adaptive features, which reinforces the generalist lifestyle of this species and highlights the relevance of the animal reservoir and potential role of the food-chain in the transmission of optrA-positive strains. On the other hand, we provide pieces of evidence that can easily acquire and exchange optrA-carrying platforms or plasmids with other Gram-positive pathogens, besides a variety of other accessory genetic information from bacteria of the phylum. The finding of optrA in strains sharing identical adaptive features and genetic backgrounds across different hosts and countries suggests the occurrence of common and independent genetic events occurring in distant regions and might explain the easy de novo generation of optrA-positive strains. The usual association of optrA with Tn554-like or IS-like elements will certainly aid in a global and dramatic increase of optrA carriage and spread with a serious impact on the efficacy of linezolid for the treatment of Gram-positive infections.

Data Bibliography

The genome assemblies of all the genomes in this study were downloaded from the NCBI genome database, and identifiers are listed in Table 1.

Click here for additional data file.

Click here for additional data file.