Prevalence and Genetic Analysis of Resistance Mechanisms of Linezolid-Nonsusceptible Enterococci in a Tertiary Care Hospital Examined via Whole-Genome Sequencing

(1) Background: Linezolid plays an important role in the treatment of invasive infections caused by vancomycin-resistant enterococci after its introduction to clinical practice. However, a detailed examination of linezolid-nonsusceptible enterococci (LNSE) is required. In this study, we attempted to analyze the mechanisms of LNSE strains isolated from a tertiary care hospital. (2) Methods: From 2019 to 2020, 18 Enterococcus faecalis, 14 E. faecium, and 2 E. gallinarum clinical isolates were collected at Severance Hospital. Agar dilution was performed to evaluate precise linezolid minimum inhibitory concentrations (MICs). Short-read whole-genome sequencing (WGS) was used to analyze resistance determinants. (3) Results: The presence of the optrA gene was likely the primary resistance mechanism in these three species, typically demonstrating a MIC value of 8 μg/mL. The co-existence of the cfr(D) and poxtA2 gene was the second major mechanism, primarily predicting a phenotype showing intermediate susceptibility to linezolid. G2576U mutation on 23S rRNA was only found in E. faecium; it mediated the most significant increase in linezolid MIC. (4) Conclusion: This is the first report examining poxtA2–cfr(D) co-harboring clinical enterococcal isolates in Korea and demonstrating the poxtA EF9F6-harboring clinical E. gallinarum strain worldwide. The comparison with resistance-gene-containing fragments in the isolates obtained from different countries and different sources demonstrated the spread of linezolid-resistance genes and suggested the possibility of foodborne transmission.


Introduction
Enterococcus species have been recognized as one of the major pathogens responsible for many infections including healthcare-associated infections in humans for decades. In addition to their intrinsic resistance to a broad spectrum of antibiotics, they can also acquire plasmids and other genetic elements that harbor antimicrobial-resistance genes [1], rendering them multi-drug-resistant pathogens, which might be difficult to treat even with vancomycin. Linezolid, the first oxazolidinone antibacterial agent approved for commercial use by the US Food and Drug Administration (FDA) in 2000, has been considered a useful therapeutic option for various infections caused by Gram-positive bacteria. However, since linezolid was introduced for clinical use, the presence of linezolid-nonsusceptible enterococci (LNSE) has continuously been reported over the years [2].

Evolution of Linezolid Nonsusceptibility in Severance Hospital over the Past Five Years
The five-year linezolid-susceptibility surveillance data of Severance Hospital are presented in Table 1. Linezolid resistance increased by 3.3% (from 3.4% to 6.7%) in E. faecalis and 1.5% (from 3.5% to 5.0%) in E. faecium. The increase in the proportion of E. faecalis strains demonstrating intermediate resistance to linezolid (from 0.5% to 2.6%) was greater than that of E. faecium (from 0.1% to 1.4%). Overall, a higher proportion of linezolidnonsusceptible E. faecalis (LNSEFA) strains than that of linezolid-nonsusceptible E. faecium (LNSEFM) strains was observed.

Linezolid Minimum Inhibitory Concentration (MIC) Distribution and Resistance Mechanisms
While the proportion of LNSEFAs was higher than that of LNSEFMs, LNSEFMs demonstrated overall higher MICs (Table 2) scattered. The two LNSEFM strains with a MIC of 32 µg/mL and one with a MIC of 16 µg/mL were attributed to the G2576U mutation on 23S rRNA. Similar to LNSEFA, the optrA genes in LNSEFMs were associated with a MIC of 8 µg/mL. An LNSEFM strain carrying the cfr(D) gene alone showed a MIC of 8 µg/mL; however, when the strain coharbored the poxtA2 gene, a tendency of intermediate susceptibility was observed. In addition, four LNSEFM strains did not have mutations on the linezolid ribosomal binding site and did not carry any known linezolid-resistant genes. As for the E. gallinarum isolates carrying poxtA EF9F6 genes, the MICs of both isolates were 8 µg/mL. Overall, the presence of optrA was the most common mechanism (50.0%), and the associated MIC was typically 8 µg/mL. The co-existence of poxtA2 and cfr(D) was the second common mechanism (14.7%), and the associated MIC was 4 µg/mL, except for one strain with a MIC of 8 µg/mL. G2576U mutation was not found in LNSEFA. However, it was the primary reason for the significantly higher MIC values of LNSEFM; 75% of LNSEFMs with G2576U mutation showed MICs from 8 µg/mL to 32 µg/mL.

Genetic Environment of Different Resistance Determinants
Resistance gene-harboring regions were divided into four groups according to their characteristics, such as the cfr(D) gene, RNA-polymerase-encoding gene, and ferredoxincoding gene. The resistance genes of strains EFM 30, EFM 31, EFA 39, EFA 62, and EFA 64 were located on the plasmids.

Genetic Environment of Different Resistance Determinants
Resistance gene-harboring regions were divided into four groups according to their characteristics, such as the cfr(D) gene, RNA-polymerase-encoding gene, and ferredoxincoding gene. The resistance genes of strains EFM 30, EFM 31, EFA 39, EFA 62, and EFA 64 were located on the plasmids.
Group I included all the cfr(D)-positive isolates (four E. faecium and three E. faecalis strains). The resistance gene-involving contigs of strains EFM 30, EFM 31, EFA 39, EFA 62, and EFA 64 were highly similar, co-harboring a poxtA2 gene upstream. It was confirmed that both the cfr(D) and poxtA2 genes were located on the plasmid. This was also evidenced by the presence of the repR gene 1495 bp downstream of the cfr(D) gene. Two contigs from E. faecium chromosome carrying the cfr(D) gene alone were also observed, which showed a GMP synthase gene downstream of the cfr(D) gene in the opposite direction ( Figure 1). Isolates termed EFM belong to E. faecium and isolates termed EFA belong to E. faecalis. The asterisk * indicates that the contig was located on the plasmid. The nucleotide sequence of the cfr(D) gene was 99.9% identical to NG067192. The nucleotide sequence of the poxtA2 gene was 100% identical to NZ_MZ603802.
Group II included five E. faecalis strains (EFA 35, EFA 3, EFA 63, EFA 20, and EFA 26) and two E. faecium strains (EFM 6 and EFM 2) ( Figure 2a). These strains showed a ferredoxin-coding gene downstream of the optrA gene in the same direction. Except for EFM 2, a fexA gene that mediates resistance to chloramphenicol and florfenicol was located 742 bp upstream of the optrA gene, and a 23S rRNA (adenine (2058)-N (6))-dimethyltransferase-encoding gene ermA was located 1417 bp downstream of the optrA gene in the opposite direction. Due to the contig length limitation, more information about the environment of the optrA gene of EFM 2 could not be obtained. However, it was found that EFA 599799 shared high similarity to the EFM2 optrA-containing contig and additionally carried a gene encoding a NAD(P)H oxidoreductase. Isolates termed EFM belong to E. faecium and isolates termed EFA belong to E. faecalis. The asterisk * indicates that the contig was located on the plasmid. The nucleotide sequence of the cfr(D) gene was 99.9% identical to NG067192. The nucleotide sequence of the poxtA2 gene was 100% identical to NZ_MZ603802.
Group II included five E. faecalis strains (EFA 35, EFA 3, EFA 63, EFA 20, and EFA 26) and two E. faecium strains (EFM 6 and EFM 2) ( Figure 2a). These strains showed a ferredoxincoding gene downstream of the optrA gene in the same direction. Except for EFM 2, a fexA gene that mediates resistance to chloramphenicol and florfenicol was located 742 bp upstream of the optrA gene, and a 23S rRNA (adenine (2058)-N (6))-dimethyltransferaseencoding gene ermA was located 1417 bp downstream of the optrA gene in the opposite direction. Due to the contig length limitation, more information about the environment of the optrA gene of EFM 2 could not be obtained. However, it was found that EFA 599799 shared high similarity to the EFM2 optrA-containing contig and additionally carried a gene encoding a NAD(P)H oxidoreductase.
Group III included EFA 25 alone (Figure 2b). These strains demonstrated a unique genetic environment in this study, carrying a gene encoding RNA polymerase upstream of the optrA gene.
Group IV included ten optrA-harboring isolates, one of which was E. faecium and the remaining nine were E. faecalis. They were associated with Tn554 components, followed by fexA, optrA, ribonuclease J1-encoding gene, and two heterodimeric efflux ABC transporter genes efrA and efrB. EFA 10 and EFA 28 both showed an insertion of approximately 900 bp between the Tn554 and the fexA gene (Figure 2c). For a detailed examination of Tn554 and Tn554-containing gene fragments, EFM 5 and EFA 4 were selected as representative isolates for comparison with the Tn554 reference sequence (GenBank accession no. X03216.1) (Figure 3). The original Tn554 comprises three transposase genes tnpA, tnpB, and tnpC, ant(9)-Ia (also known as the spc gene), and the erythromycin resistance methylase-encoding gene ermA. In EFM 5, an optrA-containing fragment of approximately 7800 bp replaced Group III included EFA 25 alone (Figure 2b). These strains demonstrated a unique genetic environment in this study, carrying a gene encoding RNA polymerase upstream of the optrA gene. no. X03216.1) (Figure 3). The original Tn554 comprises three transposase genes tnpA, tnpB, and tnpC, ant(9)-Ia (also known as the spc gene), and the erythromycin resistance methylase-encoding gene ermA. In EFM 5, an optrA-containing fragment of approximately 7800 bp replaced the ant(9)-Ia gene. Although EFM 5 harbored the Tn554 transposase-encoding genes tnpA, tnpB, and tnpC, when compared to the Tn554 reference sequence, the sequence similarity only reached 70%. The poxtA EF9F6 gene, which was first reported very recently, was found in four strains in our study: EGA 11, EGA 12, EFM 2, and EFM17. Due to the limited contig length, the genetic environment of the poxtA EF9F6 gene in EFM 2 and EFM17 did not yield sufficient information. The poxtA EF9F6 gene of EGA 11 and EGA 12 shared the same genetic environment. It was bracketed by IS1380-like transposase-coding gene upstream and downstream.

Clonal Relatedness and Other Molecular Features
Two novel E. faecalis sequence types (ST) (ST1166 and ST1167) and four alleles (aroE113, gdh107, xpt102, and yqiL105) were identified for the first time. Closely related isolates were divided into six groups, as shown in Figure 4. In any of Group 1, Group 2, Group 3, and Group 4, despite the fact that they were isolated from different patients in different departments, they had close clonal relatedness and a very high degree of genome similarity, suggesting nosocomial infections caused by LNSEFA. All the strains in Group 5 and Group 6 were also clonal. They were all isolated from inpatients in the Hematology Department, revealing the existence of LNSEFM in the ward at least from June 2019 to November 2019. The poxtA EF9F6 gene, which was first reported very recently, was found in four strains in our study: EGA 11, EGA 12, EFM 2, and EFM17. Due to the limited contig length, the genetic environment of the poxtA EF9F6 gene in EFM 2 and EFM17 did not yield sufficient information. The poxtA EF9F6 gene of EGA 11 and EGA 12 shared the same genetic environment. It was bracketed by IS1380-like transposase-coding gene upstream and downstream.

Clonal Relatedness and Other Molecular Features
Two novel E. faecalis sequence types (ST) (ST1166 and ST1167) and four alleles (aroE113, gdh107, xpt102, and yqiL105) were identified for the first time. Closely related isolates were divided into six groups, as shown in Figure 4. In any of Group 1, Group 2, Group 3, and Group 4, despite the fact that they were isolated from different patients in different departments, they had close clonal relatedness and a very high degree of genome similarity, suggesting nosocomial infections caused by LNSEFA. All the strains in Group 5 and Group 6 were also clonal. They were all isolated from inpatients in the Hematology Department, revealing the existence of LNSEFM in the ward at least from June 2019 to November 2019.
A consistency in the resistance mechanisms and MIC pattern in E. faecalis was found: 93.3% (14/15) of the isolates carrying the optrA gene showed MICs of 8 µg/mL and the isolates co-harboring poxtA2 and cfr(D) showed MICs of 4 µg/mL (Figure 4a). Compared with E. faecalis, the clonal relatedness between E. faecium isolates was high (all strains belonged to CC17); however, the resistance mechanisms were complex, and the MIC values were scattered (Figure 4b). For both E. faecalis and E. faecium, it can be inferred that linezolid susceptibility was not remarkably related to ST and was only associated with the resistance mechanism.
Antibiotics 2022, 11, 1624 7 of 12 isolates co-harboring poxtA2 and cfr(D) showed MICs of 4 μg/mL (Figure 4a). Compared with E. faecalis, the clonal relatedness between E. faecium isolates was high (all strains belonged to CC17); however, the resistance mechanisms were complex, and the MIC values were scattered (Figure 4b). For both E. faecalis and E. faecium, it can be inferred that linezolid susceptibility was not remarkably related to ST and was only associated with the resistance mechanism.

Discussion
In Korea, relatively few reports have been published examining linezolid resistance in clinical E. faecalis isolates [21]. However, some trends have been observed from other surveillance data examining E. faecium. According to a Kor-GLASS report published in 2017, 0% of E. faecium strains were linezolid-resistant, while approximately 3% of the strains showed intermediate susceptibility to linezolid [24]. In our study, the proportion

Discussion
In Korea, relatively few reports have been published examining linezolid resistance in clinical E. faecalis isolates [21]. However, some trends have been observed from other surveillance data examining E. faecium. According to a Kor-GLASS report published in 2017, 0% of E. faecium strains were linezolid-resistant, while approximately 3% of the strains showed intermediate susceptibility to linezolid [24]. In our study, the proportion of linezolid-resistant E. faecium was higher, and the proportion of E. faecium strains showing intermediate susceptibility to linezolid was reduced. Even though the difference in linezolid susceptibility can be influenced by many factors, this shift from intermediate susceptibility to resistance may be attributed to the continued widespread use of linezolid [25] and the high adaptability of enterococci to antibiotic pressure. Hence, there is an urgent need for constant LNSE monitoring and regulation for each hospital through wide surveillance networks.
Before 2012, the primary oxazolidinone resistance mechanism known in both E. faecalis and E. faecium involved mutations on 23S rRNA and ribosomal proteins, which could be spread by clonal dissemination [3]. However, since the discovery of plasmid-mediated linezolid resistance, the influence on the resistance mechanism has demonstrated a gradual shift in the presence of optrA, poxtA, and cfr genes. The presence of optrA has now become the dominant mechanism of resistance in the strains isolated from 2014 to 2016 [26]. This trend has also been demonstrated in a study on the clinical E. faecalis isolates obtained in Korea; all linezolid-nonsusceptible isolates were optrA-positive [21]. In this study, the presence of G2576U mutation on 23S rRNA led to a more significant increase in linezolid MIC values than other mechanisms. The strong influence of G2576U mutation on the increase in the MIC value has also been reported in Korea, where the MIC values of both linezolid-resistant E. faecium isolates were higher than 64 µg/mL [23]. The four linezolidresistant strains EFM 13, EFM 22, EFM 29, and EFM 46 did not demonstrate any existing resistance mechanism. A similar conclusion was drawn in a Korean study conducted in 2017 that no corresponding mechanism was found in two-thirds of linezolid-resistant E. faecium strains [22].
When comparing the genetic environment of resistance genes, not only the strains isolated in this research but also published sequences were included if a high similarity was found. A good consistency was observed in the sequence of the optrA gene between previous studies and this study, but the genetic environment of optrA-flanking regions displayed relatively low homology. Two E. faecalis strains examined in an international surveillance study (EFA 838523 and EFA 599799) [26] carrying a ferredoxin-coding gene downstream of the optrA gene belonged to Group II. A Korean foodborne E. faecalis isolate [27] and a Chinese E. faecalis isolate of human origin [28] harboring the optrA gene were categorized as Group III. As for EFA4 and other strains in Group IV, the fexA-optrA fragment was inserted at the end of Tn554. One possible explanation is that the resistance gene-containing fragment was acquired in the past few generations, and such a result was obtained after continuous gene rearrangement.
Thus far, no cfr-carrying clinical enterococcal isolates have been reported in Korea. However, several cfr-mediated linezolid resistance mechanisms have been identified in the S. aureus strains isolated from pig carcasses [29] and the E. faecalis strains isolated from food-producing animals [30], indicating the possibility of foodborne transmission. The co-location of optrA and cfr has been reported in human E. faecium isolates observed in Ireland [31] and Italy [32], providing evidence of their prevalence worldwide. The gene fragments in Group I carrying both poxtA2 and cfr(D) are highly similar to those of E. faecalis V386 isolated from Italy (NCBI Reference Sequence: NZ_MZ603802.1), with a repR gene located downstream. Compared with the 37 coding DNA sequences (CDSs) of v386, 34 CDSs of the two E. faecium strains were the same, and 13 CDSs of the three E. faecalis strains were the same. Hence, it can be inferred that the resistance was transmitted by the plasmid, even across species. A study has revealed the co-existence of poxtA2 and cfr(D) gene in two E. faecalis strains with different genetic environments. Both strains were isolated in Korea in 2018 [30].
As confirmed by MobileElementFinder, strains EFM 30, EFM 31, EFA 39, EFA 62, and EFA 64 had genes located on plasmids and were at risk of widespread horizontal transmission. However, one of the limitations of this study was that we only used the second-generation sequencing method. If MobileElementFinder did not show that the resistance gene was located on the plasmid, it could not be concluded that the resistance gene was located on the chromosome. Long-read sequencing methods such as Oxford Nanopore MinION could be applied to help with defining gene location and plasmid reconstruction.
In this study, only one of the strains carrying both poxtA2 and cfr(D) genes (EFM 30) was resistant to linezolid, while the other four strains only had intermediate resistance.
According to previous studies, a single poxtA2 gene was responsible for the resistance to linezolid in enterococci with a MIC of 8 µg/mL [18]. That is to say, lower linezolid MICs could be observed in the strains co-harboring poxtA2 and cfr(D) genes in comparison to those strains carrying only the poxtA2 gene. On the one hand, there is no one-to-one correspondence between bacterial antibiotic-resistant phenotypes and genotypes. A similar phenomenon has been reported in other studies. A pig-origin E. casseliflavus strain coharboring poxtA2 and cfr(D) gene has been reported to present a MIC of 8 µg/mL [33]. The same research group later reported a poxtA2-cfr(D) co-harboring plasmid pV386 in Enterococcus species. Of the three strains carrying pV386, two had MICs of 4 µg/mL and the other had a MIC of 2 µg/mL [19], which means that none of the three strains was resistant to linezolid. In another study of enterococci isolated from pigs, one E. faecalis strain was reported that carried both poxtA2 and cfr(D) genes. It was susceptible to linezolid with a MIC of 2 µg/mL [34]. On the other hand, the phenotype of resistance genes varies when expressed in different species. For example, the cfr(D) gene alone generally does not confer linezolid resistance when expressed in an enterococcal background but confers linezolid resistance when expressed in Escherichia coli [8]. To date, not much research has been conducted on poxtA2-cfr(D) co-harboring enterococci, and follow-up studies are needed to specifically explain the reason why low linezolid MICs (≤4 µg/mL) were still observed in a strain co-harboring two linezolid-resistant genes.

Bacterial Strains
All the enterococcal isolates were recovered from 2016 to 2020 from clinical specimens collected from patients visiting or admitted at Severance Hospital, a tertiary-care hospital with 2437 beds in Seoul, South Korea. The clinical samples were very diverse, including urine, wound secretions, blood, catheter tips, tissue, bile, bronchoalveolar lavage, and peritoneal fluid. Species identification was performed using a MALDI-TOF mass spectrometry instrument (Microflex ® LRF, Bruker Daltonik, Bremen, Germany). Antimicrobial susceptibility was determined by Vitek 2 automated system (bioMériux, Marcy l'Étoile, France) with AST-P600 cards (bioMériux). All linezolid-nonsusceptible enterococcal isolates were confirmed through disk diffusion (Thermo-Fischer Scientific, Waltham, PA, USA) or antibiotic gradient diffusion methods (bioMériux). The results of the antimicrobial susceptibility testing were interpreted according to the Clinical and Laboratory Standards Institute (CLSI) Guidelines, the 32nd edition [35]. Linezolid susceptibility data were retrieved from the laboratory information system for analysis.
To investigate linezolid resistance mechanisms in the clinical enterococcal isolates, from June 2019 to June 2020, one or two E. faecalis and E. faecium were randomly selected from the linezolid-nonsusceptible isolates collected each month. Two rarely recovered E. gallinarum were also analyzed for resistance mechanisms for linezolid nonsusceptibility. In total, 18 E. faecalis, 14 E. faecium, and 2 E. gallinarum were tested for further investigation, including MIC determination and whole-genome sequencing (Table S1).

Linezolid Susceptibility Testing
In order to determine the minimum inhibitory concentrations (MICs) against linezolid among the 34 isolates selected for further investigation, Agar dilution was used to reconfirm the MIC of the selected strains following CLSI Guideline M7, 11th edition [36]. The linezolid powder used to prepare agar plates was from Dong-A Pharmaceutical (Seoul, South Korea). The MIC range tested was from 0.25 µg/mL to 64 µg/mL. All MICs were interpreted according to CLSI M100-ED32. The isolates were determined as susceptible to linezolid when MIC ≤ 2 µg/mL, intermediate resistant to linezolid when MIC = 4 µg/mL, and resistant to linezolid when MIC ≥ 8 µg/mL. The isolates for which the linezolid MICs were greater than or equal to 4 µg/mL were defined as linezolid-nonsusceptible. E. faecalis ATCC 29212 and S. aureus ATCC 29213 served as quality control (QC) strains.

Conclusions
In conclusion, this study is the first to report the co-existence of poxtA2 and cfr(D) genes in clinical E. faecalis and E. faecium isolates in Korea and poxtA EF9F6-harboring clinical E. gallinarum globally. The similarities in the genetic context between Korean clinical isolates and the isolates from other countries and resources strongly indicate the spread of linezolid-resistant determinants between countries and the possibility of foodborne transmission. A reasonable explanation for four linezolid-resistant E. faecium strains with unknown mechanisms remains a problem that should be examined in future studies.