Two enterococcal species, Enterococcus faecium and Enterococcus faecalis, are frequently associated with human disease outbreaks characterized by high morbidity rates in the context of both hospital- and community-acquired infections (Cetinkaya et al., 2000; Salgado, 2008). Vancomycin-resistant enterococci (VRE) are of particular concern. According to the Antibiotic Resistance Threats Report of the US Centers for Disease Control and Prevention (CDC), approximately 54,500 VRE infections occurred among hospitalized patients in 2017, resulting in an estimated 5,400 estimated deaths and healthcare costs of 539 million dollars (CDC, 2019).

Enterococci have high potential to acquire antimicrobial resistance (AMR) through the transfer of mobile genetic elements or the development of mutations, which constitute advantages over nosocomial pathogens (Arias and Murray, 2012). Glycopeptide antibiotics, such as vancomycin and teicoplanin, are frequently used to treat serious infections caused by multidrug-resistant Enterococcus spp. (Munita and Arias, 2016; Ahmed and Baptiste, 2018). However, resistance to glycopeptides, especially vancomycin, has exacerbated the problems caused by enterococcal infections since VRE was first documented in the late 1980s (Heintz et al., 2010). Vancomycin resistance is mediated by various types of van gene clusters (vanA, vanB, vanC, vanD, vanE, vanG, vanL, vanM), among which vanA and vanB are the most frequently implicated in VRE outbreaks. VanA is responsible for high levels of resistance to vancomycin and teicoplanin, while vanB and vanC are only associated with vancomycin resistance (Werner et al., 2008; Hollenbeck and Rice, 2012; Munita and Arias, 2016). The transfer of vanA and vanB occurs through the acquisition of chromosomal transposons between E. faecium and E. faecalis, while vanC is intrinsically present in Enterococcus gallinarum and E. casseliflavus (Eliopoulos and Gold, 2001). Unfortunately, the situation is worsened by the emergence of resistance to last-resort antibiotics like linezolid and daptomycin which means clinicians often have no viable alternative treatments (Palmer et al., 2011; Klare et al., 2015; Bender et al., 2018).

The widespread AMR of enterococci necessitates strategies for clinical management of infections, including active surveillance and rapid recognition of outbreaks. To elucidate the mechanisms underlying AMR and prevent the dissemination of bacteria, molecular epidemiological analysis of infections is needed. Conventional molecular diagnostic approaches to the analysis of the genetic relatedness of strains include pulsed-field gel electrophoresis (PFGE) and multilocus sequence typing (MLST). PFGE was among the techniques ushering in the era of molecular epidemiology, and has long been considered the gold standard for bacterial strain typing (Goering, 2010; Neoh et al., 2019). However, PFGE is time-consuming, and the choice of restriction enzymes and electrophoresis conditions is limited (Johnson et al., 2007; Pinholt et al., 2015). MLST is another widely used approach in phylogenetic studies, food safety surveillance, and outbreak investigations (Ruiz-Garbajosa et al., 2006). However, MLST has low discriminatory power because it only relies on defining allelic profiles of the seven housekeeping genes (Homan et al., 2002; Carter et al., 2016).

Several recent studies have demonstrated the usefulness of whole-genome sequencing (WGS) for investigating outbreaks. WGS has high discriminatory power for clonal isolates and can reveal the transmission route of hospital infections caused by various species, including methicillin-resistant Staphylococcus aureus and carbapenem-resistant Enterobacterales (Madigan et al., 2018; Al Fadhli et al., 2021). Next-generation sequencing (NGS) studies of bacterial pathogens tend to use short-read sequencing, which involves the fragmentation of genomic DNA into short segments of a few hundred nucleotide bases. Rogers et al. (2021) performed a scoping review revealing that Illumina (San Diego, CA, USA) sequencers, which currently dominate the short-read NGS market, are also the most commonly used platforms for sequencing Enterococcus spp. Long-read (also known as third-generation) sequencing platforms are designed to yield much longer reads (tens of kilobases) without the need for polymerase chain reaction (PCR) amplification and are currently represented by Single-Molecule Real-time sequencing (SMRT) of Pacific Biosciences (Menlo Park, CA, USA) and Nanopore sequencing (Oxford Nanopore Technologies, Oxford, UK) (Rhoads and Au, 2015). However, their drawbacks include greater sequencing error rates compared to Illumina platforms, although the rates have decreased over recent years (Rang et al, 2018). Especially, the PacBio Sequel system generates assemblies with considerably fewer errors, but the high cost of instruments renders them less accessible (Hrabak et al., 2020; Lang et al., 2020). Nanopore sequencing (Oxford Nanopore Technologies) is cost-efficient and uses portable devices facilitating previously impossible in-field sequencing to be undertaken; thus, this method is suitable for analyzing microbial genomes (Bayley, 2015; Leggett and Clark, 2017).

In this study, we performed bacterial WGS using both short-read and long-read next-generation sequencing (SR-NGS and LR-NGS, respectively) for epidemiological analysis of VRE. We developed pipelines involving three-step analyses of SR- and LR-NGS data for strain typing and AMR profiling, respectively. Moreover, noting the differences between SR- and LR-NGS, the strain typing and AMR detection methods were evaluated using clinical VRE isolates.

Enterococci are one of the most problematic nosocomial pathogens because they usually affect patients already suffering from debilitating diseases. Enterococcal infection typically necessitates prolonged hospitalization, imposes an economic burden, and increases mortality and morbidity (Cosgrove, 2006). Bacterial resistance to commonly used antibiotics is widespread. For example, enterococci are intrinsically resistant to β-lactams like cephalosporins, penicillin, and (in the case of E. faecium) ampicillin, as well as clindamycin and low-level aminoglycosides. They also have the capacity to acquire and disseminate genetic determinants of AMR such as vancomycin, erythromycin, and high-level aminoglycosides (Eliopoulos and Gold, 2001; Gagetti et al., 2019). Thus, the World Health Organization has designated VREfm as a priority pathogen to encourage national surveillance programs (Tacconelli et al., 2018).

To prevent the spread of VRE in the hospital setting, the epidemiology of nosocomial infections should be studied, and that requires understanding the genetic relatedness of pathogens. Molecular approaches to the diagnosis and epidemiological analysis of nosocomial infections aid identification of infection sources and distinctive AMR patterns (Singh et al., 2006; Humphreys and Coleman, 2019). We used WGS for epidemiological analysis of VRE in an attempt to establish a standardized workflow. We performed strain typing of VRE isolates using SNP-based methods, along with genetic profiling of genes and mutations conferring AMR.

The popularity of NGS in clinical microbiology has increased dramatically for several reasons. WGS enables high-resolution analysis of entire genome sequences, and the cost is continually falling. Moreover, the availability of graphical user interface-based software for bioinformatic analysis is increasing (Allard et al., 2012; Tan et al., 2018). Although several studies have demonstrated the effectiveness of WGS for bacterial strain typing, including of VRE, they mostly focused on the discriminatory power of strain typing methods used in conjunction with sequencing analysis. For example, one study found that SNP-based strain typing was slightly more discriminatory than whole-genome MLST (Pearce et al., 2018). To our knowledge, few studies have compared different sequencing platforms in the context of the molecular epidemiology of bacteria. Neal-McKinney et al. (2021) compared the efficacy of MiSeq, MinION, and hybrid genome sequencing for the analysis of Campylobacter jejuni, but their comparison focused only on the preprocessing of sequencing data and read assemblies.

Despite the clear advantages of WGS-based bacterial epidemiology, some challenges remain. The main priority is to standardize workflows and pipelines, including sequencing methods and bioinformatic analyses, to ensure sensitivity and specificity. Moreover, many variables in the bioinformatic analysis are likely to affect the utility and overall performance of pipelines for QC and data trimming, bioinformatic tools and algorithms, reference databases, and thresholds for positive results. One of them is to select a proper read assembly technique based on the sequencing platform as well as downstream analysis to proceed. Herein, we chose to perform variant detection following reference-based assembly to assess genetic relatedness, and de novo assembly to enable the identification of important loci in the accessory genome like AMR genes (Scheunert et al., 2020);. In the case of a genome assembly using error-prone long reads, post-sequencing correction can increase the accuracy of the sequences. The correction algorithms include a hybrid method, which in addition uses short reads for assembly and polishing, and the long-read-only method such as consensus polishing via POA graphs (Rang et al., 2018; Lee et al., 2021). Thus, pipelines should be validated based on real-world data before being applied in clinical laboratories.

Our in-house bacterial WGS pipelines take account of the variation among sequencing platforms. Essentially, our pipelines constitute three-step analyses of SR- and LR-NGS data for the purposes of strain typing and AMR profiling. The three steps are as follows: read preprocessing and assembly (primary analysis), variant calling (secondary analysis), and strain typing and AMR detection (tertiary analysis). In this study, the steps were evaluated using clinical VRE isolates, and the results were compared between SR- and LR-NGS. For strain typing, we performed reference-based assemblies to analyze the SNPs of VRE isolates and constructed SNP-based phylogenies using a maximum-likelihood algorithm. SR-NGS detected significantly more SNPs than LR-NGS ( Figure 2 ); this was attributed to its higher resolution for detecting nucleotide substitutions of bacterial sequences, which led to differences in the phylogenetic trees generated by SR- and LR-NGS. Branch length reflects the substitution rate, and the phylogenetic tree generated by SR-NGS generally had longer branches than that of LR-NGS. For AMR profiling, reads were subjected to de novo assemblies to determine genetic determinants of AMR in each isolate, and we analyzed the concordance between genotypic and phenotypic resistance. In particular, we examined resistance to daptomycin, which is a last-resort antibiotic for VRE infection. A previous study proposed that mutations in the liaF and gdpD genes were responsible for daptomycin resistance (Palmer et al., 2011), but these mutations were not detected by the SR- or LR-NGS conducted in this study; every isolate was phenotypically susceptible to daptomycin (data not shown). In general, the concordance between the genotypic and phenotypic resistance of the antibiotics was better for LR-NGS than SR-NGS. We attributed the few discrepancies to the fact that the relationship between AMR genes/mutations and phenotypic resistance is influenced by multiple factors, including genes not directly involved in known drug-resistant machineries but required to express resistance, and an interplay among the mechanisms of AMR also exists (Yeh et al., 2006; Girgis et al., 2009; Suzuki et al., 2014). Moreover, the AMR databases used for NGS analyses are incomplete, because many genes involved in genotypic and phenotypic resistance remain to be discovered.

The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found below: https://www.ncbi.nlm.nih.gov/, PRJNA798443.

SO and KP conceived and planned the experiments. HC and SN performed DNA extraction and next-generation sequencing. SO, KP, and HC analyzed the data and interpreted the results. SO performed the statistical analysis and wrote the manuscript. SO, KP, HC, and SN contributed to the final version of the manuscript and approved the submitted version.

This work was supported by the research fund No. 2019-ER5402-01 from the Korea Disease Control and Prevention Agency.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.