Whole-genome sequencing of a large collection of Myroides odoratimimus and Myroides odoratus isolates and antimicrobial susceptibility studies

The genus Myroides comprises several species of Gram-negative, non-motile, and non-fermenting bacteria, which have been regarded as non-pathogenic for decades. Multiple recent reports, however, underscore the pathogenic potential that Myroides sp. possesses for humans. These bacteria seem to be resistant to a wide range of antibiotics (including ß-lactams and aminoglycosides). Therefore, treatment options are limited. Knowledge of antimicrobial resistance, however, is based on only one meaningful comprehensive study and on data published from case reports. This lack of data motivated us to test 59 strains from our Myroides collection (43 M. odoratimimus and 16 M. odoratus) for resistance against 20 commonly used antibiotics. We also performed molecular analyses to reveal whether our bacteria harbor the genus-specific M. odoratimimus metallo-ß-lactamase (MUS-1) or the M. odoratus metallo ß-lactamase (TUS-1), and other ß-lactamases, which may provide an explanation for the extended antimicrobial resistance.


Introduction
The bacteria known as Myroides today were first described as Bacterium faecale aromaticum by Stutzer in 1923 1 . Six years later, in 1929, these bacteria were renamed Flavobacterium odoratum 2 . Finally, in 1996, Vancanneyt et al. undertook a new reclassification based on modern techniques that included DNA-rRNA hybridization, DNA-DNA hybridization, analysis of whole-cell protein patterns, fatty acid composition, and the phenotype 3 . Based on their results, they introduced a new genus, Myroides, with the species M. odoratus (formerly F. odoratum) and a new species, M. odoratimimus, into the bacterial taxonomy 3 . The bacteria are Gram-negative, non-motile, non-fermenting, and rod-shaped. Their natural habitat is soil and water [3][4][5] . Infections caused by Myroides sp. comprise necrotizing fasciitis, soft tissue infections, ventriculitis, pneumonia, and sepsis. Immunocompromised patients are mostly affected, although being immunocompromised does not seem to be absolutely prerequisite to acquiring a Myroides infection 6-12 . In addition, two studies on outbreaks have been published 13,14 . Furthermore, the outcome was fatal in two case reports 9,[13][14][15] . Since Myroides sp. are resistant to a wide range of antibiotics, including ß-lactams and aminoglycosides, choosing an appropriate empirical antimicrobial therapy is challenging. Moreover, knowledge about the specific resistance profile of Myroides sp. is based on data mainly obtained from case reports and two publications 16,17 . In the present study, we therefore compare minimal inhibitory concentration (MIC) results for 20 important antibiotics against 59 strains from our Myroides collection (43 M. odoratimimus and 16 M. odoratus). The data were evaluated according to the guidelines published by EUCAST in 2017. Furthermore, PCR analysis and whole-genome sequencing approaches were applied to detect corresponding genes of important ß-lactamases of the OXA-types, VIM-types, IMP-types, AmpC-types, KPC-types, and NDM-types and for the Myroides sp.-specific enzymes MUS-1 and TUS-1.

Antimicrobial susceptibility
The susceptibility profile and the MIC distribution of all strains tested are summarized in Tables 1 and 2. The MIC results of each strain are additionally provided in Table S1. Only a few M. odoratimimus and M. odoratus strains are susceptible to ampicillin (M. odoratimimus, n = 1; M. odoratus, n = 5) and piperacillin/tazobactam (M. odoratimimus, n = 2; M. odoratus, n = 0). Interestingly, more of the tested strains showed intermediate susceptibility to ampicillin (M. odoratimimus, n = 19; M. odoratus, n = 2) than to piperacillin/tazobactam (M. odoratimimus, n = 9; M. odoratus, n = 0). No strain was susceptible to ceftazidime, cefepime, and aztreonam. In total, 4 M. odoratimimus strains and no M. odoratus strain were susceptible to imipenem. In contrast to this observation, 32 M. odoratimimus and 8 M. odoratus strains were susceptible to meropenem. The MIC results of the quinolones-ciprofloxacin, levofloxacin, and moxifloxacin-were also compared in this study. Most strains were susceptible to moxifloxacin (M. odoratimimus, n = 39; M. odoratus, n = 15). In contrast to this, 5 M. odoratimimus strains and 7 M. odoratus strains tested were susceptible to levofloxacin. Only 1 M. odoratimimus strain and 1 M. odoratus strain were susceptible to ciprofloxacin. Of all quinolones tested, moxifloxacin showed the lowest MIC values in this antibiotic class. In addition, only 1 M. odoratimimus strain and 1 M. odoratus were susceptible to tigecycline.
No breakpoints were available for trimethoprim/sulfamethoxazole, fosfomycin, colistin, gentamicin, amikacin, erythromycin, azithromycin, daptomycin, and rifampicin. The MIC results determined for fosfomycin, colistin, gentamicin, amikacin, and daptomycin are all at a high range. Therefore, antimicrobial resistance to these antibiotics may be assumed. Regarding trimethoprim/sulfamethoxazole, erythromycin, azithromycin, and rifampicin, the MIC results ranged from low to high levels.
Detection of relevant β-lactamase genes All strains were negative for genes coding for OXAtype, VIM-type, IMP-type, and NDM β-lactamases, and for KPC. The AmpC β-lactamase gene was detected in one isolate (DSM 100899, Table S2).

Homologies of bla MUS-1 and bla TUS-1
Genes homologous to bla MUS-1 were only detected in M. odoratimimus; genes homologous to bla TUS-1 , only in M. odoratus isolates. Figure 1 is a dendrogram that presents the nucleotide sequence similarity of all genes of our M. odoratimimus strains that show homology to the carbapenemase MUS-1 gene of M. odoratimimus CIP 103073 (GenBank accession #AF441286.1). Figure 2 is also a dendrogram. It shows the nucleotide sequence similarities of our M. odoratus strains compared to the carbapenemase TUS-1 gene of M. odoratus CIP 103105 (GenBank accession #AF441287.1) in the same fashion. No DNA sequence homologies to bla TUS-1 were found in 11 out of 16 isolates (Table S3).

Comparison of nucleotide and protein sequences
Nucleotide sequences were transferred into protein sequences using Lasergene software version 14.0.10 (DNASTAR, Madison, WI, USA). Alignments of all MUS-1 and all TUS-1 genes, as well as their protein translations, are illustrated in Figs. S1-4. When the sequences are compared, alterations are only observed at certain positions in the genes homologous to bla MUS-1 and bla TUS-1 . The active centers of both enzymes, however, were not involved. This observation is confirmed by the translated protein sequences. The maximum divergence between the sequences of genes homologous to bla MUS-1 was 5.6% in the nucleotide composition and 6.5% in the amino acid (AA) composition. Additionally, the maximum divergence of genes homologous to bla TUS-1 was 5.1% at the nucleotide level and 2.8% at the AA level.

Discussion
Bacteria of the genus Myroides are rare human pathogens that cause severe infections (including blood stream infections, necrotizing fasciitis, ventriculitis, and pneumonia) and affect mainly immunocompromised patients 6,7,9,18,19 [20][21][22][23][24][25][26][27] . M. odoratimimus and M. odoratus are described to be resistant to a large number of antibiotics. The available antimicrobial susceptibility data on Myroides sp. are mostly based on case reports and four systematic studies 16,28,29 . Two of these studies included only a few isolates 28,29 and therefore are limited. Holmes et al. tested 28 F. odoratum isolates against 15 antibiotics 16 , whereas Hu et al. summarized and compared data published in the literature and added data that previously were available only in China 17 . However, those authors did not distinguish between M. odoratimimus and M. odoratus. As far as we  (Tables 1  and 2). Additionally resistance to piperacillin/tazobactam 6,7,9,11 . However, none of our strains were susceptible to ceftazidime and cefepime (Tables 1 and 2). Our finding is also in accordance with previous reports 6,11,15 (Tables 1 and 2), which has also been reported by others 6,9 . Therefore, meropenem may be more useful for the treatment of Myroides infections than imipenem. While investigating quinolones, we found a gradual change in susceptibility. Most strains were resistant to ciprofloxacin (Tables 1 and 2): 83.7% (n = 36) of the M. odoratimimus strains and 68.7% (n = 11) of the M. odoratus strains were resistant, whereas only 2.3% (n = 1) of the M. odoratimimus strains and 6.3% (n = 1) of the M. odoratus strains were susceptible. However, resistant, as well as susceptible, isolates have been reported before 6,7,9,18 . A larger number of isolates were susceptible to levofloxacin compared to ciprofloxacin (Tables 1 and 2), namely, 11.6% (n = 5) of the M. odoratimimus and 43.8% (n = 7) of the M. odoratus strains. Interestingly, moxifloxacin testing revealed significantly lower MIC values compared to levofloxacin and ciprofloxacin (Tables 1 and 2). In fact, only one M. odoratus isolate was resistant, but 90.7% (n = 39) of our M. odoratimimus isolates and 93.8% (n = 15) of our M. odoratus isolates were susceptible ( Table 2). Based on these in vitro data, moxifloxacin may therefore be the best quinolone to be used. Ali 31 . Based on sequence analyses, both MUS-1 and TUS-1 could be identified as members of the B1 subclass family according to the classification by Ambler 32 , which has been expanded by Galleni et al 33  This ability is characteristic of enzymes belonging to the subclass 3a of metallo-ß-lactamases according to the revised functional classification proposed by Bush and Jacoby in 2010 34 . In contrast, our strains were mostly resistant to aztreonam, and many were susceptible to meropenem (Tables 1 and 2). However, meropenem is characteristically hydrolyzed by group 3a ß-lactamases 34 . For this reason (resistance against aztreonam and susceptibility to meropenem), we assume that neither MUS-1 nor TUS-1 is expressed at high quantities. Therefore, additional, yet unresolved mechanisms, such as porinmutations or other ß-lactamase hydrolyzing enzymes,  (Table S1). A Myroides strain, assigned to M. odoratimimus based on 16S rDNA sequencing, without bla MUS-1 has previously been reported by Dharne et al 35 . Yet this isolate was reported to be highly resistant to β-lactam antibiotics 35 . Based on molecular and phenotypic analyses, Al-Bayssari et al. proposed a new variant of a MUS ß-lactamase in 2015, which was named MUS-2 and possessed 98.78% AA homology to MUS-1 36 . However, in contrast to the work by Mammeri et al. 31,36 , these authors did not perform an in-depth analysis of the enzyme kinetics.
A comparison of the genomic data from all 48 Myroides strains carrying bla MUS-1 /bla TUS-1 to the sequences published by Mammeri et al. 31,36 show that the encoding genes have a fairly high homology, ranging from 94.7 to 100% for bla MUS-1 and from 94.9 to 99.3% for bla TUS-1 (Table S4 and Figs. S1, S3). At the protein level, MUS-1 shares 93.5-100% homology and TUS-1 97.2-99.6% (Table S4 and Figs. S2, S4). Notably, all 43 M. odoratimimus strains harbor a MUS-1 encoding gene, whereas, of the 16 M. odoratus isolates, 11 were negative for bla TUS-1 . Furthermore, different clusters of "MUS" or "TUS" genes can be defined (Figs. 1, 2). Only 30.2% (n = 13) of our M. odoratimimus strains harbor genes that are genetically identical to bla MUS-1 . We could not detect a gene that was 100% identical to the originally described nucleotide sequence for bla TUS-1 . In other words, the variability of M. odoratus with respect to the gene bla TUS-1 is large. Only five genes were homologous to the sequence initially described by Mammeri et al. 31,36 , with variation ranging from 94.9 to 99.3%. Comparison of the nucleotide and AA sequences, however, suggests that genomic differences may not have an effect on the active center of the enzymes, and therefore may not affect the hydrolytic activity. Additional data from a similar comparative study on enzyme kinetics, however, would elucidate whether these sequences encode for new variants. Therefore, the results and the conclusions drawn by Al-Bayssari et al. 36 regarding a newly described MUS-2 have to be corroborated. Notably, our M. odoratimimus isolate DSM 100899 harbors the gene encoding the ß-lactamase AmpC. It can be located chromosomally or on specific resistance (R-) plasmids 37 . We believe this is the first report of an AmpC encoding gene detected in Myroides sp. Furthermore, the R-plasmids seem to play an important role in antimicrobial resistance mechanisms in Myroides sp. Kono et al., for instance, reported on resistance genes against ampicillin, carbenicillin, and erythromycin which were located on an R-plasmid 38 . In addition, Kuai et al. reported on a KPC-2-positive F. odoratum strain in 2011 39 . In general, a variety of genes for antimicrobial resistance seems to be characteristic for Myroides sp. Elucidating the exact mechanisms of antimicrobial resistance, however, is a matter of ongoing and future research.

Bacterial isolates
A collection of clinical isolates of Myroides strains was created over the course of 4 years (Table S5). The strains were collected during the routine diagnostics at the Institute for Medical Microbiology and Hygiene, TU Dresden, Germany. Additional strains were provided by Bodo R. Eing and Sebastian Bertram (Synlab Medizinisches Versorgungszentrum Augsburg, Augsburg, Germany). A few isolates were of veterinary origin (Table S5). The bacteria were confirmed as M. odoratimimus or M. odoratus by two independent methods: 16S rRNA gene sequencing and MALDI-TOF MS (Bruker Daltonik, Bremen, Germany). We have addressed the suitability of this identification strategy, as well as the performance and reliability of both methods, in previous studies 40,41 . In the present investigation, we included 59 strains, with 43 being identified as M. odoratimimus and 16 as M. odoratus. All strains were also deposited in the "Open Collection" of the Leibniz Institute DSMZ-German Collection of Microorganisms and Cell Cultures (Braunschweig, Germany).

Antimicrobial susceptibility testing
The MIC tests were performed, and the results were evaluated by applying the guidelines for PK/PD (nonspecies related) breakpoints according to the criteria published by EUCAST (the European Committee on Antimicrobial Susceptibility Testing). EUCAST breakpoint tables were created for the interpretation of the MICs and zone diameters using Version 7.1, 2017 (http:// www.eucast.org/clinical_breakpoints/). In brief, a McFarland standard of 0.5 was created for each bacterium using NaCl and a DensiCHEK densitometer (bioMérieux, Nürtingen, Germany). The suspended bacteria were plated with the help of a cotton swab on Müller-Hinton Agar (Oxoid Deutschland, Wesel, Germany). Then, E-Test strips (bestbion, Cologne, Germany) for each antibiotic were placed on the agar plates. The plates were incubated for 18 ± 2 h at 37°C and 5% CO 2 . Then, the MIC results were determined. The ATCC strains Escherichia coli ATCC ® 25922™, Pseudomonas aeruginosa ATCC ® 27853™, and Staphylococcus aureus ATCC ® 25923™ served as the controls. All E-Test strips used in this study and their MIC ranges are listed in Table S6.

Real-time PCR analysis for β-lactamase gene detection
All samples were cultivated on blood agar plates (25°C, 24 h), and DNA was extracted with a QB-EX-50 Quick-Blue extraction kit using magnetic nanoparticles (Q-Bioanalytic, Bremerhaven, Germany). Real-time PCR analysis was carried out on all isolates to evaluate the prevalence of the bla OXA , bla VIM , bla IMP , bla AmpC , and bla KPC carbapenemase genes and the bla NDM metallo-ßlactamase gene. QuickBlue RealQuick PCR master mix-SYBR Green (QB-RT-ESBL-50-SYBR, Q-Bioanalytic) for carbapenemase gene detection was prepared in 15 µl of master mix that included Taq polymerase, buffer, dNTPs, the specific primer set, and SYBR Green. The amplification and detection of the metallo-ß-lactamase bla NDM amplicon was carried out using QuickBlue RealQuick PCR master mix (QB-RT-ESBL-50, Q-Bioanalytic), which included Taq polymerase, buffer, dNTPs, specific primer set, and probe (Fam-labeled). Two microliters of DNA template (5-10 ng genomic DNA) was added to each PCR application, bringing the total volume to 17 µl. Amplification reactions were carried out using a real-time thermal cycler (LightCycler 480 II, Roche Deutschland, Grenzach-Wyhlen, Germany) with the following program: one cycle of 5 min at 95°C, 35 cycles with denaturation at 95°C for 15 s and annealing at 60°C for 20 s; and an extension at 72°C for 15 s. Data were analyzed using the Roche LightCycler Software 1.5. Primer sets were designed with Integrated DNA Technologies "Primer-Quest Tool" (https://eu.idtdna.com/PrimerQuest/Home/ Index) using genomic sequences from NCBI GenBank. Multiple alignments from several ß-lactamase genes were analyzed with the Clustal Omega program (https://www. ebi.ac.uk/Tools/msa/clustalo/), and the primer sets were optimized to ensure detection of a broad range of bacterial resistance genes. Klebsiella pneumoniae NCTC 13443 (bla NDM ), K. pneumoniae NCTC 13440 (bla VIM-1 ), K. pneumoniae NCTC 13442 (bla OXA-48 ), K. pneumoniae BAA 1705 (bla KPC ), E. coli NCTC 13476 (bla IMP ), and Enterobacter cloacae BAA 1143 (bla AmpC ) strains were used as negative and positive controls. The results are illustrated in Table S2.
Detection of MUS-1 and TUS-1 encoding genes using a whole-genome sequencing approach Libraries for whole-genome sequencing were prepared with Nextera XT DNA Library Prep Kit (Illumina, San Diego, CA, USA). Sequencing was performed on the Illumina HiSeq 2500 (Illumina) for 100 cycles in both directions at the Helmholtz Centre for Infection Research (Braunschweig, Germany). Hereby, coverages of approximately 100× were obtained for each of the isolates. Assembly of short read genome data was performed using Velvet 1.2.10 42 with a kmer size of 61 for all isolates. Resulting contig sets were screened for known antibiotic resistance genes querying the ResFinder 2.1 database 43 in an automated approach. Hereby, minimum values of 60% query coverage and 90% identity have been used as search parameters. A comparative analysis of the DNA sequences obtained was performed using the Molecular Evolutionary Genetics Analysis software package version 7 44 . Protein sequences were created using DNASTAR Lasergene 14.0.10. (DNASTAR). Average nucleotide identity (ANI) values have been calculated on the whole-genome level using the software FastANI (https://github.com/ ParBLiSS/FastANI) (Table S3). Therefore, all M. odoratimimus strains used in this study were compared with M. odoratimimus CCUG39352 T (draft assembly ASM148541v1, https://www.ncbi.nlm.nih.gov/assembly/ GCF_001485415.1), and all M. odoratus strains were compared with M. odoratus DSM 2801 T (GenBank accession #CM001437).