Genomic Background and Phylogeny of cfiA-Positive Bacteroides fragilis Strains Resistant to Meropenem-EDTA

Background: Bacteroides fragilis shows high antimicrobial resistance (AMR) rates and possesses numerous AMR mechanisms. Its carbapenem-resistant strains (metallo-β-lactamase cfiA-positive) appear as an emergent, evolving clade. Methods: This work examines the genomes, taxonomy, and phylogenetic relationships with respect to other B. fragilis genomes of two B. fragilis strains (CNM20180471 and CNM20200206) resistant to meropenem+EDTA and other antimicrobial agents. Results: Both strains possessed cfiA genes (cfiA14b and the new cfiA28), along with other AMR mechanisms. The presence of other efflux-pump genes, mexAB/mexJK/mexXY-oprM, acrEF/mdtEF-tolC, and especially cusR, which reduces the entry of carbapenem via the repression of porin OprD, may be related to meropenem–EDTA resistance. None of the detected insertion sequences were located upstream of cfiA. The genomes of these and other B. fragilis strains that clustered together in phylogenetic analyses did not meet the condition of >95% average nucleotide/amino acid identity, or >70% in silico genome-to-genome hybridization similarity, to be deemed members of the same species, although <1% difference in the genomic G+C content was seen with respect to the reference genome B. fragilis NCTC 9343T. Conclusions: Carbapenem-resistant strains may be considered a distinct clonal entity, and their surveillance is recommended given the ease with which they appear to acquire AMR.


Introduction
Bacteroides fragilis is a common bacterium of the human gut. It generally behaves as a commensal species, but under certain conditions it can cause severe intra-abdominal infections, skin and soft tissue infections, brain abscesses, surgical site infections, and anaerobic bacteraemia [1]. The plasticity of the B. fragilis genome allows it to incorporate antimicrobial resistance and virulence determinants via horizontal gene transfer (HGT), and to turn specific resistance genes on or off as needed [2,3]. Its behavior as a reservoir of resistance is well-established [1][2][3].
The B. fragilis population can be separated into two divisions, I and II, depending on the mutually exclusive presence of the chromosomal cephalosporinase gene cepA or the chromosomal metallo-β-lactamase (MBL) gene cfiA, together with other characteristics [4,13,14]. cfiA confers a high level of resistance to carbapenems but also to other β-lactams and β-lactamase inhibitor combinations. In addition, the expression levels of cfiA, nim, and erm (providing resistance to metronidazole and macrolides, respectively) are partly controlled by upstream insertion sequences (IS) found in multiple copies throughout B. fragilis genomes [3,10].
The aim of the present study was to characterize two cfiA-positive B. fragilis strains resistant to meropenem+EDTA. The phenotypic detection of MBLs in B. fragilis is usually undertaken using the meropenem-EDTA double-ended Etest [15]. However, the studied strains were resistant to this, and the presence of cfiA could not be inferred.

Species Identification and Susceptibility Phenotype
B. fragilis CNM20180471 and CNM20200206 strains were initially identified by MALDI-TOF/MS as B. fragilis with a score of 2.11 and 2.38 (mean of duplicates), respectively. They were confirmed by 16S rDNA analysis: BLAST examinations returned 99.01% and 98.75% identities with respect to the sequence of the type strain B. fragilis NCTC 9343T (GenBank accession no. KP326374.1) [16]. According to EUCAST and CLSI criteria, the strains showed resistance to amoxicillin-clavulanate, piperacillin-tazobactam, carbapenems, aminoglycosides, fluoroquinolones, erythromycin, clindamycin, and tetracycline; they were susceptible to metronidazole, linezolid, and tigecycline (Table 1). No differences in categorization were detected between both criteria. No MBL activity was detected for either strain by the meropenem double-ended Etest ( Figure S1); no reduction in the MIC of meropenem was seen when EDTA was present. However, MBL activity was detected when imipenem and EDTA were present. The last combination was more effective in detecting these MBLs than the recommended meropenem double-ended Etest in the studied strains [15].
A clear pheno-to-genotype correlation was found for strain CNM20180471, with the following exceptions: (i) the lack of an effect of the combination meropenem-EDTA on MBL activity despite the presence of cfiA; (ii) high fluoroquinolone MICs despite the conserved gyrA and gyrB genes (with motifs Asp81-Ser82-Gly478 and Leu415, respectively), explained by the presence of fluoroquinolone-resistant, active multidrug and toxic compound extrusion (MATE) efflux-pumps, bexA/bexB [19]; (iii) and susceptibility to tigecycline despite the presence of tetX [20]. The first two exceptions (i) and (ii) were also detected for strain CNM20200206, together with high resistance to macrolide-lincosamide-streptogramin compounds, with no detection of the genes involved.
KEGG orthology, used to search for gene set resistance modules, detected cusR in both strains, which is a phosphate regulon response regulator of the two-component system of the OmpR family providing copper resistance. cusR reduces the entry of carbapenems via the repression of the porin OprD, thus affording imipenem resistance [21]. The MBLs cfiA14b and cfiA28 were inhibited by EDTA when combined with imipenem but also with meropenem. In the studied strains, the exact mechanism underlying the lack of susceptibility to meropenem-EDTA was not detected on the basis of functional annotation, but the above finding of cusR may be related to a reduction in meropenem permeation, perhaps due to the inactivation of a porin different from OprD. In both strains, other gene set resistance modules found were linked to resistance to vancomycin (vanX), daptomycin (dltA), and multidrug resistance efflux pumps (mexAB-oprM, mexJK-oprM, mexXY-oprM, acrEF-tolC, and mdtEF-tolC). The operon identified as mexAB-oprM is equivalent to bmeABC for the resistance-nodulation-division RND efflux pumps described in B. fragilis (involved in beta-lactam resistance). Efflux pumps have an important role in multiresistance in B. fragilis [22]. Exposure to several beta-lactams, carbapenems, fluoroquinolones, metronidazole, and certain chemical agents all induce the over-expression of efflux pump genes, thus promoting MDR in B. fragilis [23]. Other regulators of multiple antibiotic resistance such as the MarR family may be involved, as recently seen when the abolishment of the expression of MarR homologs increased susceptibility to several antimicrobials in B. fragilis [24]. The MarR sequences of the studied strains showed full identity, as they did with those of strains O21, BF8, and DCMSKEJBY0001B [4]. The interplay of the efflux pumps and the regulators of multiple antibiotic resistance, together with changes in porins, may be responsible for the lack of effect of the meropenem-EDTA combination in the studied strains, which may be another step in the evolution of the emergent clade of the carbapenem-resistant B. fragilis strains.
A putative integrative conjugative element INT_ICEBs1_C_like (C-terminal catalytic domain of integrases from bacterial phages and conjugate transposons) was also annotated in CNM20180471 (contig = 01286) [2], and was identical to the nucleotide sequence of a site-integrase tyrosine recombinase XerD type of B. fragilis DCMOUH0067B (GenBank accession no. CP036553) [10]. B. fragilis acquires and efficiently disseminates traits of its pan-genome (AMR, virulence, and the rapid adaptation to a changing environment) via conjugative and mobilizable elements [2,3]. Although the difference in the G+C content was <1% (Table S1), the average nucleotide identity (ANI) and average amino acid identity (AAI) percentage values and the in silico genome-to-genome distance similarity (GGDH) or DNA-DNA hybridization TYPS analysis [31] returned two different clusters: a main one that grouped 87.7% of the B. fragilis genomes and a smaller one encompassing 12.3% of the strains; CNM20180471 and CN20200206 fell into this latter cluster with 18 other B. fragilis genomes (Figure 1). The same clustering was observed with the chewBBACA (BSR-Based Allele Calling Algorithm) minimum spanning tree (MST) [32] (Figure 2; blue balls for the larger cluster and red balls for the smaller cluster). The node of this cluster is represented by the HMW_610 genome (GCF_000297695.1) [2]. The genomes of strains AF14-14AC (GCF_003465265.1) and 4g8B (assembly no. GCF_001373095) were located at intermediate points between the central node and that occupied by the genome of CCUG 4856T (GCF_005706655.1), analogous to the reference genome B. fragilis NCTC 9343T (NC_003228.3) and the HMW_610 genome.
The ANI, AAI, DDH-estimate, and G+C content difference values were determined with respect to the NCTC 9343T for all the genomes included in the above smaller cluster and for selected genomes of the large cluster. The ANI value was also determined with respect to the YCH46 and CNM20180471 genomes (Table S2). The strains belonging to the smaller cluster had the lowest ANI values with respect to NCTC 9343T and YCH46 (~88%) but not with respect to CNM20180471 (~98.44%). The members of the smaller cluster therefore showed less identity (and coverage) to the type strain than did the members of the larger cluster. The examined strains belonging to the larger cluster returned high ANI values with respect to NCTC 9343T (~99%) and YCH46 (~99%) and lower values with respect to CNM20180471 (~87%). The genomes of the reference strains of other species, such as Bacteroides finegoldii DSM 17565 (GCF_000156195.1), Bacteroides xylanisolvens ASM654696v1 (GCF_006546965.1), and Bacteroides ovatus ATCC 8483 (GCF_001314995.1), showed ANI/AAI values of around 75% with respect to the NCTC 9343T and YCH46 genomes (Table S2). Antibiotics 2021, 10, x FOR PEER REVIEW 8 o Figure 1. B. fragilis whole-genome, sequence-based phylogenetic tree constructed using FastME v.2.1.6.1 software (which calculates genome BLAST distance phylogeny (GBDP) distances); the branch lengths are scaled in terms of the GBDP distance formula. The numbers above the branches are GBDP pseudo-bootstrap support values (all >60%; 100 replicates); average branch support = 27.9%. The trees were rooted at the midpoint. The results were provided by the Type Strain Genome Server (TYGS) (https://tygs.dsmz.de, accessed on 19 October 2020) [33]. The same color code was used in the chewBBACA analysis.  Phylogenetic tree (constructed by the multiple sequence alignment (MAFFT) and neighbor joining using the Clustal W2 algorithm) based on the core genome Multilocus Sequence Type associations among the B. fragilis genomes and those of the related examined species. The tree was built using chewBBACA software [34] and was based on 2362 loci. The ANI, AAI, and DDH-estimate results suggest B. fragilis CNM20180471 and CNM20200206, and other members of its cluster with carbapenem-resistance via metalloβ-lactamase cfiA, support the previous hypothesis, according to which they are genomospecies "related to, but distinct from, B. fragilis" [35] or in a "quasi-transition state between the two divisions of B. fragilis" [4]. The same has been stated for carbapenem-resistant B. fragilis strains by 16S rRNA analysis (as can be observed in Figure S2), without the bias of their resistome in the whole-genome analysis.
Over the last decade, a world-wide increase in these carbapenem-resistant strains has been reported [26,36,37]; they are favored for their efficient pan-genome dissemination [2,38]. Their evolution may be guided by the impact on the gut ecosystem of carbapenems and other β-lactams [2] used to treat infections caused by extended-spectrum β-lactamase-producing Enterobacteriaceae as well as exposure to other antimicrobial and chemical agents. The lack of effect of the meropenem and EDTA combination in the studied B. fragilis strains may be another step in the evolution of this complex entity.
In conclusion, the loss of susceptibility against meropenem+EDTA in the studied cfiA-positive strains may be due to the presence of multiple efflux-pumps such as MexAB/ MexJK/MexXY-OprM, and AcrEF/MdtEF-TolC, but may be especially due to CusR, which can reduce meropenem entry by porin inactivation. No IS was seen upstream of the metalloβ-lactamase cfiA14b and cfiA28 genes responsible for carbapenem resistance. B. fragilis CNM20180471 and CNM20200206, and the other carbapenem-resistant strains examined, may be considered distinct species or novel genomospecies. At the least, they are a different and clear clonal entity of the B. fragilis group. The surveillance of this clonal lineage is recommended on the basis of whole genome analyses, including the characterization of MBLs, ISs, and plasmids; antimicrobial susceptibilities observed in this study; and previous indications of the lineage's ability to acquire and spread antimicrobial-resistance traits.

Case Reports
In case no. 1, a 56-year-old woman undergoing chemotherapy plus peripheral blood precursor treatment for renal amyloidosis was admitted to the intensive care unit at the Hospital Virgen de las Nieves (Granada, Spain) because of a rupture of the spleen with hematoma. After splenectomy, bilateral pleural effusion led to symptoms of respiratory distress. Six months later, the patient developed severe polyneuropathy and became dependent on mechanical ventilation before exitus. In case no. 2, a 52-year-old man with myelodysplasia underwent allogenic transplant at the same hospital. Two days later, he suffered diarrhea caused by toxigenic Clostridium difficile; he was treated with oral vancomycin and fidexomicin, and he was cured. Eighteen days later, however, he experienced a febrile episode with hemodynamic instability, and was admitted to an intensive care unit where he was successfully treated with meropenem over a period of 10 days. The clinical samples were taken as part of standard patient care and also for the purpose of this study. This study was focused on bacteria and no identifiable human data were used. Therefore, ethical approval was exempted.

Identification and Antimicrobial Susceptibility Testing
B. fragilis CNM20180471 and CNM20200206 were isolated from pleural liquid and blood samples of case no. 1 and case no. 2, respectively. Samples were collected depending on the type of infection. B. fragilis growths in anaerobiosis cultures with a single morphology were identified by MALDI-TOF-MS (Bruker Biotyper, Billerica, MA, USA) running v. 9 propriety software (8468 msp). Strain identification was performed by 16S rRNA gene sequence analysis, with primers fD1 and rP2 [39] for amplification, and E781 and U1115 for sequencing [33], using BLAST (http://www.ncbi.nlm.nih.gov/BLAST (accessed on 13 October 2020)).
Susceptibility to antimicrobials was examined using E-test gradient strips (Liofilchem, Teramo, Italy) on Brucella blood agar supplemented with haemin and vitamin K1, using an inoculum of 0.5 McFarland according to the manufacturer's instructions. MICs were determined after 48 h of incubation at 37 • C in an anaerobic atmosphere. The control strains B. fragilis ATCC 25285, Clostridium perfringens ATCC 13124, and Peptostreptococcus anaerobius ATCC 27377 were used for monthly quality control tests. Resistance to carbapenems were further studied using MBL meropenem-EDTA [15] and imipenem-EDTA double-ended strips (bioMérieux). Results were interpreted following EUCAST (http://www.eucast.org/ clinical breakpoints/ (accessed on 13 October 2020)) and CLSI criteria [40,41].

Whole Genome Sequencing and de novo Assembly
Genomic DNA was extracted from single subcultured colonies using the QIAamp DNA Mini Kit (Qiagen). Paired-end libraries were prepared using the Nextera-XT DNA Library Preparation Kit (Illumina 1.9) and sequencing was performed using the Illumina NextSeq 500 High platform (output flow cell ran at 2 × 150). The mean depth of coverage was 243 and 26 for the CNM20180471 and CNM 20200206 strains, respectively.
The Whole Genome Shotgun project PRJNA656918 was deposited at DDBJ/ENA/ GenBank under accessions JACLQC000000000 and JADDIJ000000000.
Supplementary Materials: The following are available online at https://www.mdpi.com/2079 -6382/10/3/304/s1. Figure S1. Phenotypic detection of cfiA in B. fragilis CNM20180471 with the meropenem-EDTA (left, negative) and imipenem-EDTA (right, positive) double-ended E-tests. Figure  S2: 16S rRNA sequence-based phylogenetic tree constructed using FastME v. 2.1.6.1 software (which calculates genome BLAST distance phylogeny distances; the branch lengths are scaled in terms of the GBDP distance formula). The numbers above the branches are GBDP pseudo-bootstrap support values (all are >60%; 100 replicates); average branch support = 21.3%. The trees were rooted at the midpoint. The results were provided by the Type Strain Genome Server (TYGS, https://tygs.dsmz.de (accessed on 13 October 2020)). The same color code was used in the chewBBACA analysis (Table S1). IS elements and plasmid identities of the two cfiA-positive B. fragilis strains resistant to meropenem-EDTA. Table S2: Comparison of the genome sequences of B. fragilis CNM20180471 and CNM20200206 and other strains against the genome of the B. fragilis NCTC 9343T (GCF_000025985.1) type strain and the B. fragilis YCH46 (GCF_000009925.1) representative genome, in terms of ANI, AAI, in silico genome-to-genome distance similarity (GGDH; DDH-estimate), and differences in G+C content.