Genomic analysis of cardiac surgery-associated Mycobacterium chimaera infections in Italy

One hundred and twenty-two Mycobacterium chimaera strains isolated in Italy from cardiac surgery-related patients, cardiac surgery-unrelated patients and from heater-cooler units, were submitted to whole-genome sequencing and to subsequent SNP analysis. All but one strains isolated from cardiac surgery-related patients belonged to Subgroup 1.1 (19/23) or Subgroup 1.8 (3/23). Only 28 out of 79 strains isolated from heater-cooler units belonged to groupings other than 1.1 and 1.8. The strains isolated from cardiac surgery-unrelated patients were instead distributed across the phylogenetic tree. Our data, the first on isolates from Italy, are in agreement with a recent large genomic study suggesting a common source, represented by strains belonging to Subgroups 1.1 and 1.8, of cardiac surgery-related Mycobacterium chimaera infections. The strains belonging to groupings other than 1.1 and 1.8 isolated from heather-cooler units evidently resulted from contaminations at hospital level and had no share in the Mycobacterium chimaera outbreak. One Mycobacterium chimaera strain investigated in this study proved distant from every previously known Mycobacterium chimaera Groups (1, 2, 3 and 4) and we propose to assign to a novel group, named “Group 5”.


Introduction
A global outbreak of Mycobacterium chimaera (M. chimaera) infections associated with open heart cardiac surgery is ongoing. Since 2013, when the first cases, dated back to 2011, were discovered [1] more than 140 cases of severe M. chimaera infection have been identified worldwide in patients who had undergone cardiothoracic surgery with extracorporeal circulation [2]. Very early it emerged that the specific heater-cooler units (HCUs) used in the operatory rooms were contaminated by M. chimaera and likely represented the source of infection [3]. Whole  a specific brand and model (Livanova 3T, Germany), both in hospitals and at the factory, has revealed high level of genetic similarity making the most plausible hypothesis a point source contamination of the devices during manufacture [4]. Very little is known about the relation of the Italian clinical and environmental isolates to the global epidemic.
We report here the results of WGS analysis conducted on 122 M. chimaera isolates from cardiac surgery (CS) related and unrelated patients and from HCUs in different centres in Italy.

Materials and methods
A total of 122 Italian isolates of M. chimaera were included in this study; 79 from HCUs and 43 from patients, collected between 2015-2019. All clinical M. chimaera isolates were received from different hospitals together with medical records of the patients. All the isolates and the patients' medical reports were totally anonymized before our access. A written informed consent to publish the data was obtained from all involved hospitals.
Among the 43 patients, 23 had a history of cardiothoracic surgery (CS-related) and 15 had never undergone open-heart surgery (CS-unrelated). For five additional patients no information was available confirming or excluding previous open-heart surgery.
In order to compare our isolates with those already reported as part of the global epidemic we included five published environmental isolates from water supply and from new built 3T-HCUs at LivaNova production site (n = 2; Subgroup 1.8; GenBank accession nos. ERR1463901, ERR1463898), 3T-HCUs in use in Switzerland (n = 2; Subgroup 1.1 and Branch 1; GenBank accession nos. ERR1464041, ERR1463965) and at Maquet (Rastatt, Germany) HCU production site (n = 1; Subgroup 1.8; GenBank Accession no. ERR1464127).
WGS was performed on the Illumina NextSeq 500 platform and the reads were mapped on M. chimaera DSM-44623 as reference genome (GenBank accession no. LQOO00000000) using Burrows-Wheeler Aligner [5]. All datasets reached a mean coverage >50 fold, with at least 80% of the reference genomes positions complying with the thresholds of variant detection (minimum depth of coverage of 10x and 75% allele frequency). Variant calling was done using the widely employed programs Samtools [6] and the Genome Analysis Toolkit (GATK) [7].Custom perl scripts were used to filter the variants with a minimum coverage of 10 reads in both forward and reverse orientation, 10 reads calling the allele with a phred score of �20, and 75% allele frequency. The combined set of detected high quality SNP positions was used to construct a Maximum parsimony tree using RAxML version 8 [8] with a general time reversible substitution model, 1,000 re-samples and Gamma20 likelihood optimization to account for rate heterogeneity among sites. The resulting phylogenetic tree was then visualized and annotated using the online program GrapeTree [9].
In order to detect the mixed populations in the samples we reduced the threshold of variant detection (at least 2 reads calling the allele with a phred score of �20 and 5% allele frequency).
Average Nucleotide Identity (ANI) were calculated based on OrthoANIu algorithm [10], using M. chimaera DSM-44623 as reference genome. Clustal W/X software were used for multiple sequence alignment of the sequences [11] and subsequent calculation of phylogenetic networks and visualization were done with SplitsTree V4.16.1 [12] Single genomes were located to Groups, Subgroups and Branches in the basis of signature SNPs described previously [4]. Non-Group 1 and mixed genomes with the major subpopulation < 75% were excluded from phylogenetic analysis (n = 19 genomes).

Results and discussion
Out of 122 genomes only three belonged to groups other than Group 1. Their distribution among the different groupings is as reported in Table 1.
All M. chimaera isolates from CS-related patients fitted Group 1; of these, 19 belonged to Subgroup 1.1, three to Subgroup 1.8 and one to Branch 1. Subgroups 1.1 and 1.8 were also the most frequent among the HCU isolate with 44 and 7 isolates, respectively. Moreover from 10 HCUs a mixed population of M. chimaera from Subgroups 1.1 and 1.8 were isolated (Tables 1  and S1). WGS analysis confirmed nine isolates from HCUs and one from a CS-related patient, all belonging to Subgroup 1.1, were identical to one strain isolated from HCU in Switzerland [4]. Furthermore, four of our isolates belonging to Subgroup 1.8, isolated from HCUs (n = 2) and CS-related patients (n = 2), were closely related (3-4 SNPs different) to the strains grown from water supply and 3T-HCU at LivaNova production site [3].
A common source of M. chimaera infection has been recognized on the basis of the remarkable similarity between almost all the isolates from patients with a history of cardiac surgery and the ones recovered from most HCUs [4,13]. Three distinct strains of M. chimaera, belonging to Subgroups 1.1, 1.8, and 2.1, have been reported responsible of contamination of Liva-Nova HCUs at the production site [4]. This finding is in agreement with our results showing large prevalence of Subgroups 1.1 and 1.8 among the isolates from CS-related patients and HCUs (Fig 1). Only in one CS-associated patient we isolated M. chimaera belonging to Branch 1; this strain was isolated from a sputum sample and the patient did not present any of signs or symptoms related to deep infection by M. chimaera.
The isolates from CS-unrelated patients were distributed across the phylogenetic tree, mostly belonging to Branch 1. Overall, the isolates within Subgroup 1.1 showed comparatively little diversity, with a median pairwise distance of only 4 SNPs (range 0-20). These results are in agreement with clonal M. chimaera isolates, described in HCU water samples and CSrelated patient from Australia, New Zealand and also US patient strains [14,15]. Unaccountably, two isolates from two CS-unrelated patients, in different hospitals, grouped with Subgroup 1.1 and were genetically very similar to those from CS-related patients and HCUs (mean pairwise distance of <5 SNPs).
A very peculiar M. chimaera strain was isolated from lower respiratory infection in a CSunrelated patient. The INNO-LiPA MYCOBACTERIA v.2 line-probe assay ascribed the strain The group/subgroups were assigned based on the specific SNP signatures [4].  [4], the strain proved different from all four known groups, with the groups 1, 2, 3 and 4 being 23793, 6837, 9431 and 23818 SNPs distant respectively (Fig 2). Present study has some limitations related to the partial coverage of the Italian M. chimaera epidemic and to lack of epidemiological data linking individual patients to specific HCUs.

Conclusions
In conclusion, our data are consistent with the hypothesis that disseminated M. chimaera in CS patient are related with the use of a specific HCU; more the similarity of isolates in different part of the world pinpoint a common, single source of infection.
Clinicians should monitor patients who have had cardiac surgery using HCUs for signs and symptoms of M. chimaera infection to enable early diagnosis and treatment. Finally, we shown that WGS is the preferred method to distinguish whether a clinical strain is related to the HCU outbreak strain. The M. chimaera genomes from HCUs, patients and Published genome data are indicated in different colours. The two major outbreak subgroups 1.1, 1.8 and also other subgroups are indicated by red labels. A cut-off of 5 SNPs was used to collapse the branches. For this analysis, we combined all group 1 isolates from our study with five other published genomes from LivaNova HCUs, water supply at production site and Maquet HCU production site. We excluded isolates for which a mixed strain population was detected based on signature SNPs [4]. https://doi.org/10.1371/journal.pone.0239273.g001