Introduction

Mycobacterium tuberculosis complex molecular typing can improve our understanding of disease transmission dynamics and hence support public health systems in outbreak investigations and contact tracing [1]. Mycobacterium tuberculosis (TB) whole genome sequencing (WGS) is progressively replacing previous PCR-based molecular typing methods including MIRU-VNTR due to its increased resolution [2]. WGS is hence already implemented in a systematic and prospective approach in several public health laboratories (e.g., England, Netherlands, New York State) [3,4].

Oxford Nanopore Technology (ONT) next-generation sequencing (NGS) platforms, including the portable MinION sequencer, are increasingly used in TB diagnostics. Hall et al. recently compared ONT-based TB sequencing to the benchmark reference Illumina technology for its ability to predict drug resistance and cluster genomically related isolates [5]. This sequencing platform head-to-head comparative analysis did not include clinical isolates for which person-to-person transmission or an epidemiological outbreak context had been confirmed by conventional clinical and epidemiology investigations.

In this study, we report on the clinical validation of TBpore, a new publicly available bioinformatic analysis pipeline enabling molecular clustering of putative transmission isolates. TBpore is designed as a companion tool to the ONT compatible genotypic drug susceptibility testing pipeline Mykrobe enabling simulataneous drug resistance and clustering testing [6]. This is a retrospective case-control molecular epidemiology study to measure the clustering ability and resolution of TBpore compared to conventional MIRU-VNTR and established Illumina NGS-based molecular typing methods. This study includes isolates from domestic TB outbreaks in Canada, which were previously well characterized by routine clinical and public health investigations [7].

Results

In the M. africanum outbreak, both WGS technologies and pipelines identified 4 isolates as lineage 5 (L5) and 12 isolates as lineage 6 (L6) using SNP-typing. Four L6 isolates had identical MIRU-type (2 2 4 4 2 2 2 4 3 5 2 2 3 5 4 5 6 2 4 4 3 3 2 4) and among those, respectively, three (MB066247, MB077499, MB086368) and one (MB085690) had a strong and moderate probability of being epidemiologically related according to field investigations [7]. In the M. tuberculosis lineage 4 (L4) outbreak, 21 isolates shared a MIRU-type (2 2 4 3 2 5 1 4 3 3 2 5 2 3 4 5 3 4 4 2 3 4 7 3). All patients were from the same Montreal area and shared one or more risk factors, including alcohol and/or drug use, supervised drug injection sites visits, homelessness or prostitution. In this putative outbreak context, it was not possible to establish direct nominal epidemiological links between all patients.

The TBpore (Nanopore) and psdm (Illumina) derived SNP distances matrices showed a high level of agreement (Fig 1, S1 in S1 File). For the M. africanum outbreak, all 4 isolates previously clustered by MIRU-VNTR and conventional epidemiological investigation remained within the established thresholds when analyzed with both sequencing methods. Although one of those four genotypically clustered isolates (MB085690) was not as strongly related to the others according to public health investigation, all three typing approaches suggested its molecular relatedness. Similarly with the M. tuberculosis outbreak and control isolates, all pairwise comparison between the 21 isolates with matching MIRU-types fell within the established SNP threshold when analyzed with both methods, except for one isolate pair (MB086444, L00208036) (Fig 2, S1 in S1 File). This pair exceeded the established threshold by one SNP (13 SNP distance) when sequenced with Illumina and analyzed using COMPASS + psdm, but fell within the 12-SNP threshold when sequenced using the Nanopore platform and TBpore cluster pipeline. Except for one isolate pair (MB090711, MB090710), all M. tuberculosis control isolates were shown to be genotypically unrelated to each other as well as with isolates from the known person-to-person transmission event (Figs 1 & 2, S1 in S1 File). Those two isolates had distinct MIRU-VNTR profile and had not been identified as a potential domestic transmission event by public health teams but were respectively 7 and 20 SNP appart when sequenced and analysed with Illumina and ONT platforms and pipelines.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Whole genome sequencing and MIRU-VNTR typing results of M. africanum lineage 5 and lineage 6 outbreak.

WGS and MIRU-VNTR typing results of M. africanum outbreaks, depicting the clustering of lineage 5 (L5) and lineage 6 (L6) isolates. Pairwise SNP distances were calculated using both Illumina (psdm) and Nanopore (TBpore) sequencing methods. Isolates from the same outbreak cluster within the SNP thresholds of ≤5 and ≤12 transmission clusters, aligning with MIRU-VNTR genotyping and epidemiological data. Reference M. africanum (ATCC25420) and M. tuberculosis (ATCC27294) are included as controls.

https://doi.org/10.1371/journal.pone.0325914.g001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Illumina and Nanopore Whole Genome Sequencing-based clustering comparative analysis.

Between isolate pairwise SNP distances using Illumina (COMPASS + psdm; y-axis) and Nanopore (TBpore; x-axis) data for WGS-based typing of M. africanum (panel A) and M. tuberculosis (panel B) retrospective outbreak isolates. Isolate pairs with identical MIRU-types (triangles) either fall below (green) or above (pink) the 12 SNPs threshold.

https://doi.org/10.1371/journal.pone.0325914.g002

In the M. africanum outbreak and control isolates dataset, the Illumina SNP distances ranged from 2 to 2059, while Nanopore’s ranged from 1 to 1931 (Fig 2). The absolute differences in pairwise SNP distances between Illumina and Nanopore ranged from 0 to 432 SNPs, with an average absolute difference of 127 SNPs. Using a paired permutation statistical test and a significance threshold of 0.05, the pairwise SNP distances were found to be significantly higher with Illumina sequencing and the COMPASS analysis pipeline. Similar observations were made with the M. tuberculosis data (see S2 in S1 File).

Discussion

WGS-based molecular typing is increasingly used by public health departments due to the comprehensive insights into TB transmission dynamics it provides as well as its increasing affordability and turnaround time. The diversification and democratization of sequencing technologies and analysis pipelines is ongoing and will help better address the specific needs of different settings. ONT platform acquisition costs are lower than those of Illumina, and ONT protocols and workflows are better suited for smaller sample processing volumes. ONT may hence represent a better strategy for laboratories and public health programs in limited resource settings or those with low TB incidence without other sequencing needs. The impact of those technology-specific relative advantages remains underexplored and merit further evaluation once WGS-based molecular investigations will be performed at scale in reference laboratory and public health workflows including those of resource-limitted settings.

This study confirms the ability of ONT sequencing coupled with the novel open access TBpore pipeline to successfully support molecular epidemiology investigations. This was established using well characterized clinical isolates from two and three distinct outbreaks and sublineages (L4, L5, L6) of the M. tuberculosis complex with MIRU and Illumina WGS molecular typing results. Further validation of this novel ONT-based clustering approach on larger datasets with increased geographic catchment, lineage diversity and robust epidemiological metadata should be performed to confirm these results.

Compared to the established standard Illumina NGS platforms, we observed a trend towards lower between isolates SNP distances using ONT and TBpore clusters although, except in one case, this did not affect isolates clustering when using previously accepted transmission thresholds [8,9]. Those between isolates SNP distances differences can be due to the underlying sequencing technology, data pre-processing, processing pipelines, or variant calling parameters [5,10]. Although this finding should also be validated in larger international datasets including samples with more diverse epidemiological and genomic backgrounds, it is consistent with previous work on the validation of Nanopore technology for TB genomic clustering analyses in which this platform was also found to cluster more isolates together [5]. In this previous work, the SNP calling features and clustering pipeline were not the same as those included in TBpore clusters which we evaluated in this study. TBPore uses a more focused decontamination database and optimises variant calling parameters to allow execution of the tool on a standard laptop hence increasing portability of the Nanopore companion analysis pipeline. The previously published analysis used a Nanopore-adapted pipeline which was tailored to a high-performance computing environment where resources were not limiting. Also, clinical epidemiology data from field investigations was not available to contextualize between-sequencing technologies discrepancies.

In our study, the ONT platform and novel pipeline led to the clustering of two isolates which would not have been linked together using standard relatedness thresholds and Illumina technology. Field epidemiology data from public health authorities confirmed that those two isolates were indeed clinically related. Those two isolates also shared the same MIRU-type. Conversely, one pair of isolate was clustered together using Illumina technology but the two isolate did not share an identical MIRU-type and patients had not been identified as being epidemiology related by public health investigations. This potential increased ability of ONT to accurately identify clinically related isolates using previously proposed SNP distance thresholds should be confirmed on larger TB WGS datasets with robust epidemiological metadata.

Methods

Supporting information

Acknowledgments

We thank the Laboratoire de Santé Publique du Québec (LSPQ) for contributing to the study isolates. We thank the clinical and epidemiology teams who conducted the detailed investigations.