1. Introduction
Tuberculosis (TB) is the number one cause of human death due to an infectious disease, with 1.7 million deaths per year worldwide [
1]. The causative agents of TB are a group of closely related bacteria known as the
Mycobacterium tuberculosis (Mtb) complex (MTBC), which has been thought to have low DNA sequence diversity [
2]. This limited diversity, however, is influenced by selective pressures and background selection [
2]. Various human-adapted MTBC variants are known to differ in virulence, progression of disease and transmission potential.
TB surveillance of highly-virulent and multi-drug resistant (MDR) strains is paramount for adequate diagnosis and treatment [
1,
3]. Traditional phenotypic drug susceptibility testing (DST) through culture-based methods has multiple caveats, amongst them being that TB culturing can take days to weeks [
4]. To reduce the time to obtain test results, alternative methods like real-time PCR-based Xpert MTB/RIF testing have been recommended by the World Health Organization [
5]. These methods, however, are unable to detect drug-resistance mutations outside of the selected target loci [
6], or they can produce false positive results [
7]. In addition, clinical management where TB risk is high is often challenged by a lack of resources such as facilities for chest X-rays or laboratories for Mtb isolation and culture. To address these challenges, a whole-genome sequencing (WGS) approach can generate antibiotic susceptibility profiles, detect MDR-TB, and discover other MTB virulence factors [
3,
4]. This method, however, is also limited by resources, hospital–laboratory infrastructure and personnel training in bioinformatic analysis. A hybridization-based system (reverse line probe assay) has been recently proposed as an alternative in cost to WGS, but since this methodology is based on hybridization, it is also limited to the genomic region of Mtb examined [
8]. Furthermore, although the cost per sample is much less than for other assays, it still requires laboratory equipment.
Development of a diagnostic assay that can be used at the point of care to rapidly and accurately diagnose TB and to include multidrug-resistant tuberculosis (MDR-TB) or extensively drug-resistant tuberculosis (XDR-TB) should be given a high priority. MDR-TB characterization typically requires costly machinery and handling in a specialized reference laboratory, not to mention the time required for shipping and processing the sample. A portable sequencing system that could be taken to the field, would not only reduce the cost of TB testing, but will also speed up the diagnoses. A rapid direct sample sequencing device would significantly reduce the time to obtain test results.
The MinION -Oxford Nanopore Technologies Limited (ONT), is a pocket sized (10 × 3 × 2 cm), portable, USB-powered, long-read sequencing instrument [
9]. Among the existing sequencing platforms, it has the potential to be the best suited method to investigate the chain of transmission of TB and to determine the susceptibility of anti-TB drugs in the near future. This platform is particularly useful in remote settings or with limited infrastructure [
9]. A careful evaluation of MinION as a potential methodology for the surveillance of TB was first reported in 2017 [
10]. In this investigation, authors used both Illumina and ONT platforms for the diagnosis of Mtb infection. Utilization of the MinION in this study was conducted only with simulated Mtb infection using Ziehl–Neelsen (ZN)-negative sputum DNA combined with
Mycobacterium bovis BCG strain DNA, not direct sputum sample. Despite the advantage of a portable sequencer in MDR-TB testing, so far there is no peer-reviewed published protocol of ONT-WGS based, rapid MDR-TB testing of patient sputum samples. It is unknown if this portable DNA sequencing system would be effective in providing information on Mtb genotype, drug-susceptibility in a sputum sample.
In this pilot study, we evaluated the performance of this portable sequence system for Mtb species identification and detection of genes related to drug resistance, as a means of MDR-TB testing in a diverse set of samples, including DNA isolated from sputum samples and from clinical microbiological isolates.
3. Discussion
In this study we have evaluated the genomic identification and drug mutation gene profiling of Mtb isolates utilizing the MinION portable sequencer. Our findings endorse the need of further research regarding the practical use of MinION for the detection and characterization of Mtb in clinical isolates and in sputum samples. Our sample set consisted of Mtb genomic DNA obtained by different extraction methods. Recently, low-cost DNA extraction methods for Mtb WGS directly from patient samples have been reported [
10], allowing the bypass of laboratory equipment requirements for genomic DNA obtainment.
The portable WGS-based detection system utilized here proved to be fast, relatively inexpensive, with rapid and simple library preparation, and automated real-time analysis tools [
10]. The most innovative aspect of this sequencing system is its portability. Its small size and use of a USB port are ideal as they reduce the infrastructure required for WGS sequencing, such as a climate-controlled building, instead requiring only a laptop computer for the system to be operational [
9,
12].
The MinION has several advantages that make it uniquely suited for TB surveillance (
Supplement Table S1). Amongst its features, the MinION provides long-read sequencing data, which are ideal for the detection of antimicrobial resistance genes [
13], and some authors suggest that this can be achieved even without the need of a high amount of reads [
14]. The real time monitoring allows the analysis of metagenomes from complex samples, which could save the 14 days of culture required for drug susceptibility testing in TB. In our set of DNA samples obtained directly from sputum, the presence of host DNA was far more abundant than Mtb DNA, but bacterial DNA could be discriminated and drug resistance related genes were detected, albeit at low sequencing depth. Although identification at the MTBC level provided by Xpert and other fast methods is usually enough for the diagnosis of TB, direct species assignment from sputum samples is an advantage to highlight. Another big challenge in the clinical setting is the bioinformatics analysis, as most clinical labs do not have trained personnel. Real time antimicrobial resistance profiling is indeed, a crucial advantage to highlight. The steps from raw data acquisition to analysis completion are fairly simple and easy to follow in their user-friendly EPI2ME platform [
15]. Furthermore, the analysis can be performed in real time even from the moment data acquisition begins, potentially minimizing the results waiting time even more.
The number of mutations in drug resistance related genes overly surpassed those detected in previous WGA. This may have several explanations. First, a high error rate has been acknowledged as a limitation of Nanopore technology [
16], thus, some of these could correspond to sequencing errors, in spite of the overall accuracy of around 90%, according to the automated results. The initial high error rates reported for the MinION [
17], have improved over the past few years [
18], currently over 95% raw read accuracy and 99.9% consensus read accuracy is achievable. Incorporation of complementary short read sequences [
18], and the use of short DNA target sequences, circularized and then amplified via rolling-circle amplification to produce high fidelity accurate repeats [
19], are new proposed ways to reduce the error rate. Additionally, recent statistical methods have been reported to aid in the accurate detection of true mutations [
20] Long read sequencing has a superior advantage over short read WGS approach, especially in homopolymeric regions where indel is commonly used by bacteria as a drug resistance strategy [
21]. Therefore, although the higher number of drug resistance (DR) related genes found in this study using MinION may be due in part a high error rate, it is also reasonable to think that more genes were detected by the long read sequencing compared to the traditional short read WGS. Further investigation is needed to clarify this. Additionally, it would be interesting to follow up on a newer version (R10) of Nanopore’s Flowcell compared to the version used in this study (R9), as improved accuracy with longer barrel and dual reader head in the sequencing pores shall provide better accuracy especially in homopolymer regions. Alternatively, the higher number of detected DR genes by the MinION could correspond to false positive hits detected by the automated ARMA pipeline. The WHO endorses the use of next-generation sequencing analysis for drug-resistance profiling, only for a limited number of genes (
rpoB,
katG and
inhA for first line drugs and
gyrA,
gyrB,
rss and
eis promoters for second line drugs) and for specific point mutations within them [
16]. However, the reference database used by the ARMA pipeline includes genes that lack empirical support for their clinical relevance in TB [
16,
22]. Almost half of the “hits” corresponded to this category (see
Appendix A,
Appendix B,
Appendix C,
Appendix D,
Appendix E,
Appendix F,
Appendix G), indicating that the reference database needs further curation. In addition, some mutations in known resistance conferring genes could correspond to polymorphisms with no functional impact depending on the mutated codon (this is not disclosed in the automated analysis) [
16], which could explain the detection of resistance related genes in susceptible isolates. The same could be said for genes like
mtrA or AAC(2’)-IC, which were detected in 5 out of 7 isolates irrespective of their resistance profile and could correspond to polymorphisms.
Nevertheless, the sensitivity of the MinION sequencing for the detection of drug resistance mutations was good. Isolates 410 and 6548 belong to the extensively studied MDR M strain [
23,
24] which accumulated resistance to several drugs. The ARMA pipeline detected three of the four drug resistance conferring mutations and an additional mutation in isolate 410, and all six resistance mutations of isolate 6548. Interestingly, the
gidB mutation, which confers resistance to streptomycin, is not the most frequent among clinical isolates but is characteristic of this cluster and was acquired four decades ago when the expansion of this cluster began [
24]. In addition, a rifampicin resistance conferring mutation was found in the metagenome of the sputum sample 1766, which belongs to the Ra cluster, another conspicuous MDR strain of Argentina [
25]. These phenotypically confirmed drug-resistance conferring mutations were identified with two to 17 reads depending on the gene, with similar accuracy values. This indicates that although it is usually regarded a critical variable in the analysis of next-generation sequencing data, sequencing depth was not the main constraint in our work. Prompt and accurate information on Mtb strains would have implications for management to minimize transmission of drug-resistant TB and start the most appropriate TB control and anti-TB therapy. Various phylogenetic lineages of the Mtb complex are distributed differently around the world [
2]. In Latin America, both drug susceptible and drug resistant TB are mainly related to the Euro–American Lineage [
26,
27,
28,
29] and the Beijing strain has a minor impact, in contrast to what is reported in other regions. Drug resistance databases mostly rely on the genome H37Rv strain. It is interesting to challenge this sequencing system with samples sets with diverse genetic backgrounds like ours to assess its impact in the performance.
Overall, our findings indicate that the improvements in the future should focus on: (1) recovering higher number of reads corresponding to Mtb from sputa; (2) lowering MinION sequencing error rates; (3) improving the drug-resistance conferring mutation detection algorithms for automated analysis and (4) curating the reference database to include only those hits that have a strong correlation with Mtb drug resistance phenotype.
Although our data relies on a short number of DNA samples, our findings suggest that this portable DNA sequencing system could be effective in reducing time and providing information on Mtb genotype and drug-susceptibility from direct sputum samples. As larger studies—evaluating parameters such as the minimal number of reads for a complete reliable drug susceptibility profiling, optimization in software and database accuracy for the prediction of new drug resistance genes, and reduction in false positive drug detection—are conducted, this system could potentially revolutionize current TB testing procedures, especially in genomic surveillance for MDR-TB in the clinical setting.