NGS for whole genome sequencing to diagnose MDR TB infection

Currently, next-generation sequencing (NGS) is widely used for drug resistance rapid diagnosis, so that prospective treatment and prognosis of patients with MDR-TB can be determined. Moreover, NGS can be used to sequence the whole-genome of MTB. However, there are some challenges in using NGS i.e. sample preparation is elaborated, so the methodology needs to be simplified and optimized for direct sputum samples; in some developing countries, there is lack of capacity, and high cost may hamper its use in global surveillance [3].

NGS provides options to detect the whole characteristics of MDR-TB. Unlike other molecular tests for MDR-TB, which relies on indirect MTB identification by a series of limited resistance mutations through hybridization of a probe to a specific genetic sequence, the NGS may provide detailed whole genome information. In NGS, each genome parts are sequenced several times, so this assessment provides accurate data and insight into the presence of genetic polymorphisms [26]. Therefore, NGS can confirm and assess genome level mutations that cannot be detected by other molecular tests. Moreover, NGS is flexible and can be programmed for various applications, such as for assessment of genetic information of co-existing micro organisms other than MTB that is present in a clinical sample [10].

Despite the challenges, the benefits of NGS remain more compared to other methods, when they are used for diagnosis, to determine appropriate treatment, and epidemiology. The benefits of NGS include time for faster diagnosis of MTB (from weeks in other methods to hours in NGS), the amount of information provided (information about strains, whether single or multiple infection; distinguishing recurrence from re-infection; drug resistance profile; and  chain of transmission) and comfort (potential to combine diagnosis, drug resistance, and epidemiological analysis) that can be collected in real-time.  In addition, NGS can be done on direct clinical specimens or cultured MTB isolates [27].

Apart from the benefits over other molecular methods, the use of NGS is hindered in low and middle-income countries, as NGS needs high cost, trained operator, special means for data analysis/storage, and plug-and-play device to directly sort information from primary clinical samples, which are not readily available in those countries [10, 27].

Both phenotype and genotype tests have been used to detect MTB drug resistance to various types of drugs. The phenotype test is obtained by culture, which requires a long time from several weeks to months, while genotyping may shorten the time, so this method, especially those using NGS is highly needed [28].

Whole Genome Sequencing (WGS) is a method to determine the nucleotide sequence of certain genomes, which are the genetic material of whole cells or microbes.28 This method might obtain  genetic and epidemiological data, as well as indications of cross contamination risk, with high sensitivity [29, 30]. The WGS provides more precise and accurate information and improves understanding of MTB pathogenesis, immunology, and transmission [31].

WGS requires a genome as a reference to show various genetic mutations/sample variations. The genome, which is the most used as a reference, is MTB H37Rv genome with worldwide distribution and wide usage. The H37Rv genome is one of the best for curating MTB, due to transcriptomic and proteomic data sources that are available in the world. In molecular epidemiology and phylogenetic studies, mutations are associated with anti-tuberculosis drug resistance in MTB [23].

Various NGS platforms

NGS, which is a "high throughput, massive and parallel" sequencing method,  can be used to assess the sequence of a whole genom where all genes are  ordered along the whole target genom in a single biochemical reaction. This technique sort multiple DNA fragments in parallel, assemble and map them according to the reference genome using bioinformatic analysis [10, 32].

To obtain the maximum potential of NGS in a clinical setting to improve TB diagnosis and management, the NGS workflow and platform should be understood, so that optimal integration of NGS into existing laboratory workflows can be achieved [6, 33]. There are two main NGS platforms, i.e. that works by producing relatively short DNA sequences (<1000 base pairs), or longer DNA sequences (long reads of > 5,000 base pairs in average). The speed to get detailed information from NGS holds the potential to increase the accuracy of MDR-TB diagnosis, speed up TB treatment, as well as better understand TB epidemiology. NGS provides information in the form of thousands of millions of varying length reading to be assembled into longer sequences or even the whole genome, using bioinformatics. In addition, NGS platforms might use different sequencing chemistry (Table 1) [34, 35].

All NGS platforms use similar basic workflow to get sequence readings from clinical samples, which consist of: DNA extraction, purification, target specific enrichment, and data analysis. DNA extraction from clinical samples (sputum/sediments) or cultured isolates should have sufficient yield, integrity, and purity. High quality DNA is crucial in NGS, especially WGS, thus efficient DNA extraction and purification methods are highly important. The DNA extraction method should be able to break the MTB cell wall to release the DNA to be purified. After purification, the next step is target specific enrichment. The resulting DNA fragments are then sorted in parallel, followed by data analysis, where the fragments are mapped according to the reference genome using bioinformatics special software [10].

Whole genome sequencing can be used as a tool to determine sequence variations in epidemiology and to determine the source of infection and transmission pattern between various patients. WGS uses NGS technology that can generate millions of brief readings from the entire genome, which is then analyzed by a special software [32].

Specific target enrichment

Next generation sequencing needs to do specific target enrichment that aims to eliminate genomic DNA regions that are not of interest in a certain experiment. By targeting specific areas, a greater sequence information is obtained for the areas of interest. In addition, time, resources and cost can be reduced, so that more individual samples can be analyzed. The specific target enrichment strategy can be carried out by various approaches to capture specific sequences of interest, i.e: hybrid capture, selective circularization, and PCR amplification [36].

Hybrid Capture is conducted using  nucleic acid sequences, which are derived from input samples, to be specifically hybridize to pre-prepared DNA fragments  that are complementary to the target area of interest, so that the target area of interest can be captured, isolated, and enriched, either on a solid support or in solution [36].

Selective circularization is conducted using molecular inversion probes (MIP), or gap-fill padlock or selector probes in the form of single strand (ss) DNA circles, which contain the  target areas to be enriched. The ss DNA circles, which are formed by gap filling and ligation in such a way, are used for selective amplification of the target area of interest [36].

PCR Amplification is conducted using polymerase chain reaction (PCR) method, which is directed to the target areas of interest, by carrying out many remote PCRs in parallel, in limited quantities. The PCR method is very multiplex that amplifies a very large number of short fragments [36].

Next generation sequencing methods

There are several NGS methods that need to be known and are often used for diagnostic tools. They are WGS, targeted and shotgun NGS. The first method is WGS.

WGS is a method to analyse the entire sequence of a cell whole genome at a single time to provide the most comprehensive information of the genome.  Therefore, WGS provides the most comprehensive data on a certain organism, due to its ability to show the profile of an entire genome; thus WGS enables scientists to find genes or variants that were previously unknown, and can provide large amounts of data in a short time. WGS can even predict the likelihood of genetic disorders,  which can be mediated through lifestyle or environmental changes,  to help in prevention of the occurrence of the disorders [37].

The second method is the targeted NGS. The targeted NGS is a type of NGS that focuses on specific genome areas, and scrutinize genome specific areas for deeper analysis. Therefore, it is faster and more cost-effective than WGS, but produce less data compared to WGS, as it analyse only selected specific areas [38].

The third method is the shotgun NGS, which is a technique to determine the sequence of all chromosomes and the whole genome by producing random DNA fragments with overlapping ends, followed by using a computer to assemble and order the fragments according to the overlapping ends into the whole genome. Shotgun NGS begins with the construction of a genome library. The genome library is produced by fragmenting the whole genome, the fragments are cloned into vectors, partially sorted and using a computer the data are collected to produce an entire genome sequences that consists of all fragments, which form known old sequences that are called contigs.  “A contig is a set of overlapping DNA segments that together represent the consensus region of DNA” (cited from Yin and Zheng) [39].  Many contigs are often separated by unknown sequences. Unknown sequences can be determined either by hybridization protocols using known contig ends, or random PCR primers to 

reveal the sequence of the gap region between known contigs. In hybridization protocol, the sequence of the gap between contigs is determined by filtering the hybridized clone library with a probe that is generated from the end of previously identified contigs. In random PCR primers’ method, the primers are annealed to the ends of two known contigs, and thus fill the gap and reveal the unknown sequence [39].

A special approach, which can be incorporated into the three methods are multiplexing, also called pooling, so that several samples can be processed simultaneously, to save costs and time. Multiplexing requires the addition of a barcode (index) to the sample so that they can be identified after being sorted. Samples used for hybridization catches can be duplicated after library preparation, but before target capture (enrichment). Samples used for sequential amplification must be transformed into libraries and enriched through individual PCR amplification before multiplexing to be sorted [39].

In a population-based multi-country study conducted by WHO using various NGS platforms (e.g. Illumina, Ion Torrent) and targeted methods, the results showed that NGS was very accurate in diagnosing MDR-TB from isolates that are cultured. 9 The study assessed various genes, including katG, promoter inhA and fabG to assess isoniazid, pncA to asses pyrazinamide, rpoB to asses rifampicin, gyrA and gyrB to asses ofloxacin and moxifloxacin, rrs to asses amikacin and capreomycin, and rrs and eis to asses kanamycin resistance. The study was conducted on 7.000 patients and found that there was good prediction accuracy of genetic sequences to show resistance to isoniazid, rifampicin, fluoroquinolones, and injections. This study strengths included the fact  that the tested isolates came from a whole population of TB cases with a high TB or MDR-TB burden [10, 40]. Table 2 shows various studies that were conducted in Asian countries and used NGS and previous sequencing methods, which showed sensitivity and specificity of above 90% [41-46]. Most studies found that almost all specimens contained resistance to a minimum of two anti-TB drugs (isoniazid and rifampicin).

Many studies have been carried out using NGS method to test drug susceptibility of MTB, in which MTB susceptibility testing using NGS for clinical application gave faster results compared to the phenotypic testing.

Ko et al [47] studied eighteen susceptible strains with benign polymorphisms, to assess phenotypic and genotypic resistance patterns for isoniazid, rifampicin, ethambutol, and pyrazinamide. Thirty-nine variants were found in this study, and six mutations were regarded as causing resistance i.e.: katG c.C944G, rpoB c.C1349T, embB c.A916G and c.G918A, gyrA c.C269T and rpsL c.A128G. There were two main causes of isoniazid resistance, which were in katG and inha, but a most common variant, which was katG R463L, was regarded as a benign polymorphism. On the other hand, the most common resistant mutation, which was in S315T katG promoter area, was mostly found in Indonesia.The rifampicin rpoB mutation occurred in codon 407-533, and gave rise to S450L variant. For ethambutol resistance, mutation occurred in codon 306 of embB, while pyrazinamide resistance was due to two pncA mutations that might cause frame-shift mutations.

A recent study compared phenotypic testing with NGS Illumina platform to assess various TB drug resistance on overall isolates, and provided interpretations of mutations in katG, inha, fabG promoters, rpoB, pncA, gyrA, gyrB, rrs, and eis genes. The NGS was carried out by standard approaches to assess MTB mutations and provide drug resistance genetic sequencing data.  The data was obtained by sorting all relevant genomes/genes to the target genome region of katG, inha, and fabG promoters for isoniazid, rpoB for rifampicin, pncA for pyrazinamide, and gyrA/gyrB for fluoroquinolones. The results on isolates from 7094 TB patients showed that overall sensitivity value of NGS to predict genetic resistance varies between various drugs. Good accuracy of sensitivity values was found for first-line drugs (katG promoter, inha, fabG, and rpoB) compared to second-line drug resistance, but lowest for detection of pyrazinamide resistance due to pncA mutation [10, 13].

Several studies showed that the main cause (96%) of rifampicin resistance was due to mutations, which occurred in the centre  of a 81-base pair  region in rpoB gene, and rpoB S531L was the most common mutation in clinical MTB isolates. In countries around the world, 40%-93% of all rifampicin resistant MTB isolates was due to rpoB S450L mutations. In addition, the most extensive rifampicin resistance was due to rpoB gene single amino acid mutation in codon 445 that changed His [CAC] to Tyr [TAC] [15-19].

Falzon et al [23]  did a meta-analysis of individual MDR-TB patient data and concluded that the most common fluoroquinolone resistance was due to GCG90GTG and GAC94GGC mutations; where isolates with a GAC94GGC mutation usually showed a Minimum Concentration (MIC) that is higher compared to a GCG90GTG mutation.

A study concluded that NGS provided information that enable easier differentiation between various strains, and assessment of MTB genetic heterogeneity, by which MDR-MTB Beijing strain complex evolutionary pattern was revealed. Moreover, NGS showed more accuracy than other genotype methods, namely Mycobacterial interspersed repetitive unit- variable number tandem repeat (MIRU-VNTR) and Xpert MTB/RIF assay. So that the complex evolutionary patterns of Beijing MDR MTB strains in patients have been revealed using NGS [33].

Maningi et al study [21] using NGS showed that pncA gene mutation in Pyrazinamide resistance revealed additional diagnostic values that were accurate and might be considered in routine testing for MDR-TB management strategies. Eighty-eight isolates were sequenced using NGS Ion Torrent (PGM) platform, and 55 wild type and 33 pncA gene mutation containing isolates were found. PncA mutations in 33 of these isolates represent nucleotide substitution (missense mutations), deletions that cause amino acid substitution or frame shift mutation, and insertion. Using the NGS, additional mutations in 6 isolates were detected that were undetected in previous Sanger sequencing.

A study by Daum et al [17] in Ukraine used NGS to analyze MDR mutations in inhA, katG, and rpoB genomes and showed that all clinical isolates (100%) showed inhA wild type gene, among which 23% had mutation in inhA promoter region, while the most frequent mutation (89%) that caused isoniazid resistance occurred in katG S-315-T. For rifampicin resistance, 95% mutations occurred in rpoB ‘81-bp nucleus region’, which was rifampicin resistance determining region (RRDR), and most mutations occurred in positions 516 (D-516-V), 526 (H-526-Y or D) and 531 (S -531-L). Further, two rifampicin resistant isolates were isoniazid sensitive, and five rifampicin resistant strains showed heteroresistance, i.e.: two strains of S-531-S/F /L, which are rare mutations  with one wild type and two mutant residues, two strains of S-531-S / L and one strain of H-526-H / D, where the three later strains were double mutants.

For amikacin and kanamycin resistance, rrs A1401G mutation, which caused high level resistance to both drugs, was the most common in Thailand (21 of 29 strains). This mutation showed cross-resistance with capreomycin. In addition, mutations in eis promoter caused low-level resistance to amikacin, which was found in five out of 29 strains that were kanamycin resistant [24]. Miotto et al [18] concluded that the mutations, which were associated with amikacin resistance, were only rrs A1401G and G1484T, and rrs mutations were associated with high MIC in a variety of different media.

Conclusions