The genome of the strain H37Rv of Mycobacterium tuberculosis, the reference strain used for many applications in TB research, was published in 1998.1 It took several years and millions of euros to finish the genome, which was the second obtained after Haemophilus influenzae. Nowadays we can sequence not a reference strain, but any clinical strain of M. tuberculosis for less than a hundred euros. We can do that in less than a week if enough resources are available, and usually in batches of dozens or hundreds of strains.
This leap forward in sequencing and computational capacities, together with democratization of sequencing, is mediating the transition of sequencing from the research environment to clinical and public health. The transition is fueled by three major advances in the case of M. tuberculosis. First, the information in the genome of M. tuberculosis holds clinical and public health actionable results.2 It provides data on the genotype of the strain (as exemplified in the manuscript published in this issue3); data on transmission, so it can be used as a high-resolution epidemiological marker; and data on mutations potentially associated with drug resistance. Thus, with a sole determination we can obtain information previously provided by multiple different methods, and usually with higher resolution.
Second, we now have catalogues of mutations that allow us to predict from the genome sequence the drug susceptibility profile of a strain. Those catalogues have very high sensitivity and specificity for first-line drugs.4 For example, in a recent retrospective study in Comunitat Valenciana, our TB consortium showed a high concordance in drug resistance profiles of hundreds of strains between the phenotype determined in the hospitals and the genotype prediction from genomic sequences.5 Additional examples and exceptions will be discussed below, but the point is that catalogues of mutations are mature enough to be used at the clinical and surveillance level, with limitations. In that sense, the release of the WHO catalogue version 1.0 in 2022 represented a landmark in tuberculosis.6
Third, the democratization of sequencing and costs allows us to implement sequencing at a population scale in almost real time, if needed, and closer to the bedside. For example, as part of a research project, more than 70% of culture positive isolates have been sequenced by the TB consortium of the Comunitat Valenciana. This is helping, depending on the setting and the clinical or public health question, to personalize treatment,7 estimate drug resistance burden,8 or carry out transmission studies at a scale not available before.9 Fast turnaround times have helped us to advise on mutations acquired during treatment in real time or to assist public health in ongoing outbreak investigations.
Below I will detail some selected examples of how genomic data is making an impact on different aspects of the clinical management of TB.
Genotyping to understand local M. tuberculosisIn this issue of EIMC, isolates from TB cases from Hospital Virgen del Rocío, Seville, Spain are genotyped using genomics. As usually happens, a blend of lineages and sublineages is found. Globally, Lineage 4 is the most common, but Lineage 4 is itself complex and not all sublineages are everywhere.10 In Spain, for example, L4.1.2, L4.3.2 and L4.10 are usually found, and most locally born cases are infected with strains of those sublineages. As there is a strong association between lineages and geography, the identification of the infecting strain's lineage can help to understand if the case is imported and suspect whether it may be associated with drug resistance.11 Lineage 2 strains from former Soviet republics are the most typical example, as they tend to be associated with MDR-TB, as exemplified in this issue in the isolates from Seville.3 But evidence is emerging that different clinical phenotypes are associated with different lineages. The most recent example is the association of Lineage 1 with extrapulmonary disease. Lineage 1 is mainly found in the Rim of the Indian Ocean. This association was first discovered after the epidemiological analysis of an outbreak in North Carolina, where most of the cases were extrapulmonary, an unusual feature compared to the much more common pulmonary TB.12 Research in the laboratory identified that Lineage 1 produces a protein, esxM, not present in the lineages circulating in Europe and elsewhere (Lineages 4, 3, 2). The new protein mediates the capacity of Lineage 1 strains to cause extrapulmonary TB. Later studies in Europe, where Lineage 1 is rare, show an association of cases infected with Lineage 1 with extrapulmonary forms of the disease. While the mutation does not account for the full clinical manifestation, it is an example of how understanding what is circulating may help to pinpoint unusual clinical phenotypes of Mtb strains. Importantly, in Spain, initiatives such as the one mentioned in the Comunitat Valenciana, Andalucia,3 Catalonia and elsewhere are generating a wealth of genomic data from circulating strains, which will help to understand our local epidemic at multiple geographic levels.
Transmission and outbreak investigation using the genome as an epidemiological markerWhile traditional molecular epidemiology markers interrogate 1% of the genome, access to the whole genome allows us to interrogate almost 100%. This has promoted whole genome sequencing as the gold standard molecular epidemiology marker in tuberculosis.13 We and others have shown that the whole genome can assist in outbreak investigations in low-burden countries. For example, by differentiating the co-existence of different outbreaks at the same time. Importantly, when applied in a universal manner, as we do in the Comunitat Valenciana since 2014, it can expand the outbreak investigation by identifying additional cases linked to the outbreak, thus complementing conventional contact tracing. In addition, when applied at a population level, it can help to survey the burden of recent transmission. For example, we have shown that recent transmission – happening in the last years and which is the realm of public health investigations – can be very different from country to country. In Comunitat Valenciana, around 35% of the cases are in a recent transmission cluster and associated with Spanish-born individuals. However, in countries like the United Kingdom or Germany, the clustering percentage is much lower and TB cases are associated with imported infected cases rather than transmission.9 Thus, even countries with similar burden can have very different transmission profiles, and the TB epidemic must be handled in a tailored manner. Genomic transmission has been key also in other areas, for example, to revealed the role of prisons in the spread of TB in different settings14 or to understand the emergence and spread of multidrug resistance. For several decades now, we know that in countries with high MDR-TB burden, transmission – rather than acquisition during treatment – is the major driver of the epidemic. Mirroring what happened in the past, we are starting to see bedaquiline-resistant strains in transmission, suggesting poor TB control is already impacting the efficacy of one of the major landmarks in TB treatment in the last 40 years.15 Bedaquiline is a drug that is key for second-line treatment shortening and will probably be part of a future, brand-new, first-line treatment. We need to protect it and not replay the errors from the past.
Genome as a diagnostic marker for drug resistanceThe whole genome sequence of the bacteria also harbours key diagnostic information. First, the genomic information can identify if the isolate belongs to the M. tuberculosis complex or to other non-tuberculous mycobacteria. But more importantly, it harbours clinically actionable results in the form of drug resistance-associated mutations. For a very long time, molecular diagnosis of drug resistance in M. tuberculosis lagged behind other major pathogens like HIV. The advent of whole genome sequencing at a large scale, however, allows us to compare the genome of phenotypically resistant bacilli and identify, via statistical associations, the mutations behind. By analyzing collections of thousands of isolates with paired genotypic and phenotypic information, it has been possible to build catalogues of mutations. The most widely used and standardized is that developed by the WHO, currently in its second version.6 The catalogue defines more than 1300 mutations significantly associated with antibiotic resistance to 15 different drugs, making the genome the most comprehensive diagnostic tool available. Importantly, it also defines almost 13,000 mutations commonly found that are not associated with resistance and another 33,000 which are still under evaluation. The catalogue has high sensitivity and specificity for first-line drugs, although it is still limited for second-line, particularly for new and repurposed drugs. For example, sensitivity to Bedaquiline is 49.4% and to Delamanid 14.7%. The numbers do not only reflect a limitation of the catalogue, but also the limited available data we have on circulating drug-resistant strains for those drugs, for which phenotypic testing is also not fully standardized.
Applying the catalogue to the Comunitat Valenciana setting (years 2014–2016), we found a very good prediction of pan-susceptibility (96.4%) and genotype–phenotype concordance (98.7%).5 However, specificities were lower than those found in the catalogue for a global collection. This reflects two important aspects. One is that the catalogue needs to be enriched in terms of drug-resistant strains and geographic breadth, something that future versions of the catalogue are contemplating. And second, in a low-burden MDR-TB setting like ours, there can be a small number of discrepancies that importantly affect sensitivities. In fact, a detailed analysis of those discrepancies shows that some are due to false negatives in culture, while others are false negatives in genomes. For example, we described in 2019 a patient carrying a borderline mutation to rifampicin.16 That mutation is known to generate false negatives in the BACTEC/MGIT system and was not targeted by any molecular assay in the hospital. Failure to recognize resistance to rifampicin early led to an amplification of resistance, including the emergence of rare mutations to isoniazid, ethambutol, capreomycin, and fluoroquinolones. Complementing the work at the microbiology unit of the hospital, genome sequencing allowed clinicians to adjust the treatment regimen of this complicated case, and as a result, treatment was finally successful.
Importantly, large-scale genomic datasets have enabled the development of mutation catalogues like the one described above. This has facilitated alternative next-generation sequencing-based diagnostic approaches known as targeted sequencing (tNGS). Targeted sequencing allows simultaneous interrogation of multiple bacterial regions that harbor the most prevalent drug resistance mutations, generating a drug susceptibility profile. Thus, tNGS represents an intermediate step between molecular approaches such as Xpert/Ultra, currently used in clinics, and whole-genome sequencing.17
In fact, thanks to interim evidence of high sensitivity for first-line drugs, targeted sequencing was conditionally recommended by the WHO in 2023 guidelines.18 Its main advantage over whole-genome sequencing is that it does not require sequencing the entire genome. Instead, it uses PCR to enrich targeted regions coupled with deep sequencing, which can be applied directly to clinical specimens and can potentially detect heteroresistance at very low levels (∼3%).17 Some approaches, though not all, also include targets for typing major lineages.
However, because tNGS does not interrogate the entire genome, it cannot be used for epidemiological purposes and still depends on sequencing equipment and expertise.
Limitations and future perspectivesThe examples above reflect the breadth of applications that whole genome sequencing can play in TB control, from public health to diagnostics. But also, each country should identify the role of whole genome sequencing in their national health ecosystem. Whole genome sequencing is not cheap, is not as standardized as other techniques in hospitals, and requires expertise in bioinformatics often missing in clinical settings. Also, it is not a replacement for rapid, point-of-care molecular tests like Xpert/Ultra, which have had a major impact on TB management. Although it is slowly replacing culture in certain low burden countries. In those countries the challenge is how to make it cost-effective as numbers are low and thus cannot be sequenced in a cost-effective manner. One proposed way is by pooling samples from multiple suspected pathogens detected in clinical microbiology units or by centralized sequencing, although the latter usually implies delays in results reports. In other countries, sequencing can play a role for surveillance and as an evaluation of point-of-care tests. For example, in Mozambique we have detected the circulation of rifampicin and isoniazid-resistant strains not detected by Xpert/Ultra, which is the frontline test of the national control programme.19 Also in Mozambique, another group was able to describe transmission of bedaquiline resistant strains.8 Importantly, one major limitation is that whole genome sequencing depends on culture, so ironically, it still suffers from the same problem of slow response. There is work in progress that will allow us to sequence the genome from the diagnostic sample, but it still has limited sensitivity compared to Xpert and is not standardized.20 However, around 38% of all TB-diagnosed cases today are diagnosed with molecular tests, particularly Xpert Ultra. As technology, cost, and miniaturization develop, there is no reason to think this could not be the case for genome sequencing in the future, while it has proven in the present that it has a role in TB control and management.
FundingThis project received funding from the Spanish Ministry of Science and Innovation (PID2022-137607OB-I00) and the Generalitat Valenciana (Project CIPROM/2023/30).
The author thanks the Tuberculosis Genomics Unit for their support. Most of the work described here was carried out as part of the Valencian Community TB Consortium, which involves microbiology units and public health offices from the Comunitat Valenciana. I. Comas is part of the CSIC Global Health Platform.
