We selected 107 RR-TB strains harboring 34 unique rpoB mutations, including 30 within the RRDR (Multimedia Appendix 1), from the Belgian Coordinated Collections of Microorganisms and the World Health Organization Tropical Disease Research mycobacteria collection in the Institute of Tropical Medicine, Antwerp, Belgium. We prepared thermolysates of these strains as previously described, stored them at –20°C [14], and subjected them to RDTs, namely Xpert MTB/RIF, GenoType MTBDRplus v2.0, and Genoscholar NTM + MDRTB II.

We measured the DNA concentration of the thermolysates through an Invitrogen Qubit 2.0 Fluorometer (Thermo Fisher Scientific, Waltham, MA, USA) following the manufacturer’s instructions. The calculated weight of DNA per M tuberculosis bacillus served as the divisor for each DNA concentration, the quotient of which was the equivalent colony-forming units (CFUs) per milliliter of each thermolysate.

We subjected thermolysates with initial concentrations of 10⁸ to 10⁹CFU/mL to 1:2 and 10-fold serial dilutions until we obtained 10⁶ to 10⁷CFU/mL, respectively. We then dispensed a 1:2 mixture of diluted thermolysate and sample reagent in the Xpert MTB/RIF cartridge following the manufacturer’s instructions.

We subjected undiluted thermolysates to the amplification and hybridization steps of GenoType MTBDRplus v2.0 and Genoscholar NTM + MDRTB II following the manufacturers’ instructions.

We validated RDT results against the strains’ rpoB gene sequences, allowing for correlation between probe reactions and specific rpoB mutations. We then performed a comprehensive comparison and analysis of rpoB mutants and associated RDT results using the consensus numbering system based on M tuberculosis strain H37Rv [2]. We visualized results of the analysis using Geneious version 10.1.2 (Biomatters Limited) [15] and Affinity Designer 1.5.3 (Serif (Europe) Ltd) [16]. The percentage average reproducibility of the RDTs was analyzed through mutation profiles found in more than one strain.

Our laboratory validation of rifampicin-resistant strains and visualization of results revealed the specific relationship between unique RDT probe reactions and the majority of documented rpoB mutations. We also found RDT probe reactions representing low-frequency mutations in particular settings that have a high potential for use in evaluating transmission. We are proposing a novel approach to the optimal use of readily available diagnostic data through the shift to fully automated and digitized surveillance for real-time detection of RR-TB transmission hotspots.

To our knowledge, this work is the first pragmatic and scalable response to the consensus of world-renowned TB experts, published in 2016 [12], that diagnostic connectivity has the potential to eliminate TB. This is evidenced by the ability of our proposed approach to facilitate comparison of probe reactions between different RDTs used in the same setting, potentially linked to an open source software platform, by the Connected Diagnostics Initiative [12], for instance. Our findings affirm the important role of diagnostic connectivity platforms in accelerating TB control efforts and highlight the importance of fully understanding the relationship between TB diagnostic test results and drug resistance-conferring mutations. Our proposed approach will maximize the use of connected RDTs and associated data shared in repositories through a plug-in module that will automatically translate RDT results to useful information for RR-TB outbreak investigations. This will potentially aid laboratory personnel, clinicians, and national TB control programs in closing detection gaps of RR-TB cases by revealing abnormal figures or trends in particular settings, as well as real-time probable horizontal transmission and epidemiologic linkage between patients.

The comprehensive analysis of claimed and observed probe reactions revealed the ability of the RDTs to detect documented RRDR mutations. Most of the RRDR mutations analyzed yielded the expected result, concordant with manufacturers’ claims. We observed the highest concordance between observed and claimed probe coverage in Genoscholar NTM + MDRTB II, exemplified, for instance, by probe S2 detecting mutation N437D, which was missed by Xpert MTB/RIF probe C and GenoType MTBDRplus v2.0 probe WT5 (Figure 2,Multimedia Appendix 1, and Figure 1). Remarkably, the mutations missed by one RDT probe but captured by another were situated in the regions where the probes overlap. This affirms the role of overlapping probes in improving the usefulness of RDTs for RR-TB detection. Critically, mutations M434T missed by Xpert MTB/RIF and M434V missed by Genoscholar NTM + MDRTB II are not yet catalogued in the Tuberculosis Drug Resistance Mutation Database, nor in RefSeq [20,21]; hence, their clinical relevance is yet to be determined. Alarmingly, missed mutations, including N437D (Xpert MTB/RIF), S428R, and D435F (GenoType MTBDRplus v2.0), are known to confer rifampicin resistance [22]. These undetected rifampicin resistance-conferring mutations within the RRDR, together with those outside the RRDR, most notably the I491F mutation [2], may be untraceably circulated through chains of transmission, as samples with these mutations alone will be falsely classified as rifampicin susceptible, rendering treatment ineffective. Additionally, the impact of silent mutations such as T444T must be further assessed to ensure that recommended probes generate results with high specificity for rifampicin resistance. Accordingly, we suggest that manufacturers review and modify claims of mutation coverage based on the findings of this work. Additionally, for future versions of the line probe assays, manufacturers might consider synthesizing mutation probes that would also capture low-frequency mutations. These proposed modifications would improve identification of underlying mutations and thus increase the usefulness of linked RDTs for detection of rifampicin resistance and for epidemiologic surveillance.

We also gathered probe reactions representing low- to high-prevalence RRDR mutations [4,12,13,18,19,22] in the most comprehensive comparison of rpoB mutants and associated RDT probe reactions to date and propose their potential application for transmission studies. The use of RDT probe reactions in defined geographic settings for evaluating transmission is grounded in the need to easily deduce the underlying RRDR mutation to identify genotype clustering in time and space. When probe reactions between 2 strains correspond, even between distinct RDTs, the prevalence of probe reactions and associated RRDR mutations in a given setting must be considered when assessing the probability of a transmission event taking place. This probability will be relatively higher for probe reactions representing low-frequency mutations. This is explained by the principle that, in a normal distribution, the probability of plotting a low-frequency mutation in the tail region is low. Fitting it in the exact same location for the second time has much lower probability. Hence, when low-frequency probe reactions representing less-prevalent mutations correspond between 2 strains, an epidemiologic link might exist between the 2, and the occurrence of transmission can possibly be suspected. For example, Xpert MTB/RIF probe C, with the lowest frequency in Pakistan and Nigeria [20,21], is a potentially valid marker of S441L/Q transmission. To differentiate between the 2 mutations, RDT probe reactions must be combined with complementary genotyping test results. This proposed approach is evidenced by mutations L452P [23] and D435G [18] detected in strains epidemiologically linked to an extensively drug-resistant TB outbreak in KwaZulu-Natal, South Africa. Accordingly, we assert that Xpert MTB/RIF probes A, Bdelayed, C, and Edelayed; GenoType MTBDRplus v2.0 probes WT2, WT5, and WT6; and Genoscholar NTM + MDRTB II probes S1 and S3, which have a low frequency in specific settings [20,21] and detect less-prevalent mutations (emphasized in Figure 2), have a high potential for use in transmission studies. On the contrary, the probability of detecting a highly prevalent probe reaction in the distribution is high, but fitting it into the exact same point the second time would have lower probability. It would be unlikely that this is a random event solely attributable to high-frequency mutations being enriched by evolutionary convergence [24]. Therefore, when this is observed, further investigation of suspected cases of transmission is necessary. Particularly in settings of low RR-TB incidence, probe reactions corresponding with highly prevalent mutations such as Xpert MTB/RIF probes E, D, and B [20,21], if repeatedly obtained in the same health facility, may be potentially useful for finding an epidemiologic link between strains. The potential usefulness of RDT probe reactions for transmission studies lies in accounting for the frequency of probe reactions and RRDR mutations associated with RR-TB prevalence in a specific geographic setting. Linking this novel approach to a connectivity software can potentially bring about real-time reporting of suspected RR-TB transmission cases, which national TB control programs and public health officials may then investigate and confirm or exclude based on epidemiologic linking.

In addition to analyzing routine RDT data for evaluating transmission in detail, we found that RDT probe reactions corresponding with disputed mutations, supplemented with rpoB gene sequences linked to connectivity software, may assist in resolving discordance between phenotypic and genotypic rifampicin susceptibility results. Our observed probe reactions for disputed mutations were concordant between Xpert MTB/RIF and GenoType MTBDRplus v2.0, and were consistent with previous reports [13]. In contrast with discordant Xpert MTB/RIF and Mycobacterium growth indicator tube results due to the disputed mutation L430P [13,25], reliable probe reactions between RDTs were observed in this work. Consequently, detecting Xpert MTB/RIF probe A, GenoType MTBDRplus v2.0 probe WT2, or Genoscholar NTM + MDRTB II probe S1 would suggest an L430P mutation, despite obtaining a rifampicin-susceptible Mycobacterium growth indicator tube result. For other cases, disputed mutation D435Y can be distinguished from other variations at position 435, particularly from D435V, by a higher proportion of delayed results for probe B (Multimedia Appendix 3). Another delayed reaction observed was for probe E, associated with the disputed mutation L452P. Delayed results [3] for Xpert MTB/RIF probes B and E could be attributed to a reduced intensity of fluorescence in the end-probe codons. Alternatively, the specific nucleotide substitution, associated with the average proportion of delayed results and ∆Cts (Multimedia Appendix 3), may have caused the delays. These inferred associations may aid in better differentiation of Xpert MTB/RIF probe reactions that have a high potential use for assessing transmission and resolving discordance between phenotypic and genotypic drug susceptibility results. On the contrary, disputed mutations H445L and H445N associated with high in vitro resistance [26] could not be distinguished from undisputed mutations in codon 445; complementary rpoB sequences would be very beneficial in this case. We strongly suggest that each specific mutation and RDT result be assessed case-by-case for discordant results when compared with phenotypic drug susceptibility testing and their potential for transmission clustering studies.

The limitation of the study was the lack of strains with RRDR mutations outside labeled codon positions in Figure 1. However, to our knowledge, the uncovered mutations have not yet been encountered in clinical isolates and thus are likely to be extremely rare. This work is not a field evaluation of the potential of connected TB diagnostic data, but rather a pragmatic laboratory validation and analysis of test results.

We propose a novel approach for potential real-time detection of RR-TB transmission through fully using connected TB diagnostics and a shared repository of test results. To our knowledge, this is the first pragmatic and scalable work in response to the consensus of world-renowned TB experts in 2016 on the potential of diagnostic connectivity to accelerate efforts to eliminate TB [12]. This is evidenced by the ability of our proposed approach to facilitate comparison of probe reactions between different RDTs used in the same setting.

While less-prevalent RDT probe reactions corresponding with low-frequency mutations would be most suitable for evaluating transmission, combining this approach with complementary genotyping tests such as membrane-based spoligotyping, MIRU-VNTR, deep targeted sequencing, or whole genome sequencing on cultured isolates [9] would increase the resolution of the approach in identifying RR-TB transmission hotspots. Supplementing RDT test results with spoligotyping data, for instance, not only illustrates data pooling [12], but also enhances the discriminatory power of this combined approach to confirm suspected cases of TB transmission. This proposed prospective approach, inspired by the SpoNC tool, which combined pncA sequencing and spoligotyping data for detection of pyrazinamide-resistant TB transmission clusters [9], will be validated against conventional approaches through strains with specific mutations and originating from documented clusters.

Integrating the combined approach as an additional module in the connectivity platform will not only automate the analysis of shared RDT results and its translation into crucial clinical and public health information, but also allow more precise real-time estimation of transmitted RR-TB proportions in a population. Detection of RR-TB transmission hotspots would initiate a timely outbreak response and appropriate investigation of suspected transmission cases based on epidemiologic linking that would prevent further spread of TB.