Emerging resistance to the nitroimidazole drugs delamanid (DLM) and pretomanid (Pa) threatens treatment outcomes for rifampicin-resistant and multidrug-resistant tuberculosis (RR/MDR-TB). Rapid molecular diagnostics are limited by incomplete knowledge of resistance-conferring mutations, and phenotypic drug susceptibility testing (pDST) is challenged by standardization of critical concentrations (CC). This study aimed to correlate minimum inhibitory concentrations (MICs) determined by a Resazurin Microtiter Assay (REMA) for DLM and Pa with whole genome sequencing (WGS) data from clinical Mycobacterium tuberculosis isolates that were phenotypically DLM-resistant by routine MGIT testing.
The investigation used baseline isolates collected through routine National TB Program (NTP) practice in Georgia between February 11, 2021 and July 31, 2024. The project had institutional review board approval at the National Center for Lung Health (NCLH) in Tbilisi; biobanked isolates and genomic data from a prior nationwide project were used under existing approvals. Clinical data were pseudonymized and no patient-identifying information was accessible to the study team.
Routine DLM pDST at the NCLH used the BACTEC MGIT 960 system with a WHO-recommended CC of 0.06 μg/ml. During the study period, 469 RR-TB patients were identified; DLM pDST was performed for 425 (90.6%). Thirty-two (7.5%) isolates were initially classified as DLM-resistant by routine testing. Repeat MGIT testing reclassified 11 of those isolates as susceptible, resulting in 21 reported as DLM-resistant; four isolates were then excluded (one failed regrowth, three lacked WGS data), leaving 17 isolates for REMA MIC testing and WGS analysis.
All available baseline DLM-resistant isolates were subcultured and tested by REMA. DLM concentration series spanned 0.008–2 μg/ml; Pa concentrations spanned 0.125–32 μg/ml. H37Rv was included as a control and yielded MICs ≤0.008 μg/ml for DLM and ≤0.125 μg/ml for Pa within the tested ranges. Plates were incubated at 37 °C and MIC endpoints were defined visually after Alamar blue color change at 21 days as the lowest concentration inhibiting growth in triplicate wells.
DNA was extracted and sequenced using Illumina platforms at sequencing centers in Basel. Reads passing quality filters were taxonomically classified and non-MTBC reads removed. Reads were aligned to an inferred MTBC ancestor reference and variants called with GATK Mutect2 in microbial mode. Analyses focused on nonsynonymous mutations, frameshifts, indels, and nonsense mutations in established DLM/Pa-associated genes: ddn, fgd1, fbiA, fbiB, fbiC, fbiD, and additionally dprE2. Variants in repetitive PE/PPE/PGRS regions were excluded. Resistance profiles were assigned relative to the WHO mutation catalogue where possible.
A multiple alignment of non-redundant SNPs (minimum intra-sample allele frequency 90%) was used to infer a maximum-likelihood phylogeny with IQ-TREE 2, rooted on M. canettii. The dataset comprised predominantly lineage 2 isolates (14/17, 82%) and lineage 4 (3/17, 18%). Resistance-associated variants showed predominance in the L2.2.1 sub-lineage; the premature ddn W88* stop codon was detected exclusively within the W148 outbreak cluster.
REMA testing showed that 8 of 17 isolates (47%) had DLM MICs below the WHO CC of 0.06 μg/ml, indicating susceptibility despite initial MGIT-based resistance calls. Seven isolates had MICs at or below the lowest tested concentrations (≤0.008 μg/ml for DLM and ≤0.125 μg/ml for Pa). One isolate with DLM MIC 0.016 μg/ml had a Pa MIC of 2 μg/ml. The nine isolates with REMA DLM MICs above the CC had DLM MICs of 0.125–2 μg/ml (six at 2 μg/ml, two at 0.125 μg/ml, one at 0.25 μg/ml) and corresponding Pa MICs were elevated (2–8 μg/ml).
WGS identified mutations in DLM/Pa-associated genes in 9 of 17 isolates (53%). Mutations in ddn were the most frequent, notably the premature stop codon W88* present in five isolates; these isolates displayed elevated DLM MICs (0.125–2 μg/ml) and Pa MICs (4–8 μg/ml). Mutations in fbiC were detected in three isolates (including FbiC C565W and W678G). Two isolates carried fgd1 substitutions (Q317R, Q279H). No fbiA, fbiB, fbiD, or dprE2 mutations were observed in this dataset.
Notably, several variants were not previously described in public resources including the WHO catalogue: ddn G35fs, fgd1 Q317R and Q279H, and fbiC C565W. Allele frequencies varied across isolates; some variants were present at high allele fractions. Among the isolates with DLM MICs below the CC, all were wild-type in the canonical resistance genes except one with fbiC W678G (96.7% allele frequency).
The dataset supports associations between certain variants (notably ddn W88* and fbiC W678G) and elevated DLM and Pa MICs, consistent with cross-resistance between the nitroimidazoles. The detection of previously undescribed mutations located exclusively within genes linked to DLM/Pa susceptibility indicates candidate resistance determinants that warrant experimental validation and inclusion in mutation catalogues when supported by further evidence. The observed discordance between routine MGIT-based resistance calls and REMA-derived MICs in nearly half of the isolates highlights challenges in pDST interpretation and the potential for differing assay performance or critical concentration thresholds.
Isolate sequencing data are deposited in the European Nucleotide Archive under study PRJEB39561. The study reported IRB approval at the NCLH and used biobanked isolates from a prior genomic project with existing approvals. Financial support included a Fogarty International Center (NIH) grant; the funders did not influence study design, data collection, analysis, or publication decisions.
Linking REMA MIC data with WGS in this set of Georgian M. tuberculosis isolates identified known and novel genomic variants within DLM/Pa-associated genes. The ddn W88* premature stop codon correlated with higher MICs to both delamanid and pretomanid in multiple isolates. Newly observed variants in ddn, fgd1, and fbiC require further study to determine causal roles in resistance and to inform molecular diagnostic development and WHO/EUCAST interpretive frameworks.