Alphafold3 to Model Conformational Changes in the Mycobacterium Tuberculosis Mycobacterial Membrane Protein Regulator (MMPR) to Enhance the Accuracy of Targeted Deep Sequencing for Predicting Phenotypic Resistance to Bedaquiline

Abstract Background Accurate and timely detection of drug resistance in Mycobacterium tuberculosis is essential for the effective treatment of multidrug-resistant TB (MDR-TB). Current sequencing methods often fall short in sensitivity and specificity, particularly because mutations in mmpR5 (Rv0678), the transcriptional repressor of the Mycobacterial Membrane Protein efflux pump—the gene most frequently linked to clinical bedaquiline resistance—lead to variable resistance phenotypes. Leveraging AlphaFold3, we aim to model the conformational changes in MmpR5 that correspond to different levels of bedaquiline resistance. These structural insights could improve targeted deep sequencing accuracy, enhancing resistance prediction and treatment outcomes. Methods In this study, we conducted a literature review to identify genotype-phenotype correlations in mmpR5 pinpointing resistance-associated variants (RAVs) that span a spectrum of phenotypic resistance to bedaquiline. Using sequence libraries, we retrieved corresponding mmpR5 sequences for each RAV. We then modeled these variants with AlphaFold3 to predict conformational changes and examined their impact on biologically important regions of the protein, such as DNA binding and dimerization domains. Structural comparisons between each RAV and the wild-type protein provided insights into how these mutations may contribute to variable resistance levels, enhancing the potential of targeted deep sequencing to predict clinical resistance profiles. Results Through our literature review, we identified three mmpR5 resistance-associated variants (RAVs) representing a spectrum of phenotypic resistance to bedaquiline (BDQ). These include: Q22R, V85A, and V92fs with low or non-raised MIC (≤ 0.03 µg/mL); G66 and M139fs with an intermediate MIC at the critical concentration (0.25-0.5 µg/mL), and L142R, C46fs, and M10fs, a frameshift mutation resulting in high-level resistance (≥ 2 µg/mL). Conclusion RAVs linked to low-level BDQ resistance, such as Q22R, V85A, and Y92fs, exhibit minor structural changes within DNA binding and dimerization domains compared to wild-type. Conversely, high-resistance mutations like L142R and C46fs cause substantial alterations in these domains, supporting a correlation between structural disruption and elevated phenotypic BDQ resistance. Utilizing AlphaFold3, we can model the impact of RAVs on critical MmpR5 domains, potentially improving the accuaracy of sequencing to predict phenotypic BDQ resistance. Comparison of Alphafold predictions of wild-type and mutant MmpR5. Wild-type MmpR5 is depicted in red and mutant MmpR5 proteins in green. Low or no MIC elevations for BDQ for Q22R, V85A, and Y92fs; intermediate resistance for G66E and M10fs; and high MICs for L142R, C46fs, and M139fs.