Skip to main page content
U.S. flag

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2019 Nov 12;10(6):e02627-19.
doi: 10.1128/mBio.02627-19.

Transcriptomic Signatures Predict Regulators of Drug Synergy and Clinical Regimen Efficacy against Tuberculosis

Shuyi Ma  1   2 , Suraj Jaipalli  3 , Jonah Larkins-Ford  4   5   6 , Jenny Lohmiller  1 , Bree B Aldridge  4   5   7   8 , David R Sherman  9   2 , Sriram Chandrasekaran  10   11
Affiliations

Transcriptomic Signatures Predict Regulators of Drug Synergy and Clinical Regimen Efficacy against Tuberculosis

Shuyi Ma et al. mBio. .

Abstract

The rapid spread of multidrug-resistant strains has created a pressing need for new drug regimens to treat tuberculosis (TB), which kills 1.8 million people each year. Identifying new regimens has been challenging due to the slow growth of the pathogen Mycobacterium tuberculosis (MTB), coupled with the large number of possible drug combinations. Here we present a computational model (INDIGO-MTB) that identified synergistic regimens featuring existing and emerging anti-TB drugs after screening in silico more than 1 million potential drug combinations using MTB drug transcriptomic profiles. INDIGO-MTB further predicted the gene Rv1353c as a key transcriptional regulator of multiple drug interactions, and we confirmed experimentally that Rv1353c upregulation reduces the antagonism of the bedaquiline-streptomycin combination. A retrospective analysis of 57 clinical trials of TB regimens using INDIGO-MTB revealed that synergistic combinations were significantly more efficacious than antagonistic combinations (P value = 1 × 10-4) based on the percentage of patients with negative sputum cultures after 8 weeks of treatment. Our study establishes a framework for rapid assessment of TB drug combinations and is also applicable to other bacterial pathogens.IMPORTANCE Multidrug combination therapy is an important strategy for treating tuberculosis, the world's deadliest bacterial infection. Long treatment durations and growing rates of drug resistance have created an urgent need for new approaches to prioritize effective drug regimens. Hence, we developed a computational model called INDIGO-MTB that identifies synergistic drug regimens from an immense set of possible drug combinations using the pathogen response transcriptome elicited by individual drugs. Although the underlying input data for INDIGO-MTB was generated under in vitro broth culture conditions, the predictions from INDIGO-MTB correlated significantly with in vivo drug regimen efficacy from clinical trials. INDIGO-MTB also identified the transcription factor Rv1353c as a regulator of multiple drug interaction outcomes, which could be targeted for rationally enhancing drug synergy.

Keywords: Mycobacterium tuberculosis; computer modeling; drug combinations; drug interactions; drug synergy; gene expression; transcription factors; transcriptomics; tuberculosis.

PubMed Disclaimer

Figures

FIG 1
FIG 1
Schematic of INDIGO-MTB. INDIGO uses drug-gene associations inferred from transcriptomic data and experimentally measured drug-drug interactions as inputs to train a computational model that can infer interactions between new combinations of drugs. It does this by learning patterns in the drug-gene associations that are correlated with synergy and antagonism. In the example above, MTB upregulation of both gene 1 and gene 3 in response to the drugs measured in monotherapy is predictive of antagonism when the drugs are combined. By perturbing individual genes and known targets of transcription factors (TFs) in the model, we can infer the impact of individual gene and TF activity, respectively, on drug interactions and subsequently engineer interaction outcomes.
FIG 2
FIG 2
INDIGO-MTB accurately predicts novel drug interactions. (A) Drug combinations chosen for experimental testing span the entire range of drug interaction predictions by INDIGO. The histogram and box plot above it show the distribution of pairwise drug interaction scores for the 35 high-interest TB agents (the edges of the box plot demarcate the 25th and 75th percentile, and the dashed line extends between the 1st and 99th percentile). The interaction scores of the combinations chosen for testing are shown as red dots. The 35 high-interest agents contain drugs either currently used to treat TB or have been used in the past to treat TB (58). (B) Comparison of INDIGO-MTB interaction scores with experimental in vitro interaction scores. Each symbol indicates a specific drug combination. Dark red dots mark two-drug regimens (R = 0.62, P = 9.3 × 10−3), and blue dots mark three-drug regimens (R = 0.64, P = 8.81 × 10−2). The specific combinations mentioned in the text are highlighted in the plot. For both experimental and INDIGO-MTB scores, values less than 0.9 indicate synergy, values between 0.9 and 1.1 denote additivity, and values greater than 1.1 indicate antagonism. (C) Dot plot of experimentally measured drug interaction scores versus the INDIGO-MTB predicted drug interaction type. The dots labeled in red font denote outlier combinations that were misclassified by INDIGO-MTB. The interaction scores were significantly different between predicted synergistic and antagonistic combinations (P = 0.0009, Komolgorov-Smirnov test). The horizontal lines in the box plot represent the median and the first and third quartiles. (D) Receiver operating curves (ROC) plotting sensitivity versus specificity for INDIGO-MTB predictions of synergy and antagonism for both two-drug and three-drug combinations in the validation set. Sensitivity measures the true positive rate, which is the fraction of true positive interactions correctly identified; specificity measures the true negative rate. The area under the ROC (AUC) values provides an estimate of the sensitivity and specificity of model predictions over a range of thresholds. The AUC values are 0.89 and 0.91 for synergy and antagonism, respectively. (sensitivity = 90.9% and specificity = 84.6% for synergy; sensitivity = 66.6% and specificity = 91.7% for predicting antagonism).
FIG 3
FIG 3
INDIGO-MTB drug interaction scores correlate with sputum culture negativity at 2 months. (A) Comparison of model predictions with sputum conversion rates in human patients after 8 weeks of treatment in clinical trials (R = −0.55, P ∼ 10−5). Higher patient negative percentages indicate more-effective regimens. Each dot indicates a specific drug combination reported from a specific clinical trial. Dots highlighted in the legend are drug combinations of interest mentioned in the text. (B) Dot plot of sputum conversion rates against the INDIGO-MTB-predicted drug interaction type. The dots labeled in red font denote outlier combinations that were misclassified by INDIGO-MTB. The horizontal lines represent the first quartile, third quartile, and median (the widest horizontal line). The colored dots correspond to combinations highlighted in the legend.
FIG 4
FIG 4
Rv1353c influences interactions between drug combinations. The in vitro experimentally measured drug interaction scores are quantified for the three selected drug interactions, plotted as the difference in FIC score of the gene perturbation relative to the wild-type strain (H37Rv). The red bars denote values for the knockout (KO) strain, and the blue bars show values for the strain with Rv1353c induced. Negative values indicate shifts toward synergy, and positive values indicate shifts toward antagonism. The asterisk and dagger indicate that differences are significantly greater or less than zero, respectively (P < 0.05, one-tailed one-sample t test). The error bars represent the standard deviations between replicates.
See this image and copyright information in PMC

Similar articles

See all similar articles

Cited by

See all "Cited by" articles

References

    1. World Health Organization. 2016. Global tuberculosis report 2016. World Health Organization, Geneva, Switzerland.
    1. Mdluli K, Kaneko T, Upton A. 2015. The tuberculosis drug discovery and development pipeline and emerging drug targets. Cold Spring Harb Perspect Med 5:a021154. doi:10.1101/cshperspect.a021154. - DOI - PMC - PubMed
    1. Zumla A, Chakaya J, Centis R, D’Ambrosio L, Mwaba P, Bates M, Kapata N, Nyirenda T, Chanda D, Mfinanga S, Hoelscher M, Maeurer M, Migliori GB. 2015. Tuberculosis treatment and management–an update on treatment regimens, trials, new drugs, and adjunct therapies. Lancet Respir Med 3:220–234. doi:10.1016/S2213-2600(15)00063-6. - DOI - PubMed
    1. Silva A, Lee BY, Clemens DL, Kee T, Ding X, Ho CM, Horwitz MA. 2016. Output-driven feedback system control platform optimizes combinatorial therapy of tuberculosis using a macrophage cell culture model. Proc Natl Acad Sci U S A 113:E2172–E2179. doi:10.1073/pnas.1600812113. - DOI - PMC - PubMed
    1. Lee BY, Clemens DL, Silva A, Dillon BJ, Maslesa-Galic S, Nava S, Ding X, Ho CM, Horwitz MA. 2017. Drug regimens identified and optimized by output-driven platform markedly reduce tuberculosis treatment time. Nat Commun 8:14183. doi:10.1038/ncomms14183. - DOI - PMC - PubMed

Publication types

MeSH terms