Biology subjects

Aldridge, B. B.

Publications and source records attributed to Aldridge, B. B..

17 recordsLinked to original sources

bioRxiv2026

Modeling development of tertiary lymphoid structures in pulmonary tuberculosis by 3D profiling and trajectory analysis

Tertiary lymphoid structures (TLSs) are sites of immune organization in peripheral tissues that arise from chronic inflammation. They play important roles in infection control and cancer but the mechanisms controlling their formation remain only partly understood. Here, we combine high-plex imaging, serial-section 3D reconstruction, and optimal transport trajectory modeling to reconstruct TLSs in human lungs infected with Mycobacterium tuberculosis. We find that the extended and irregular shape of TLSs is poorly captured by 2D histopathology or spatial profiling, making assessment of developmental stage (early, primary, or secondary) error prone. In contrast, modeling TLS development in 3D using optimal transport reveals two trajectories that differ in the timing and coordination of follicular dendritic cell association, germinal center consolidation, and position within the tissue. These findings highlight the value of volumetric analysis in understanding immune organization and provide new insights into TLS biology. Graphical Abstract O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=98 SRC="FIGDIR/small/725201v2_ufig1.gif" ALT="Figure 1"> View larger version (32K): org.highwire.dtl.DTLVardef@11b4daorg.highwire.dtl.DTLVardef@7871feorg.highwire.dtl.DTLVardef@1840c2dorg.highwire.dtl.DTLVardef@3b6028_HPS_FORMAT_FIGEXP M_FIG C_FIG HighlightsO_LI3D reconstruction of TB-infected lung reveals TLSs to be extended, heterogeneous structures that are frequently misclassified in 2D C_LIO_LITLS maturation follows continuous trajectories rather than discrete histological stages C_LIO_LIOptimal transport modeling identifies distinct maturation paths with divergent follicular dynamics C_LIO_LIGerminal center organization and Tfh-FDC interactions emerge along maturation trajectories C_LIO_LITLS maturation states are shaped by spatially localized microenvironmental niches. C_LI

immunology↗

bioRxiv2026

In vitro dormancy models improve ability to predict treatment response in severe marmoset tuberculosis lesions

Tuberculosis lesions are structurally and physiologically heterogeneous, creating microenvironments that restrict antibiotic penetration and alter bacterial susceptibility. This heterogeneity remains a major impediment to shortening tuberculosis treatment, as it is difficult to predict which regimens will sterilize bacteria in these hard-to-treat lesion niches. Many in vitro systems and commonly used animal models fail to recapitulate the necrotic, caseous lesion pathology associated with treatment failure and relapse. Here, we show that resource-efficient in vitro assays engineered to mimic key lesion microenvironments, including those with lipid-rich, caseum-like matrices with controlled oxygenation and pH, generate drug-response metrics that predict lesion-level treatment responses in a non-human primate model. We stratified individual marmoset lesions by baseline 2-deoxy-2-[18F] fluoro-D-glucose ([18F] FDG) positron emission tomography/computed tomography (PET/CT) features, then quantified drug potency and interactions across three lipid-induced dormancy (LIDs) conditions under both equipotent and lesion pharmacokinetic-informed dosing schemes. LIDs-based and pharmacokinetic-informed measurements aligned more closely with treatment responses in severe lesions than conventional in vitro data and recapitulated interaction patterns consistent with known regimen performance. Integrating imaging-derived and in vitro features in multivariate, sequential computational models improved predictive accuracy over baseline imaging alone and identified a subset of informative in vitro metrics. Together, these results establish a scalable framework linking lesion-mimicking assays to lesion-specific outcomes, enabling earlier, more cost-effective prioritization of pre-clinical regimens aimed at sterilizing hard-to-treat tuberculosis lesions and shortening therapy.

molecular biology↗

bioRxiv2025

Strain diversity drives heterogeneous responses to tuberculosis combination therapy

BackgroundStrain diversity in Mycobacterium tuberculosis (Mtb) underlies distinct clinical presentations and outcomes, but the range of drug susceptibility phenotypes among clinical isolates is poorly understood. We aimed to identify drug response patterns in phylogenetically diverse clinical isolates to combination treatment. MethodsOut of 641 drug-sensitive clinical isolates, we selected 13 strains that capture local and global phylogenetic diversity and included Erdman ATCC-35801 as a reference. We selected ten antibiotics with diverse mechanisms of action to study phenotypic responses to combination therapy. We treated each strain with 10 single drugs, 45 drug pairs, and 20 three-way combinations in standard and cholesterol-rich media. To compare combination treatment responses across strains and conditions that have varying doubling times, we computed normalized growth rate inhibition metrics (GRmax). FindingsMtb clinical strains displayed a broad range of drug response phenotypes across the 65 drug combinations and two metabolic conditions tested. The most effective drug pairs (based on potency and synergy) varied both by strain and metabolic condition. Within our 14-strain panel, strains that were less sensitive to single drugs were also less sensitive to combination treatment, with very few exceptions. For all drug combinations tested, the magnitude of GRmax variation across all strains was driven primarily by variation among genetically related strains, rather than between genetically distant strain groups. InterpretationPreclinical studies should reflect the diversity of Mtb clinical strains; our data suggest that selecting strains based on the range of drug response phenotypes displayed, rather than by genetic diversity alone, may better account for the effects of strain variation. Our findings also support the understanding that constituent drug pairs of high-order combinations target metabolically heterogeneous Mtb. Selection of these pairs should likely involve multiple factors including the infecting strain, metabolic niche, and drug response metrics. FundingGates Foundation INV-027276; NIH P01AI143575&1F32AI174653; Wellcome 206724/Z/17/Z

microbiology↗

bioRxiv2025

Increased proportion of growth-arrested bacilli in acidic pH adaptation promotes Mycobacterium tuberculosis treatment survival

The ability of Mycobacterium tuberculosis (Mtb) to adapt its growth behavior in response to host environments promotes survival against immune and drug stressors. However, how these behaviors shift at the single-cell level remains poorly understood. Here, we show that Mtb adapts to acidic conditions by increasing the proportion of bacteria in a growth-arrested state, rather than uniformly slowing the growth rate of the entire population. This non-growing subpopulation exhibits greater tolerance to ethambutol. Clinical strains displayed higher proportions of growth-arrested cells under both neutral and acidic conditions, suggesting that growth arrest serves as a bet-hedging strategy during infection. Though the PhoPR two-component system contributes to regulating this non-growing state, we show that it is a partial regulator of the non-growing bacterial subpopulation and that additional transcriptional regulators are involved. Our study demonstrates that non-growing subpopulations of Mtb provide fitness benefits and are an active adaptation to environmental cues and not a passive consequence of stressors.

microbiology↗

bioRxiv2025

Fleming: An AI Agent for Antibiotic Discovery in Mycobacterium Tuberculosis

Antibiotic development is challenged by high costs and failure rates. Artificial intelligence (AI) holds promise to overcome these challenges by predicting inhibitory properties of novel compounds, generating new candidates, and contextualizing property predictions in the biological background. Fleming is an integrative AI agent that explores novel chemical space to identify lead compounds meeting multiple criteria. The discriminative and generative AI models for Mycobacterium tuberculosis (Mtb) inhibition were trained on a set of 114,900 diverse compounds and fragments based on in vitro growth inhibition. We combined both models as well as molecular optimization, ADMET prediction and literature search functions to make Fleming an integrated agent for Mtb preclinical lead identification. Fleming has 17% higher discrimination between known Mtb leads and leads for other diseases than a generic LLM agent along with 13% higher discrimination than molecular property prediction alone on challenging ADMET tasks. Fleming demonstrates an 83% in vitro hit rate of predicted inhibition and a 100% hit rate of de novo generative design. Flemings generative designs also demonstrate an 83% rate of favorable ADMET profiles. Fleming is an integrative AI agent able to explore new regions of the chemical space to select lead compounds that simultaneously meet several desirable criteria.

bioinformatics↗

bioRxiv2025

Multiomic analysis identifies suppressive myeloid cell populations in human TB granulomas.

Tuberculosis (TB) remains a major global health challenge, particularly in the context of multidrug-resistant (MDR) Mycobacterium tuberculosis (Mtb). Host-directed therapies (HDTs) have been proposed as adjunctive therapy to enhance immune control of infection. Recently, one such HDT, pharmacologic modulation of myeloid-derived suppressor cells (MDSCs), has been proposed to treat MDR-TB. While MDSCs have been well characterized in cancer, their role in TB pathogenesis remains unclear. To investigate whether MDSCs or other myeloid suppressor populations contribute to TB granuloma microenvironments (GME), we performed spatial transcriptional profiling and single-cell immunophenotyping on eighty-four granulomas in lung specimens from three individuals with active disease. Granulomas were histologically classified based on H&E staining, and transcriptional signatures were compared across regions of interest (ROIs) at different states of granuloma maturation. Our analysis revealed that immune suppression within granuloma was not primarily driven by classical MDSCs but rather by multiple myeloid cell subsets, including dendritic cells expressing indoleamine 2,3 dioxygenase-1 expressing (IDO1+ DCs). IDO1+ DCs were the most frequently observed suppressive myeloid cells, particularly in cellular regions, and their spatial proximity to activated T cells suggested localized immunosuppression. Importantly, granulomas at different stages contained distinct proportions of suppressor myeloid cells, with necrotic and cellular regions showing different myeloid phenotypes that may influence granuloma progression. Gene set enrichment analysis (GSEA) further indicated that elevated IDO1 expression was associated with a complex immune response that balanced suppressive signaling, immune activation, and cellular metabolism. These findings suggest that classical MDSCs, as defined in tumor microenvironments, likely play a minor role in TB, whereas IDO1+ DCs may be key regulators of immune suppression in granulomas influencing local Mtb control in infected lung. A deeper understanding of the role of IDO1+ suppressive myeloid cells in TB granulomas is essential to assessing their potential as therapeutic targets in TB treatment.

pathology↗

bioRxiv2024

Integration of multi-modal measurements identifies critical mechanisms of tuberculosis drug action

Treatments for tuberculosis remain lengthy, motivating a search for new drugs with novel mechanisms of action. However, it remains challenging to elucidate the direct targets of a drug, and even more so, to determine which disrupted cellular processes lead to bacterial killing. We developed a computational tool, DECIPHAER (DEcoding Cross-modal Information of PHarmacologies via AutoEncodeRs), to select the important correlated transcriptional and morphological responses of Mycobacterium tuberculosis to drug treatments. By finding a reduced feature space from these measurements, DECIPHAER highlighted essential features of Mtb cellular damage such as phosphosugar stress and inhibition of translation and DNA replication. After training, DECIPHAER provides cell-death-relevant insight into single-modal datasets, enabling interrogation of drug treatment responses for which transcriptional data are unavailable. Using morphological data alone with DECIPHAER, we discovered that respiration inhibition by the poly-pharmacological drugs, SQ109 and BM212, can influence cell death more than their effects on the cell wall. This study demonstrates that DECIPHAER can extract the critical shared information from multi-modal measurements to identify cell death-relevant mechanisms of TB drugs.

microbiology↗

bioRxiv2024

Massively parallel combination screen reveals small molecule sensitization of antibiotic-resistant Gram-negative ESKAPE pathogens

Antibiotic resistance, especially in multidrug-resistant ESKAPE pathogens, remains a worldwide problem. Combination antimicrobial therapies may be an important strategy to overcome resistance and broaden the spectrum of existing antibiotics. However, this strategy is limited by the ability to efficiently screen large combinatorial chemical spaces. Here, we deployed a high-throughput combinatorial screening platform, DropArray, to evaluate the interactions of over 30,000 compounds with up to 22 antibiotics and 6 strains of Gram-negative ESKAPE pathogens, totaling to over 1.3 million unique strain-antibiotic-compound combinations. In this dataset, compounds more frequently exhibited synergy with known antibiotics than single-agent activity. We identified a compound, P2-56, and developed a more potent analog, P2-56-3, which potentiated rifampin (RIF) activity against Acinetobacter baumannii and Klebsiella pneumoniae. Using phenotypic assays, we showed P2-56-3 disrupts the outer membrane of A. baumannii. To identify pathways involved in the mechanism of synergy between P2-56-3 and RIF, we performed genetic screens in A. baumannii. CRISPRi-induced partial depletion of lipooligosaccharide transport genes (lptA-D, lptFG) resulted in hypersensitivity to P2-56-3/RIF treatment, demonstrating the genetic dependency of P2-56-3 activity and RIF sensitization on lpt genes in A. baumannii. Consistent with outer membrane homeostasis being an important determinant of P2-56-3/RIF tolerance, knockout of maintenance of lipid asymmetry complex genes and overexpression of certain resistance-nodulation-division efflux pumps - a phenotype associated with multidrug-resistance - resulted in hypersensitivity to P2-56-3. These findings demonstrate the immense scale of phenotypic antibiotic combination screens using DropArray and the potential for such approaches to discover new small molecule synergies against multidrug-resistant ESKAPE strains. Significance StatementThere is an unmet need for new antibiotic therapies effective against the multidrug-resistant, Gram-negative ESKAPE pathogens. Combination therapies have the potential to overcome resistance and broaden the spectrum of existing antibiotics. In this study, we use DropArray, a massively parallel combinatorial screening tool, to assay more than 1.3 million combinations of small molecules against the Gram-negative ESKAPE pathogens, Acinetobacter baumannii, Klebsiella pneumoniae, and Pseudomonas aeruginosa. We discovered a synthetic small molecule potentiator, P2-56, of the antibiotic rifampin effective in A. baumannii and K. pneumoniae. We generated P2-56-3, a more potent derivative of P2-56, and found that it likely potentiates rifampin by compromising the outer membrane integrity. Our study demonstrates a high-throughput strategy for identifying antibiotic potentiators against multidrug-resistant bacteria.

microbiology↗

bioRxiv2024

Unique growth and morphology properties of Clade 5 Clostridioides difficile strains revealed by single-cell time-lapse microscopy

Clostridioides difficile is a gastrointestinal pathogen of both humans and agricultural animals and thus a major One Health threat. The C. difficile species consists of five main clades, with Clade 5 currently undergoing speciation from Clades 1-4. Clade 5 strains are highly prevalent in agricultural animals and can cause zoonotic infections, suggesting that these strains have evolved phenotypes that distinguish them from Clade 1-4 strains. Here, we compare the growth properties of Clade 5 strains to those of Clade 1-4 strains using anaerobic time-lapse microscopy coupled with automated image analysis. Our analyses indicate that Clade 5 strains grow faster and are more likely to form long chains of cells than Clade 1-4 strains. Using comparative genomic and CRISPRi analyses, we show that the chaining phenotype of Clade 5 strains is driven by the orientation of the invertible cmr switch sequence, with chaining strains exhibiting a bias to the cmr-ON state. Interestingly, Clade 5 strains with a bias towards the cmr-ON state shifted to a largely cmr-OFF state during murine infection, suggesting that the cmr-OFF state is under positive selection during infection. Collectively, our data reveal that Clade 5 strains have distinct growth properties, which may allow them to inhabit diverse ecological niches. Author SummaryThe Clade 5 strains of the Clostridioides difficile species are so phylogenetically divergent that they almost meet the threshold of being a distinct species. Although these strains are ubiquitously isolated from agricultural and environmental settings and an important source of zoonotic and community-acquired infections, it is unclear whether they have distinct phenotypic properties that allow them to colonize diverse hosts or persist in the environment. By combining a novel anaerobic time-lapse microscopy method with automated image analysis, we discovered that Clade 5 strains grow faster than strains from other C. difficile clades and that they frequently form long chains. These chaining properties are driven by the environmentally responsive expression of a non-canonical signal transduction system, which our analyses suggest is selected against during murine infection. Collectively, our analyses reveal that Clade 5 strains have distinct growth properties that may promote their persistence in the environment.

microbiology↗

bioRxiv2023

Mycobacterium tuberculosis grows linearly at the single-cell level with larger variability than model organisms

The ability of bacterial pathogens to regulate growth is crucial to control homeostasis, virulence, and drug response. Yet, we do not understand the growth and cell cycle behaviors of Mycobacterium tuberculosis (Mtb), a slow-growing pathogen, at the single-cell level. Here, we use time-lapse imaging and mathematical modeling to characterize these fundamental properties of Mtb. Whereas most organisms grow exponentially at the single-cell level, we find that Mtb exhibits a unique linear growth mode. Mtb growth characteristics are highly variable from cell-to-cell, notably in their growth speeds, cell cycle timing, and cell sizes. Together, our study demonstrates that growth behavior of Mtb diverges from what we have learned from model bacteria. Instead, Mtb generates a heterogeneous population while growing slowly and linearly. Our study provides a new level of detail into how Mtb grows and creates heterogeneity, and motivates more studies of growth behaviors in bacterial pathogens.

microbiology↗

bioRxiv2022

Novel synergies and isolate specificities in the drug interactions landscape of Mycobacterium abscessus

Mycobacterium abscessus infections are difficult to treat and are often considered untreatable without tissue resection. Due to the intrinsic drug-resistant nature of the bacteria, combination therapy of three or more antibiotics is recommended. A major challenge in treating M. abscessus infections is the absence of a universal combination therapy with satisfying clinical success rates, leaving clinicians to treat infections using antibiotic lacking efficacy data. We systematically measured drug combinations in M. abscessus to establish a resource of drug interaction data and identify patterns of synergy to help design optimized combination therapies. We measured approximately 230 pairwise drug interactions among 22 antibiotics and identified 71 synergistic pairs, 54 antagonistic pairs, and four potentiator-antibiotics not previously reported. We found that commonly used drug combinations in the clinic, such as azithromycin and amikacin, are antagonistic in lab reference strain ATCC19977, whereas novel combinations, such as azithromycin and rifampicin, are synergistic. Another challenge in developing universally effective multidrug therapies for M. abscessus is the significant variation in drug response between isolates. We measured drug interactions in a focused set of 36 drug pairs across a small panel of clinical isolates with rough and smooth morphotypes. We observed highly strain-dependent drug interactions that cannot be predicted from single-drug susceptibility profiles or known drug mechanisms of action. Our study demonstrates the immense potential to identify synergistic drug combinations in the vast drug combination space and emphasizes the importance of strain-specific combination measurements for designing improved therapeutic interventions.

microbiology↗

bioRxiv2022

Critical role of growth medium for detecting drug interactions in Gram-negative bacteria that model in vivo responses

The rise in infections caused by multidrug-resistant (MDR) bacteria has necessitated a variety of clinical approaches, including the use of antibiotic combinations. Antibiotic susceptibility is affected in part by the growth state of bacteria within various tissues. Here we tested the hypothesis that drug-drug interactions vary in different media, and hence, using a medium that reflects tissue environments will better predict in vivo outcomes. We systematically studied pair-wise antibiotic interactions in three different media (CAMHB, a urine mimetic, and a lung mimetic) using three Gram-negative ESKAPE pathogens, Acinetobacter baumannii (Ab), Klebsiella pneumoniae (Kp), and Pseudomonas aeruginosa (Pa). There were pronounced differences in responses to antibiotic combinations between the three bacterial species grown in the same medium. However, within species, Pa responded to drug combinations similarly when grown in all three different media, whereas Ab responded similarly when grown in CAMHB and a lung mimetic medium. By contrast, drug interactions in Kp were poorly correlated across three different media. To assess whether distinct media were predictive of antibiotic interactions in Kp in the lungs of mice, we developed a treatment strategy and tested three antibiotic combination pairs. Measurements obtained in vitro from lung mimetic medium, but not rich medium, predicted in vivo outcomes. This work demonstrates that antibiotic interactions are highly variable when comparing across three gram-negative pathogens and highlights the importance of growth medium by showing a superior correlation between in vitro interactions in a growth medium that resembles the tissue environment and in vivo outcomes.

microbiology↗

bioRxiv2022

Tuberculosis treatment failure associated with evolution of antibiotic resilience

Antibiotics are a cornerstone of medicine, placing bacterial pathogens under intense pressure to evolve new survival mechanisms. Analysis of 51,229 Mycobacterium tuberculosis (Mtb) clinical isolates identified an essential transcriptional regulator, Rv1830 (here named resR) as a frequent target of positive (adaptive) selection. resR mutants do not demonstrate canonical drug resistance or drug tolerance but instead have significantly faster recovery after drug treatment across all antibiotics and combinations tested, a phenotype which we term antibiotic resilience. ResR acts in a regulatory cascade with other growth-controlling transcriptional regulators WhiB2 and WhiA, which are also under positive selection in Mtb clinical isolates. Mutations of these genes are associated with treatment failure and the acquisition of canonical drug resistance.

microbiology↗

bioRxiv2021

Design principles to assemble drug combinations for effective tuberculosis therapy using interpretable pairwise drug response measurements

A challenge in designing treatment regimens for tuberculosis is the necessity to use three or more antibiotics in combination. The combination space is too large to be comprehensively assayed; therefore, only a small number of possible combinations are tested. We narrowed the prohibitively large search space of combination drug responses by breaking down high-order combinations into units of drug pairs. Using pairwise drug potency and drug interaction metrics from in vitro experiments across multiple growth environments, we trained machine learning models to predict outcomes associated with higher-order combinations in the BALB/c relapsing mouse model, an important preclinical model for drug development. We systematically predicted treatment outcomes of >500 combinations among twelve antibiotics. Our classifiers performed well on test data and predicted many novel combinations to be improved over bedaquiline + pretomanid + linezolid, an effective regimen for multidrug-resistant tuberculosis that also shortens treatment in BALB/c mice compared to the standard of care. To understand the design features of effective drug combinations, we reformulated classifiers as simple rulesets to reveal guiding principles of constructing combination therapies for both preclinical and clinical outcomes. One example ruleset is to include a drug pair that is synergistic in dormancy and another pair that is potent in a cholesterol-rich growth environment. These rulesets are predictive, intuitive, and practical, thus enabling rational construction of effective drug combinations based on in vitro pairwise drug synergies and potencies. As more preclinical and clinical drug combination data become available, we expect to improve predictions and combination design rules.

microbiology↗

bioRxiv2021

C25-modified rifamycin derivatives with improved activity against Mycobacterium abscessus

Infections caused by Mycobacterium abscessus are difficult to treat due to its intrinsic resistance to most antibiotics. Formation of biofilms and the capacity of M. abscessus to survive inside host phagocytes further complicate eradication. Herein, we explored whether addition of a carbamate-linked group at the C25 position of rifamycin SV blocks enzymatic inactivation by ArrMab, an ADP-ribosyltransferase conferring resistance to rifampicin. Unlike rifampicin, 5j, a benzyl piperidine rifamycin derivative with a morpholino substituted C3 position, is not modified by purified ArrMab. Additionally, we show that the ArrMab D82 residue is essential for catalytic activity. Thermal profiling of ArrMab in the presence of 5j, rifampicin or rifabutin shows that 5j does not bind to ArrMab. We found that the activity of 5j is comparable to amikacin against M. abscessus planktonic cultures and pellicles. Critically, 5j also exerts potent antimicrobial activity against M. abscessus in human macrophages and shows synergistic activity with amikacin and azithromycin.

microbiology↗

bioRxiv2021

A competitive activity-based protein profiling platform yields cell wall synthesis inhibitors active against replicating and non-replicating Mycobacterium tuberculosis

The identification and validation of a small molecules targets is a major bottleneck in the discovery process for tuberculosis antibiotics. Activity-based protein profiling (ABPP) is an efficient tool for determining a small molecules targets within complex proteomes. However, how target inhibition relates to biological activity is often left unexplored. Here we studied the effects of 1,2,3-triazole ureas on Mycobacterium tuberculosis (Mtb). After screening ~200 compounds, we focused on two inhibitors active against both exponentially replicating and hypoxia-induced drug-tolerant Mtb that form part of a four-compound structure-activity series. The compound with negligible activity revealed potential false positive targets not addressed in other ABPP studies. Biochemistry, computational docking, and morphological analysis confirmed that active compounds preferentially inhibit serine hydrolases with cell wall and lipid metabolism functions and that disruption of the cell wall underlies biological activity. Our findings showed that ABPP identifies the targets most likely relevant to a compounds antibacterial activity.

microbiology↗

bioRxiv2021

Systematic measurement of combination drug landscapes to predict in vivo treatment outcomes for tuberculosis

A lengthy multidrug chemotherapy is required to achieve a durable cure in tuberculosis. Variation in Mycobacterium tuberculosis drug response is created by the differing microenvironments in lesions, which generate different bacterial drug susceptibilities. To better realize the potential of combination therapy to shorten treatment duration, multidrug therapy design should deliberately explore the vast combination space. We face a significant scaling challenge in making systematic drug combination measurements because it is not practical to use animal models for comprehensive drug combination studies, nor are there well-validated high-throughput in vitro models that predict animal outcomes. We hypothesized that we could both prioritize combination therapies and quantify the predictive power of various in vitro models for drug development using a dataset of drug combination dose responses measured in multiple in vitro models. We systematically measured M. tuberculosis response to all 2- and 3-drug combinations among ten antibiotics in eight conditions that reproduce lesion microenvironments. Applying machine learning to this comprehensive dataset, we developed classifiers predictive of multidrug treatment outcome in a mouse model of disease relapse. We trained classifiers on multiple mouse models and identified ensembles of in vitro models that best describe in vivo treatment outcomes. Furthermore, we found that combination synergies are less important for predicting outcome than metrics of potency. Here, we map a path forward to rationally prioritize combinations for animal and clinical studies using systematic drug combination measurements with validated in vitro models. Our pipeline is generalizable to other difficult-to-treat diseases requiring combination therapies. One Sentence SummarySignatures of in vitro potency and drug interaction measurements predict combination therapy outcomes in mouse models of tuberculosis.

microbiology↗