Digitally Barcoding Mycobacterium tuberculosis Reveals In Vivo Infection Dynamics in the Macaque Model of Tuberculosis

Abstract

Infection with Mycobacterium tuberculosis causes a spectrum of outcomes; the majority of individuals contain but do not eliminate the infection, while a small subset present with primary active tuberculosis (TB) disease. This variability in infection outcomes is recapitulated at the granuloma level within each host, such that some sites of infection can be fully cleared while others progress. Understanding the spectrum of TB outcomes requires new tools to deconstruct the mechanisms underlying differences in granuloma fate. Here, we use novel genome-encoded barcodes to uniquely tag individual M. tuberculosis bacilli, enabling us to quantitatively track the trajectory of each infecting bacterium in a macaque model of TB. We also introduce a robust bioinformatics pipeline capable of identifying and counting barcode sequences within complex mixtures and at various read depths. By coupling this tagging strategy with serial positron emission tomography coregistered with computed tomography (PET/CT) imaging of lung pathology in macaques, we define a lesional map of M. tuberculosis infection dynamics. We find that there is no significant infection bottleneck, but there are significant constraints on productive bacterial trafficking out of primary granulomas. Our findings validate our barcoding approach and demonstrate its utility in probing lesion-specific biology and dissemination. This novel technology has the potential to greatly enhance our understanding of local dynamics in tuberculosis.IMPORTANCE Classically, M. tuberculosis infection was thought to result in either latent infection or active disease. More recently, the field has recognized that there is a spectrum of M. tuberculosis infection clinical outcomes. Within a single host, this spectrum is recapitulated at the granuloma level, where there can simultaneously be lesional sterilization and poorly contained disease. To better understand the lesional biology of TB infection, we digitally barcoded M. tuberculosis to quantitatively track the fate of each infecting bacterium. By combining this technology with serial PET-CT imaging, we can dynamically track both bacterial populations and granuloma trajectories. We demonstrate that there is little constraint on the bacterial population at the time of infection. However, the granuloma imposes a strong bottleneck on dissemination, and the subset of granulomas at risk of dissemination can be distinguished by physical features.

Keywords: Mycobacterium tuberculosis; bacterial barcode; granuloma; infection mapping; lung infection; macaque.

FIG 1

Barcode structure and pipeline validation. (A) Barcode structure in the M. tuberculosis genome (top) and after preparation for sequencing (middle). Schema of the BARTI pipeline for barcode identification, thresholding, and counting. (B) Abundance of individual barcodes in a single library. (C) Number of barcodes counted using BARTI from a known number of barcoded Mycobacterium smegmatis colonies at a read depth of >100,000. (D) Ratios of barcodes counted using BARTI (actual) from known mixtures of barcode-containing plasmids mixed at indicated ratios (expected) and sequenced at a read depth of >100,000. The dashed line indicates linear regression; R² is the Spearman correlation.

FIG 2

Contained and disseminated lesions from 15- to 20-week-infected macaques. (A) Thoracic CT scan from animal 180-14, with barcode composition in lymph nodes (pie charts) and granulomas (bubbles). The inset shows 4 and 5 weeks postinfection (p.i.) (WPI) with granulomas identifiable at those time points. The white arrow marks dissemination of the purple barcode from the founder lesion at 4 weeks p.i. to new lesion at 5 weeks p.i. The red arrow marks dissemination of the olive barcode from founder lesion at 4 weeks p.i. to a new lesion at 5 weeks p.i. (B) Quantification across all animals of number of unique barcode sequences found in granulomas and lymph nodes. (C to F) Contained and disseminated barcode sequences in all recovered CFU⁺ lesions arrayed according to a spatial distribution approximately from top to bottom of each lung lobe for the 4 macaques. (The macaque ID number is given in the top left of each graph.) Blue shading reflects the percentage of molecular counters out of the total for a lesion for that barcode. Numbered lesions were identified prenecropsy by scan, while lettered lesions were found during necropsy and often do not have corresponding xyz coordinates. Gastric aspirates in panel E were taken 13 days apart. Barcodes are arbitrarily numbered. RLL, right lower lobe; LLL, left lower lobe; ACC, accessory lung lobe; EP, extrapulmonary; LN, lymph node; AIR, airway.

FIG 3

Spatial and temporal characteristics of dissemination. (A) Density histogram of the Euclidean distance of each disseminated lesion (red) to contained (blue) lesions with the same barcode sequence across all animals. The blue line is the Euclidean distance of all contained lesions to other lesions that do not contain the same barcode sequence. Data are only reflective of lesions for which xyz coordinates are known. The dashed line is the mean for each distribution, *; P < 0.05 by Welch’s t test. (B) Number of lung disseminating events across all animals at the indicated time points (weeks postinfection [WPI]). The timing reflects definitive dissemination events matched with barcode and serial imaging.

FIG 4

Granuloma size early during the early phase differentiates dissemination. (A) Bacterial burden (CFU) from granulomas obtained at necropsy (15 to 20 weeks p.i.) of contained (n = 41) and disseminated (n = 42) lesions from four barcoded macaques. Data are reflective of lesions initiating dissemination as determined by temporal PET/CT analysis. In instances where the precise founder is not known, all involved lesions were classified as disseminated. (B) Total bacterial load (live plus dead [CEQ]) of contained (n = 20) and disseminating (n = 28) lesions. (C) Bacterial killing determined by the CFU/CEQ ratio for contained (n = 20) and disseminating (n = 28) lesions. (D) Granuloma FDG avidity (SUVR) and (E) size (millimeter) of contained (n = 36) and disseminating (n = 30) lesions at 4 to 5 weeks postinfection as assessed by PET/CT (*, P = 0.0154). In panels A to E, each symbol represents a granuloma. Statistics for panels A to E were determined by the Mann-Whitney test.