A Novel Tool to Identify Bactericidal Compounds against Vulnerable Targets in Drug-Tolerant M. tuberculosis found in Caseum

ABSTRACT

Caseous necrosis is a hallmark of tuberculosis (TB) pathology and creates a niche for drug-tolerant persisters within the host. Cavitary TB and high bacterial burden in caseum require longer treatment duration. An in vitro model that recapitulates the major features of Mycobacterium tuberculosis (Mtb) in caseum would accelerate the identification of compounds with treatment-shortening potential. We have developed a caseum surrogate model consisting of lysed and denatured foamy macrophages. Upon inoculation of Mtb from replicating cultures, the pathogen adapts to the lipid-rich matrix and gradually adopts a nonreplicating state. We determined that the lipid composition of ex vivo caseum and the surrogate matrix are similar. We also observed that Mtb in caseum surrogate accumulates intracellular lipophilic inclusions (ILI), a distinctive characteristic of quiescent and drug-tolerant Mtb. Expression profiling of a representative gene subset revealed common signatures between the models. Comparison of Mtb drug susceptibility in caseum and caseum surrogate revealed that both populations are similarly tolerant to a panel of TB drugs. By screening drug candidates in the surrogate model, we determined that the bedaquiline analogs TBAJ876 and TBAJ587, currently in clinical development, exhibit superior bactericidal against caseum-resident Mtb, both alone and as substitutions for bedaquiline in the bedaquiline-pretomanid-linezolid regimen approved for the treatment of multidrug-resistant TB. In summary, we have developed a physiologically relevant nonreplicating persistence model that reflects the distinct metabolic and drug-tolerant state of Mtb in caseum.

INTRODUCTION

In pulmonary tuberculosis (TB), the necrotic core of closed granulomas and cavities can harbor high burdens of Mycobacterium tuberculosis (Mtb) bacilli and is very difficult to sterilize (1–5). These necrotic foci are made of caseum, a largely acellular matrix that essentially results from necrosis of infected foamy macrophages (FM) starting from the center of the lesion (6, 7), or from a bronchiolar obstruction progressing into caseating cavitary disease (8). Preclinical and clinical studies have shown that these niches constitute reservoirs of extracellular Mtb that can persist throughout antibiotic treatment (5, 9–12) and are prone to the emergence of drug resistance (13–15). Thus, drug regimens that efficiently eradicate this bacterial subpopulation would have a positive impact on the global control of TB by improving cure rates, reducing treatment duration, and curbing the emergence of drug resistance.

In a previous study, we demonstrated that caseum-resident Mtb is profoundly tolerant to first- and second-line TB drugs by using caseum explanted from rabbit cavities (16). In this native environment, Mtb adopts a nonreplicating state and accumulates intrabacterial lipid inclusions (ILIs). ILIs are neutral lipid storage vesicles that are strongly associated with dormancy and phenotypic drug resistance in mycobacteria (17–21). Decreased susceptibility is compounded by suboptimal drug distribution in the nonvascularized caseous core of necrotic granulomas, creating pockets of subinhibitory drug concentrations and de facto monotherapy (22–25). The field of TB drug discovery has spent significant effort developing single and multistress nonreplicating Mtb culture models that mimic the diverse physiological states that the pathogen adopts during infection. While these models reveal that nonreplicating Mtb is tolerant to many antibiotics (18, 26–28), patterns of drug tolerance tend to be different in each assay, and a consensus is lacking as to which conditions are most predictive of preclinical and clinical efficacy (29, 30). A model that faithfully mimics the specific physiology and drug tolerance of Mtb in native caseum would be adequate to identify both active compounds and vulnerable Mtb targets in this distinctive, mostly acellular lipid-rich matrix.

Experimental findings support the notion that caseum is made primarily of necrotized FM. The foamy phenotype results from intracellular lipid accumulation in response to Mtb infection, exceeding the host cell’s capacity to maintain lipid homeostasis. Host lipids are sequestered in lipid droplets (LDs), contributing to the characteristic foamy appearance. Infected FM directly surround the caseous core of TB lesions in patients and animal models (7, 31–33), becoming gradually necrotic as infection progresses, thus expanding the caseous center. Biochemical and molecular analyses revealed that FM and caseum have comparable protein and lipid compositions (7). Mtb infection of macrophages in vitro produces the foam cell phenotype via tumor necrosis factor receptor signaling and downstream activation of caspase pathways and the mammalian target rapamycin complex 1. In addition, reduced oxygen tension, associated with poor vascularization, and exposure to very-low-density lipoproteins and selected fatty acids trigger FM development (34–36).

Here, we attempted to deliver a predictive assay that replicates the phenotypic drug resistance of Mtb found in caseum without relying on animal infection models, starting from a caseum surrogate previously developed to estimate nonspecific drug binding in caseum (37, 38). Properties of the matrix and of the inoculated pathogens were optimized to reproduce major lipid contents of caseum, Mtb’s nonreplicating status, accumulation of ILIs, regulation of specific dormancy and lipid metabolic pathways, and phenotypic resistance to major TB drugs, as seen in native caseum Mtb. To establish the utility and predictive value of the model, we characterized the bactericidal activities of novel and potent new bedaquiline analogs currently in clinical development in the ex vivo and surrogate caseum models. We propose that the caseum surrogate model is adequate to screen drug discovery compounds, identify vulnerable drug targets in dormant caseum Mtb, and inform the development of shorter and more effective treatment regimens.

RESULTS

To reproduce the drug-tolerant phenotype of caseum-resident Mtb (16) and develop an appropriate drug screening tool, we designed a surrogate assay consisting of THP-1-derived FM lysate inoculated with Mtb until the pathogen acquires physiological, genetic, and phenotypic properties that recapitulate those of caseum Mtb.

Reproducing the fatty acid profile, cholesterol and glyceride abundance, and pH of native caseum.

To reproduce the lipid-rich environment of caseum, we first optimized lipid droplet formation and accumulation in macrophages. THP-1-derived macrophages (THPMs) were exposed to oleic acid (OA, 18:1), stearic acid (SA, 18:0), and irradiated Mtb (iMtb) for 24 h (Fig. 1), followed by staining for neutral lipid deposits and flow cytometry analysis of mean fluorescence intensities (MFI). OA-treated THPMs produced the greatest increase in lipid staining compared to nonfoamy cells (Fig. 2A; also, see Fig. S1C in the supplemental material), while both SA and iMtb treatments resulted in comparable staining. These quantitative results were confirmed visually by microscopic analyses, with the most abundant lipid droplets found within the cytosol of OA-induced macrophages (Fig. 2B). Using fluorescence microscopy, we observed that the lipophilic dye localizes to cytosolic spherical organelles surrounding the nucleus (Fig. 2B).

FIG 1.

Illustration of the overall experimental design. THP-1 derived FM are used to generate a caseum surrogate matrix, which is spiked with M. tuberculosis HN878 to reproduce the typical physiologic, genetic, and phenotypic properties of caseum-resident Mtb. Image created with BioRender.

FIG 2.

Quantitative and qualitative analyses of induced lipid droplets and neutral lipids in THP-1 macrophages and macrophage lysates. (A) Lipid droplet abundance in THP-1 macrophages after exposure to various foamy-cell inducers. THP-1 macrophages were treated with irradiated Mtb (iMtb; MOI, 1:50), stearic acid (SA; 100 μM), or oleic acid (OA; 100 μM) for 24 h. Lipid bodies were stained with BODIPY 493/503 prior to analysis by flow cytometry. Data are expressed as the geometric mean fluorescence intensity (MFI) for each group. Error bars represent standard deviations for three biological replicates. *, P < 0.05. (B) Bright-field and fluorescence microscopy images of iMtb-, SA-, and OA-induced THP-1 macrophages. Lipid droplets and nuclei were stained with BODIPY and DAPI, respectively. (C) Fatty acid quantification in rabbit caseum and surrogate caseum generated using three different inducers. The six most abundant fatty acids are shown.

FIG S1

Gating strategy of BODIPY 493/503-based flow cytometry for detecting foamy macrophages. (A) Viable cells were gated using the forward (FSC-A) and side (SSC-A) scatter areas. (B) Single cells were gated using FSC-A and FSC height (FSC-H). (C) Histogram displaying the distribution of BODIPY signals in THP-1 macrophages relative to cells unstained with the dye. Download FIG S1, PDF file, 0.1 MB ^{(107.3KB, pdf)} .

This is a work of the U.S. Government and is not subject to copyright protection in the United States. Foreign copyrights may apply.

We then compared the glyceride, cholesterol, and fatty acid contents of the three caseum surrogates to those of native rabbit caseum. The combined concentration of mono-, di- and triglycerides was 3.5-, 7.8-, and 29-fold higher in iMtb-, SA-, and OA-induced caseum surrogates, respectively (Table 1). The combined free cholesterol and cholesterol ester concentrations were similar in all surrogates and in native caseum (Table 1). To generate quantitative information on specific free fatty acid (FFA) species present in the various surrogates, we measured the concentrations of 23 individual species by liquid chromatography coupled to mass spectrometry (LC-MS) (Table S1). The 6 most abundant species are shown in Fig. 2C. Qualitatively, the FFA content of ex vivo caseum was recapitulated in the caseum surrogates, but stimulation of THPM with OA resulted in a marked enrichment of this fatty acid in the surrogate matrix (Fig. 2C and Table S1). We also showed that the pHs of iMtb-, SA-, and OA-induced surrogates (pH 6.6 to 6.9) fall within the range of pHs measured in caseum excised from rabbit cavities, lesions from C3HeB/FeJ mice and guinea pigs, and resected clinical samples (6, 39) (Table 1). Overall, the major FFA species found in rabbit caseum are best reproduced in iMtb- and SA-induced surrogate caseum.

TABLE 1.

Lipid and pH profile of native rabbit caseum and caseum surrogates

Caseum or surrogate	Concn (mM) (mean ± SD) of:
	Total glyceridesa	Total cholesterolb	pH
Rabbit caseum	3.2 ± 0.2	9.2 ± 0.1	6.5–7.1
iMtb-induced surrogate	11.2 ± 1.9	6.6 ± 0.3	6.6–6.9
SA-induced surrogate	24.8 ± 7.0	6.8 ± 0.9	6.6
OA-induced surrogate	92.9 ± 9.1	6.5 ± 0.7	6.6

^aInclusive of tri-, di- and monoglycerides.

^bInclusive of free cholesterol and cholesterol esters.

TABLE S1

Quantification of 23 different free fatty acids (FA) in rabbit caseum and iMtb-, stearic acid-, and oleic acid-induced surrogate caseum preparations. Each FA species is denoted using standard nomenclature (i.e., number of carbon atoms:number of double bonds). The 6 most abundant species (blue text) are illustrated in Fig. 3. BLOD, below the limit of detection. Download Table S1, PDF file, 0.08 MB ^{(80.3KB, pdf)} .

This is a work of the U.S. Government and is not subject to copyright protection in the United States. Foreign copyrights may apply.

Mtb in caseum surrogate is viable and nonreplicating.

We first demonstrated that native Mtb in ex vivo rabbit cavity caseum exists in a nonreplicating state and maintains a stable bacterial burden for up to 4 weeks (Fig. 3A), in agreement with previous 8-day growth kinetic data (16). We then observed that replicating Mtb adopts the nonreplicating state when inoculated into ex vivo caseum and that bacterial burden remained stable for up to 6 weeks (Fig. 3B). To identify the caseum surrogate matrix that best supports bacterial survival and induces nonreplication of Mtb introduced exogenously, logarithmically growing Mtb was inoculated into the three caseum surrogates (Fig. 1), generating stable CFU kinetics in iMtb- and SA-induced preparations (Fig. 3C and D), which suggests that Mtb arrests replication without compromising bacterial viability. In the OA-induced surrogate, a rapid and transient reduction in recoverable bacterial burden (1 to 2 log, statistically significant) was observed within a few days postinoculation. The CFU counts were restored to starting levels by day 14 postinoculation (Fig. 3E). Given the high OA content of this matrix (~6 mg/mL) (Fig. 2D and Table S1), we tested whether this long-chain FFA may have contributed to the transient loss of culturability. We observed >90% growth inhibition at 0.5 mg/mL OA against Mtb in liquid medium (Fig. S2A). This may have contributed to the transient inability of Mtb to produce recoverable colonies on solid medium. Incubating, or “resting,” the SA- and iMtb-induced surrogates at 37°C for 3 days after the denaturation step prior to Mtb inoculation greatly improved Mtb recovery in this model (Fig. S2B to D). To test the hypothesis that fresh macrophage lysates may contain microbicidal factors such as reactive oxygen and nitrogen species, we measured the lipid peroxidation product malondialdehyde as a thiobarbituric acid-reactive substance (TBARS) in each matrix. We found significantly more TBARS in SA and iMtb surrogates postrest, indicating that reactive species are dissipated via the oxidative deterioration of lipids within these matrices during the 3-day resting period (Fig. S2E).

FIG 3.

Growth kinetics of Mtb in caseum and surrogates. (A) Native Mtb in rabbit caseum remains in the nonreplicating state postexplantation. (B) Replicating Mtb inoculated into low-burden ex vivo caseum quickly adopts a viable and nonreplicating state. (C to E) Replicating Mtb was inoculated into caseum surrogates generated using (C) iMtb, (D) SA, and (E) OA as foamy-cell inducers. Error bars represent standard deviations for three technical replicates. The Mtb CFU drop observed between day 1 and day 4 or 7 in response to exposure to OA-induced caseum surrogate was statistically significant (two-tailed t test; ***, P ≤ 0.005). (F) Genome sequencing coverage patterns of native Mtb in rabbit caseum and Mtb inoculated in caseum surrogates over time. Normalized genome coverage of native and inoculated Mtb are shown for a single sample in each panel. The red line is a LOESS (locally estimated scatterplot smoothing) filter smoothing curve (span = 10) with standard error shown in gray (confidence level = 0.95). The replication start and end sites are indicated by dashed blue lines at positions 4386870 and 2077870, respectively.

FIG S2

(A) Growth inhibitory activities of oleic acid (hollow circles) and triolein (filled circles) against replicating Mtb culture. Concentrations are presented in milligrams per milliliter on a log scale. (B) Replicating Mtb inoculated into caseum surrogates generated using oleic acid, stearic acid, and iMtb as lipid droplet inducers loses viability to various degrees (hollow circles). Incubating the surrogates for 3 days at 37°C prior to inoculation partially restores bacterial viability in the OA-induced surrogate and fully restores Mtb viability in SA- and iMtb-induced matrices (filled circles). (C) A three-day resting period appears to dissipate reactive oxygen species in SA- and iMtb-induced surrogates, measured as a TBARS to detect the lipid oxidation product malondialdehyde. ***, P < 0.005. Download FIG S2, TIF file, 0.7 MB ^{(750.3KB, tif)} .

This is a work of the U.S. Government and is not subject to copyright protection in the United States. Foreign copyrights may apply.

To determine whether the apparently stable CFU kinetic profile was the result of true nonreplication or rather a balance between cell growth and death, we analyzed the pattern of metagenomic sequencing read coverage to assess genome replication (40), where nonuniform genome coverage with an overrepresentation of sequences at the origin of replication is indicative of active genome replication. Upon introduction of Mtb from replicating cultures into native caseum or caseum surrogates, we observed increased uniformity of the microbe’s genome coverage over time, indicative of gradual progression to the nonreplicating state in all matrices (Fig. 3F). Thus, both SA- and iMtb-induced caseum surrogates trigger transition to true nonreplication in Mtb.

Residence in caseum surrogate induces physiologic signatures of dormancy in Mtb.

To determine whether Mtb accumulates ILIs in the surrogate caseum environment as it does in native caseum, we used dual Nile red-auramine O staining and confocal microscopy (16). In ex vivo caseum and in SA- or iMtb-induced surrogate caseum, a large proportion of Mtb bacilli accumulated neutral lipids and stained red (Fig. 4A). In contrast, very few bacilli appeared Nile red positive in replicating Mtb cultures. Magnification of individual Nile red-positive bacilli revealed the formation of distinct ILIs. Quantification of Nile red- and auramine O-stained bacilli confirmed a dramatic increase in neutral lipid accumulation in caseum and the surrogates compared to replicating Mtb cultures (Fig. 4B). In summary, neutral lipid accumulation is a defining feature of nonreplicating Mtb in caseum and caseum surrogates. Next, we compared the expression of 20 genes (18 selected genes, plus sigA and 16S rRNA as references) with diverse functions in native caseum Mtb and Mtb preadapted in caseum surrogate (Table 2). These genes are involved in cell wall biosynthesis, lipid metabolism, respiration, virulence and detoxification, transmembrane transport, and iron storage and were selected for their putative roles in nonreplicating persistence models. Relative gene expression in caseum surrogate was quantified by quantitative reverse transcription-PCR (qRT-PCR), normalized to sigA expression, and calibrated against expression levels in ex vivo caseum. Relative gene expression between caseum and the surrogate was not significantly different (Student’s t test; P > 0.05) for 15 of the 18 genes studied (Fig. 4C). Additionally, when average normalized quantification cycle (C_q) values between caseum and surrogate samples were compared for 19 genes (16S excluded), there was a significant correlation (Pearson’s correlation; R = 0.74; P < 0.005) (Fig. S3). Altogether, these data indicate that the surrogate model recapitulates selected features of the physiologic and metabolic state of Mtb in ex vivo caseum.

FIG 4.

Intrabacterial neutral lipid accumulation in native caseum-resident Mtb, in bacilli preadapted in caseum surrogate, and in replicating (control) Mtb culture. All smears were stained with auramine O (green, acid-fast stain) and Nile red (red, neutral lipid stain) and examined by confocal laser scanning microscopy at the same laser intensity for all samples with Z-stacking to get the depth of the scan field. Bar, 5 μm. (B) Venn diagrams showing the distribution of auramine-positive and Nile red-positive bacilli in caseum, caseum surrogate, and replicating Mtb culture. (C) Plot showing Mtb gene expression across 20 loci of interest. qRT-PCR was performed on RNA extracted from Mtb-infected rabbit caseum or Mtb adapted for 6 weeks in iMtb-induced caseum surrogate. Rounded points show individual C_q values normalized to sigA, triangles show mean normalized C_q values, and error bars show standard deviations. Groups of normalized samples were compared using an unpaired Student’s t test; the P values were adjusted using the Holm-Bonferroni method. ns, not significant (P > 0.05); *, P ≤ 0.05; **, P ≤ 0.01.

TABLE 2.

Nonreplicating Mtb gene expression in freshly excised rabbit caseum and iMtb-induced caseum surrogate

Category and gene identifier	Gene name	Mean normalized expression value in:
		Caseum	Caseum surrogate
Cell wall biosynthesis
Rv3825c	pks2	1.43	1.24
Rv2245	kasA	1.16	1.04
Lipid metabolism
Rv1169c	lipX	1.00	1.03
Rv3130c	tgs1	0.98	1.02
Respiration
Rv0467	icl1	1.33	1.14
Rv2194	qcrC	1.12	1.09
Rv1854c	ndh	1.01	1.03
Rv3145	nuoA	1.09	1.04
Rv2583c	relA	1.09	1.10
Rv1131	gltA1/prpC	1.12	1.05
Rv3341	metA	1.10	1.02
Rv1130	prpD	1.20	1.05
Transcription
Rv2710	sigB	1.14	1.19
Rv3260c	whiB2	1.07	1.06
Virulence, detoxification, adaptation
Rv2031c	hspX	1.05	0.98
Rv3499c	mce4A	1.24	1.09
Transporters
Rv1737c	narK2	1.09	1.11
Iron storage
Rv1876	bfrA	1.14	1.13

FIG S3

Average normalized ΔΔC_q values for 20 target genes in the iMtb-induced surrogate (x axis) compared to those from rabbit caseum (y axis). The line shows the predicted linear model, and the strength of correlation was calculated using the Pearson correlation method (R and P values displayed). Download FIG S3, TIF file, 0.3 MB ^{(324.4KB, tif)} .

This is a work of the U.S. Government and is not subject to copyright protection in the United States. Foreign copyrights may apply.

Mtb gradually develops phenotypic resistance to major TB drugs in caseum surrogate.

Exponentially growing Mtb was inoculated into caseum surrogate and preadapted for 4, 6, and 8 weeks prior to incubation with a panel of TB drugs (Fig. 1). The assay was designed to generate the 90% minimal bactericidal concentration in caseum (casMBC₉₀) or in surrogate (srgMBC₉₀) to infer potency and Δlog CFU_max or maximum killing effect (E_max) to infer sterilizing potential at the highest concentration (Table 3). Isoniazid and delamanid were excluded from the study because of significant instability in both matrices (Fig. S4A). We found that drug tolerance in caseum was best reproduced after 6 weeks of Mtb preadaptation in surrogate, as measured by MBC₉₀ (Table S2). After 6 weeks, the srgMBC₉₀/casMBC₉₀ ratio for all drugs tested clustered most closely around 1 (Fig. S4B). Given bedaquiline’s slow onset of action (41, 42), its bactericidal activity was evaluated after 7 (standard) and 14 days of incubation. Bedaquiline’s casMBC₉₀ was 14-fold lower in the 14-day assay than in the 7-day assay (Fig. S4C). A robust correlation between srgMBC₉₀ and casMBC₉₀ was obtained (R² = 0.78) (Fig. 5A). Similarly, we obtained a good correlation between Δlog CFU_max in the ex vivo and in vitro assays (R² = 0.82) (Fig. 5B), indicating comparable phenotypic drug resistance in both matrices. Rifamycins were among the most potent drugs in caseum and exhibited the highest sterilizing activity (Fig. 5B). We also compared the magnitude of the drug tolerance induced by caseum versus the widely used nutrient and oxygen starvation conditions (Fig. 5C) (26, 27, 43, 44). The intracaseum potencies of rifampicin, rifapentine, kanamycin, and clofazimine are best reproduced during nutrient starvation, while bedaquiline’s is recapitulated under hypoxia. The fluoroquinolones display a distinct potency profile, performing better in caseum and caseum surrogate (MBC₉₀s ≤ 10 μM) than in the other nonreplicating (NRP) models, with the exception of moxifloxacin being equally active under hypoxic conditions. The casMBC₉₀s and srgMBC₉₀s of rifampicin and moxifloxacin are achieved in caseum in patients receiving clinically approved doses, whereas linezolid and clofazimine are inactive within their respective intracaseum exposure windows (Fig. S5) (25).

TABLE 3.

Bactericidal activity of TB drugs against replicating Mtb cultures, native bacteria in caseum, and Mtb preadapted in iMtb-induced surrogate caseum for 6 weeks

Drug	Replicating Mtb MBC90 (μM)a	Caseum		Caseum surrogate
		MBC90 (μM)	ΔLog CFUmax	MBC90 (μM)	ΔLog CFUmax
Pyrazinamide	>80	>512	0.5	>512	0.1
Ethambutol	2.5–5	>512	0.2	90	1.1
Rifampicin	0.07	10	3.5	2.5	4.6
Rifapentine	0.08	10	3.1	2.2	4.4
Rifabutin	0.04	1	4.0	1.5	4.3
Moxifloxacin	0.3–0.6	1.6	1.8	1.5	1.8
Levofloxacin	0.3–2.5	9	1.8	9	1.3
Gatifloxacin	0.6	1.5	1.9	3.5	1.4
Linezolid	10	>512	0.2	>512	0.5
Sutezolid	1.4–5.7	16	1.4	8	1.6
Pretomanid	0.6	70	1.6	63	2.0
SQ109	0.8–2.3	260	1.8	200	1.8
Bedaquilineb	3.6–10	4.7	1.9	5.4	1.6
Kanamycin	5	>512	0.0	512	1.0
Clofazimine	40	>128	0.7	>128	0.1

^aMBC₉₀s in replicating bacteria were measured after 5 days of incubation (66, 80).

^bBedaquiline was tested in 14-day assays in caseum and surrogate instead of the standard 7-day incubation period.

FIG 5.

Comparative bactericidal activity of TB drugs against Mtb in caseum and in iMtb-induced caseum surrogate. (A) Correlation between bactericidal activity in caseum (casMBC₉₀) and surrogate caseum (srgMBC₉₀). Drug abbreviations are provided in panel C. (B) Correlation between maximal CFU reduction at the highest assay concentration (Δlog CFU_512μM) in caseum (cas) and surrogate caseum (srg). (C) Heat map of potencies of first- and second-line drugs and compounds in clinical development against Mtb in replicating cultures, in two in vitro models of nonreplicating persistence (66), and in native and surrogate caseum. Efficacies in the C3HeB/FeJ mouse model of necrotic lesions (12, 24, 67–69) and in phase IIa early bactericidal activity (EBA) trials (70–79) are shown as in vivo references. Image created with BioRender.

TABLE S2

Influence of duration of preadaptation on Mtb drug tolerance in caseum surrogate. Mtb was inoculated in caseum surrogate and allowed a 4-, 6-, or 8-week adaptation period prior to MBC assays. Download Table S2, PDF file, 0.08 MB ^{(80.3KB, pdf)} .

This is a work of the U.S. Government and is not subject to copyright protection in the United States. Foreign copyrights may apply.

FIG S4

(A) Instability of isoniazid (INH) and delamanid (DLM) in caseum homogenate over time. Data are expressed as the percentage of drug remaining at specific time points. (B) Relationship between the duration of Mtb preadaptation period in caseum surrogate and drug susceptibility. The latter is expressed as a ratio between antibiotic activity in the surrogate matrix (srgMBC₉₀) and caseum (casMBC₉₀). (C) Bactericidal activity of bedaquiline in 7- and 14-day caseum MBC assays. Dots and error bars represent means and standard deviations for three technical replicates. Dotted lines indicate the cutoffs for 1-log reductions in bacterial burdens. Download FIG S4, TIF file, 0.5 MB ^{(470.9KB, tif)} .

This is a work of the U.S. Government and is not subject to copyright protection in the United States. Foreign copyrights may apply.

FIG S5

(A) casMBC and (B) srgMBC plots of rifampicin (RIF), moxifloxacin (MXF), linezolid (LNZ), and clofazimine (CFZ) with overlaid drug concentration windows in human caseum from the work of Strydom et al. (25). Dots and error bars represent means and standard deviations, respectively, for three technical replicates. Dotted lines indicate the cutoffs for 1-log reductions in bacterial burdens. Download FIG S5, TIF file, 1.1 MB ^{(1.1MB, tif)} .

This is a work of the U.S. Government and is not subject to copyright protection in the United States. Foreign copyrights may apply.

Screening of development compounds identifies TBAJ876 and TBAJ587 as potent drug candidates in caseum.

To validate the surrogate assay as a new screening platform, we measured the MBCs of a series of discovery compounds in caseum surrogate and identified the bedaquiline analogs TBAJ876 (srgMBC₉₀ = 0.3 μM) and TBAJ587 (srgMBC₉₀ = 3.2 μM) as potent agents against nonreplicating Mtb persisters in 7-day assays (Fig. 6A). We then repeated the measurements in ex vivo caseum and found similar MBC₉₀s and comparable E_max values (Fig. 6B). We replaced bedaquiline with either TBAJ876 or TBAJ587 in the bedaquiline-pretomanid-linezolid (BPaL) regimen, which was highly successful against drug-resistant TB in the pivotal Nix-TB trial (45). To directly compare the contribution of the TBAJ analogs and bedaquiline to the killing of caseum Mtb in the Nix-TB background, the concentrations of pretomanid (Pa) and linezolid (L) were fixed at their estimated average concentrations achieved in caseum over the 24-h dosing interval (C_ave[0–24]), and the concentration of bedaquiline, the TBAJ compounds, and their active metabolites were increased in 4-fold increments, centered on the C_ave[0–24]. We measured a combined 1-log killing at TBAJ concentrations that were approximately 16-fold lower than that of bedaquiline. The E_max of the TBAJ-based combinations was greater than that of Nix-TB at TBAJ concentrations 4-fold lower than that of bedaquiline (Fig. 6C). These results show that the improved potency of TBAJ analogs against replicating Mtb (46, 47) translates into increased bactericidal activity against caseum persisters compared to bedaquiline and may guide dose selection in future clinical trials. Thus, the surrogate model enables screening for compounds that are active against drug-tolerant caseum Mtb and may have treatment-shortening potential.

FIG 6.

Bactericidal activity of bedaquiline (BDQ) analogs against Mtb persisters. TBAJ876 (blue), TBAJ587 (purple), and bedaquiline (black) were tested in (A) caseum surrogate and (B) rabbit caseum MBC assays for 14 days. The MBC₉₀ of each compound is shown. (C) Bactericidal activity of bedaquiline and the TBAJ analogs in combination with pretomanid (Pa) and linezolid (L) in rabbit caseum. The des-methyl active metabolites of BDQ (M2) and TBAJ compounds (M3) are included at concentrations that reproduce their proportion relative to the parent drug in caseum. Concentration ranges used for each drug are shown in the table (T876, TBAJ876; T587, TBAJ587). The dotted line indicates a 1-log reduction in bacterial burden compared to day 1. Data points and error bars represent the means and standard deviations for three technical replicates each.

DISCUSSION

Phenotypic screens that mimic host-relevant environmental conditions to accurately reflect the multiple physiological states of Mtb during infection are invaluable tools for understanding Mtb’s dormancy adaptations and discovering drugs that kill these populations. Animal models of TB infection reveal that caseous necrosis is associated with high bacterial burdens and limited drug-mediated sterilization (39, 48, 49). Positron emission tomography-computed tomography (PET-CT) scans from TB patients illustrate the bronchial spread of infection from caseous cavities through the lungs and show that cavitary lesion dynamics is a major determinant of treatment outcome (5). Hence, predictive screening platforms that model the profound drug tolerance of caseum-resident Mtb are needed to identify promising therapeutic options. We have successfully optimized a caseum model which reproduces the extreme phenotypic drug resistance levels of caseum-resident Mtb for a range of TB drugs with diverse targets, and in which the starting inoculum is fully controlled. In contrast, caseum specimens excised from rabbits typically contain 10⁵ to 10⁹ Mtb bacilli per g but include occasional very-low-burden cavities (16, 39).

Given the critical link between Mtb adaptations to lipid-rich environments, intrabacterial lipid accumulation, nonreplicating persistence, and drug tolerance (19, 31, 50), we focused on producing a surrogate with a caseum-like lipid profile by using biochemical and biological triggers known to induce lipid body accumulation in macrophages. Based on triacylglycerol (TAG), FFA, and cholesterol quantitation, we determined that while OA most effectively stimulates lipid body accumulation, the OA-induced surrogate least resembled the biochemical composition of ex vivo rabbit caseum due to overrepresentation of TAG and OA. It is also inherently bactericidal to Mtb at high concentrations, like other long-chain FFAs (51, 52), causing a dip in CFU upon Mtb inoculation. SA, on the other hand, is one of the most abundant FFAs in FM and caseum (7, 19), does not affect the growth or survival of Mtb, and is internalized and incorporated unchanged into phthiocerol dimycocerosate (PDIM) (53). SA- and iMtb-induced surrogate matrices adequately reproduce the host lipid profile of ex vivo caseum and halt Mtb replication while preserving viability, key features of nonreplicating culture models of Mtb dormancy. Both iMtb and SA can be used as lipid droplet inducers, providing the option to avoid confounding bacterial proteins, genetic material, or lipids present in iMtb, which could interfere with -omics analyses.

In mycobacteria, intrabacterial TAG accumulation is associated with low metabolic activity, growth arrest, loss of acid fastness, and phenotypic drug resistance (54). These ILI TAG stores serve as long-term energy sources during periods of stress and enable mycobacterial persistence (55). Accordingly, we observed ILI accumulation in Mtb in the ex vivo and surrogate models and corresponding expression of lipid metabolism genes. The gradual increase in Mtb drug tolerance up to 8 weeks postinoculation in caseum surrogate is the result of continuous adaptations of the pathogen in response to changes in its environment. The transition to a lipid-rich diet triggers rapid metabolic and physiologic adaptations, such as accumulation of ILIs. We hypothesize that this is followed by a slower process of matrix/pathogen coevolution, whereby gradual nutrient and oxygen depletion (the plates are sealed and incubated without agitation) as well as spontaneous hydrolysis of glycerides, oxidation of fatty acids and other lipids, and the decay of macromolecules collectively trigger a slow but profound dormancy program inducing phenotypic drug resistance. Drug tolerance reaches a “Goldilocks” state around 6 weeks postinoculation and exceeds caseum-induced drug tolerance for a subset of drugs after 8 weeks of preincubation in the surrogate (Fig. S4B). Our model provides a unique tool to investigate the matrix/pathogen coevolution using biochemical and transcriptome sequencing (RNA-Seq) methods and identify the determinants of the excessive drug tolerance observed at 8 weeks.

Overall, the caseum surrogate model better recapitulates the drug tolerance profile of Mtb in ex vivo caseum than other commonly employed NRP models and provides one answer to the long-standing question of identifying a clinically relevant assay of nonreplicating persistence (56). Rifamycins and fluoroquinolones are most potent, and the caseum model accurately ranks potencies within drug classes (rifabutin > rifapentine ≈ rifampicin; moxifloxacin > gatifloxacin > levofloxacin, as observed in native caseum), making it an excellent screening tool for lead optimization in TB drug discovery. As measured by Δlog CFU_max or E_max, rifamycins are the most effective sterilizers in native and surrogate caseum, highlighting the potential of novel RpoB inhibitors against dormant Mtb. The robust correlation between Δlog CFU_max in caseum and the surrogate further validates the model, which was recently exploited to confirm CinA as a major determinant of drug tolerance in Mtb (57). Comparing the bactericidal activity of the study drugs in native and model caseum to clinical concentrations achieved in the cavity caseum of TB patients, measured for a subset of 6 drugs (15, 25), revealed that rifampicin and moxifloxacin achieve bactericidal concentrations in caseum, consistent with recent promising results of the first ever 4-month regimen containing rifapentine and moxifloxacin in the treatment of drug-susceptible TB (58). TBAJ876 and TBAJ587 were identified as potent bactericidal drug candidates in the caseum surrogate model and improved the bactericidal activity of the recently approved bedaquiline-pretomanid-linezolid regimen for multidrug-resistant TB against caseum Mtb (45). Upcoming clinical trials will confirm whether this promising result translates into shortening of treatment for multidrug-resistant TB.

Potential limitations of this study are the use of rabbit caseum to calibrate the model, which may not fully recapitulate the properties and drug tolerance of clinical caseum Mtb, and the use of THP-1 cells, which, while easy to handle and robust, are not completely representative of primary macrophages from which foamy macrophages emerge in human lesions. On the other hand, the choice of THP-1 cells will facilitate implementation of the assay by other investigators, the lab-to-lab reproducibility of which will be the focus of follow-up studies.

The assay is amenable to medium-throughput screening to identify compounds active against nonreplicating persistent Mtb in cavity caseum, is cost-effective, and circumvents the need to infect animals. In addition to measuring bactericidal activity, the tool can be adapted to identify vulnerable bacterial targets and study chemical genetic interactions in caseum Mtb, using high-density insertional mutagenesis or CRISPRi genetic knockdown libraries (59, 60). In combination with DiaMOND, an emerging platform to predict high-order drug interactions (61) and in vivo treatment outcomes (30), the model can also be leveraged to identify translatable synergistic drug combinations that sterilize cavity caseum and may shorten TB chemotherapy duration.

MATERIALS AND METHODS

Chemicals and cell lines.

Moxifloxacin and linezolid were purchased from Sequoia Research Products (Berkshire, UK) while rifapentine, rifabutin, pyrazinamide, sutezolid, and clofazimine were purchased from Sigma-Aldrich (St. Louis, MO). Rifampicin and levofloxacin were purchased from Gold Biotechnology (St. Louis, MO) and ChemImpex Intl Inc. (Wood Dale, IL), respectively. Kanamycin sulfate, gatifloxacin, and ethambutol dihydrochloride were purchased from Fisher Scientific (Hampton, NH). Pretomanid was obtained from BioDuro (San Diego, CA). Bedaquiline was provided by Janssen Research and Development (Raritan, NJ). TBAJ876 and TBAJ587 were provided by the TB Alliance (Pretoria, South Africa). SQ109 was provided by Sequella, Inc. (Rockville, MD). Oleic acid (OA) and stearic acid (SA) were obtained from MP Biomedicals (Irvine, CA) and Alfa Aesar (Haverhill, MA), respectively. Mycobacterium tuberculosis HN878 and gamma-irradiated whole HN878 cells were obtained from BEI Resources (Manassas, VA). The human monocytic leukemia cell line THP-1 was obtained from the American Type Culture Collection (ATCC) (Manassas, VA).

Rabbit infection model and caseum collection.

New Zealand White (NZW) rabbits were used for aerosol infection by M. tuberculosis HN878, as previously described (62). Briefly, rabbits were exposed to Mtb-containing aerosol using a nose-only delivery system. The infection was allowed to progress for 12 to 40 weeks prior to necropsy and the collection of caseum from both closed and cavitary lesions. Animal studies were carried out with approval from the Institutional Animal Care and Use Committees of the National Institute of Allergy and Infection Disease, NIH, MD (protocol number LCIM-3), or Hackensack Meridian Health, NJ (protocol number 270).

Generation of foamy macrophages and surrogate caseum.

The THP-1 monocytic cell line was cultured in RPMI 1640 supplemented with 10% fetal bovine serum (FBS) and 2 mM l-glutamine. Cells were seeded on 150-mm dishes (1 × 10⁶ cells/mL) and differentiated with 100 nM phorbol 12-myristate 13-acetate (PMA). Intracellular lipid accumulation was stimulated overnight using OA, SA, or iMtb. OA was added at 0.4 mM as described previously (37). SA was dissolved in ethanol and added to the culture at 0.1 mM. iMtb was used to treat THP-1 macrophages at a multiplicity of infection (MOI) of 1:50. Incubation concentrations of OA and SA above 400 μM and 100 μM, respectively, and treatment with iMtb at MOI greater than 1:50 compromised monolayer adherence and subsequent surrogate caseum recovery. After 24 h, lipid body production was confirmed using a light microscope, and foamy macrophages were detached with 5 mM EDTA, washed three times with PBS, and pelleted by centrifugation. These pellets were subjected to three freeze-thaw cycles to achieve cell lysis, followed by a 30-min incubation period at 75°C to achieve protein denaturation. Caseum surrogate was rested at 37°C for 3 days and stored at −20°C.

Flow cytometry.

THPMs were cultured in 12-well plates at 1 × 10⁶ cells per well and treated with OA, SA, or iMtb as described above. The cells were fixed with 4% paraformaldehyde, treated with 5 μM BODIPY 493/503 for 30 min and rinsed with 0.5% bovine serum albumin (BSA) in PBS. Samples were analyzed on a BD LSR Fortessa flow cytometer. BODIPY staining was assessed in the fluorescein isothiocyanate (FITC) channel. A total of 35,000 events were measured from each sample and gated as described in Fig. S1A and B.

Fluorescence microscopy.

THP-1 cells were cultured on Lab-Tek chamber slides with 8 chambers each and treated as described above. The slides were washed with PBS, fixed in 4% paraformaldehyde for 30 min, and stained with 10 μM BODIPY 493/503 (Invitrogen, Waltham, MA) in PBS in the dark for 20 min. After three washes, cells were stained with 300 nM 4′,6-diamidino-2-phenylindole (DAPI) in PBS for 5 min. After 3 final PBS washes, slides were mounted with ProLong Gold antifade reagent (Invitrogen) and left to cure overnight at room temperature. The stained cells were observed with a Nikon Eclipse Ti fluorescence microscope under a 40× objective.

TBARS assay.

Malondialdehyde contents in each surrogate caseum preparation were measured using the TBARS assay kit (R&D Systems). Caseum surrogates were homogenized in water (1:9 [wt/vol]), precipitated with equal volumes of acid reagent, and incubated with TBA reagent as instructed by the manufacturer.

Genomic DNA sequencing to assess bacterial growth dynamics.

Total DNA was extracted from rabbit caseum homogenates containing native Mtb or caseum surrogate homogenates containing preadapted Mtb using the QIAamp Fast DNA tissue kit (Qiagen, Germantown, MD) and heat inactivation at 90°C for 30 min. Sequencing libraries were prepared using the Nextera XT DNA library preparation kit (Illumina, San Diego, CA). Samples were pooled and sequenced on the Illumina NextSeq platform (paired end, 2 × 75 bp). Low-quality bases and Illumina-specific reads were trimmed (Trimmomatic-0.32, LEADING:3 TRAILING:3 SLIDINGWINDOW:4:15 MINLEN:30) (63). Reads were merged using FLASH-1.2.7 and aligned (BWA-mem) against the rabbit genome (GCF_000003625.3) (64). Unaligned reads were extracted and aligned to the circularized Mycobacterium tuberculosis H37Rv genome (edited from NC_000962.3). Bacterial genome replication dynamics were visualized using the approach described by Burnham et al. (65).

Host lipid quantitation.

Total glyceride and cholesterol quantification were achieved using colorimetric kits (BioVision, Inc., Milpitas, CA) as instructed by the manufacturer. For TAG quantification, caseum and surrogate samples were homogenized in a 5% NP-40 solution (1:9 [wt/vol] dilution), heated to 90°C and cooled to completely solubilize all TAG. For total cholesterol quantification, samples were homogenized in chloroform–isopropanol–NP-40 (7:11:0.1). Dissolved lipids were dried down completely and dissolved in the assay buffer.

Specific FFA species in all matrices were quantified by LC-MS. Twenty-five microliters of 3×-diluted homogenates were extracted with 225 μL of acetonitrile, and insoluble material was removed by centrifugation. Palmitic acid-D₃₁ (Avanti Polar Lipids) was included in the analysis as an internal standard. Calibration standards for quantification were prepared in the range of 1 to 50,000 ng/mL for each FFA listed in Table S1 (Avanti Polar Lipids). LC-MS analysis was performed on a Q Exactive high-resolution mass spectrometer (Thermo Fisher Scientific, MA) coupled to a Thermo Scientific Dionex UltiMate 3000 binary system. Chromatography was performed with a Kinetex C₁₈ column (2.1 by 50 mm; particle size, 1.7 μm; Phenomenex, Torrance, CA) using reverse-phase gradient elution. FFA species were quantified as [M-H]⁻ adducts and identified based on retention time and exact mass referenced to the Lipid Maps database.

pH determination for caseum and surrogates.

pH measurements were made as described previously (39). A thin layer of each caseum surrogate was applied on prewetted pH 5.1 to 7.2 pH indicator strips (VWR Chemicals BDH).

Nile red staining and confocal microscopy.

Fixed smears were stained with auramine O (Remel, Lenexa, KS, USA) for 15 min, rinsed, decolorized with 3% acid alcohol (Remel) for 3 min, rinsed again, stained with 10 μg/mL Nile red in ethanol for 15 min, and rinsed again. After a 1-min incubation with potassium permanganate and a final rinse, smears were mounted with ProLong Gold antifade reagent. Images were acquired by confocal laser microscopy at 561 nm (excitation) and 582 to 622 nm (emission) for the red channel and 489 nm (excitation) and 522 to 542 nm (emission) for the green channel, using an A1 confocal microscope (Nikon, Tokyo, Japan). Three-dimensional reconstruction and deconvolution of areas of interest were performed on 19 to 25 sections. Images were analyzed using NIS Elements software (Nikon). For quantitative comparison, we counted the number of Nile red- and auramine-stained Mtb in >350 individual bacilli from multiple microscopic scans of each experimental arm.

Gene expression profiling.

Specific gene expression in native Mtb in rabbit caseum and in preadapted bacteria in caseum surrogate was compared by qRT-PCR. Approximately 100 mg of fresh rabbit caseum specimens from a 12- to 40-week Mtb-infected rabbit or 100 mg of caseum surrogate with an approximate bacterial burden of 10⁸ CFU/g was used in each RNA isolation preparation. Caseum specimens were processed immediately upon excision from pulmonary granulomas and cavities. Mtb genomic DNA (gDNA) was used as a positive control for amplification. All samples were resuspended in 1 mL of TRIzol (Invitrogen). The suspensions were initially disrupted at maximum speed on a Fisherbrand Bead Mill 24 homogenizer along with 0.1-mm zirconia/silicon beads (BioSpec Products), followed by further disruption of the Mtb using Lysing Matrix B tubes (MP Biomedicals). Total RNA was extracted using phenol-chloroform in a heavy-phase Lock Gel I tube (QuantaBio) followed by column purification using the Qiagen RNeasy kit (Qiagen). An on-column DNase digest was performed using DNase I (Qiagen) in RDD buffer. RDD buffer is a component of the RNase-Free DNase Set. After column purification, an additional DNase digest was performed using Ambion Turbo DNase treatment. Reverse transcription was carried out using SuperScript III reverse transcriptase (Invitrogen). Matched samples were included without the addition of reverse transcriptase as controls. Quantitative PCR was performed on a Bio-Rad CFX384 real-time PCR detection system using iTaq Universal SYBR green Supermix (Bio-Rad). Each plate used included the reference gene (sigA) to account for any variability across days or plates. Primers are listed in Table S3. Reverse PCR amplification conditions consisted of 30 s at 95°C, followed by 40 cycles of 5 s at 95°C and 30 s at 60°C, followed by melting curve analysis from 65 to 95°C with 0.5°C increments at 5 s/step. Data were analyzed using R version 4.1.2 using the tidyverse, ggpubr, and ggrepel packages. The sigA locus was used as the endogenous control for normalization; for each target gene, the C_q values were divided by those from the matched sigA control to generate normalized C_q values. Groups of normalized C_q values from either mimic or rabbit samples were compared using an unpaired Student’s t test, and the P values were adjusted using the Holm-Bonferroni method.

TABLE S3

Primers used in this study. Download Table S3, PDF file, 0.10 MB ^{(100.9KB, pdf)} .

This is a work of the U.S. Government and is not subject to copyright protection in the United States. Foreign copyrights may apply.

MBC assays.

The minimum bactericidal concentration assay against M. tuberculosis found in rabbit caseum was performed as described previously (37). casMBC₉₀ is defined as the minimum concentration that killed 90% of bacteria residing in caseum. The bactericidal assay was adapted for use with caseum surrogate. HN878 was grown in 7H9 Middlebrook medium supplemented with 10% albumin dextrose catalase supplement (ADC), 0.2% glycerol, and 0.05% Tween 80 to an optical density at 600 nm (OD₆₀₀) of 0.6 to 0.9. The culture was centrifuged and resuspended in water to an OD of 0.8. This cell suspension was added to the surrogate caseum pellets in a ratio of 2:1 (vol/wt). Suspensions were briefly homogenized with 1.4-mm zirconia beads. Mtb adapted to the surrogate matrix for 6 weeks at 37°C, unless specified otherwise. Subsequently, the mixture was exposed to test compounds in the range of 0.031 μM to 512 μM for 7 days, unless specified otherwise. srgMBC₉₀ is defined as the minimum concentration that killed 90% of bacteria residing in the caseum surrogate. ΔLog CFU_max is defined as the logarithmic function of the net decrease in bacterial burden between the untreated (vehicle-only) condition and the highest drug concentration tested. The drug combination bedaquiline-pretomanid-linezolid was designed by centering the concentration range on the casMBC₉₀ of individual drugs with 4-fold increments between data points (Fig. 6C, right).