Mycobacterium tuberculosis (Mtb) is the world’s most deadly pathogen. Unlike less virulent mycobacteria, Mtb produces 1-tuberculosinyladenosine (1-TbAd), an unusual terpene nucleoside of unknown function. In the present study 1-TbAd has been shown to be a naturally evolved phagolysosome disruptor. 1-TbAd is highly prevalent among patient-derived Mtb strains, where it is among the most abundant lipids produced. Synthesis of TbAd analogs and their testing in cells demonstrate that their biological action is dependent on lipid linkage to the 1-position of adenosine, which creates a strong conjugate base. Furthermore, C20 lipid moieties confer passage through membranes. 1-TbAd selectively accumulates in acidic compartments, where it neutralizes the pH and swells lysosomes, obliterating their multilamellar structure. During macrophage infection, a 1-TbAd biosynthesis gene (Rv3378c) confers marked phagosomal swelling and intraphagosomal inclusions, demonstrating an essential role in regulating the Mtb cellular microenvironment. Although macrophages kill intracellular bacteria through phagosome acidification, Mtb coats itself abundantly with antacid.
Mycobacterium tuberculosis kills more humans than any other pathogen1. Whereas most bacterial pathogens cause acute disease, Mtb usually undergoes a years-long infection cycle. Mtb persists in humans in part through parasitism of macrophage phagosomes. Survival in this intracellular niche is accomplished by slowing phagosomal maturation and reducing intracellular killing mechanisms2,3,4, while offering partial cloaking from immune cells and access to lipids and other host nutrients5,6. As Mtb interactions with the host play out over years and at diverse anatomical sites, pinpointing specific events that determine tuberculosis (TB) disease outcome is challenging. However, a successful approach has been the comparative profiling of mycobacteria of varying virulence to discover factors selectively present in highly virulent species. Mycobacterium species naturally differ in their potential to infect, persist and cause TB, and transmit among hosts. With an estimated 1.7 billion infections worldwide1, only Mtb has broadly colonized the human species, and humans represent its only natural host. These observations highlight the need to identify factors selectively expressed in Mtb but not in other mycobacterial species.
Comparative genomics and transcriptomics of Mtb and Bacille Calmette-Guèrin (BCG) have isolated factors selectively present in Mtb, such as the ESX-1 transporter7. Whereas genetic techniques are widely used, comparative chemical biology screens are uncommon in mycobacteria. An HPLC–mass spectrometry (MS)-based lipidomics platform was developed for analysis of all chloroform/methanol-extractable mycobacterial lipids8,9. Comparative lipidomics of Mtb and BCG identified a previously unknown, Mtb-specific lipid missed by genomics approaches: 1-tuberculosinyladenosine (1-TbAd, 1)10. Cyclization of geranylgeranyl pyrophosphate into tuberculosinyl pyrophosphate occurs via the enzyme, Rv3377c, and tuberculosinyl transferase (Rv3378c) generates 1-TbAd, which can chemically rearrange to N6-TbAd (2)10,11,12. So far 1-TbAd has been detected only in Mtb12, so its expression correlates with evolved virulence. However, 1-TbAd has been studied only in laboratory-adapted strains12,13, and the extent to which it is produced by patient-derived Mtb strains remains unknown.
Furthermore, 1-TbAd’s function remains unknown. Transposon inactivation of Rv3377c or Rv3378c reduced Mtb uptake, phagosomal acidification and killing of Mtb in mouse macrophages14. Therefore, 1-TbAd might influence some aspects of these processes in host cells. However, any host receptor, receptor-independent mechanism or other target of 1-TbAd in host cells remains unknown. Commonly used bioinformatic predictors were not helpful for understanding 1-TbAd function, because it was not possible to identify orthologous biosynthetic genes or similar 1-linked purines in other species. Therefore, diverse candidate mechanisms of 1-TbAd action were tested on human cells. Unexpectedly, 1-TbAd acts as an antacid that directly protects Mtb from acid pH and physically remodels Mtb phagolysosomes. Whereas Mtb was previously known to resist acidification via exclusion of lysosomal fusion with infected phagosomes4,6,15, in the present study it is proposed that Mtb also resists its normally acidic microenvironment by shedding massive quantities of antacid.
High prevalence of 1-TbAd among clinical Mtb strains
1-TbAd was identified in the laboratory strain H37Rv10,11,12. To determine whether 1-TbAd is produced in patient-derived Mtb strains, 52 sputum isolates from Peruvian TB patients were cultured. Complex lipid extracts were subjected to positive-mode HPLC–MS, in which ion chromatograms matching the mass (m/z 540.4) and retention time values of 1-TbAd (23 min) and N6-TbAd (6 min) tracked key compounds (Fig. 1a, and see Supplementary Fig. 1a). Using phosphatidylinositol (3) (PI, m/z 870.6) as a loading control, strong signals for 1-TbAd were detected in 50 of 52 patient-derived Mtb strains (Fig. 1b), leading to several conclusions: 1-TbAd is highly prevalent (96%) but not universally present among strains. Among the 50 1-TbAd+ strains, mean 1-TbAd signals were higher than for PI, an abundant membrane lipid. Patient-derived strains expressed 1-TbAd at higher mean intensity (78 × 106) than strain H37Rv (31 × 106), indicating that prior studies on laboratory strains10,12 underestimated production by clinical strains.
Extending a prior study that detected 1-TbAd only in Mtb12, a basic local alignment search tool analysis identified orthologs of Rv3377c and Rv3378c only in species highly related to Mtb. We did not identify species with orthologs among environmental mycobacteria distant from the Mtb complex or any other bacterium, including common non-mycobacterial lung pathogens (see Supplementary Fig. 1b). Although these genetic results do not rule out convergent evolution of TbAd-like molecules based on unrelated genes, the probable conclusion is that TbAd biosynthesis is restricted to virulent mycobacterial species. Mtb-specific expression of TbAd biosynthesis genes and broad clinical prevalence of 1-TbAd provided a strong rationale to determine TbAd’s function.
TbAd biosynthetic genes in pH regulation
Macrophages are host cells for Mtb residence and play key roles in killing Mtb4. The known roles of adenosine and other purinergic receptors on macrophages and the adenosine moiety in TbAd16,17 led to testing of its activating properties on mixed human myeloid cells. However, 1-TbAd did not alter markers of monocytes (CD14), activation (CD80, CD86) or maturation (CD1a, CD83), which tested MyD88, mitogen-activated protein kinase and other activation pathways (Fig. 1c). Quorum sensing was considered, but no correlation was found between 1-TbAd concentrations and growth (see Supplementary Fig. 1c).
Based on the observed high 1-TbAd production (Fig. 1b) and reinterpretation of a transposon screen showing that Rv3377c and Rv3378c controlled Mtb phagosomal acidification14, it was hypothesized that 1-TbAd might act as an exotoxin that controls the pH of phagolysosomes. This hypothesis was plausible because only Mtb blocks phagosomal acidification4, and only Mtb is known to produce 1-TbAd12 (see Supplementary Fig. 1b). Two landmark studies previously showed that Mtb blocks vesicular ATPase (vATPase) fusion with infected phagosomes2,3. This deacidification mechanism is now considered a central means of evasion of cellular killing by macrophages. Resetting the pH from 5 to 6.2 inhibits downstream antibacterial effector functions15,18, including acid hydrolases, and reactive oxygen and nitrogen intermediates4,19, as well as autophagy, which is a major cellular pathway controlling Mtb survival20.
In contrast to deacidification via vATPase exclusion2,3, two observations indicated that TbAd probably operated via a previously unknown mechanism. First, Rv3377c–Rv3378c was transferred into M. kansasii, which conferred biosynthesis of a molecule with the mass, retention time and collision-induced dissociation (CID) mass spectrum (Fig. 1d) of 1-TbAd. Gene transfer did not affect growth in media at neutral pH, but did increase survival under pH stress (Fig. 1e). Increased growth was observed at pH 5.4–5.1, which can be achieved in activated macrophages, but is not found during virulent Mtb infection, where the pH is 6–6.2 (ref. 6). Decreased Mtb survival after knockout of Rv3377c–Rv3378c14 and increased survival after Rv3377c–Rv3378c knock-in to M. kansasii both represented strong links of TbAd biosynthetic genes with mycobacterial growth and survival. Importantly, the M. kansasii experiment indicated that TbAd biosynthetic genes act on Mtb itself, so key effects occurred independently of vATPases and all other host cell factors. A second clue to a possible function is that 1-TbAd is a strong conjugate base, so it has intrinsic antacid properties. Model compounds21,22 indicate that the pKa of 1-TbAd is ~8.5. The lipid linkage to the 1-position of adenosine renders the molecule acidic and therefore in equilibrium with its conjugate base, which is a strong base (Fig. 1f). Thus, the abundant shedding of 1-TbAd into the extrabacterial space10 could act as an antacid, locally neutralizing the acidic phagolysosomal microenvironment.
Massive biosynthesis of 1-TbAd by Mtb
Unlike receptor-mediated amplification, base-mediated pH neutralization is stoichiometric: one molecule of basic 1-TbAd captures one proton (Fig. 1f). De novo synthesis of 1-TbAd requires 26 steps (see Supplementary Fig. 2), so neutralization of pH by ~0.7 pH units would be metabolically expensive. Mitigating this concern, Mtb strains showed high absolute MS signals (Fig. 1b) and produce 1-TbAd constitutively under many conditions10,11, and at low and high bacterial density (see Supplementary Fig. 1). The pH-dependent effector molecules in phagolysosomes are directly adjacent to intracellular mycobacteria, and phagosomes have a small volume (~10−15 l) (23). Thus, effective pH neutralization could plausibly be generated in close proximity to Mtb.
To assess whether 1-TbAd biosynthesis was quantitatively sufficient for its proposed cellular effect, 1-TbAd was measured among additional patient-derived Mtb strains. Thin-layer chromatography of total lipids from three TbAd+ but not TbAd− strains showed dark spots with the same retention factor as the 1-TbAd standard (0.51). This spot was among the darkest spots seen in total Mtb lipid extracts (Fig. 2a). After synthesis of 1-[13C5]TbAd (4) (m/z 545.374) as an internal standard (Fig. 2b), quantitative analysis demonstrated that 1-TbAd comprised ~1% of total Mtb lipids in clinical strains (Fig. 2c). Approximately 7 ng of 1-TbAd was measured per 108 bacteria, with 91% cell-associated and 9% shed (Fig. 2d). Assuming one bacterium per phagosome23, intraphagosomal concentrations could plausibly reach micromolar concentrations and cause ~0.7 pH unit effect (see Supplementary Fig. 3)6,14. Overall, these experiments document massive biosynthesis and accumulation, establishing 1-TbAd as one of the most abundant Mtb lipids.
Influence on lysosomal pH in cells
Next, 1-TbAd’s effect on phagocytosis and lysosomal pH was tested in macrophage-like (THP-1) cells. As trace contaminants in Mtb-derived 1-TbAd can confound cellular assays, 1-TbAd was synthesized as well as nine analogs (compounds 5–12) for testing in parallel with Mtb-derived material24 (Fig. 3a, and see Supplementary Fig. 4). LysoTracker is a red fluorescent dye that accumulates in acidic compartments. Pre-feeding cells with green fluorescent beads allows concomitant measurement of phagocytosis (see Supplementary Fig. 5). Green beads were rarely seen outside the perimeter of cells in diffraction interference contrast images and could be excluded when present (see Supplementary Figs. 5, left panels, 6a). To assess cellular entry, a pilot study showed that ~81% of beads were ringed with lysosomal-associated membrane protein 1 (LAMP-1) staining and that 1-TbAd pre-treatment did not alter this ratio (see Supplementary Fig. 6a). For the larger study with seven tested compounds (see Supplementary Fig. 5), fluorescence results from five fields (>100 cells) showed similar results in three assays (see Supplementary Fig. 6b). Using total bead uptake as a measure of phagocytosis, no statistically significant changes were found in any condition. For measurement of lysotracker+ compartments, two negative controls, phosphatidylcholine (13) and isotuberculosinol (14), showed no effect. Both natural and synthetic 1-TbAd notably reduced the number of lysotracker+ compartments (Fig. 3b, and see Supplementary Figs. 5 and 6b). 1-Phytyladenosine (5), which contains a straight-chain polyprenyl group substituted for the ringed tuberculosinyl unit in 1-TbAd, showed a similar effect, indicating that the halimane core is not essential. Adenosine (15) and N6-TbAd, which lack the 1-linkage that confers proton capture (Figs. 1f and 3b), had no effect on lysotracker staining.
Extending this preliminary study of THP-1 cells, studies of primary mouse bone marrow macrophages treated with carboxyfluorescein were undertaken. The latter approach allows more direct assessment of pH through comparison of fluorescence at pH-dependent (excitation 490 nm, emission 520 nm) and pH-independent (excitation 450 nm, emission 520 nm) wavelengths25. Feeding carboxyfluorescein silica beads to fresh macrophages provides a readout of relative pH as phagosomes mature over time26. 1-TbAd and chloroquine (16), a positive control for a lysosomotropic base, showed similar outcomes. Chloroquine provided a more rapid effect, but slightly lower peak effect on baseline fluorescence (Fig. 3c). Control wells with beads treated with 1-TbAd demonstrated the lack of direct effect on fluorescence (Fig. 3d). In wells with macrophages, addition of 1-TbAd and 1-phytyladenosine (5) delayed and reduced acidification of phagosomes in a concentration-dependent manner, whereas adenosine had little effect (Fig. 3d). Overall, in human and mouse cells, the outcomes of two pH assays matched the prediction that 1-linked but not N6-linked adenosines could raise the pH. The role of C20 lipid moieties was hypothesized to provide the hydrophobicity needed to traverse membranes. However, assays in live cells do not establish this conclusion because compounds could have been actively ingested.
Lysosomotropism predicts TbAd behavior
These clear structure–activity relationships (SARs) were combined with descriptions of lysosomotropic drug behavior by de Duve et al.27 to generate a detailed model for 1-TbAd function (Fig. 4a)—TbAd is in equilibrium with its uncharged conjugate base (Fig. 1f), which is proposed to permeate mycobacterial membranes to reach the phagosomes (Fig. 4a, and see Supplementary Fig. 7). In an acidic environment (pH 5.5–6.2), 1-TbAd (pKa ≈ 8.5) but not N6-TbAd (pKa ≈ 3.8) neutralizes pH, creating 1-TbAd+, a charged, membrane-impermeable species that is trapped (see Supplementary Fig. 7a). Intraphagosomal protonation is predicted to generate a concentration gradient of uncharged base across the mycobacterial membrane, where permeation and trapping continue until equilibrium is reached. This process is proposed to create a large intraphagosomal pool of 1-TbAd+, the relative size of which is predicted by the Henderson–Hasselbalch equation (Fig. 4a, and see Supplementary Fig. 7b). Although extracellular drugs, such as chloroquine, cross membranes into cells and enter lysosomes27, the reverse topology is proposed here: natural molecules made in the mycobacterial cytosol escape outwards and are trapped within maturing phagosomes (see Supplementary Fig. 7c).
High-throughput screens28 and lysosomotropic models27 emphasize that a pKa ≈ 8, as in the case of 1-TbAd, is optimal for lysosomotropism. Compounds with substantially higher pKa values remain charged and impermeant, whereas those with lower pKa values, such as N6-TbAd, permeate membranes but do not efficiently capture protons (see Supplementary Fig. 5). These models led us to more extensively compare 1-TbAd function with chloroquine, a known lysosomotropic base27. Chloroquine is widely used against malaria, systemic lupus erythematosus and other autoimmune diseases. These medical indications rely on its lack of signaling, low toxicity and tropism to lysosomes29. In experimental medicine, chloroquine’s antacid effects block antigen presentation by major histocompatibility complex class II (MHC-II), toll-like receptor activation and autophagy29.
EM of human macrophages
In EM, macrophage lysosomes appear as highly electron-dense structures (Fig. 4a, and see Supplementary Fig. 8a,b). Chloroquine or 1-TbAd transformed these small, electron-dense compartments into large, electron-lucent compartments that, despite complete loss of their multilamellar and electron-dense appearance, could still be recognized as lysosomes based on immunogold staining of CD63, a lysosome marker (Fig. 4b,c, and see Supplementary Fig. 8). This process was widespread (Fig. 4b, and see Supplementary Fig. 8a,c,e), such that cells were designated as having swollen lysosomes when the electron-lucent compartments involved more than one-third of the cytoplasm (Fig. 4b). Both the nature of intraphagosomal inclusions and the broad extent of cellular involvement are visualized through side-by-side comparison of low- and high-magnification images with pseudocoloring of lysosomes (see Supplementary Fig. 8). By analysis of >100 cells per condition, 1-TbAd showed statistically significant effects on lysosomes compared with no treatment (Fig. 4b).
For chloroquine this structural transformation is known to involve its accumulation in acidic compartments and secondary osmotic effects20,27,29. 1-TbAd showed stronger effects than chloroquine on human macrophages (Fig. 4b). The lysosomotropic model predicts that non-acidic organelles, such as early endosomes, Golgi bodies and mitochondria, would be relatively unaffected27, as observed in the assays in the present study (Fig. 4d). For chloroquine and 1-TbAd the appearance of small, mildly electron-dense inclusions was noted, including membrane-containing intralysosomal bodies (Fig. 4e, and see Supplementary Fig. 8), which, for chloroquine, results from autophagy blockade20,27,29. Last, a negative control, tuberculosinol (17), produced no discernible effects on lysosomes, again suggesting the importance of the 1-adenosine linkage (see Supplementary Fig. 9).
Testing vesicle permeation
Lysosomotropism requires membrane permeation (Fig. 4a), but lysotracker suppression in THP-1 cells (Fig. 3b) and macrophages (Fig. 3d) might have occurred through active cellular uptake. To measure transmembrane diffusion, the present study took a look at model membrane systems. For inverted membrane vesicles (IMVs)30, inversion orients proton pumps so that the interior spontaneously acidifies in response to added energy substrates, ATP or succinate. The reporter, 9-amino-6-chloro-2-methoxyacridine, is quenched by acid. Quenching is reversed by penetration of lysosomotropic substances into the IMV. Compounds that disrupt membranes cause decreased fluorescence, providing a control for membrane leakage30. Unlike cells (Figs. 3 and 4, and see Supplementary Fig. 5), IMVs have no active cytoskeleton-mediated drug uptake. As a positive control, chloroquine showed reporter quenching with half-maximal effect at ~20 μM, matching its expected function and potency27,28,29 (Fig. 5a).
All four 1-linked adenosines (1, 5–7) carrying C20 lipids were more potent than chloroquine, and none of the N6-linked adenosines (2, 11, 12) showed suppression (Fig. 5a). These results suggest direct membrane permeation and rule in the 1-linkage as the essential chemical feature (Fig. 4a). Stereochemical changes in the 1-linked lipid ((Z)1-TbAd) (7), or an acyclic lipid (1-phytyladenosine, 5), had little effect, as long as a C20 lipid was present. However, C10, C5 or C1 lipid analogs (8–10) showed that potency declines as chain length decreases, consistent with the lipid moiety generating a hydrophobic effect (Figs. 3a and 5a). Carbohydrate-modified, 2′-deoxy-1-TbAd (6), was more potent than 1-TbAd and chloroquine. Thus, hydrophobicity and potency are correlated among analogs. This result suggests that hydrophobicity drives biological action and provides an approach to future synthesis of yet more potent compounds. Similar results were observed when using ATP or succinate as energy substrates, and using Mycobacterium smegmatis or Escherichia coli IMVs (Fig. 5a). The SARs in IMVs generally matched the patterns seen in human (Figs. 3b and 4) and mouse cells (Fig. 3c,d).
Prior studies suggested that beads coated with the free alcohol component of TbAd could alter the pH of phagosomes31. However, isotuberculosinol (14) and tuberculosinol (17) are not predicted to have basic properties, and they showed few effects in lysotracker studies (Fig. 3b), EM images (see Supplementary Fig. 9) and E. coli IMVs (Fig. 5a). However, IMV assays with M. smegmatis membranes (Fig. 5a) did show some effect with synthetic tuberculosinol. These somewhat differing results might be explained if the tuberculosinyl moiety mediates some unknown but specific interaction with mycobacterial membranes.
Protein-free model membranes
To exclude artifacts from inhibitors binding to protein targets in IMVs, testing of protein-free membranes was required. Therefore, we generated liposomes with differing interior pH values (5.0 and 7.4) (Fig. 5b). Lysosomotropism predicts that externally applied compounds selectively penetrate acidic but not neutral liposomes. Fulfilling this prediction, chloroquine co-migrated with acidic but not neutral liposomes on a size-exclusion column. A negative control, adenosine, failed to bind to either liposome type. 1-TbAd and N6-TbAd showed some adhesion to neutral and acidic liposomes (Fig. 5c), which was probably mediated by their identical lipid moieties (see Fig. 1a). Only 1-TbAd was preferentially captured by liposomes with an acidic interior (Fig. 5c). This reductionist system rules in a purely chemical mechanism: 1-TbAd is a membrane-permeable antacid, where the 1-linked lipid moiety provides intrinsic transmembrane tropism for acidic compartments (Fig. 5d). Overall, 1-TbAd is a naturally evolved phagolysosome disrupter, with an action that mimics the widely used lysosomotropic drug chloroquine.
Generation of Mtb lacking Rv3378c and 1-TbAd
To identify non-redundant functions of the 1-TbAd biosynthesis pathway during cellular infection, the tuberculosinyl transferase gene, Rv3378c, was deleted in the H37Rv strain (MtbΔRv3378c) (see Supplementary Fig. 10a). Gene deletion and replacement with the hygromycin-resistance cassette were confirmed through PCR analysis, as well as the complete abrogation 1-TbAd biosynthesis (see Supplementary Fig. 10b). For genetic complementation (MtbΔRv3378c::Rv3378c), a single-copy chromosomal integration of Rv3378c was used under the control of a mycobacterially optimized promoter. Full restoration of 1-TbAd production to wild-type levels was observed after subculture and selection of a high producing Mtb clone (see Supplementary Fig. 10b).
Role of biosynthetic genes in live macrophage infection
After 4 d of infection, human macrophages were examined by transmission EM (TEM). High magnification images depicting transbacterial sections revealed infected phagosomes. Transverse sections show that mycobacterial cytosol (*) is surrounded by the cell wall, phagosomal space, limiting phagosomal membrane and macrophage cytosol (Fig. 6a). For MtbΔRv3378c, most intraphagosomal bacteria were surrounded by an electron-lucent ring of uniform thickness (~20 nm), the mycobacterial polysaccharide capsule (Fig. 6a, small arrows)32. For MtbΔRv3378c the phagosomal membrane is typically tightly wrapped around the 20-nm mycobacterial capsule (tight phagosome).
In contrast, for wild-type Mtb, phagosomal membranes typically showed numerous large (~20–250 nm) ectopic blebs (Fig. 6a, large arrows) outside the capsular ring. In some cases these blebs almost surround the bacterium, creating the appearance of a loosely wrapped (swollen) phagosome with many non-bacterial, intraphagosomal inclusions. Inspection of >150 phagosomes in each of two experiments showed that Rv3378c expression resulted in a 4.5- or 4.9-fold increase in the ratio of swollen to tight phagosomes (Fig. 6b). The swollen phagosomes provided clear evidence for compartment autonomous effects of Rv3378c expression (Fig. 4a). Low-power analysis revealed that expression of Rv3378c was associated with many swollen phagosomes that lacked visible bacteria (Fig. 6c). The broader involvement of compartments was somewhat surprising, because it suggested that 1-TbAd produced in one compartment could more broadly affect lysosomes in cells.
Therefore, more detailed and quantitative analysis was undertaken of low-magnification images of infected cells with visible bacteria (Fig. 6c, red pseudocolor). Individual cells were scored as having swollen lysosomes when one-third of the cytosol was involved (Fig. 6c, yellow pseudocolor). Compared with uninfected macrophages, Mtb-infected human macrophages showed marked increases with swollen lysosomes (Fig. 6d). Rv3378c deletion significantly reduced swelling, and complementation restored the phenotype to the baseline value. Although undetected bacteria cannot be ruled out in swollen lysosomes, the multiplicity of infection used (two bacteria per cell) and the large number of swollen compartments suggested non-compartment autonomous effects. It was concluded that effects downstream of Rv3378c play an essential role in the physical remodeling of the local intraphagosomal growth niche of Mtb. Phenotypes from wild-type Mtb (Fig. 6) mimicked key aspects seen after treatment with pure 1-TbAd (Fig. 4c–e), and they match the predicted effects of release of any lysosomotropic substance27 and the known effects of chloroquine29.
Based on 1-TbAd’s shed nature and restriction to virulent mycobacteria in the Mtb complex, initially comparisons were considered with bacterial endotoxins and exotoxins, such as lipopolysaccharide, which mediates rapid and extreme host cellular response. Although no experiment can rule out receptor-mediated signaling, no evidence for generalized cellular activation was found in response to 1-TbAd. Mechanistically, increased growth of 1-TbAd+ M. kansasii at acidic pH is critical, because it rules in macrophage-independent effects on mycobacterial growth and specifically connects survival to pH. Rather than a generalized cellular toxin, 1-TbAd is a lysosomotrope, causing gross phagosome disruption and raising the pH in THP-1 cells and macrophages. This conclusion is strengthened by comparison to the known lysosomotrope, chloroquine27,29, as well as consistent patterns among TbAd analogs that directly implicate the 1-linkage, which confers its antacid property.
Thus, 1-TbAd mediates a previously unknown effect involving neutralization and local remodeling of Mtb’s intracellular acidic growth niche within macrophages. As 1-TbAd is initially released only at the surface of live mycobacteria, the model predicts the strongest effects within infected phagolysosomes. Therefore, finding broadly swollen lysosomes, including many compartments with no detectable bacilli, was initially surprising. However, the reverse lysosomotropic model allows that steady production over time by intracellular bacteria could lead to 1-TbAd penetration to all acidic compartments, a hypothesis that can now be tested with kinetic studies.
Massive production of 1-TbAd, among 96% of tested patient strains, is notable. Extending work in which TbAds were detected in mice12, these findings support the feasibility of developing 1-TbAd as a marker of Mtb infection and TB disease. Also, the high production among patient strains highlights one surprising aspect of this work12, which is the failure of such an abundant molecule to be detected during decades of TB research. Overall, the high biosynthesis is well matched to an unamplified, stoichiometric mechanism of action that involves proton capture.
Modified adenosines and related purines have evolved repeatedly in eukaryotic and prokaryotic cells, but 1-linked adenosines are rare in nature10,33. Such adenosines can non-enzymatically rearrange to N6-variants12, which occurs in vivo during infection in mice11. The instability of 1-linked purines might account for their rarity in nature, but they are stable in acid environments, as in the mechanism proposed here. SAR studies with synthetic analogs directly demonstrate that the 1-linkage, the unusual and defining chemical feature of 1-TbAd, controls its biological activity. Thus, the revised model is that, among the last two steps of the natural TbAd biosynthetic pathway, 1-TbAd is the active metabolite that controls lysosomal function. N6-TbAd largely lacks this function, but might have use as a species-specific diagnostic marker of infection.
Phagolysosomal acidification is a key outcome of the interferon-γ-mediated interactions between macrophages and T cells. The role of acidification as an upstream controller of phagolysosome maturation and intracellular killing has been recognized for decades4,15, as has evidence for Mtb’s manipulation of this pathway2,3. Acidification blockade by Mtb is currently considered an immunoevasion mechanism resulting from bacterial blockade of vATPase delivery to infected phagosomes6. Although the previously known mechanism limits proton pumping into infected compartments, it is proposed that antacid release is a complementary mechanism that acts within phagosomes to scavenge protons that do arrive.
Lysosomal acidification controls the final mechanisms of intercellular lipid and protein degradation that are common to autophagy pathways. Autophagic degradation is co-opted during Mtb infection to generate intracellular inclusions29,34, and chloroquine is perhaps the most widely used autophagy inhibitor in experimental settings29. In the present study similar pKa values, potency and biological function of 1-TbAd and chloroquine have been outlined, pointing to a candidate role for 1-TbAd in autophagy inhibition. These observations support future development of synthetic 1-TbAd and analogs as lysosome-disrupting drugs. Related to this, both chloroquine and Mtb acid blockade limit MHC-II peptide loading29, so 1-TbAd now becomes a candidate to influence MHC-II-mediated human T cell response.
Selective expression of 1-TbAd by Mtb, but not less virulent species, provides a correlative basis for proposing that 1-TbAd could be an evolved virulence factor. The Rv3377c–Rv3378c pathway is known only in an obligate human pathogen that continually grows under host immune pressure. Escape from acid-mediated killing could plausibly outweigh the metabolic costs of 1-TbAd biosynthesis. This hypothesis is supported by experimental data: transposon deletion of either Rv3377c or Rv3378c diminishes Mtb survival in mouse macrophages14, and knock-in of these two genes confers a growth advantage in M. kansasii. Thus, these biological studies are consistent with the conclusion that the pH neutralization mechanism identified in the present study controls some aspect of Mtb growth and survival. The overarching questions going forward will be the extent to which this acid neutralization mechanism controls Mtb outcomes during natural infection in vivo, and whether it acts in the distinct phases of acute infection, persistence or transmission.
Patient-derived Mtb strains
Mtb strains were cultivated from the sputum of human TB patients recruited in or near Lima Peru by Socios en Salud under oversight from the Institutional Committee of Ethics in Research of the Peruvian Institutes of Health, the Institutional Review Board of the Harvard Faculty of Medicine and the Partners Healthcare IRB. Peruvian patients provided oral and written informed consent in Spanish.
Flow cytometry of human myeloid cells
Monocyte-derived dendritic cells were prepared by plating human monocytes in a 24-well plate (1 × 106) with 30 ng ml–1 of granulocyte–macrophage colony-stimulating factor (GM-CSF) (Peprotech 300-03) and 40 ng m–1 of interleukin-4 (200-04; Peprotech) for 3 d. The cells were then treated with synthetic 1-TbAd, natural 1-TbAd, Pam3Cys-SKKKK (L2000, EMC Microcollections) or medium for another day. Cells were then washed and collected for FACS analysis. For cell-surface protein detection, cells were stained with mouse anti-human CD1a (OKT6; in-house purified), CD80 (557223), CD86 (555655), CD14 (550376; BD Pharmingen), and isotype controls for immunoglobulin G (IgG)1 (P3; in-house purified) or IgG2a (14-4724-B1, eBioscience) followed by a fluorescein isothiocyanate-conjugated, goat anti-mouse IgG F(ab′)2 (A-10683, Thermo Fisher Scientific), then measured by a FACS Canto Flow Cytometer (BD) and analyzed using FlowJo.
Acidification and phagocytosis
THP-1 cells were obtained from the American Type Culture Collection (ATCC) and cultured in RPMI 1640 medium supplemented with 10% fetal bovine serum (FBS), l-glutamine, penicillin–streptomycin, 2-mercaptoethanol and 10 mM 4-(2-hydroxyethyl)-1-piperazine-ethanesulfonic acid (Hepes). Green fluorescent beads (silica, 3 μM, excitation/emission: 485/510 nm, Kisker Biotech). The beads were washed five times with phosphate-buffered saline (PBS) and stored in 0.5% bovine serum albumin (BSA) in PBS (108 beads ml–1) at 4 °C. Beads were then resuspended in complete media plus 10% human serum (Gemini) for 10 min before the phagocytosis assay. To generate macrophage-like cells, THP-1 cells were plated on cover slips in 24-well plates (3 × 105) and treated with phorbol 12-myristate 13-acetate (p1585; Sigma-Aldrich), 50 nM, for 72 h. The differentiated cells were washed twice, then rested in fresh media for 4 h, followed by lipid treatment for 2 h. Lipid samples were vortexed and sonicated for 2 min before being added to the cells. Cells were then fed with excess fluorescent beads (4 beads per cell) and centrifuged at 1,000g for 1 min before incubation at 37 °C. After 30 min, cells were washed with PBS three times to remove extracellular beads, then treated with 250 nM LysoTracker-red DND-99 (L7528, Thermo Fisher Scientific) at 37 °C for 60 min. To assess washing and surface adherence, pilot studies showed that beads were rarely found outside the margin of cells (<2%) and more than 80% of beads penetrated to LAMP-1+ compartments. Cells were washed, then fixed with 4% paraformaldehyde at room temperature for 20 min before the cover slips were mounted on the slides. Slides were analyzed on a Nikon Eclipse TE2000-UC1 confocal microscope by counting 5 low power fields (~200 cells).
Measurement of pH in mouse macrophages
To generate carboxyfluorescein beads for analysis of phagosomal pH, 50 μg carboxyfluorescein succinimidyl ester (Invitrogen) was added to 12.5 mg carboxylated, 3-μm silica beads (Kisker Biotech) that had been covalently linked to human IgG (Sigma-Aldrich) and defatted BSA26. After incubation on a nutator for 90 min at room temperature, the carboxyfluorescein beads were washed and stored in PBS at 4 °C. Bone marrow-derived macrophages were isolated from C57BL/6J mice (Jackson Laboratories), and maintained in Dulbecco’s modified Eagle’s medium containing 10% FBS (Gibco), 15% L-cell-conditioned media, 2 mM l-glutamine (Sigma-Aldrich), 1 mM sodium pyruvate (Gibco) and antibiotics (penicillin–streptomycin) (Gibco), at 37 °C in a 5% CO2 atmosphere. Macrophages, 2 × 105 per well, were seeded into 96-well, clear-bottom, black plates for assays (Corning Costar). Assays were performed 1–2 d after seeding in the 96-well plates. Macrophages were washed three times with assay buffer (PBS, pH 7.2, 5% FBS, 5 mM dextrose, 1 mM CaCl2, 2.7 mM KCl, 0.5 mM MgCl2), and assay buffer containing indicated concentrations of compounds, the identities of which were blinded to the technician, added back to each well as appropriate. Carboxyfluorescein beads at about two to five beads per macrophage in assay buffer containing compounds were then added, and bottom reads at 450/520 nm and 490/520 nm acquired every 2 min for 2 h on a Biotek Synergy H1 microplate reader. A total of five to six replicate wells per condition were used, with temperature maintained at 37 °C. The ratio of the carboxyfluorescein fluorescence signal at excitation 490 nm (pH sensitive) versus 450 nm (pH insensitive) provides a readout of relative pH26. Animal procedures adhered to the National Institutes of Health’s Guide for the Care and Use of Laboratory Animals, with animal protocol (no. B2016-37) approval by the Institutional Animal Care and Use Committee at Tufts University, in accordance with the Association for Assessment and Accreditation of Laboratory Animal Care, US Department of Agriculture, and US Public Health Service guidelines.
For EM studies, human monocytes (5 × 106) were treated with M-CSF (25 ng ml–1) and GM-CSF (2.5 ng ml–1) for 6 d and treated with lipids or infected with Mtb. The infected cells were incubated 4 h and washed to remove extracellular bacteria. The cells were incubated for 4 d and fixed with 2% formaldehyde and 0.2% glutaraldehyde in 0.4 M PHEM buffer (240 mM 1,4-piperazinediethanesulfonic acid (Pipes), 100 mM Hepes, 40 mM ethylene glycol-bis(β-aminoethyl ether)-N,N,N′,N′-tetraacetic acid (EGTA), 8 mM MgCl2). For the lipid treatment experiments, the cells were treated for 2 h with 20 μM lipid and the fixed cells were processed for EM as published35. For infection experiments, infected cells were incubated 4 h and washed to remove extracellular bacteria. The cells were incubated for 4 d and fixed with 2% formaldehyde and 0.2% glutaraldehyde in 0.4 M PHEM buffer (240 mM Pipes, 100 mM Hepes, 40 mM EGTA, 8 mM MgCl2 (ref. 35). For EM, samples were embedded in gelatin blocks and plunge frozen in liquid nitrogen, and 60-nm ultrathin sections were produced at −120 °C. Then immunogold labeling was performed using antibodies against CD63 (Sanquin), CD107A (Biolegend), EEA1 (Thermo Fisher Scientific), rabbit anti-mouse bridging antibody and 10-nm gold particles conjugated to protein A (Utrecht University). Sections were stained with uranyl acetate and analyzed in a blinded manner using TEM (FEI Tecnai G2 Spirit Biotwin) at 100 kV.
Genomic analysis and statistics
To discover possible orthologs, sequences of Rv3377c or Rv3378c from the National Center for Biotechnology Information Mtb H37Rv reference genome (NC_000962.3) were used to interrogate whole-genome sequences with the basic local alignment search tool (blastn) set at these parameters: base match score of 2, base mismatch score of −3, E-value threshold of 10, minimum word size of 11, gap penalty of 5 and gap extension penalty of 2.
Phosphatidylcholine (Sigma-Aldrich P5394, 1.8 μmol) and cholesterol (C8667; Sigma-Aldrich, 0.73 μmol) were mixed in chloroform in a 50-ml glass tube. The solvent was evaporated under nitrogen to yield a thin film, which was then hydrated in citrate buffer (500 μl) with the desired pH (5.0 or 7.4) by vortexing, followed by six freeze–thaw cycles. Vesicle size was homogenized using a liposome extruder (Avanti Polar Lipids) through polycarbonate membranes (0.2 μm and 0.1 μm). Liposomes were dialyzed against PBS (pH 7.4, 1 l) overnight using a Slide-A-Lyzer MINI Dialysis Device with a 10-kDa cutoff value (100-μl device, Thermo Fisher Scientific). After dialysis, two batches of liposomes were adjusted to equal concentration (absorbance = 0.25 at 600 nm). The final liposomes were examined by TecnaiG2 Spirit BioTWIN EM at the Harvard Medical School Core EM Facility. Liposomes (25 μl) and chloroquine (Sigma-Aldrich, C662825, 50 μM PBS solution) were mixed in a small glass insert, placed in an Eppendorf tube and incubated in a Thermomixer at 37 °C. After 2 h, the mixtures were loaded on to a MicroSpin G-50 column (GE Healthcare) and eluted with PBS, and nine 100-μl fractions were collected. The amount of liposomes in each fraction was determined by optical density. The liposomes were dispersed with 3 μl of Triton X-100 (10%) and the amount of chloroquine in each fraction was measured by the absorption at 342 nm. For adenosine (Sigma-Aldrich, A9251) uptake, 25 μl of a PBS solution was mixed with 25 μl of liposomes, as described above except the final quantification steps. The collected fractions (100 μl each) were mixed with 0.5 ml of chloroform/methanol (1/2, v/v), vortexed and dried under nitrogen. Each dried fraction was re-dissolved in a mobile phase of hexane/isopropyl alcohol (70/30, v/v) and loaded to a 1200 series HPLC system using a normal phase column (Varian MonoChrom Diol: 3 μm × 150 × 2 mm2) and analyzed using an Agilent 6520 Accurate Mass Q-TOF mass spectrometer, based on published methods8. Phosphatidylcholine (m/z 760.58, used as a surrogate for liposome content) and adenosine (m/z 268.10) were quantified by comparing the peak area of extracted-ion chromatograms with external standards. For 1-TbAd and N6-TbAd uptake, the experiment was performed in triplicate. Of lipid sonicate (25 μl), 1 μg was mixed with 25 μl of liposomes and incubated at 37 °C. After 2 h, the mixtures were loaded on to a MicroSpin G-50 column. Due to lipid adherence to the liposomes, the eluate consisted of two fractions: a liposome-associated lipid fraction (lipids co-elute with liposomes) and a liposome-free lipid fraction (lipids that stay on the column). The PBS eluate (500 μl) and column contents (Sephadex G-50) were treated with chloroform/methanol (1/2) 2 ml and 0.5 ml, respectively. The PBS eluate and the column extracts were analyzed by liquid chromatography (LC)–MS–quadruple time of flight (QToF) as described above. Both 1-TbAd and N6-TbAd were detected as m/z 540.35, at retention times of 23 and 5 min, respectively.
Synthetic TbAd-like compounds were tested in triplicate in a blinded fashion in four types of IMV assays using succinate or ATP as energy substrates and E. coli or M. smegmatis membranes in two independent experiments as described previously16.
Knock-in of genes to M. kansasii
M. kansasii ATCC 12478 was transformed with an integrative vector (pMV306) that either carried no insert (empty vector), or a 2.4-kilobase PCR fragment containing Rv3377-8c under the control of a constitutive hsp60 promoter (Rv3377–Rv3378c). Cell-associated lipids were extracted and subjected to MS analyses10. To measure M. kansasii survival at different pH, organisms were maintained at mid-log phase (OD600 0.2–0.5), passed through 22-G and 25-G needles, centrifuged at 450g for 5 min to remove clumps, and resuspended in 20 ml medium with altered pH (prepared using 2 M HCl) at a theoretical OD600 of 0.01. Cultures were grown for 16 d in biological triplicates. At each time point, a volume of 100 μl was taken in duplicate from each culture to measure the OD600 on a 96-well, tissue-culture plate (Falcon) using the Infinite M200 Pro NanoQuant spectrophotometer (Tecan).
Deletion and complementation of Rv3378c
Mtb strain H37Rv was used for complete deletion of Rv3378c using recombineering gene-replacement strategy36. A targeting construct consisting of 500 basepair-flanking regions of Rv3378c and the loxP–hygromycin–LoxP cassette was synthesized and cloned into a pUC57 vector. The linear DNA substrate was amplified from the vector, and electrophoretically transformed into the Mtb H37Rv strain carrying the pNit–recET–SacB–kan plasmid and induced to express recombinase. Transformed bacteria were plated on to 7H10 agar plates containing 50 µg ml–1 of hygromycin for selection of recombinants. The recombinants were further selected for the absence of pNit–recET–SacB–kan plasmid by growing the colonies on 7H10 plates containing 5% sucrose and hygromycin (50 µg ml–1) and subsequently testing the colonies for absence of growth on 7H10 plates with kanamycin (25 µg ml–1). The recombinant colonies were screened by PCR for target gene deletion and replacement by the hygromycin cassette. The PCR screening was performed using primers that amplify the 5′-junction, 3′-junction. The entire target gene locus is positive for the PCR product, indicating target gene replacement and negative for the target gene. Complementation was performed by integrating a single copy of the gene under the control of the MOP promoter in the pJEB402 vector37.
Further information on research design is available in the Nature Research Reporting Summary linked to this article.
Institutional review boards require confidentiality of patient data and biological material. Distribution of Mtb strains is subject to biosafety approvals. Otherwise, all data and reagents are available.
Global Tuberculosis Report (WHO, 2018).
Armstrong, J. A. & Hart, P. D. Response of cultured macrophages to Mycobacterium tuberculosis, with observations on fusion of lysosomes with phagosomes. J. Exp. Med. 134, 713–740 (1971).
Sturgill-Koszycki, S. et al. Lack of acidification in Mycobacterium phagosomes produced by exclusion of the vesicular proton-ATPase. Science 263, 678–681 (1994).
Vandal, O. H., Nathan, C. F. & Ehrt, S. Acid resistance in Mycobacterium tuberculosis. J. Bacteriol. 191, 4714–4721 (2009).
McKinney, J. D. et al. Persistence of Mycobacterium tuberculosis in macrophages and mice requires the glyoxylate shunt enzyme isocitrate lyase. Nature 406, 735–738 (2000).
Russell, D. G. Phagosomes, fatty acids and tuberculosis. Nat. Cell Biol. 5, 776–778 (2003).
Behr, M. A. et al. Comparative genomics of BCG vaccines by whole-genome DNA microarray. Science 284, 1520–1523 (1999).
Layre, E. et al. A comparative lipidomics platform for chemotaxonomic analysis of Mycobacterium tuberculosis. Chem. Biol. 18, 1537–1549 (2011).
Galagan, J. E. et al. The Mycobacterium tuberculosis regulatory network and hypoxia. Nature 499, 178–183 (2013).
Layre, E. et al. Molecular profiling of Mycobacterium tuberculosis identifies tuberculosinyl nucleoside products of the virulence-associated enzyme Rv3378c. Proc. Natl Acad. Sci. USA 111, 2978–2983 (2014).
Pan, S. J. et al. Biomarkers for tuberculosis based on secreted, species-specific, bacterial small molecules. J. Infect. Dis. 212, 1827–1834 (2015).
Young, D. C. et al. In vivo biosynthesis of terpene nucleosides provides unique chemical markers of Mycobacterium tuberculosis infection. Chem. Biol. 22, 516–526 (2015).
Layre, E., de Jong, A. & Moody, D. B. Human T cells use CD1 and MR1 to recognize lipids and small molecules. Curr. Opin. Chem. Biol. 23c, 31–38 (2014).
Pethe, K. et al. Isolation of Mycobacterium tuberculosis mutants defective in the arrest of phagosome maturation. Proc. Natl Acad. Sci. USA 101, 13642–13647 (2004).
Vandal, O. H., Pierini, L. M., Schnappinger, D., Nathan, C. F. & Ehrt, S. A membrane protein preserves intrabacterial pH in intraphagosomal Mycobacterium tuberculosis. Nat. Med. 14, (849–854 (2008).
Heyl, A., Riefler, M., Romanov, G. A. & Schmulling, T. Properties, functions and evolution of cytokinin receptors. Eur. J. Cell Biol. 91, 246–256 (2012).
Cekic, C. & Linden, J. Purinergic regulation of the immune system. Nat. Rev. 16, 177–192 (2016).
MacMicking, J. D. Cell-autonomous effector mechanisms against Mycobacterium tuberculosis. Cold Spring Harb. Perspect. Med. 4, a018507 (2014).
Rohde, K., Yates, R. M., Purdy, G. E. & Russell, D. G. Mycobacterium tuberculosis and the environment within the phagosome. Immunol. Rev. 219, 37–54 (2007).
Deretic, V. et al. Immunologic manifestations of autophagy. J. Clin. Invest. 125, 75–84 (2015).
Kapinos, L. E., Operschall, B. P., Larsen, E. & Sigel, H. Understanding the acid–base properties of adenosine: the intrinsic basicities of N1, N3 and N7. Chemistry 17, 8156–8164 (2011).
Martin, M. G. & Reese, C. B. Some aspects of the chemistry of N(1)- and N(6)-dimethylallyl derivatives of adenosine and adenine. J. Chem. Soc. Perkin 1, 1731–1738 (1968).
Winterbourn, C. C., Hampton, M. B., Livesey, J. H. & Kettle, A. J. Modeling the reactions of superoxide and myeloperoxidase in the neutrophil phagosome: implications for microbial killing. J. Biol. Chem. 281, 39860–39869 (2006).
Buter, J. et al. Stereoselective synthesis of 1-tuberculosinyl sdenosine: a virulence factor of Mycobacterium tuberculosis. J. Org. Chem. 81, 6686–6696 (2016).
Tan, S., Yates, R. M. & Russell, D. G. Mycobacterium tuberculosis: readouts of bacterial fitness and the environment within the phagosome. Methods Mol. Biol. 1519, 333–347 (2017).
Podinovskaia, M., Lee, W., Caldwell, S. & Russell, D. G. Infection of macrophages with Mycobacterium tuberculosis induces global modifications to phagosomal function. Cell Microbiol. 15, 843–859 (2013).
de Duve, C. et al. Commentary. Lysosomotropic agents. Biochem. Pharmacol. 23, 2495–2531 (1974).
Nadanaciva, S. et al. A high content screening assay for identifying lysosomotropic compounds. Toxicology In Vitro 25, 715–723 (2011).
Plantone, D. & Koudriavtseva, T. Current and future use of chloroquine and hydroxychloroquine in infectious, immune, neoplastic, and neurological diseases: a mini-review. Clin. Drug Invest. 38, 653–671 (2018).
Feng, X. et al. Antiinfectives targeting enzymes and the proton motive force. Proc. Natl Acad. Sci. USA 112, E7073–E7082 (2015).
Mann, F. M. et al. Edaxadiene: a new bioactive diterpene from Mycobacterium tuberculosis. J. Am. Chem.Soc. 131, 17526–17527 (2009).
Sani, M. et al. Direct visualization by cryo-EM of the mycobacterial capsular layer: a labile structure containing ESX-1-secreted proteins. PLoS Pathog. 6, e1000794 (2010).
Samanovic, M. I. et al. Proteasomal control of cytokinin synthesis protects Mycobacterium tuberculosis against nitric oxide. Mol. Cell 57, 984–994 (2015).
Deretic, V. Autophagy in tuberculosis. Cold Spring Harb. Perspect. Med. 4, a018481 (2014).
van der Wel, N. et al. M. tuberculosis and M. leprae translocate from the phagolysosome to the cytosol in myeloid cells. Cell 129, 1287–1298 (2007).
Murphy, K. C., Papavinasasundaram, K. & Sassetti, C. M. Mycobacterial recombineering. Methods Mol. Biol. 1285, 177–199 (2015).
Guinn, K. M. et al. Individual RD1-region genes are required for export of ESAT-6/CFP-10 and for virulence of Mycobacterium tuberculosis. Mol. Microbiol. 51, 359–370 (2004).
The authors thank H. van Veen and W. Tigchelaar for EM, P. Reinink for phylogenetic graphs and S. Suliman for advice. Work was supported by grant nos. AI116604 (to D.B.M. and N.N.v.d.W.), AI111224 (to D.B.M. and M.M.), GM065307 (to E.O.), CA158191 (to E.O.) and AI114952 (to S.T.), the Dutch Science Foundation NWO-VICI 70.57.443 (to A.J.M.) and a Canadian Institute of Health Research Foundation grant 148362 (to M.A.B.).
The authors declare no competing interests.
Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
About this article
Cite this article
Buter, J., Cheng, TY., Ghanem, M. et al. Mycobacterium tuberculosis releases an antacid that remodels phagosomes. Nat Chem Biol 15, 889–899 (2019). https://doi.org/10.1038/s41589-019-0336-0
This article is cited by
Nature Reviews Microbiology (2022)
Cellular & Molecular Immunology (2020)
Nature Chemical Biology (2019)