Rv2780 inhibits the expression of AMPs
HEK293T cell is widely used to study the function of pathogenic bacteria secretory proteins on the activation of host NF-κB and MAPKs signal, indicating the existence of integral immune molecules of these two pathways in HEK293T cells23,24. We also detected endogenous β-Defensin 4 (DEFB4) mRNA level in HEK293T cells to verify AMP DEFB4 expression at baseline (Supplementary Fig. 1A). To identify M. tuberculosis proteins that inhibit the expression of AMPs, we transfected HEK293T cells with plasmids encoding 201 M. tuberculosis secreted proteins or lipoproteins25 and examined their effects on the expression of DEFB4 using reverse transcription (RT)-PCR (Supplementary Fig. 1B; Supplementary Table 1). Rv2780, a secreted alanine dehydrogenase26,27 of M. tuberculosis, was found to reduce the mRNA levels of several AMPs including not only DEFB4 but also β-Defensin 3 (DEFB3) and Cathelicidin Antimicrobial Peptide (CAMP), as measured by RT-PCR assay (Supplementary Fig. 1C-E). Rv2780 was detected in both the supernatants and lysates of M. tuberculosis cultures (Supplementary Fig. 1F, G), illustrating that Rv2780 is a secreted protein. In addition, Rv2780 was detected in the cytoplasm of mice peritoneal macrophages (MPMs) and A549 cells during Mtb infection (Supplementary Fig. 1H, I), suggesting Rv2780 could be secreted to host cells. However, compared to A549 cells, much more abundant Rv2780 protein was detected in H37Rv infected macrophages (Supplementary Fig. 1H), suggesting a more powerful function of Rv2780 in macrophages.
To further evaluate whether Rv2780 inhibits the expression of AMPs during M. tuberculosis infection, we deleted Rv2780 from an M. tuberculosis H37Rv strain, thus generating an H37RvΔRv2780 strain (Supplementary Fig. 1F, G). Consistent with previous report28,29, Rv2780 did not significantly change in vitro H37Rv growth in aerobic condition or fitness to hypoxic condition (Supplementary Fig. 1J, K). Electronic scanning microscopy analysis showed the similar morphology of H37RvΔRv2780 and H37Rv strain (Supplementary Fig. 1L). Rv2779c is an Lrp/AsnC family transcriptional factor that binds amino acid ligands to regulate Rv2780 expression30,31. Deletion of Rv2780 in H37Rv strain dramatically decreased Rv2780 expression but did not significantly change Rv2779c expression (Supplementary Fig. 1M, N). Besides, alanine level was significantly increased in H37RvΔRv2780 strain (Supplementary Fig. 1O), suggesting that Rv2780 may function as an alanine dehydrogenase in M. tuberculosis.
Macrophages, which serve as both habitats for and the first line of defense against M. tuberculosis, were infected with the H37Rv or H37RvΔRv2780 strain. Primary peritoneal macrophages infected with H37Rv showed limited increases in the expression of Defb4 (9.63-fold), Defb3 (5.67-fold) and Camp (3.79-fold) at 24 hours post infection (Fig. 1A and Supplementary Fig. 2A, B). However, H37RvΔRv2780 was associated with much higher induction of the mRNA of Defb4 (21.08-fold), Defb3 (16.94-fold) and Camp (10.41-fold) than in cells infected with wild-type H37Rv for 24 hours (Fig. 1A and Supplementary Fig. 2A, B). Complementation of H37RvΔRv2780 with Rv2780 restored the ability of M. tuberculosis to suppress the expression of Defb4, Defb3 and Camp (Fig. 1B and Supplementary Fig. 2C, D). Taken together, these results suggest that M. tuberculosis Rv2780 may inhibit the expression of AMPs.
Antimicrobial peptides kill bacteria directly in vitro and are crucial for macrophages to limit the intracellular survival of M. tuberculosis11–15. We also examined direct killing effects of AMPs on M. tuberculosis as described previously by Liu et al11, and found that the MIC of Defb4, Defb3 and Camp were at 0.01 µg/ml, 10 µg/ml and 0.1 µg/ml, respectively, suggesting that these AMPs may have the anti-M. tuberculosis activity in vitro (Supplementary Fig. 2E). To examine whether Rv2780 regulates the intracellular survival of M. tuberculosis, we infected primary peritoneal macrophages with H37Rv or H37RvΔRv2780 strains and measured the survival rate of intracellular M. tuberculosis using a colony forming unit (CFU) assay. H37RvΔRv2780 showed much lower CFU counts in macrophages at 12- and 24-hours post-infection than H37Rv (Fig. 1C, D), suggesting that Rv2780 may be essential for the intracellular survival of M. tuberculosis. ROS production and xenophagy were also shown to restrict the growth intracellular M. tuberculosis32, however deletion of Rv2780 did not significantly change ROS production and xenophagy during Mtb infection in macrophages (Supplementary Fig. 2F-H). To further investigate the functional relevance of Rv2780 in the in vivo pathogenesis of M. tuberculosis infection, we challenged C57BL/6 mice with wild-type H37Rv, H37RvΔRv2780 or H37Rv(ΔRv2780 + Rv2780) for 28 days. The bacterial burden in the lung tissues of mice infected with H37RvΔRv2780 was much lower (decreased 1.26-fold in log10) than mice infected with H37Rv (Fig. 1E, F). Consistent with this, lung tissues from mice infected with H37RvΔRv2780 showed less immune-cell infiltration and fewer inflammatory lesions than those from mice infected with H37Rv (Fig. 1F, G). Together, these results suggest that Rv2780 is an essential virulence factor of M. tuberculosis.
Rv2780 dehydrogenates L-alanine
Rv2780 encodes L-alanine dehydrogenase, an enzyme that catalyzes the NAD+-dependent interconversion of alanine and pyruvate26,27 (Fig. 2A). The enzymatic kinetics of Rv2780 was assesses by analyzing the enzymatic product pyruvate. The Km and Vmax were found to be 0.964 mM and 111.8 M/s, respectively (Fig. 2B). Another in vitro alanine dehydrogenation assay showed that the addition of purified recombinant wild-type Rv2780 led to the greater production of NADH from alanine (Fig. 2C), suggesting that Rv2780 has the alanine dehydrogenase activity.
By performing gas chromatography-mass spectroscopy analysis of metabolites in sera of C57BL/6 mice infected with M. tuberculosis H37Rv, we found that the level of alanine was markedly reduced in sera of infected mice (Fig. 2D; Supplementary Table 2). By contrast, other amino acids such as methionine, phenylalanine and aspartic acid were not significantly changed in response to H37Rv infection (Fig. 2E and Supplementary Fig. 3A-C), suggesting that the decreased alanine level may be specifically caused by M. tuberculosis infection rather than food intake or metabolism. Moreover, smear-positive patients with TB had much lower level of alanine in their plasma than healthy people (Fig. 2F). This is consistent with a previous report showing that alanine was one of the metabolites showing the greatest decrease in a 1H nuclear magnetic resonance spectroscopy-based metabolomic analysis of sera from TB patients33. Host alanine aminotransferase (ALT) can catalyze the reversible interconversion of L-alanine and 2-oxoglutarate to pyruvate and L-glutamate34. Therefore, we next analyzed the relationship between alanine level and ALT in sera of TB patients. However, as shown in Supplementary Fig. 3D, no significant correlation between alanine and ALT was noted in patients with TB. Together, the decrease of alanine level in Mtb-infected mice and TB patients might be mediated by Mtb infection.
Structural analysis of Rv2780 revealed two typical alanine dehydrogenase activity sites at histidine 96 (H96) of the catalytic domain and aspartic acid 270 (D270) of the NAD+ binding domain, which are highly conserved across different bacterial species (Supplementary Fig. 3E). Mutation of two active sites on Rv2780 (Rv2780DM, with H96A and D270A) impaired its alanine dehydrogenase activity (Fig. 2C). Overexpression of wild-type Rv2780, but not its inactive mutant Rv2780DM markedly decreased the level of L-alanine in both HEK293T and A549 cells (Supplementary Fig. 3F, G). Moreover, the level of alanine was gradually reduced in H37Rv or H37Rv(ΔRv2780 + Rv2780) infected macrophages, but infection of H37RvΔRv2780 or H37Rv(ΔRv2780 + Rv2780DM) led to much more abundant alanine in the infected cells (Fig. 2G, H). Especially, complementation of H37RvΔRv2780 with wild-type Rv2780, rather than Rv2780DM mutant significantly decreased alanine in lung tissues and sera from H37Rv infected mice at 7- and 28-days post-infection (Fig. 2I, J). These results suggest that Mtb may have evolved a metabolic ability to dehydrogenate L-alanine via Rv2780 in host cells and can therefore reduce the alanine level in eukaryotes.
Rv2780 suppress AMPs by dehydrogenating alanine.
Given that Rv2780 decreased both the level of L-alanine and expression of AMPs, we hypothesized Rv2780 might suppress AMPs expression through L-alanine dehydrogenation. To verify the effect of L-alanine on AMPs expression, we supplemented macrophages with L-alanine before Mtb infection. Addition of L-alanine significantly increased mRNA levels of Defb4 (27.95-fold), Defb3 (9.97-fold) and Camp (8.57-fold) in macrophages infected with M. tuberculosis H37Rv for 24 hours (Fig. 3A and Supplementary Fig. 4A, B). Rv2780 also shows glycine dehydrogenase activity in vitro35,36. We supplemented Rv2780-overexpressed HEK293T cell with L-alanine or glycine, and ELISA analysis was performed to determine the protein level of Defb4 and Camp37,38. Administration of alanine rather than glycine rescued the Rv2780-mediated inhibition of AMPs expression (Supplementary Fig. 4C-E). Moreover, only supplementation with L-alanine, but not D-alanine or glycine increased Defb4 and Camp protein level in response to H37Rv infection (Fig. 3B and Supplementary Fig. 4F). These results suggest that Rv2780 may inhibit AMPs expression through its alanine dehydrogenase activity
To further examine whether Rv2780 suppresses the AMPs by its dehydrogenase activity, we infected murine peritoneal macrophages with H37Rv(ΔRv2780 + Rv2780) or H37Rv(ΔRv2780 + Rv2780DM) and examined the protein level of Camp and Defb4. Only the H37RvΔRv2780 strain complemented with wild-type Rv2780, but not with Rv2780DM, restored the ability of M. tuberculosis to suppress the production of Defb4 and Camp (Fig. 3C and Supplementary Fig. 4G). Consistently, H37Rv(ΔRv2780 + Rv2780), but not H37Rv(ΔRv2780 + Rv2780DM), rescued an Rv2780-mediated increase in the intracellular survival of M. tuberculosis within primary macrophages (Fig. 3D, E). In addition, deletion of Rv2780 also reduced M. tuberculosis survival in alveolar macrophages or neutrophils, and the reduced survival of H37RvΔRv2780 was rescued by the complementation of Rv2780, but not Rv2780DM (Supplementary Fig. 4H-K). These results suggest that M. tuberculosis Rv2780 may suppress the expression of AMPs, thus promoting M. tuberculosis intracellular survival by its alanine dehydrogenase activity.
To examine whether Rv2780 increased M. tuberculosis intracellular survival via inhibiting AMPs, we infected Defb4−/− mice peritoneal macrophage with H37Rv and H37RvΔRv2780. We found that deletion of Defb4 markedly increased intracellular survival of H37Rv, and eliminated the enhanced effects of Rv2780 on intracellular survival of H37Rv (Fig. 3F, G). Moreover, we infected Defb4 knockout mice with H37Rv or H37RvΔRv2780 to further validate in vivo relevance of Rv2780 and Defb4. Knockout of Defb4 markedly increased bacterial burden and pathological damages in lung tissues of the M. tuberculosis H37Rv-infected mice, and abolished the increased bacterial burden and pathological damages by Rv2780 in lung tissues of the Mtb-infected mice (Fig. 3H-J). Above all, these data suggested that Rv2780 may increase the survival of M. tuberculosis through suppressing the expression of AMPs.
L-Alanine interacts with PRSS1 to induce AMPs
We next investigated the mechanism underlying the induction of AMPs by L-alanine. By performing biotin-streptavidin pull-down combined with mass spectrometry analyses39,40 (Fig. 4A and Supplementary Fig. 5A; Supplementary Table 3), we found that cationic trypsinogen (protease serine 1, PRSS1), encoded by a susceptibility gene associated with chronic pancreatitis41, interacted with L-alanine, but not with D-alanine (Fig. 4B, C), and non-biotinylated L-alanine could competitively elute biotinylated L-alanine from PRSS1 (Supplementary Fig. 5B). PRSS1 is a serine protease composed of the N-terminal alpha-trypsin chain 1 and C-terminal chain 2 that are linked by a disulfide bond42,43 (Supplementary Fig. 5C). Only the N-terminal alpha-trypsin chain 1, not C-terminal chain 2, of PRSS1 interacted with L-alanine (Supplementary Fig. 5D). Surface plasmon resonance (SPR) assay revealed that L-alanine strongly interacted with PRSS1 (KD = 2.6×10− 3 M) (Fig. 4D). These results suggest that L-alanine may interact with PRSS1.
It has been shown that M. tuberculosis infection induces the expression of AMPs through the TLR2/NF-κB signaling pathway11. Upon stimulation, the ubiquitin ligase, TRAF6, which is downstream of the TLR2 receptor, induces TAK1 oligomerization-dependent auto-phosphorylation and TAK1 subsequently activates the IKK-mediated NF-κB signaling pathway44,45. We next examined whether PRSS1 had any effect on activation of NF-κB using a luciferase reporter gene assay. As shown in Supplementary Fig. 5E, the overexpression of PRSS1 markedly suppressed the activation of NF-κB by TRAF6 or TAK1, but not that mediated by IKKα/β, suggesting that PRSS1 may block the activation of NF-κB signaling by acting at downstream of the TAK1 complex and upstream of IKKα/β.
To elucidate the mechanism underlying the inhibition of NF-κB signaling by PRSS1, we examined the interactions between PRSS1 and TLR pathway signaling molecules. PRSS1 was found to interact with TAK1, which is co-expressed with TAB1 in HEK293T cells (Fig. 4E). The interaction between TAK1 and TAB1 is important for the activation of TAK146. In HEK293T cells, PRSS1 markedly impeded the interaction between TAK1 and TAB1 and consequently inhibited the enhanced phosphorylation of TAK1 by TAB1 (Fig. 4F, G). Moreover, enhanced formation of TAK1-TAB1 complex was found in Prss1+/− peritoneal macrophages (Fig. 4H), suggesting that PRSS1 may disrupt formation of the TAK1-TAB1 complex. Lastly, deletion of Rv2780 markedly increased the phosphorylation of p65, but treatment of TAK1 inhibitor ((5Z)-7-oxozeaenol, 5Z-7Ox) eliminated the reduced phosphorylation of p65 by Rv2780. (Supplementary Fig. 5F). Consistently, inhibition of p65 phosphorylation by Rv2780 was not observed in Prss1+/− macrophages (Fig. 4I). These results suggest that Rv2780 may inhibit NF-κB signaling via PRSS1 and TAK1 during Mtb infection.
We next investigated the role of PRSS1 in the regulation of AMPs. Prss1+/− macrophages had much higher mRNA levels of AMPs than wild-type cells infected with M. tuberculosis H37Rv, suggesting PRSS1 as a potent negative regulator of AMPs expression (Fig. 4J and Supplementary Fig. 5G, H). To validate the in vivo role of Prss1 in macrophages, we also generated macrophage conditional Prss1 knockout mice (Lyz2crePrss1floxp/floxp mice). Accordingly, Lyz2crePrss1floxp/floxp mice exhibited decreased lung bacterial burden and tissues damages compared with Prss1floxp/floxp mice (Fig. 4K-M). These results suggest that PRSS1 may inhibit the induction of AMPs, and negatively regulates anti-TB immunity.
To further examine the functional relevance of PRSS1 and L-alanine, peritoneal macrophages from wild-type or Prss1+/− mice were treated with L-alanine followed by infection with M. tuberculosis H37Rv. Prss1+/− MPMs infected with M. tuberculosis H37Rv had much lower intracellular CFU than those WT counterparts (Fig. 4N, O), indicating that PRSS1 may promote the intracellular survival of M. tuberculosis through suppressing AMPs. Moreover, L-alanine significantly inhibited the intracellular survival of M. tuberculosis H37Rv in WT macrophages, but not in Prss1+/− peritoneal macrophages (Fig. 4N, O), suggesting that L-alanine may restrict the intracellular growth of M. tuberculosis through PRSS1.
Supplementation of L-alanine enhances anti-TB immunity
Above all, we aim to test the effect of L-alanine on clearance of Mtb inside macrophages. The growth of the M. tuberculosis H37Rv strain in vitro was not significantly affected by L-alanine treatment (Supplementary Fig. 6A, B). However, treatment with L-alanine dramatically inhibited the intracellular survival of M. tuberculosis H37Rv at an efficient level equivalent to that of the best-in-class antibiotic rifampicin (RIF)47 and a combination of L-alanine and RIF resulted in an even lower bacterial burden compared with either agent alone (Fig. 5A, B), suggesting that L-alanine could be used to complement first-line anti-TB drugs. Moreover, L-alanine efficiently killed a clinical multiple-drug-resistant (MDR) M. tuberculosis strain in macrophages (Fig. 5C, D). No significant effect on cell viability was observed of L-alanine (Supplementary Fig. 6C). These results suggest that L-alanine may act as an efficient host-directed inhibitor of M. tuberculosis, particularly for drug-resistant M. tuberculosis for which current antibiotics are largely ineffective.
Since L-alanine was a strong inducer of AMPs that restrict the intracellular survival of M. tuberculosis, while M. tuberculosis infection substantially reduced the level of alanine in host immune cells, we next addressed the therapeutic effectiveness of L-alanine in vivo. In severe combined immunodeficient (SCID) mice model48, mice given L-alanine lived much longer, suggesting L-alanine functions in an innate immunity-dependent way (Fig. 5E). C57BL/6 mice challenged with H37Rv were given double-distilled water or that containing 30 mg/mL L-alanine or D-alanine, and their lungs examined by histopathology and for bacterial burden. Upon M. tuberculosis H37Rv infection, mice supplemented with L-alanine, but not D-alanine, had less histological damage in their lungs than mice given double-distilled water alone (mock) (Fig. 5F-H). Similarly, the bacterial burden in the lungs of H37Rv-infected mice treated with L-alanine was also much lower (decreased 1.332-fold in log10) than control mice. These results suggest that L-alanine may inhibit the pathogenesis of M. tuberculosis infection in vivo.
Targeting Rv2780 inhibits the growth of mycobacteria in vivo
The crystal structure of the M. tuberculosis Rv2780 (PDB code: 2VHX) with NAD+ binding domain was used for structure-based virtual screening of commercial databases (Locator Library and MCE Compound Library), which contain 309,800 inhibitors. As shown in Fig. 6A-C, a small-molecule compound, (S)-N-(5-(3-fluorobenzyl)-1H-1,2,4-triazol-3-yl) tetrahydrofuran-2-carboxamide (GWP-042), bound to Rv2780, forming four hydrogen bonds, one cation - π interaction and multiple hydrophobic interactions (Supplementary Table 4). Localized SPR assay revealed that GWP-042 interacted strongly with Rv2780. The equilibrium dissociation constant (KD) of GWP-042 to Rv2780 was 1.896×10− 5 M, nearly 3–10 folds lower than other reported anti-tuberculosis drug to their target protein49,50 (Fig. 6D). To further clarify whether GWP-042 inhibits the activity of Rv2780, we measured the hydrogenase activity of Rv2780 in the presence of increasing concentrations of GWP-042. By measuring the enzymatic production of pyruvate that reflects the enzyme activity of Rv2780, the IC50 of GWP-042 on Rv2780 was 0.21 ± 0.05 µM as indicated by pyruvate (Fig. 6E), which is almost 100 folds lower than the reported Rv2780 inhibitors51. These data suggested that GWP-042 may act as a powerful inhibitor of Rv2780.
Mycobacterium marinum, a pathogen of zebrafish that is the closest genetic relative of the M. tuberculosis organism complex52, possesses a conserved homologue of alanine dehydrogenase (Rv2780) (Supplementary Fig. 3E). Zebrafishes have an antimicrobial peptide system53 and have been used as a powerful host–pathogen system for characterizing anti-mycobacterial compounds54,55. From the top 15 compounds of the docking study with the best docking scores, GWP-042 was found to be the most effective inhibitor to restrict the growth of M. marinum in zebrafish larvae (Supplementary Fig. 7A, B), but showed no significant effect on the growth rate of M. marinum in vitro (Supplementary Fig. 7C, D).
One hallmark of TB is the formation of caseous necrotic granulomas56, which are organized aggregates of macrophages and other immune cells that serve as niches for the bacteria to obtain nutrients or evade anti-TB immunity, and to provide a source for mycobacteria for later reactivation and dissemination57,58. Respiration-inhibiting conditions, such as hypoxia, nitric oxide, low pH and nutrient starvation, are assumed to be characteristics of TB granulomatous lesions. Expression of the ald gene is upregulated under oxygen-limiting, nutrient starvation and nitrogen monoxide (NO) conditions28,59–61. The growth rate of M. marinum under hypoxia was not significantly affected by the treatment of GWP-042 (Supplementary Fig. 7D). Moreover, adult zebrafish treated with GWP-042 had a much lower bacterial burden of wild-type and rifampicin resistant M. marinum at 14 days post-infection (Supplementary Fig. 7E, F). These results suggest that targeting mycobacterial alanine dehydrogenase may inhibit the growth of pathogenic mycobacteria in granulomas.
In murine peritoneal macrophages infected with wild-type H37Rv, the addition of GWP-042 increased the production of Defb4 and Camp ; but the increases were not observed upon infection with H37RvΔRv2780 strains (Fig. 6F and Supplementary Fig. 7G). These results suggest that GWP-042 may increase the AMPs by targeting Rv2780. Consistent with this, GWP-042 dramatically inhibited the intracellular survival of both M. tuberculosis H37Rv and a clinical MDR strain in infected macrophages (Supplementary Fig. 7H-K). However, treatment with GWP-042 showed no significant effect on the in vitro growth curve of M. tuberculosis H37Rv or MDR M. tuberculosis (Supplementary Fig. 7L-O), suggesting that GWP-042 may exert its anti-mycobacterial effect through targeting host anti-TB pathways. Moreover, the deletion of Rv2780 almost eliminated the inhibitory effect of GWP-042 on the growth of intracellular M. tuberculosis (Fig. 6G, H), indicating that GWP-042 may exert its anti-mycobacterial activity through inhibiting Rv2780. Furthermore, GWP-042 showed no significant effect on the viability of cells even at very high concentrations (Supplementary Fig. 7P). Together, these results suggest that targeting Rv2780 has potential as a host-directed candidate for the therapeutic treatment of TB, especially drug-resistant TB.
We further evaluated the pharmacokinetic properties of GWP-042. As shown in Supplementary Table 5, the half-life of GWP-042 was 2.07 and 2.25 hours when C57BL/6 mice were treated by intravenous injection (10 mg/kg) and intragastric administration (100 mg/kg), respectively. We also observed a high maximal concentration (Cmax = 7237 ng/mL for intravenous injection and Cmax = 45425 ng/mL for o.p) and a good bioavailability of 80.79% when GWP-042 was given orally. A clearance of 8.68 mL/min/kg suggested the metabolic stability of GWP-042 was good. An in vivo toxicity study of GWP-042 was performed in C57BL/6 mice (Supplementary Table 6). No mice died after receiving 50 or 200 mg/kg by intragastric administration. When the dosage was raised up to 1000 mg/kg, two of three mice died. No significant change of bodyweight was observed for C57BL/6 mice administrated with GWP-042 at 50mg/kg once by oral gavage for 14 days (Supplementary Fig. 7Q), indicating that GWP-042 was nontoxic. Furthermore, when treated with GWP-042, the lung tissues of C57BL/6 mice infected with H37Rv had much lower bacteria burden than those mice treated with rifampicin (Fig. 6I-K), indicating that the killing effect of GWP-042 against M. tuberculosis alone is better than rifampicin. These results suggest that targeting mycobacterial alanine dehydrogenase may inhibit the growth of pathogenic mycobacteria in vivo. This is consistent with our conjecture that the mechanism of GWP-042 activity differs from that of traditional anti-TB drugs, which directly target M. tuberculosis itself; GWP-042 may resuscitate host immunity to eliminate M. tuberculosis.