Introduction
Metazoan cells inhabit environments frequently subjected to mechanical stress, such as the stress that occurs in skin, gut and muscle1, or as a consequence of interactions with pathogens2, 3, which can result in plasma membrane lesions. To ensure survival, cells rapidly repair membrane damage. Resealing of the plasma membrane is a ubiquitous and highly conserved process based on the exocytosis of endomembranes1, 4. Although Golgi-derived vesicles have been linked to membrane repair5, the most thoroughly studied secretory vesicles involved in plasma membrane repair resemble lysosomes6. The exocytosis of lysosomes is induced by calcium and depends on the function of the calcium sensor synaptotagmin 7 (Syt-7; A002565)7, 8, 9. Whereas Syt-7 is the calcium sensor of the lysosome7, 10, neuronal calcium sensor-1 (NCS-1; A000957) is the main calcium sensor of the Golgi membranes11, 12 and is involved in vesicle trafficking from the trans-Golgi network13.
Infection with Mycobacterium tuberculosis (Mtb), the causative agent of tuberculosis and the main source of mortality from chronic pulmonary bacterial infection9, occurs in the lung through phagocytosis of the pathogens by pulmonary macrophages. After infection, virulent Mtb blocks phagosome maturation by interrupting acidification and lysosome fusion, which creates a protected niche in the cell for bacterial replication14. Ultimately, intracellular infection with virulent Mtb leads to macrophage death by necrosis, a process characterized by plasma membrane lysis and escape of the pathogens into the surrounding tissue for a new cycle of infection. In contrast, avirulent strains of Mtb induce apoptosis, a process that leads to sequestration and killing of intracellular bacilli and acts as a bridge from the innate immune response to the adaptive immune response15.
The underlying mechanisms by which virulent Mtb induces necrosis or inhibits apoptosis in macrophages remain largely unknown. The host lipid mediators prostaglandin E2 (PGE2) and lipoxin A4 (LXA4) exert opposing effects on the modality of Mtb-induced cell death in macrophages16. Macrophages infected with attenuated Mtb produce only small amounts of LXA4 and instead elaborate prostanoids, including PGE2, that protect against macrophage necrosis and promote apoptosis. In contrast, virulent Mtb infection induces production of LXA4, which inhibits PGE2 synthesis and apoptosis and leads to macrophage necrosis. These eicosanoids are also important in vivo, as mice deficient in 5-lipoxygenase (5-LO; Alox5-/- mice) that are unable to synthesize LXA4 are more resistant to chronic infection with virulent Mtb17. In contrast, prostaglandin E synthase–deficient (Ptges-/-) mice, which are unable to produce PGE2, are more susceptible to virulent Mtb16. The potential importance of the 5-LO pathway in humans is emphasized by the association people with 5-LO alleles that have lower 5-LO activity and a lower risk of tuberculosis18.
Mtb is endowed with the specialized protein secretion system ESX-1, which is a type VII secretion system19. ESX-1 secretion is thought to be critical in pore formation in host cell membranes3, 20. We therefore considered that virulent Mtb induces macrophage necrosis by disruption of the plasma membrane and inhibition of lesion repair. As embryonic fibroblasts from Syt-7-deficient mice are defective in lysosomal exocytosis and resealing of plasma membrane lesions21, we further considered that Syt-7 is a lysosomal component needed for calcium-dependent exocytosis and repair of plasma membrane lesions in macrophages infected with Mtb.
Here we report that plasma membrane microdisruptions induced by attenuated Mtb were rapidly resealed by a repair mechanism that depended on recruitment of lysosomal and Golgi apparatus–derived membranes and resulted in apoptosis of infected macrophages. In contrast, virulent Mtb inhibited membrane repair and induced necrosis of the infected macrophage. Lysosome-dependent membrane repair was promoted by PGE2 and, in the absence of PGE2, infected macrophages were unable to control bacterial replication. Syt-7 is a critical gene product regulated by PGE2, as in its absence, infected macrophages underwent necrosis and were unable to control Mtb growth.
Results
Virulent Mtb causes persistent membrane microdisruptions
To determine whether Mtb induces plasma membrane disruptions, we evaluated the permeability of infected macrophages to FDX, a 75-kilodalton, inert, impermeant fluorescent molecule that enters the cytoplasm through membrane lesions22, 23. Beginning at 12 h after infection, there was significant influx of FDX into macrophages infected with virulent Mtb (strain H37Rv) and the FDX influx gradually increased with time. In contrast, there was significantly less influx of FDX into macrophages infected with avirulent Mtb (strain H37Ra; Fig. 1a). To exclude the possibility that enhanced accumulation of intracellular FDX was due to greater uptake of FDX by pinocytosis, we incubated macrophages with cytochalasin B, an inhibitor of pinocytosis; this treatment did not alter FDX influx after H37Rv infection (data not shown). These data collectively indicate that 12 h after infection with virulent H37Rv, persistent membrane lesions developed in infected macrophages.
Figure 1: Infection of human macrophages with virulent H37Rv inhibits lysosomal and Golgi-mediated plasma membrane repair.
(a) Kinetics of FDX influx through membrane lesions in macrophages left uninfected or infected with H37Ra or H37Rv. (b) Translocation of LAMP-1 to the plasma membrane lesions of human macrophages left uninfected or infected for 12 h with H37Ra (Ra) or H37Rv (Rv). Shaded histograms, isotype control; open histograms, specific LAMP-1 staining; above plots (in parentheses), multiplicity of infection (MOI). Numbers in histograms indicate the mean fluorescence intensity (MFI) of the entire cell population. (c) Kinetics of the translocation of LAMP-1 to the surface of human macrophages infected with H37Ra or H37Rv. (d) Translocation of LAMP-1 and Syt-7 to the surface of macrophages left uninfected or infected for 12 h with H37Ra or H37R. (e) Kinetics of the translocation of mannosidase II (Mannos II) as described in c. (f) Translocation of mannosidase II and the endoplasmic reticulum marker BiP as described in d; below, some macrophages were made permeable and/or stained with irrelevant antibody (Anti-BiP -) or BiP-specific antibody (Anti-BiP +). MOI, 10 (a–c,e) or as noted below graphs (d,f). *P < 0.05 (two-way analysis of variance (ANOVA; a) or t-test (c–f)). Data are representative of at least three independent experiments (error bars, s.e.m.).
Full size image (48 KB)The findings reported above could suggest that the membrane lesions inflicted by H37Ra were quickly resealed and that the lesions caused by H37Rv remained unrepaired. To determine whether lysosome recruitment is involved in membrane repair of H37Ra-infected macrophages and, if so, whether transport of lysosomal membranes to the cell surface is inhibited in H37Rv-infected macrophages, we measured translocation of LAMP-1, a specific marker of late endosomes and lysosomes, to the cell surface24. Macrophages infected with H37Ra showed significant translocation of LAMP-1 to the cell surface, which was visible as early as 3 h and was maximal by 12 h after infection (Fig. 1b,c and Supplementary Fig. 1). In contrast, we noted little or no translocation of LAMP-1 in macrophages infected with virulent H37Rv.
Lysosomal trafficking and exocytosis is dependent on Syt-7, the calcium-sensing protein located on late endosomes and lysosomes2, 25. Therefore, we investigated whether there was more cell surface Syt-7 after macrophage infection with avirulent H37Ra. Syt-7 expression on the macrophage surface was significantly greater after infection with H37Ra but not after infection with H37Rv (Fig. 1d and Supplementary Fig. 1). Because Golgi-derived membranes are also linked to plasma membrane resealing5, we next measured translocation of mannosidase II, a Golgi marker26, to the cell surface. Mannosidase II expression increased on the macrophage plasma membrane starting at 12 h after H37Ra infection, but H37Rv-infected macrophages maintained low cell surface expression of mannosidase II (Fig. 1e,f and Supplementary Fig. 1). Conversely, endoplasmic reticulum–derived membranes were not involved in membrane repair, as the endoplasmic reticulum marker GRP78 (also known as BiP)27, 28 was not recruited to the macrophage surface after infection with either H37Ra or H37Rv (Fig. 1f).
To determine whether the higher expression of lysosomal and Golgi markers on the surface of infected macrophages was due to greater protein synthesis, we measured total mannosidase II, LAMP-1 and annexin-1 protein in macrophages infected with H37Ra or H37Rv. Annexin-1 is a phospholipid-binding anti-inflammatory protein present in many cell types. The total amount of these proteins were not altered in macrophages infected with either H37Ra or H37Rv (Supplementary Fig. 2), which indicated that the differences in redistribution of the lysosomal and Golgi membrane compartments were not due to differences in total protein synthesis. These results collectively suggest that membranes from both the lysosomal and Golgi compartments are involved in macrophage plasma membrane resealing during avirulent mycobacterial infection.
Calcium sensors in plasma membrane repair
We next determined the function of calcium sensors in the recruitment of lysosomal and Golgi membranes to the cell surface of infected macrophages. Silencing of Syt-7 expression mediated by small interfering RNA (siRNA) impaired recruitment of lysosomal membranes to the macrophage surface after H37Ra infection (Fig. 2a,b). However, silencing of Syt-7 did not diminish and instead increased translocation of Golgi membranes to the macrophage surface (Fig. 2b), possibly due to a compensatory mechanism.
Figure 2: Distinct calcium sensors regulate the recruitment of lysosomal and Golgi apparatus–derived membranes in Mtb-infected human macrophages.
![Figure 2 : Distinct calcium sensors regulate the recruitment of lysosomal and Golgi apparatus|[ndash]|derived membranes in Mtb-infected human macrophages.](/ni/journal/v10/n8/thumbs/ni.1758-F2.jpg)
(a) Immunoblot analysis of the expression of Syt-7 and NCS-1 before (-) and after gene silencing in human macrophages with targeting (T) or nontargeting (NT) siRNA. (b,c) Influence of Syt-7-specific siRNA (b) and brefeldin A (BFA; c) on the translocation of LAMP-1, mannosidase II, phosphatidylserine (PS) and annexin-1 to the surface of H37Ra-infected macrophages. (d) Influence of NCS-1–specific siRNA on the translocation of mannosidase II, phosphatidylserine and annexin-1 to the surface of H37Ra–infected macrophages. (e) Influx of FDX into H37Ra-infected macrophages expressing Syt-7- or NCS-1-specific siRNA or treated with BFA. (f) Necrosis of H37Ra-infected macrophages expressing Syt-7- or NCS-1-specific siRNA, assessed by staining with 7-amino-actinomycin D (7-AAD). MOI, 10. *, P < 0.05 (t-test). Data are from one representative of three independent experiments (error bars, s.e.m.).
Full size image (95 KB)Phosphatidylserine and annexin-1 appear on the plasma membrane early during apoptosis to enable formation of the apoptotic envelope in macrophages infected with attenuated Mtb29; this process is impaired in macrophages infected with virulent H37Rv30. We therefore investigated whether translocation of Golgi membranes or lysosomal vesicles to the macrophage membrane surface was required for phosphatidylserine exocytosis and recruitment of annexin-1 to the macrophage surface. Syt-7-specific siRNA, which decreased translocation of LAMP-1 to the macrophage surface, enhanced rather than decreased expression of phosphatidylserine and annexin-1 on the macrophage surface (Fig. 2b). In contrast, brefeldin A, a highly specific inhibitor of Golgi membrane recruitment31, blocked the translocation of mannosidase II, phosphatidylserine and annexin-1 to the surface of macrophages infected with H37Ra in a dose-dependent way (Fig. 2c). In these conditions, translocation of LAMP-1-containing lysosomal membranes to the macrophage surface was not altered.
The experiments reported above suggested that Golgi apparatus–derived vesicles are recruited to the cell surface independently from lysosomal vesicles and that recruitment of the Golgi-derived membranes depends on a calcium sensor different from Syt-7. NCS-1 is a member of the EF family that has a calcium-binding motif12, 32 and is especially abundant in Golgi apparatus–derived vesicles13. NCS-1-specific siRNA (Fig. 2a) had an effect similar to brefeldin A, in that it resulted in inhibition of the translocation of Golgi membranes and significantly inhibited the translocation of phosphatidylserine and annexin-1 (Fig. 2d). These findings indicate that both lysosomal and Golgi–derived membranes move independently to the plasma membrane in infected macrophages.
To determine whether lysosome recruitment and Golgi membrane–derived vesicle recruitment are both important for the repair of plasma membrane damage, we tested whether lysosomal and Golgi membrane translocation were both required to prevent the influx of FDX into macrophages infected with H37Ra. Treatment with Syt-7-specific or NCS-1-specific siRNA or brefeldin A led to significantly greater influx of FDX into H37Ra-infected macrophages (Fig. 2e). In addition, Syt-7–specific and NCS-1-specific siRNA promoted macrophage necrosis after infection with H37Ra (Fig. 2f). Our data did not support the alternative explanation that instead of being quickly resealed, plasma membrane lesions are not generated by avirulent H37Ra (Fig. 2e,f). H37Ra caused considerable microdisruption of the plasma membrane when repair was inhibited. These data indicate that recruitment of lysosomal and Golgi membrane–derived vesicles is critical in the repair of plasma membrane damage after Mtb infection and is required to prevent necrosis.
PGE2 promotes plasma membrane repair
The membrane-resealing process in fibroblasts is dependent on cAMP33, 34. We therefore investigated whether upregulation of cAMP concentrations is sufficient to trigger membrane repair. Treatment with forskolin, an activator of adenylate cyclase, resulted in greater translocation of LAMP-1– and Syt-7–containing membranes to the cell surface of H37Rv-infected macrophages but did not affect the translocation of Golgi membranes (Fig. 3a). PGE2 exerts an important antinecrotic effect in infected macrophages by preventing perturbation of the mitochondrial inner membrane. This protective effect of PGE2 on mitochondria is mediated by engagement of the PGE2 receptor EP2, which induces the production of protein kinase A (PKA) and cAMP30. In fact, the PGE2 receptor EP2 and EP4, but not EP1 or EP3, both activate cAMP-dependent pathways35. Therefore, we sought to determine whether induction of membrane repair, which is triggered by an increase in cAMP, is activated by PGE2.
Figure 3: PGE2 reconstitutes lysosomal repair in human macrophages infected with virulent Mtb.
(a,b) Translocation of LAMP-1, Syt-7 and mannosidase II to the surface of macrophages infected with H37Rv (MOI, 10) and treated with 1–10 
M forskolin (a) or PGE2 (b). (c) LAMP-1 translocation induced by H37Rv in the presence of PGE2 after the addition of LY294002 (PI(3)K inhibitor) and/or KT5720 (PKA inhibitor). (d) Translocation of Syt-7, LAMP-1 and mannosidase II to the surface of H37Ra-infected wild-type (WT) and Ptges-/- macrophages. *P < 0.05 (t-test). Data are from one representative of three independent experiments (error bars, s.e.m.).
The exogenous addition of PGE2 to H37Rv-infected macrophages reconstituted plasma membrane repair, as measured by enhanced translocation of LAMP-1 and Syt-7 to the cell surface (Fig. 3b,c). In contrast, PGE2 did not affect the recruitment of mannosidase II–containing membranes. EP2 activates mainly PKA, whereas EP4 receptors activate phosphoinositide-3-OH kinase (PI(3)K)35, 36. To determine whether PGE2-dependent activation of plasma membrane repair is caused by activation of PKA or PI(3)K, we assessed whether the specific PKA inhibitor KT5720 (ref. 37) and/or the PI(3)K inhibitor LY294002 (ref. 38) affect PGE2-induced translocation of LAMP-1 to the plasma membrane of H37Rv-infected macrophages. LY294002 abrogated the translocation of LAMP-1 translocation to the cell surface of H37Rv-infected macrophages treated with PGE2, whereas KT5720 had no effect (Fig. 3c). Either inhibitor alone had no effect on LAMP-1 translocation. Therefore, in contrast to the protective effects of PGE2 on mitochondria, which depend on the EP2 receptor and 'downstream' activation of PKA16, PGE2-dependent lysosomal membrane translocation seems to require PI(3)K activation, which is typical of EP4 activation39, 40.
To further evaluate the function of PGE2 in stimulating lysosome-dependent repair of plasma membranes, we infected mouse wild-type and Ptges-/- splenic macrophages with H37Ra for 24 h. As expected, translocation of LAMP-1 and Syt-7 to the cell surface was significantly greater after infection with H37Ra in wild-type macrophages (Fig. 3d). In contrast, LAMP-1 and Syt-7 were not recruited to the plasma membrane of H37Ra-infected Ptges-/- macrophages, which are unable to produce PGE2 (Fig. 3d). We detected translocation of mannosidase II in both wild-type and Ptges-/- macrophages. These data independently confirm that although recruitment of lysosomal membranes is PGE2 dependent, recruitment of Golgi-derived membranes is independent of PGE2. Notably, the propensity of Ptges-/- macrophages to undergo necrosis when infected with H37Ra is reversed when exogenous PGE2 is added16. These findings show the importance of PGE2 in inducing lysosome-dependent repair of the plasma membrane.
Balance between LXA4 and PGE2 in control of Mtb infection
Infection of human macrophages with virulent Mtb induces LXA4 synthesis, which leads to inhibition of PGE2 production16. We sought to understand how PGE2 and LXA4 affect the outcome of Mtb infection. Ptges-/- macrophages infected with avirulent H37Ra underwent significantly more necrosis and significantly less apoptosis than did wild-type or Alox5-/- macrophages (which cannot produce LXA4; Fig. 4a). We obtained the converse result with H37Ra-infected Alox5-/- macrophages; these had more apoptosis and less necrosis than wild-type and Ptges-/- macrophages. We obtained similar results with wild-type, Alox5-/- and Ptges-/- cells infected with virulent H37Rv.
Figure 4: Bacterial growth and the death modality of Mtb-infected mouse macrophages is regulated by eicosanoids.
(a) Enzyme-linked immunosorbent assay of apoptosis and necrosis 3 d after infection of Alox5-/-, wild-type and Ptges-/- macrophages with H37Rv (bottom) or H37Ra (top). (b) Replication of H37Rv (bottom) or H37Ra (top) at 4 h, 4 d or 7 d after infection (MOI 10:1) of Alox5-/-, WT and Ptges-/- macrophages. *P < 0.05 (two-way ANOVA). Data are representative of three (a) and two (b) independent experiments (error bars, s.e.m.).
Full size image (75 KB)Ptges-/- and Alox5-/- macrophages also showed substantial differences in the control of Mtb growth. H37Rv grew slowly in wild-type macrophages, increasing 2.5-fold after 7 d; in contrast, there was little replication of H37Ra in wild-type macrophages (Fig. 4b). The growth of virulent H37Rv was significantly lower in Alox5-/- macrophages, whereas the growth of both H37Ra and H37Rv was enhanced in Ptges-/- macrophages.
Alox5-/- mice are more resistant17 and Ptges-/- mice are more susceptible16 to virulent mycobacterial infection. However, the question of whether the mechanisms we have delineated in vitro are reflective of in vivo pathophysiology is difficult to answer, as the fate of infected macrophages can affect host resistance in many different ways41, 42. To determine whether more apoptosis occurs in the lungs of Alox5-/- mice after virulent Mtb infection, we infected Ptges-/-, wild-type and Alox5-/- mice by the intratracheal route with 1 
106 colony-forming units (CFU) of H37Rv and obtained cells by pulmonary lavage 3 d after infection. Cells from the lungs of Alox5-/- mice underwent more apoptosis than did those from wild-type or Ptges-/- mice (Fig. 5a).
Figure 5: The fate of Mtb-infected macrophages in vitro reflects the innate control of infection in vivo.
(a) Apoptosis of adherent antigen-presenting cells in bronchoalveolar lavage fluid collected 3 d after intratracheal infection of Alox5-/-, Ptges-/- and wild-type mice (n = 3 per group) with H37Rv (1 106 CFU). (b,c) Bacterial burden in the spleen and/or lung 14 d (b) and 28 d (c) after intratracheal transfer of H37Rv-infected Alox5-/-, Ptges-/- or wild-type macrophages into Rag1-/- mice. Bacteria in lungs on day 1 after adoptive transfer: wild-type, log 10 = 1.81; Ptges-/-, log 10 = 1.79; Alox5-/-, log 10 = 1.8. *P < 0.05 (one-way ANOVA). Data are from a single experiment with two time points (n = 5 mice per group per time point; error bars, s.e.m.).
To study the consequences of macrophage function on innate immunity to Mtb, we developed an experimental model involving the adoptive transfer of Mtb-infected macrophages. The advantage of this model is that it avoids the complications of analyzing knockout mice in which the deleted gene is expressed ubiquitously and affects several physiological processes. Thus, it allows specific determination of how manipulation of the lipid mediators produced by Mtb-infected macrophages alters the outcome of infection independently of their function in other cell types. We infected wild-type, Alox5-/- and Ptges-/- macrophages in vitro with a low dose of virulent H37Rv Mtb, then transferred the cells by the intratracheal route into recipient mice deficient in recombinase-activating gene 1 (Rag1-/- mice). By transferring the cells into Rag1-/- recipient mice, we could evaluate the functional effect of eicosanoid regulation in the absence of any contribution from the adaptive immune system.
Two weeks after adoptive transfer, the pulmonary bacterial burden was significantly higher in Rag1-/- mice that received infected Ptges-/- macrophages than in recipients of infected wild-type macrophages (Fig. 5b). In contrast, the bacterial burden was significantly lower in Rag1-/- mice that received infected Alox5-/- macrophages than in those that received infected wild-type macrophages (Fig. 5b). This effect was durable and was detected 28 d after transfer of the infected macrophages in the lungs as well as in the spleen (Fig. 5c). The mean difference in the pulmonary bacterial burden of Rag1-/- mice that received Alox5-/- macrophages and those that received wild-type macrophages was 
log10 = 1.3 (P < 0.05). These experiments show that transfer of infected Alox5-/- macrophages (which are predisposed to apoptosis after infection) into Rag1-/- mice restricts virulent Mtb replication in vivo. The use of Rag1-/- recipient mice shows that this effect is determined by the macrophage genotype and is independent of adaptive immunity. Thus, the balance of PGE2 and LXA4 production by infected macrophages affects the outcome of infection in the microenvironment of the lung.
Induction of Syt-7 transcription by PGE2
We have shown that Syt-7 is a central regulator of calcium-dependent translocation of lysosomal membranes to the cell surface and is required for successful membrane repair and prevention of necrosis in human macrophages. We have also shown that PGE2 is indispensable for the induction of lysosome translocation to the macrophage surface and plasma membrane repair in Mtb-infected macrophages. We next investigated whether PGE2 is involved in Syt-7 synthesis. For this, we quantified Syt-7 mRNA transcripts in uninfected macrophages in the presence and absence of PGE2. Although we did not detect Syt-7 transcripts in uninfected macrophages, the addition of PGE2 induced Syt-7 expression in a dose- and time-dependent way (Fig. 6a). In contrast, LAMP-1 transcription was not substantially affected when exogenous PGE2 was added to uninfected macrophages (Fig. 6b). We next sought to determine whether exogenous PGE2 increases Syt-7 transcript abundance in H37Rv-infected macrophages. H37Rv infection and exogenous PGE2 synergistically increased Syt-7 transcript abundance (Fig. 6c). The abundance of LAMP-1 transcript was not affected in these conditions. These findings provide a mechanistic link between the ability of PGE2 to induce Syt-7, a critical regulator of lysosome translocation, to plasma membrane lesions and its ability to stimulate membrane repair.
Figure 6: PGE2 regulates Syt-7 expression in mouse macrophages.
(a,b) Real-time PCR analysis of the expression of Syt-7 mRNA (a) and LAMP-1 mRNA (b) in naive wild-type macrophages treated with PGE2 for various times or at various doses. (c) Real-time PCR analysis of the expression of Syt-7 and LAMP-1 in wild-type macrophages left uninfected (Media), treated with PGE2 alone (PGE2), infected with H37Rv alone (Rv), or treated with PGE2 and then infected with H37Rv (PGE2 + Rv). (d) Real-time PCR analysis of the expression of Syt-7 and LAMP-1 mRNA in Alox5-/-, Ptges-/- and wild-type macrophages 0–48 h (horizontal axis) after infection with H37Rv. (e) Real-time PCR analysis of the expression of Syt-7 and LAMP-1 mRNA in lungs of wild-type mice left uninfected or infected by the aerosol route with a low dose (
100 CFU) of H37Ra or the Erdman strain of Mtb; RNA was extracted from whole lung 7 d after infection. Expression of Syt-7 or LAMP-1 is presented relative to 
-actin expression. *P < 0.05 (one-way ANOVA). Data are from one experiment representative of two (a–c,f) or three (d,e) independent experiments (n = 3 mice per group; error bars, s.e.m.).
As expected, Alox5-/- macrophages infected with H37Rv expressed more Syt-7 than did wild-type or Ptges-/- macrophages infected with H37Rv (Fig. 6d). Syt-7 expression was induced only transiently in wild-type and Ptges-/- macrophages infected with virulent Mtb. LAMP-1 expression was not affected by infection with virulent Mtb or by the abiity of the macrophages to produce eicosanoids (Fig. 6d and Supplementary Fig. 2).
We confirmed the findings reported above with in vivo experiments. At 7 d after aerosol infection with a low dose of H37Ra, wild-type mice had a greater abundance of Syt-7 transcripts in their lungs, whereas mice infected with virulent Mtb had smaller quantities of Syt-7 mRNA that were similar to those in uninfected mice (Fig. 6e). LAMP1 expression was not affected by infection. These data collectively show that virulent Mtb evades innate immunity by suppressing the production of PGE2 (ref. 16), which is required for optimal expression of Syt-7 and lysosome-dependent plasma membrane repair.
Syt-7 is essential for control of virulent Mtb
We have established the importance of eicosanoids in determining the cellular fate of Mtb-infected macrophages, which determines whether the bacteria succumb to or evade innate immune control (Figs. 4 and 5). The finding that Syt-7 is regulated by PGE2 (Fig. 6) indicates a mechanism by which PGE2 induces membrane repair (Fig. 3). We next sought to show a direct link among Syt-7 function, the death modality of Mtb-infected macrophages and the outcome of infection. Because our data indicated that one consequence of LXA4 induction by virulent Mtb is inhibition of Syt-7 transcription, we predicted that Alox5-/- macrophages, which accumulate Syt-7 transcripts after infection (Fig. 6d), would have enhanced membrane repair after H37Rv infection.
We detected translocation of LAMP-1 to the cell surface after H37Rv infection in Alox5-/- macrophages but not in wild-type or Ptges-/- macrophages (Fig. 7a). To directly visualize whether there was enhanced plasma membrane repair in Alox5-/- macrophages, we infected wild-type, Ptges-/- and Alox5-/- macrophages with green fluorescent protein (GFP)-labeled H37Rv, then labeled the surface of infected cells with an antibody specific for the luminal (extracellular) domain of LAMP-1. We detected little LAMP-1 on the surface of GFP-H37Rv–infected wild-type or Ptges-/- macrophages (Fig. 7b). In contrast, GFP-H37Rv induced extensive recruitment of LAMP-1 to the surface of Alox5-/- macrophages (Fig. 7b). The greater LAMP-1 staining reflected a true change in LAMP-1 recruitment to the cell surface, as wild-type, Ptges-/- and Alox5-/- macrophages all expressed similar amounts of intracellular LAMP-1 (Supplementary Fig. 3). These data indicate that of the larger amounts of PGE2 in H37Rv-infected Alox5-/- macrophages enhance lysosomal plasma membrane repair.
Figure 7: Syt-7 is essential for the induction of plasma membrane repair, prevention of necrosis and control of bacterial growth in mouse macrophages.
(a,b) Flow cytometry (a) and confocal microscopy (b) showing translocation of LAMP-1 to the cell surface of Alox5-/-, Ptges-/- and wild-type macrophages left uninfected or 24 h after infection with H37Rv (MOI, 5). (b) Macrophages infected with GFP-labeled H37Rv (MOI, 10) and stained with monoclonal antibody to the lumenal domain of LAMP-1 (no permeabilization). Scale bar, 5 m. (c) Apoptosis and necrosis of Alox5-/- macrophages left untransfected (control) or transfected for 24 h with scrambled (Scr) or Syt-7-specific siRNA, followed by H37Rv infection (MOI, 5) and analysis 3 d after infection; results are presented relative to those of uninfected cells. (d) H37Rv growth in wild-type and Alox5-/- macrophages left untreated or transfected with Syt-7-specific or scrambled siRNA at 4 h, 4 d or 7 d after infection with H37Rv. *, P < 0.05 (one-way (c) or two-way (d) ANOVA). Data are representative of two independent experiments (error bars, s.e.m.).
To show that Syt-7 expression has a direct influence on the death modality of the infected macrophages, we undertook gene silencing of Syt-7 in H37Rv-infected Alox5-/- macrophages (Fig. 7c). Although H37Rv induced more apoptosis than necrosis in Alox5-/- macrophages, Syt-7 silencing reversed this phenotype and resulted in significantly less apoptosis and more necrosis (Figs. 4a and 7c). Thus, Syt-7 seems to be critical in preventing necrosis and inducting apoptosis in macrophages infected with Mtb.
Finally, as Alox5-/- macrophages have an enhanced ability to limit Mtb replication both in vitro and in vivo (Figs. 4 and 5), we sought to determine whether Syt-7 function is directly involved in innate control of Mtb infection. Although Alox5-/- macrophages limited bacterial replication more efficiently than did wild-type macrophages, Syt-7 silencing significantly impaired the ability of Alox5-/- macrophages to restrict bacterial replication and led to a significant increase in Mtb (Fig. 7d). Collectively, these data indicate that PGE2 is an essential mediator that stimulates Syt-7 production to activate lysosome-dependent membrane repair, which prevents necrosis, instead leading to apoptosis and innate protection against Mtb infection.
Discussion
Necrosis is a highly regulated irreversible loss of plasma membrane integrity. Although virulent Mtb has been shown to block the formation of the apoptotic envelope in infected macrophages and thereby lead to necrosis29, the mechanisms that facilitate necrosis in this system have remained unknown. Here we found that virulent Mtb perturbs the repair of plasma membrane microdisruptions inflicted by the pathogen. We identified two distinct and essential components involved in plasma membrane repair: lysosomal and Golgi-derived vesicles. Although macrophages infected with attenuated Mtb underwent plasma membrane repair and apoptosis, blockade of the translocation of either lysosomal or Golgi apparatus–derived vesicles to the plasma membrane resulted in significant necrosis. Silencing of the calcium sensor Syt-7 inhibited lysosome-dependent repair of plasma membranes but did not affect the translocation of mannosidase II–containing Golgi-derived membranes. In contrast, silencing of NCS-1, a calcium sensor present in Golgi apparatus, significantly inhibited the translocation of Golgi-derived membranes and downregulated the translocation of phosphatidylserine and annexin-1 to the macrophage surface. Thus, two distinct calcium sensor proteins regulate lysosomal and Golgi-dependent plasma membrane repair in Mtb-infected macrophages. Although both lysosomal and Golgi apparatus–derived vesicles were required for membrane repair, only the Golgi vesicle–dependent membrane repair facilitated exocytosis of the apoptotic marker phosphatidylserine and deposition of annexin-1 on the cell surface. Golgi-derived vesicle–mediated membrane repair was PGE2 independent, which indicates that other mediators are involved in the recruitment of these membranes to the plasma membrane lesions. Further study of the Golgi-derived vesicle membrane–repair pathway will considerably extend understanding of the biology of apoptosis.
Virulent H37Rv stimulates LXA4 production in macrophages, which inhibits PGE2 production by downregulation of the accumulation of mRNA encoding cyclooxygenase 2 (ref. 16). PGE2 protects mitochondrial membranes from damage caused by Mtb infection and thereby inhibits necrosis. Here we have reported that PGE2 production restored translocation of lysosomal membranes to the cell surface in macrophages infected with virulent Mtb, and that PGE2 upregulated synthesis of Syt-7 (ref. 7). Thus, PGE2 activates at least two independent pathways that protect Mtb-infected macrophages from necrosis. First, PGE2 acts on the PGE2 receptor EP2, which stimulates adenylate cyclase to produce cAMP by activation of protein kinase A36 and protects against Mtb-induced mitochondrial damage30. Second, PGE2 activates plasma membrane repair through PI(3)K, which most probably involves the EP4 receptor35, 40. Thus, PGE2 protects Mtb-infected cells against necrosis by preventing mitochondrial inner membrane instability and plasma membrane disruption.
Even attenuated H37Ra was more virulent when infecting Ptges-/- macrophages than when infecting wild-type macrophages. Conversely, virulent H37Rv showed an attenuated phenotype when infecting Alox5-/- macrophages, as PGE2 production was not counter-regulated by LXA4. Thus, in an intracellular milieu dominated by PGE2, infected macrophages underwent more apoptosis and restricted Mtb growth. These findings indicate that the ability of host macrophages to produce PGE2 modulates the virulence of Mtb. Therefore, the innate host response is able to modify the phenotypic expression of bacterial virulence.
Alox5-/- mice were more resistant and Ptges-/- mice were more susceptible than wild-type mice when infected by the aerosol route with Mtb16, 17. However, our transfer experiments showed that the fate of the infected macrophages was a key determinant of the relative resistance of these mice. Although Alox5-/-, Ptges-/- and wild-type macrophages were all infected to a similar degree, transfer of infected Alox5-/- macrophages resulted in a much less severe systemic infection than did transfer of infected wild-type or Ptges-/- macrophages. Therefore, the fate of transferred macrophages, whether apoptotic or necrotic, had a durable effect on the course of infection. This is the first direct demonstration to our knowledge that the death modality of infected macrophages alters the course of Mtb infection in vivo. Thus, we have provided a direct mechanistic link between the beneficial function of PGE2 and outcome of infection.
Although we have gained some knowledge about the donor vesicles involved in plasma membrane repair of Mtb-infected macrophages and about the importance of calcium sensors in the regulation of plasma membrane repair, exactly how PGE2 facilitates membrane repair remains unclear. Attenuated Mtb triggered PGE2-dependent LAMP-1 translocation to the macrophage surface in a PI(3)K-dependent way. Phagosome-lysosome fusion is thought to be inhibited in Mtb-infected macrophages43 by constant removal of phosphatidylinositol-3-phosphate from the endosomal membranes by SapM, a pathogen-derived phosphatase, in a way that is independent of cytosolic calcium44. It is therefore likely that PGE2 upregulates PI(3)K activity to generate sufficient phosphatidylinositol-3-phosphate for membrane repair. If pathogen-mediated depletion of phosphatidylinositol-3-phosphate affects both phagosome-lysosome fusion and plasma membrane repair, it could be assumed that plasma membrane repair and phagosome-lysosome fusion are mediated by related mediators. This observation is consistent with published work demonstrating that ESX-1 proteins are required for translocation of Mtb from the phagolysosome to the cytosol45, and our finding that translocation of LAMP-1 to macrophage plasma membrane lesions also depended on ESX1-encoded proteins (data not shown). In addition, other studies have reported that the early intracellular survival of Salmonella and Yersinia is inhibited in mouse embryonic fibroblasts as a consequence of Syt-7-dependent phagosome-lysosome fusion2.
Here we established a causal relationship between the capacity of macrophages to restrict mycobacterial growth and their ability to reseal membrane lesions inflicted by the pathogen and to induce apoptosis. If membrane repair is prevented, infected macrophages are doomed to become necrotic and support enhanced bacterial growth. Thus, inhibiting membrane repair by blocking PGE2 production represents a critical mechanism that allows virulent bacilli to replicate, induce necrosis and escape from the host macrophage and infect other cells. Better understanding of the mechanisms by which Mtb induces necrosis might identify new targets for drugs that modulate innate immune responses to control the initial infection as well as to enhance adaptive immunity.
Methods
The Methods and their associated references appear only online.
Accession codes.
UCSD-Nature Signaling Gateway (http://www.signaling-gateway.org): A002565 and A000957.
Note: Supplementary information is available on the Nature Immunology website.
Author Contributions
S.M.B., H.G.R. and M.D. conceived of and designed the experiments, analyzed the data and wrote the paper; M.D., M.C. and H.G. did the experiments with assistance from D.D.; T.T.H. did confocal microscopy; and D.M.L. and S.F. provided reagents and intellectual input.
-d-mannosidase II in the Golgi apparatus of rat liver
, and kill intracellular bacteria
