|Home | About | Journals | Submit | Contact Us | Français|
K.C.E.K., J.E.Q. and A.M.S. did most of the experiments. J.T.P. and R.W.T. did the T. gondii experiments. M.H.-T., R.J.B., T.K., U.S., M.-S.K., G.K., E.I.T., I.M.O. and C.B. did the infection and biochemistry experiments. K.A.F. and T.-D.K. contributed key research reagents. P.J.M. created the Arg1 conditional knockout. P.J.M. and A.M.S. bred and backcrossed the mice. T.A.W. and P.J.M. conceived and designed the project and wrote the manuscript.
Toll-like receptor (TLR) signaling in macrophages is required for antipathogen responses, including the biosynthesis of nitric oxide from arginine, and is essential for immunity to Mycobacterium tuberculosis, Toxoplasma gondii and other intracellular pathogens. Here we report a ‘loophole’ in the TLR pathway that is advantageous to these pathogens. Intracellular pathogens induced expression of the arginine hydrolytic enzyme arginase 1 (Arg1) in mouse macrophages through the TLR pathway. In contrast to diseases dominated by T helper type 2 (TH2) responses, TLR-mediated Arg1 induction was independent of the TH2-associated STAT6 pathway. Specific elimination of Arg1 in macrophages favored host survival in T. gondii infection and decreased lung bacterial load in tuberculosis infection.
In macrophages, TLRs activate protective immune responses, including recognition of pathogens, activation of antipathogen effector pathways and transition to protective adaptive responses1. Classically activated macrophages (CAMs) are important in combating infections caused by intracellular pathogens. A key antipathogen effector of CAMs is nitric oxide (NO), which is required for host control of intracellular infections, including Mycobacteria species, T. gondii, Leishmania species and Trypanosoma cruzi2, and has direct antimicrobial toxicity3. TLR and interferon pathways synergistically trigger NO production by transcriptional and post-transcriptional mechanisms that enhance expression of inducible nitric oxide synthase (iNOS), the enzyme responsible for NO production from arginine in macrophages2,4,5. As a countermeasure, some pathogens deploy NO scavengers6 or their own arginases7,8 or, in the case of M. tuberculosis, adapt to NO made by activated macrophages9 and exclude iNOS from phagosomes10. However, mechanisms by which intracellular pathogens induce host responses that reduce or bypass NO remain largely uncharted.
Although a published report has shown that arginase activity is induced in the J774 macrophage-like cell line by infection with Mycobacterium bovis bacillus Calmette-Guérin (BCG)11, the isoform of arginase induced, the mechanism of induction and the biological consequences in primary macrophages and whole-animal models remain unknown. Here we report that Arg1 was considerably induced in primary mouse macrophages by mycobacterial infection. Although macrophage Arg1 expression is commonly linked to the hypothesized antiworm functions of alternatively activated macrophages (AAMs)12,13, we found that Arg1 was induced in CAMs and functioned, in part, to suppress NO production in intracellular infection. In whole-animal models of intracellular infection, mice lacking macrophage Arg1 expression had an advantage in terms of clearance of pathogens or survival.
Our first clue that intracellular pathogens influence host pathways to counter NO came from our observation of robust expression of the host gene encoding arginase 1 (Arg1), but not Arg2 (Supplementary Fig. 1 online). Levels of Arg1 protein increased over time and peaked 48 h after infection of primary mouse macrophages with M. bovis BCG (Fig. 1a). Although Arg1 is expressed by AAMs in response to infection by extracellular pathogens such nematodes and trematodes12,14, Arg1 expression has not been generally associated with intracellular infections that critically involve CAM activity. In contrast to infection with intracellular pathogens, the AAM response to extracellular pathogens does not involve iNOS, which is normally regarded as detrimental in AAM-dominated responses15. Rather, in AAMs, interleukin 4 (IL)-4 and IL-13 stimulate host Arg1 production through the STAT6 pathway, and AAM-produced Arg1 is hypothesized to be involved in the repair and resolution of worm-induced tissue damage12,13. In determining whether the BCG-induced expression of host Arg1 was regulated through the IL-4–IL-13–STAT6 pathway, we initially reasoned that BCG infection might cause macrophages to express IL-4, IL-13 (ref. 16) or another factor that could activate STAT6. We used an assay to transfer supernatants from BCG-infected macrophage cultures to uninfected macrophages, followed by measurement of STAT6 phosphorylation. No factors were detectable in BCG-infected culture supernatants that could activate STAT6 phosphorylation (Supplementary Fig. 2 online). As STAT6 is essential for Arg1 expression in macrophages stimulated by cytokine(s) that favor AAM development17, we next infected bone marrow–derived macrophages (BMDMs) isolated from Stat6−/− mice. Arg1 was induced by BCG infection to the same extent in Stat6−/− and control BMDMs (Fig. 1a), albeit with a slight delay in the Stat6−/− mice.
Because BCG-induced Arg1 expression was STAT6 independent, we examined whether a pathogen detection pathway was involved in Arg1 expression. We tested BCG-infected macrophages lacking key components of the interferon pathway (Irf3−/− and Ifnar−/−), the Nod1 and Nod2 pathways18 (Ripk2−/−), the inflammasome (Nlrp3−/− and Pycard−/−) and the TLR and IL-1 pathways (Myd88−/− and Il1r−/−) for Arg1 expression. Only the adaptor molecule MyD88 was required for the BCG-mediated induction of Arg1 expression (Fig. 1b and Supplementary Fig. 3 online). However, the MyD88-dependent pathway was most likely to use TLR receptors, as the IL-1 receptor was not required for expression of Arg1 (Fig. 1b). We next confirmed the mRNA findings at the protein level: in BMDMs isolated from Tlr2−/− and Myd88−/− mice, Arg1 expression after BCG infection was largely dependent on MyD88 and partially dependent on TLR2, regardless of an intact STAT6-dependent Arg1 expression pathway (Fig. 1c). The partial function of TLR2 in Arg1 expression is consistent with the essential function of TLR2 in macrophage-mediated mycobacterial recognition and subsequent downstream cytokine production19. Finally, we infected mice intraperitoneally with BCG and collected their spleens 10 d later. Whereas adherent splenocytes isolated from normal mice did not express detectable Arg1, BCG-infected mice had robust Arg1 expression, confirming that mycobacteria can induce Arg1 in vivo (Fig. 1d).
The TH2 cytokine-driven increase in macrophage Arg1 expression is controlled by an enhancer that is ~3 kb upstream of the basal promoter and is active in hepatocytes20,21. The Arg1 enhancer binds STAT6 and other proteins, including the transcription factor C/EBPβ20,21. We tested whether the MyD88-dependent pathway for Arg1 expression targets the basal Arg1 promoter or the enhancer. In a reporter assay21, BCG induced the expression of the Arg1 reporter only when the upstream enhancer was present (Fig. 1e). Notably, BCG-mediated induction of Arg1 reporter activity was independent of the STAT6 binding site in the enhancer that is essential for IL-4- and IL-13-mediated expression of Arg1 (refs. 21,22; Supplementary Fig. 4 online).
We next considered that BCG-induced Arg1 expression could be linked to polyamine amounts in infection. Arginases supply substrate (ornithine) to ornithine decarboxylase (encoded by Odc1), the rate-limiting enzyme for polyamine synthesis. Studies have shown that Helicobacter pylori induces expression of both Arg2 and Odc1 as a means to perturb polyamine homeostasis23,24. We therefore tested the possibility that Odc1 mRNA amounts would be increased by BCG infection as a possible mechanism for polyamine sequestration by mycobacteria. In contrast to H. pylori infection, however, Odc1 mRNA decreased after BCG infection (Supplementary Fig. 5a online) in a MyD88-dependent, STAT6-independent way (Supplementary Fig. 5b). Collectively, our results suggest that BCG-mediated upregulation of Arg1 is unlikely to cause an increased requirement for polyamine production in our in vitro culture system.
To determine the mechanisms that link the MyD88 pathway to Arg1 regulation, we focused on transcription factors that target the Arg1 enhancer. Given that the BCG-mediated increase in Arg1 expression required the upstream enhancer but not STAT6, we tested the requirement for the C/EBPβ site, which is adjacent to the STAT6 site21. We created luciferase reporter lines in RAW macrophages in which the basal promoter (−31/− 2365) and promoter enhancer (−31/−3810) were cloned into an insulated backbone where the reporter was flanked by duplicated β-globin insulators25. An additional reporter line was made in which the C/EBPβ binding site was mutated21. After infection with BCG, the enhancer was required for reporter activity, consistent with the transient transfections described above (Fig. 2a and Supplementary Fig. 4). Deletion of the C/EBPβ site ablated BCG-mediated induction of reporter activity.
Because C/EBPβ is an essential component of Arg1 induction by the IL-4–STAT6 pathway26, we tested whether C/EBPβ is also required for Arg1 expression in mycobacterial infection. Arg1 levels were much lower in BCG-infected Cebpb−/− macrophages than in BCG-infected wild-type macrophages (Fig. 2b). We next asked whether C/EBPβ expression is increased in BCG-infected macrophages. RNA blot analysis and a quantitative RT-PCR assay revealed that BCG infection resulted in a four- to five-fold increase in C/EBPβ mRNA in Myd88+/+ macrophages but not in Myd88−/− macrophages (Fig. 2c,d). Thus, BCG induced Arg1 expression in a STAT6-independent way that was dependent on MyD88-mediated induction of C/EBPβ (Supplementary Fig. 4), but both the STAT6 and MyD88 pathways of Arg1 expression require upstream regulatory elements for full activity. These data indicate the existence of two distinct pathways to induce Arg1 expression in macrophages. One previously described pathway is linked to AAM function and regulated by STAT6. Our results identify a second pathway that is independently controlled by MyD88-dependent TLR signaling to C/EBPβ and does not require STAT6.
Although AAM Arg1 is associated with antihelminth functions13-15, we considered the findings that pathogen arginases subvert the host's NO-based antimicrobial response7,8. We hypothesized that induction of CAM Arg1 cripples CAM antimicrobial activity by hydrolyzing the substrate required for NO production. Such a mechanism would suggest that Arg1 is an essential component of the strategy used by intracellular pathogens to survive inside NO-generating macrophages. To test the physiological importance of macrophage Arg1 expression in intracellular infection, we constructed an Arg1 conditional knockout allele (Arg1flox/flox; Supplementary Fig. 6 online) and used it in the LysMcre deleter strain (B6.129P2-Lyz2tm1(cre)Ifo/J (Lysz is now called Lyz2); transgene abbreviated here as LysMcre) to generate mice lacking Arg1 in macrophages and neutrophils (Arg1flox/flox; LysMcre; Supplementary Fig. 7 online). We generated another mouse strain lacking Arg1 in all hematopoietic lineage cells using the Tie2cre deleter strain (B6.Cg-Tg(Tekcre)1Ywa (Tie2 is now known as Tek); transgene abbreviated here as Tie2cre; Fig. 3a). We also produced a complete null mutation of Arg1 (Arg1Δ/Δ) that replicated the published conventional Arg1 knockout27. The Arg1Δ allele was used as a control for absolute Arg1 deficiency (Fig. 3). As Arg1 is predominantly expressed in myeloid but not lymphoid lineage cells12, the different strains provided parallel systems for testing the function of Arg1 in macrophages in vivo. Macrophages isolated from the Tie2cre intercross showed almost complete deletion of Arg1 protein and complete ablation of enzyme activity in all macrophage types tested (Fig. 3a,b). The Arg1flox/flox; LysMcre mice also showed efficient deletion in BMDMs and >80% deletion in postmitotic macrophages (Supplementary Fig. 7).
We first tested whether Arg1 is an essential component of a pathway for arginine depletion that hinders macrophage NO production. We examined NO production through a substrate depletion assay that uses a defined order of cytokine stimulation to induce first Arg1 and then iNOS28. Applying this assay to macrophages deficient in Arg1, we found that Arg1 was essential for decreasing NO production by the substrate depletion mechanism (Supplementary Fig. 8 online). Given our finding that host Arg1 expression is stimulated by intracellular pathogens, we next tested whether Arg1 influences NO synthesis in response to lipopolysaccharide (LPS) or BCG infection. After BCG infection, macrophages lacking Arg1 produced more NO, suggesting that TLR-mediated regulation of Arg1 is a mechanism that inhibits macrophage production of NO (Fig. 3c). After stimulation with LPS, an extracellular TLR4 ligand, macrophages lacking Arg1 similarly produced more NO, suggesting that TLR-mediated regulation of Arg1 is a common mechanism that restricts macrophage production of NO (Fig. 3d).
Because NO is required for immune-mediated eradication of many intracellular pathogens, we next tested the response of Arg1flox/flox; Tie2cre mice lacking macrophage Arg1 to M. tuberculosis infection. We first measured Arg1 expression in the lungs of M. tuberculosis–infected mice, given that human peripheral blood mononuclear cells from pulmonary tuberculosis patients have elevated arginase activity29. Using mice infected with CDC1551, a rapidly growing human isolate of M. tuberculosis, we measured the expression of iNOS, Arg1 and Arg2 in the lungs, in parallel with colony counts (Fig. 4a–d). Whereas iNOS and Arg1 mRNA expression increased over the course of infection, Arg2 mRNA amounts did not change during the observation period. Next, we infected mice lacking macrophage Arg1 with M. tuberculosis H37Rv through the aerosol route with an inoculum of 200 bacilli. Compared to wild-type littermates, mice lacking macrophage Arg1 had lower bacterial counts in the lungs after the initial bacterial growth period (Fig. 4e). Lower bacterial load was also noted in the spleens of the macrophage Arg1–deficient mice at day 70 after infection (data not shown). Lung granulomas in macrophage Arg1–deficient mice occupied a smaller fraction of the total lung area (Supplementary Fig. 9 online) and were associated with a more pronounced lymphocytic infiltrate than were control mice (Fig. 4f). Collectively, these data indicate that loss of macrophage Arg1 improved host control of M. tuberculosis infection.
We next assessed whether loss of Arg1 in macrophages causes elevated NO in tissues associated with mycobacterial infection. Because NO has a half-life of seconds, we used immunohistochemistry to stain nitrotyrosine as an indirect measure of the NO ‘footprint’30. Compared to infected control sections, we found greater nitrotyrosine staining in liver granulomas from BCG-infected macrophage Arg1-deficient mice and in macrophage-rich areas in the lungs of tuberculosis-infected macrophage Arg1-deficient mice (Fig. 5). We could therefore associate the loss of Arg1 in macrophages to increased NO production in infected tissues where NO is made in abundance. It is likely that elevated NO in the absence of Arg1 is one pathway that contributes to the decrease in colony counts in M. tuberculosis–infected Arg1flox/flox; Tie2cre mice.
Another intracellular pathogen that is controlled by the host NO response is T. gondii. T gondii induces NO production from macrophages, which is essential for long-term control of the infection31,32. T. gondii injects ROP kinases that activate host STAT proteins, including STAT6 (ref. 33). We reasoned that Arg1 might be a target of the ROP kinases through T. gondii–mediated ‘hijacking’ of the STAT6 response. We therefore infected macrophages with T. gondii strain ME49 and measured STAT6 phosphorylation and Arg1 expression (Fig. 6a). Arg1 protein expression was rapidly increased by T. gondii infection (within 1 h) but was independent of STAT6, as Stat6−/− macrophages had an identical response to wild-type cells. STAT6 tyrosine phosphorylation was induced by T. gondii as expected, but occurred after the increased expression of Arg1. These data confirm that, like M. bovis BCG, T. gondii induces Arg1 in a STAT6-independent way.
Given these findings, we next considered that T. gondii may have evolved a pathway to use host Arg1 to suppress NO production. In this scenario, mice lacking Arg1 in macrophages would have a survival advantage over control mice. We infected control (Arg1flox/flox or C57BL/6 mice) or Arg1flox/flox; LysMcre mice with ME49 and measured weight and overall appearance of the mice over 7 weeks. Loss of Arg1 expression in macrophages improved host ability to combat infection with ME49. Systemic infection of C57BL/6 or Arg1flox/flox control mice with ME49 caused a wasting disease that required the mice to be killed, whereas mice lacking macrophage Arg1 did not lose weight and did not show signs of disease (Fig. 6b). Thus, the observations we made with the mycobacterial infection models can be extended to an apicomplexan intracellular parasite that also requires NO for effective immune control.
To determine whether the protective effects of macrophage Arg1 deletion are specific to intracellular pathogens or also occur in systemic models of infection and sepsis, we used two models of infection challenge known to systemically increase NO: LPS challenge and Streptococcus pneumoniae, which replicates extracellularly34,35. We challenged macrophage Arg1–deficient mice with LPS or with a strain of S. pneumoniae that spreads throughout the host after intranasal inoculation (Supplementary Fig. 10 online). In multiple experiments, control, Arg1flox/flox; LysMcre and Arg1flox/flox; Tie2cre mice showed identical survival, blood nitrates and systemic bacterial numbers (Supplementary Fig. 10). Thus, macrophage Arg1 has a specific function in host susceptibility to intracellular pathogens and does not seem to affect the response to systemic infection with a Gram-positive pathogen or the response to systemic LPS administration, even though stimulation of Arg1-deficient macrophages with LPS leads to increased NO (Fig. 3d). Our data are therefore consistent with a model in which NO made during sepsis is derived largely from sources other than macrophages36 and does not seem to be under the control of macrophage Arg1.
CAMs are required to control and kill mycobacteria and T. gondii, which grow within the macrophages2. Infected CAMs are activated by innate recognition pathways, including the TLR and interferon-γ pathways, to make NO. We found that macrophage Arg1 is involved in preventing this NO production, consistent with previous reports that arginases can compete with NO synthases for their common substrate, arginine28,37-39. Our data suggest that in CAMs, where NO is thought to be involved in directly killing pathogens, successful chronic infections are associated with pathogen-induced Arg1 expression, which in turn keeps NO production in check. Indeed, the absence of Arg1 was associated with increased macrophage NO production and was linked to enhanced control of mycobacteria and T. gondii. An implication of this finding is that transient interruption of macrophage Arg1 function by competitive inhibitors40 may augment the ability of the immune system to control or eliminate intracellular pathogens like toxoplasma and mycobacteria, which establish long-term latent infections.
The products of arginase catalysis are urea and ornithine. It remains to be determined whether obligate intracellular pathogens such as M. tuberculosis hijack the Arg1 pathway not only to suppress NO production but also to supply substrates for growth and survival, as has been proposed for Leishmania species, whose growth depends in part on host polyamines derived from ornithine41. M. tuberculosis strains made genetically deficient in urease are being tested for their ability to act as an attenuated live vaccine42. The function of M. tuberculosis urease is unknown, but the enzyme presumably uses urea as a substrate for production of compounds needed for survival within macrophages.
Our data provide a possible rationale for the previous finding that T. gondii injects ROP kinases that activate the host's STAT proteins33. It seems likely that T. gondii–activated STAT6 induces Arg1 by bypassing the IL-4 or IL-13 receptors and directly activating the Arg1 enhancer. Other toxoplasma-induced mechanisms are probably involved in Arg1 expression, as we found rapid T. gondii–induced Arg1 expression in macrophages derived from STAT6-deficient mice. T. gondii therefore uses multiple strategies to induce macrophage Arg1, potentially as a means to regulate exposure to NO. T. gondii induces a robust host inflammatory response that is considered essential to the parasite's ability to establish chronic infections. NO production from iNOS is crucial for host control of the chronic phase of toxoplasmosis31,32. This is consistent with our finding that Arg1 is advantageous to the pathogen in the CAM-dependent response to T. gondii infection. Although we attribute the elevated macrophage NO in the absence of Arg1 as a major mechanism for host survival in T. gondii infection, it is possible that macrophage Arg1 has additional functions that the parasite could use as a survival strategy.
Unlike CAMs, AAMs are a distinct macrophage population that arise in polarized TH2 responses and are not the cellular host for pathogens. AAMs are thought to be involved in immune responses associated with asthma, worm infections and pathological scenarios involving TH2 cytokines12-15. In AAMs, TH2-driven immune responses, driven by IL-4 or IL-13, induce Arg1 expression and other markers of AAM activity, such as the mannose receptor, chitinases and metalloproteinases43. AAM expression of Arg1 absolutely requires STAT6 (ref. 17), and we and others have shown that IL-4- and IL-13- mediated expression of macrophage Arg1 requires direct binding of STAT6 to an upstream enhancer element in the −3 kb region of the Arg1 gene20,21. However, AAMs are not known to require NO for antihelminth immunity, and AAMs do not express much iNOS. Thus, the functions of Arg1 in AAMs remain unknown.
We found that distinct mechanisms regulate Arg1 expression in different types of infections. AAMs require the functions of both STAT6 and C/EBPβ but are independent of MyD88. In contrast, expression of Arg1 induced by mycobacteria is independent of the STAT6 pathway but depends on C/EBPβ and MyD88. These data are consistent with studies documenting the induction of Arg1 expression by LPS35,44. We speculate that the direct or indirect activation of C/EBPβ by LPS is most likely to be responsible for Arg1 induction in LPS-stimulated macrophages. As we found no obvious protective or pathogenic function for macrophage Arg1 in acute experimental LPS challenge, TLR-induced Arg1 has a more specific function for intracellular pathogens that require NO for their control.
Not all intracellular pathogens are killed by the NO pathway. NO has no obligate function in the clearance of chlamydia45, which have evolved to parasitize various cell types in diverse anatomical niches including the eye, lungs, and genital tract. Notably, MyD88-dependent Arg1 expression was found in total lung homogenates in the early phase of Chlamydia pneumoniae infection46. The role of Arg1 in infections where NO is not essentially required for killing remains to be determined.
Our studies raise the issue of why an antipathogen response would include a component that favors intracellular pathogens. We speculate that the TLR-mediated induction of macrophage Arg1 has positive antimicrobial effector functions against other types pathogens detected by the TLR system.
We generated a conditional Arg1 allele, complete Arg1 knockout alleles and crosses to the LysMcre and Tie2cre deleter strains as described in Supplementary Methods online. Backcrossing strategies are described in Supplementary Methods. All mice in this study were used according to protocols approved by the Institutional Animal Care and Use Committees at St. Jude Children's Research Hospital, the National Institute of Allergy and Infectious Diseases, Colorado State University and University of Freiburg.
Bone marrow–derived macrophages, peritoneal inflammatory macrophages and macrophages differentiated from the livers of embryonic day 14 mice were isolated and cultured as described47.
Age-matched mice were challenged with Escherichia coli LPS as described in ref. 48). Mice were challenged with S. pneumoniae D39X strain, modified for in vivo luciferase expression and imaged with the Xenogen IVS system as described49. Mice (6–8 weeks of age) were challenged intraperitoneally with M. bovis BCG Pasteur strain. For M. tuberculosis challenges, mice were infected with 200 colony-forming units of H37Rv as described50. Two mice were killed on day 1 to measure bacterial amounts in the lungs at the initial infection. Thereafter, four or five mice of each genotype were killed at the times indicated in Figure 6 for histological evaluation of the lungs and bacterial counts in lung and spleen homogenates. T. gondii challenges were done and evaluated as described31.
Griess assays for nitrite accumulation and arginase assays were done as described17.
Immunoblotting was done as described47. We used chicken polyclonal antibody to Arg1 (1:2,000; gift from S. Morris), rabbit polyclonal antibody to iNOS (1:5,000; Chemicon International), mouse monoclonal antibody to iNOS (1: 1,000; sc-7271 from Santa Cruz Biotechnology or gift from C. Nathan) and mouse monoclonal antibody to Grb2 (1:2,000; Signal Transduction Laboratories). Immunohistochemistry for nitrotyrosine was done with rabbit polyclonal antibody to nitrotyrosine (Upstate Biochemicals).
BMDMs were plated in 12-well plates and allowed to adhere overnight. Cells were washed with DMEM containing 10% FCS and incubated in a final volume of 1 ml. BCG cultures containing ~1 ml packed bacteria were washed in PBS and sonicated in 10 ml to reduce clumping and then diluted across a range of concentrations so that the approximate infection ratio was 100, 10 or 1 bacteria per macrophage. Infection cultures were maintained for 72–96 h, with time points for protein, RNA or NO analysis.
We used a two-sided Student t test for the in vivo M. tuberculosis infection study. Survival curves were analyzed by nonparametric Kaplan-Meier statistics embedded in the Prism software package. P < 0.05 was considered significant in both tests.
We thank I. Förster (Technical University of Munich) for the LysMcre mice, M. Yanagisawa (University of Texas Southwestern) for the Tie2cre mice, P. Ney (St. Jude Children's Research Hospital) for the CMV-cre mice, D. Green (St. Jude Children's Research Hospital) and S. Akira for the MyD88-deficient mice, S. Morris (University of Pittsburgh) for the antibodies to Arg1, C. Nathan (Weill Medical School of Cornell University) for the antibody to iNOS, S. Smale (University of California, Los Angeles) for the insulated reporter constructs, D. Bush for nitrotyrosine staining of BCG-infected livers, Xenogen for construction of luciferase-bearing pneumococci, Ozgene for microinjection of the targeted Bruce4 cells into C57BL/6 blastocysts and chimera generation, A. DeFreitas for technical assistance, B. Schulman for discussions and generation of Supplementary Figure 6b, and M. Koyanagi and M. Bix for discussion and preliminary infection experiments. This work was supported by the Sandler Program for Asthma Research (to P.J.M.), the National Institutes of Health (AI062921 to P.J.M., AI27913 to E.I.T., AI66046 to G.K., CORE grant P30 CA21765 and the NIAID intramural research program to T.A.W.), the German Research Foundation (SFB620 project A9 to C.B. and U.S.) and the American Lebanese Syrian Associated Charities.
Supplementary information is available on the Nature Immunology website.