Interdependence between interleukin-1 and tumor necrosis factor controls TNF-dependent effector functions during Mycobacterium tuberculosis infection

IL-1RI on hematopoietic cell is required for host resistance to Mtb

Although WT mice or WT mice transplanted with WT bone marrow did not show susceptibility to Mtb infection, Il1r1^−/− mice, WT mice transplanted with Il1r1^−/− bone marrow cells, and Il1r1^−/− mice transplanted with Il1r1^−/− bone marrow, succumbed to Mtb infection, indicating that IL-1RI on hematopoietic cells is critical for host resistance to Mtb (Figure 1A). In agreement with previous reports (Fremond et al., 2007; Mayer-Barber et al., 2011), the analysis of expression of inflammatory cytokines and chemokines 30 days p.i. showed that TNF, IL-6, and key pro-inflammatory chemokines were expressed at much higher amounts in the lungs of Mtb-infected Il1r1^−/− or Il1ab^−/− mice, when compared to WT, Il1a^−/−, or Il1b^−/− mice (Figure 1B), demonstrating that without functional IL-1RI signaling, Mtb infection leads to exuberant inflammation and TNF expression fails to confer protection. In order to understand why IL-1RI-deficient mice succumb to Mtb in the presence of TNF, we next infected WT mice transplanted with bone marrow cells from either WT or Il1r1^−/− mice with an Mtb strain that expresses the red fluorescent protein mCherry and analyzed expression of TNF-RI and reactive oxygen species (ROS) production in Mtb-infected cells 17 days p.i. This analysis showed that Il1r1^−/− mice contained higher numbers of Mtb-infected cells in the lungs (Figures 1C and S1A). This gain in Mtb-infected cells in Il1r1^−/− mice was due to the higher numbers of cells that showed autofluorescence in the 488 channel (488-Auto⁺, Figures 1D–F and Figure S1B), and over 95% of these cells express CD64 and CD11b monocytic markers and could be further divided onto distinct CD11c⁺ (myeloid dendritic cells and alveolar macrophages) and CD11c⁻ (newly recruited monocytes, macrophages, and PMNs) populations (Figures 1D, S1C; (Wolf et al., 2007)). Using ImageStream technology, we next confirmed that all cells that appeared mCherry-positive by flow cytometry, including those autofluorescent in 488 channel, were genuinely infected with mCherry-expressing Mtb bacilli (Figure 1E). Because the gain in Mtb-infected cells in Il1r1^−/− mice, compared to WT mice, was primarily attributable to 488-Auto⁺ CD64⁺CD11c⁻ myeloid cells (Figures 1F and S1B), we next analyzed the cellular composition of this population using antibodies specific to Ly-6G and CD11b. This analysis showed that this population was composed of Ly6G^high (and CD11b^high; PMNs) and Ly6G^low (and CD11b^low; newly recruited monocytes and macrophages) cells (Figure S1C). CD64⁺CD11c⁻Ly6G^high cells represented 33% and 27% of all Mtb-infected cells in WT and Il1r1^−/− mice respectively (Figure 1G). In contrast, Mtb-infected CD64⁺CD11c⁻Ly6G^low cells were nearly absent in WT mice and their number was significantly higher in Il1r1^−/− mice (4% and 12.5%, respectively). The analysis of cell-specific properties of these two populations showed that upon MTb infection in vivo, the 488-Auto⁻ phenotype associated with uniformly low TNF-RI expression, ROS production, and pathogen burden in both WT and Il1r1^−/− mice (Figures 1H–M, 488-Auto⁻ populations). Upon acquisition of 488-Auto⁺ phenotype, in WT mice CD64⁺CD11c⁻Ly6G^high cells activated high amounts of surface TNF-RI expression, ROS production, and exhibited low MTb burden indicating effective pathogen control (Figures 1H–J). In contrast, this population in Il1r1^−/− mice activated significantly lower amounts of TNF-RI, ROS, and failed to control the burden of MTb at a single cell level (Figures 1H–J). The response of CD64⁺CD11c⁻Ly6G^low cells to Mtb infection was qualitatively different and, although activating high amounts of TNF-RI and ROS upon transition to autofluorescent phenotype, this cell population failed to control MTb burden in both WT and Il1r1^−/− mice, demonstrating that this cell type, consistent with newly recruited inflammatory monocytes, is intrinsically permissive to Mtb, and failed to establish bactericidal state even in WT mice. Importantly, this cell population was only marginally present in WT mice, while its proportion was significantly higher in Il1r1^−/− mice (Figure 1G). Analysis of freshly isolated lung mononuclear cells showed that higher numbers stained positively with necrotic cell dye in Il1r1^−/− mice, compared to WT mice (Figures S1E–F). Furthermore, administration of propidium iodide into Mtb-infected mice revealed extensive distribution of necrotic cells in granulomas of Il1r1^−/− mice but not in WT mice at 30 days p.i. (Figure S1H). At this time point, the MTb burden was 3 order of magnitude higher in Il1r1^−/− mice, compared to WT mice (Figure S1G), and acid-fast staining of lung sections revealed high number of Mtb bacilli localized in cells with interstitial monocyte morphology (Figure 1N).

Together, these data indicate that the lack of IL-1RI on hematopoietic cells results in inability of CD64⁺CD11b⁺CD11c⁻Ly6G^high Mtb-infected cells to upregulate protective amounts of TNF-RI and trigger ROS production, resulting in compromised cell-intrinsic Mtb control. This further associates with the recruitment of Mtb-permissive CD64⁺CD11b⁺CD11c⁻Ly6G^low interstitial monocyte cell pool, exuberant Mtb replication, and early lethality despite abundant TNF expression.

Monocyte-stroma cross-talk through IL-1RI and TNF-RI is required for optimal resistance to Mtb

In order to define the source of IL-1 and TNF and the directionality of host-protective IL-1RI and TNF-RI signaling, we cross-transplanted WT mice with bone marrow from TNF or TNF-RI-deficient animals. In agreement with earlier studies, mice deficient in TNF or TNF-RI in hematopoietic cells were highly susceptible to Mtb (Figure 2A) (Flynn et al., 1995). We further found that Tnf^−/− and Tnfr1^−/− mice transplanted with WT bone marrow were also more susceptible to Mtb infection, compared to WT mice transplanted with WT bone marrow, demonstrating that TNF from and TNF-RI on stromal cells were also needed for optimal protection from Mtb (Figure 2A). Treatment of bone marrow-differentiated macrophages or immortalized mouse lung epithelial cells (MLg) with IL-1β, TNF, and/or heat-killed Mtb (as likely agonists during progressive Mtb infection), showed that combination of IL-1β+TNF or heat-killed Mtb+TNF resulted in the most potent activation of TNF and IL-1α, among other cytokines and chemokines in both cell types (Figures 2B and 2C).

Because IL-1α can function as a membrane-bound cytokine, we treated bone marrow macrophages with LPS or heat-killed Mtb and stained cells with anti-IL-1α Abs. This staining showed that both treatments produced IL-1α-specific staining on non-permeabilized cells (Figure 2D). To analyze if membrane-IL-1α is biologically active, we exposed reporter cell line LA4 (that produces CXCL1 chemokine in response to recombinant IL-1 (Figure 2E)) to bone marrow macrophages from WT, Casp1^−/−, or Casp1^−/−Il1a^−/− (as negative control) mice treated to produce membrane-bound IL-1α, and confirmed that membrane-bound IL-1α on macrophage cells was biologically active and its CXCL1-stimulating activity on LA4 cells could be blocked by anti-IL-1α antibodies (Figure 2F). Moreover, and similar to cell treatment with IL-1β+TNF and heat-killed Mtb+TNF (Figure 2C), MLg cells treated with a combination of TNF and macrophage cells presenting membrane-bound IL-1α led to the most potent activation of cytokines and chemokines when compared to individual treatments (Figure 2G). To ultimately test whether simultaneous IL-1RI and TNF-RI signaling to and from stromal cells is required for host resistance to Mtb in vivo, we cross-transplanted Il1r1^−/− or Il1r1^−/−Tnfr1^−/− mice with WT bone marrow. While Il1r1^−/− mice transplanted with WT bone marrow survived up to 100 days p.i (Figure 2H), WT bone marrow failed to rescue Il1r1^−/−Tnfr1^−/− mice, which succumbed to infection with median survival of 38 days. This median survival was significantly lower than that observed in Tnfr1^−/− mice transplanted with WT bone marrow (58 days), demonstrating that both IL-1RI and TNF-RI signaling on stroma cells are synergistically required for resistance to Mtb.

IL-1α mediates host resistance to Mtb in a non-redundant fashion

To define the role of individual IL-1RI ligands during the course of Mtb infection, we infected mice deficient in IL-1RI, IL-1α (Horai et al., 1998), IL-1αβ (Horai et al., 1998), or IL-1β (Shornick et al., 1996) with Mtb. Il1r1^−/− and Il1ab^−/− mice succumbed to infection with a median survival of 28 and 35 days (Figure 3A). There was a significant divergence in the survival phenotype between Il1a^−/− and Il1b^−/− mice, which succumbed to Mtb infection with a median survival of 95 and 152 days p.i., respectively (Figure 3A). These results were reproduced with more virulent Mtb Erdman strain and with highly-virulent W-Beijing SA161 strain (Figures S2A and S2B). Bacterial burden analysis in the lungs of WT mice showed a plateau of ~10⁶ CFU per lung at 30 days p.i., and infection was controlled at this level for over 200 days. In contrast, Il1r1^−/− and Il1ab^−/− mice reached a bacterial burden of 10⁸ CFU by 30 day p.i., and Il1a^−/− mice reached 10⁸ CFU by 90 days p.i. (Figure 3B). The Mtb burden in the lungs of Il1b^−/− and Casp1^−/−11^−/− mice was not different from mice in the control group (Figure S2C). Hematoxylin and eosin and acid-fast staining of lung sections revealed that by 60 days p.i. in WT mice, defined granulomas containing few Mtb bacilli were formed. In contrast, in Il1a^−/− mice lungs contained large inflammatory lesions with highly diffuse structure and cells with numerous bacilli were dispersed throughout the lung parenchyma (Figure 3C). Evaluation of lung histopathology revealed worsening of histopathology score (Figures 3D and S2D) and progressive decline in inflammation-free airway space in the lungs of Il1a^−/− mice (Figure 3E). Collectively, our analyses show that although within the first 5 weeks p.i. IL-1RI ligands can play compensatory roles in mediating host resistance to Mtb and mice deficient in either IL-1α or IL-1β survive beyond 35 days p.i., IL-1α and IL-1β play non-redundant roles at later times p.i. when deficiency in IL-1α compromises host resistance to a significantly greater degree, compared to deficiency in IL-1β.

Mice deficient in IL-1 ligands develop functional Mtb-specific adaptive immunity

The delayed susceptibility of Il1a^−/− mice to pulmonary Mtb infection could be explained by a requirement of IL-1α for an effective adaptive immune response (Ben-Sasson et al., 2009). To test this hypothesis, we used MHC class I and II tetramers to track CD8⁺ and CD4⁺ T cells recognizing Mtb immunodominant MHC class I (TB10.4_4–11:K^b) and MHC class II (ESAT-6_4–17:I-A^b) -restricted T cell responses in the lungs of WT, Il1a^−/−, and Il1ab^−/− mice. At 30 days and 60 days p.i., this analysis revealed similar frequencies of tetramer-binding CD4⁺ and CD8⁺ T cells in all groups (Figures 4A–D, S3). To determine the frequency of IFN-γ-producing T cells in vivo, we also performed direct ex vivo intracellular IFN-γ staining on freshly isolated lung lymphocytes in the absence of in vitro stimulation (Shafiani et al., 2010). This analysis revealed significantly higher proportions of IFN-γ-producing cells within the tetramer-binding CD4⁺ and CD8⁺ populations of Il1a^−/− and Il1ab^−/− mice, compared to control group, at 30 days p.i. (Figure 4B, D). The higher frequency of IFN-γ-producing T cells in Il1a^−/− and Il1ab^−/− mice demonstrate that T cells from these strains are not intrinsically deficient in their ability to express IFN-γ in vivo. In vitro stimulation of lung lymphocytes, freshly isolated from Mtb-infected WT and Il1a^−/− mice, with the anti-CD3 and anti-CD28 Abs, ESAT-6^4–17, or TB10.4^4–11 peptides did not reveal any differences in the proportions of IFN-γ or TNF-producing CD4⁺ and CD8⁺ T cells between the groups (Figure 4E, F). Therefore, these analyses show that the development of Mtb-specific adaptive immunity does not depend on IL-1 cytokines.

The lack of IL-1α-IL-1RI signaling leads to a non-protective highly inflammatory state and premature lethality

Gross examination of the Mtb infected mice showed that by 90 days p.i., the wet lung weight for Il1a^−/− mice exceeded that for WT animals by 180% (P<0.01, Figure 5A), and the number of mononuclear cells harvested from the lungs of Il1a^−/− mice was significantly greater than from the WT mice both at 60 and 90 days p.i. (P<0.01, Figure 5B).The frequency of Ly-6G⁺Ly-6C⁺ polymorphonuclear leukocyte (PMNs) population at this late time point was lower in Il1a^−/− mice (1.88%) compared to WT animals (3.14%, P<0.05, Figure 5C), excluding the possibility that the pulmonary pathology in Il1a^−/− mice is associated with neutrophilic inflammation. This analysis also revealed that the majority of leukocytes purified from the lungs of Il1a^−/− mice expressed Ly-6C marker, associated with inflammatory monocytes, macrophages, and dendritic cells (DCs) (De Trez et al., 2009; Mayer-Barber et al., 2011; Swirski et al., 2009). To better define the in vivo activation status of monocytes accumulated in the lungs of WT and Il1a^−/− mice after Mtb infection, we assessed isolated lung cells for expression of CD11b and CD11c, as well as Ly-6C and CD40 as markers associated with cellular activation. We observed a dramatic increase in Ly-6C expression by both CD11b⁺ and CD11c⁺ cell populations isolated from Il1a^−/− mice, compared to WT animals (P<0.01, Figure 5D). In addition, the population of CD11c^highLy-6C^int monocytes observed in WT was absent in Il1a^−/− mice (population a, Figure 5D). The analysis of CD40 expression on lung leukocytes showed that after Mtb infection, the proportion of highly activated CD11c^highCD40^high monocytes was over 2-fold higher in Il1a^−/− mice compared to WT animals (Figure 5E). Furthermore, a distinct population of CD11c^highCD40^low monocytes that was observed in WT animals was completely missing from the lungs of Il1a^−/− mice (Population b, Figure 5E).

Next, we generated bone marrow chimeras in which lethally irradiated WT mice were transplanted with either WT or Il1a^−/− bone marrow cells. Five months after hematopoietic reconstitution, chimeric mice were infected with Mtb and Ly-6C and CD40 marker expression on their CD11c⁺ cells were analyzed 90 days p.i.. We observed that CD11c⁺ cells in chimeric mice recapitulated the phenotypes observed in mice from which the donor bone marrow cells were derived (Figure 5F). Collectively, these data show that in IL-1α-deficient mice, the higher inflammatory state does not correlate with improved resistance to Mtb but rather is associated with progressive non-resolving monocytic inflammation that constrains free airway space and leads to premature lethality.

Restoration of IL-1α expression in CD11c⁺ cells of Il1a^−/− mice alleviates inflammation and improves survival after Mtb infection

To determine whether and what kind of Mtb-infected cells express IL-1α in the lungs of Mtb-infected mice, we analyzed sections of lungs from Mtb-infected mice stained with fluorescent IL-1α-specific antibody and polyclonal Mtb-specific antibody or antibodies for various cellular markers by confocal microscopy (Di Paolo et al., 2009). At 60 days p.i., those cells that stained positive for Mtb antigens also stained positive for IL-1α (Figure 6A), and IL-1α-staining co-localized with CD11c⁺ and CD11b⁺cells but not with CD3⁺ or Gr1+ cells (Figure 6B).

Because the majority of lung leukocytes including alveolar macrophages, immature DC, and activated DCs, express CD11c, we next determined whether the restoration of IL-1α expression specifically in CD11c⁺ cells would suffice to rescue the susceptibility of Il1a^−/− mice to Mtb infection. We utilized a foamy virus vector-based stable gene transfer approach (Josephson et al., 2002) (Figure S4A), where mouse IL-1α gene was expressed under the control of a minimal Itgax promoter (Ni et al., 2009). CD11c^highCD40^low monocyte population was restored in Mtb infected Il1a^−/− mice transplanted stem cells transduced with Itgax-IL-1α virus (Figure 6C) but not Itgax-stop-IL-1α virus (Figure 6D). Importantly, the median survival of Il1a^−/− mice transplanted with Itgax-stop-IL-1α vector-transduced stem cells was only 55 days, while 8 out of 9 Il1a^−/− mice transplanted with hematopoietic stem cells transduced with Itgax-IL-1α virus mice survived beyond 80 days after Mtb infection (P < 0.001, Figure 6E). Therefore, our data show that cell-type-specific restoration of IL-1α expression in CD11c⁺ cells in Il1a^−/− mice results in lung cell phenotypes comparable to WT mice, and prolonged survival after Mtb infection.

IL-1α is required for intracellular control of Mtb replication at a single cell level in vivo via cell extrinsic mechanism

To reveal the mechanism of IL-1α-mediated control of Mtb resistance, we analyzed the dynamics of Mtb replication in vivo. Quantification of Mtb bacilli on acid-fast stained sections of lungs revealed that the number of bacilli per high power view field rapidly increased and peaked by 30 days p.i. in all groups (Figure 7A). Of note, although granulomatous areas in the lungs expanded continuously over time (Figure 3D, E), by 60 days p.i., the number of bacilli per view-field drastically declined in WT mice and remained low until 100 days p.i., implying the effective control of Mtb. In contrast, the number of bacilli in lung sections was significantly higher for Il1a^−/− and Il1ab^−/− mice at 30 days p.i., compared to WT animals (Figure 7A), suggesting that these mice were deficient in their ability to control Mtb replication during early infection. Of note, as in WT animals, the number of bacilli per view-field in Il1a^−/− mice declined to a similar degree between days 30 and 60 p.i., but it increased significantly again by 100 days p.i.

Detailed evaluation of lung sections of Il1a^−/− and WT mice at day 60 p.i. revealed that in Il1a^−/− mice, more than 80% of individual Mtb-infected cells contained over 4 bacilli per cell, while in WT animals, only 45% of infected cells contained over 4 bacilli (Figure 7B, C). Furthermore, electron microscopy analysis showed that Mtb bacilli in Il1a^−/− mice could be found in clusters (Figures 7D and S5), suggesting bacterial replication in vivo. Using a previously described “Mtb replication clock” approach to quantify the rate of Mtb replication in vivo (Gill et al., 2009), we calculated Mtb replication rates in Il1a^−/− and WT mice. Between days 1 and 11 post infection, the Mtb population in the lungs of Il1a^−/− mice replicated at a rate of 0.77 per day, corresponding to a 31.17 hour generation time. In contrast Mtb population in the lungs of WT mice replicated at a significantly slower rate of 0.55 per day, corresponding to 43.64 hours generation time (P<0.01, Figure 7E).

We hypothesized that if the defect in Mtb replication control in Il1a^−/− mice is entirely cell-intrinsic, it should be possible to be reproduced in vitro. We differentiated macrophages from the bone marrow of WT or Il1a^−/− mice, infected them with Mtb, and analyzed the bacterial burden over time. We also treated Mtb-infected cells with IL-1α, IFN-γ, or a combination of both, to determine whether these cells were deficient in restricting Mtb replication even in the presence of these exogenous cytokines. We found that Mtb replicated equally efficiently in both WT and Il1a^−/− macrophages. Furthermore, in agreement with an earlier study, IFN-γ was highly effective at restricting Mtb replication in vitro (MacMicking et al., 2003), and WT and Il1a^−/− macrophages were equally responsive to IFN-γ and exhibited reduced Mtb CFUs (Figures S5B, C). The addition of IL-1α to Mtb-infected cells did not restrict bacterial growth. Taken together these data show that Il1a^−/− mice fail to control Mtb replication at a single-cell level and this inability to control Mtb replication is not cell intrinsic.

Experimental Animals

All studies were conducted in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals and the Institutional Animal Care and Use Committee guidelines of the University of Washington and Seattle Biomedical Research Institute where all the work involving animals was conducted. C57BL/6 mice were purchased from Charles River, Wilmington, MA. Il1r1^−/− and Tnf^−/− mice were purchased from Jackson Laboratory. Casp1^−/−11^−/− mice were kindly provided by Dr. R. Flavell (Yale University) and described in (Kuida et al., 1995); Il1a^−/−, Il1ab^−/− mice were described in (Horai et al., 1998). Il1b^−/− mice were described in (Shornick et al., 1996). All mice were on C57BL/6 genetic background, matched by age and housed in specific-pathogen-free facilities.

Bacterial Infection

Stock of Mtb strains H37Rv, H37Rv::pBP10 (Gill et al., 2009), Erdman, or W-Beijing SA161 were sonicated before use, and mice were infected in an aerosol infection chamber (Glas-Col). To determine the Mtb burden, at indicated times the left lung of each mouse was homogenized in PBS with 0.05% Tween 80. Ten-fold serial dilutions were made in PBS with 0.05% Tween 80 and plated on Mitchison 7H10 plates. Colonies were counted after 21 d of incubation at 37°C, and CFUs per organ were determined. Approximately 100 CFUs were deposited in the lungs of each mouse upon initial infection. Mice were sacrificed when the body weight declined by 20%, compared to the body weight of each animal at the day of infection.

Histology and immunofluorescence staining

For hematoxylin and eosin, acid-fast Kinyoun's, and electron microcopy studies, the right lung was excised, fixed in 4% paraformaldehyde (PFA) for 1 week at room temperature, embedded in paraffin, and sectioned by microtome in consecutive sections of 5 µm. For confocal immunofluorescence, staining was performed on frozen sections. For PI analyses, dry frozen sections were evaluated under a fluorescent microscope without further processing.

Cell Isolation, Staining and Flow Cytometry

Briefly, single-cell suspensions of intraparenchymal lung lymphocytes were stained at saturating conditions using antibodies specific for CD3 (145-2C11), CD4 (RM4-5), CD8 (53-6.7), CD11c (HL), CD11b (M1/70), Gr-1 (RB6-8C5), CD45.2 (104), CD45.1 (A20), Ly-6C (HK1.4), Ly-6G (1A8), CD40 (1C10) all from BD Biosciences. IFN-γ and TNF intracellular staining was performed using a kit as instructed by the manufacturer (BD), with minor modifications. For direct ex-vivo detection of IFN-γ, lung cells were isolated in the presence of Brefeldin A (Sigma) and staining for surface markers and intracellular IFN-γ was performed as described, without in vitro restimulation. A minimum of 100,000 live cells per sample was acquired on an LSRII instrument using the FACSDiva software (BD Biosciences). The samples were then analyzed by Flowjo Software (Tree Star, Inc.). TNF-RI and mCherry analyses were done using CD64 and CD11c-gating approach (Figure S1A). ROS staining was analyzed using alternate CD11b and CD11c-gating approach (Figure S1D) to avoid spectral overlap between ROX reagent and CD64 Ab staining.

Detection of ESAT-6^4–17– and TB10.4^4–11–specific cells

PE-labeled MHC class II tetramers (I-A^b) containing the stimulatory residues 4 to 17 (QQWNFAGIEAAASA) of the early secreted antigenic target 6 kD (ESAT-6) of Mtb and APC-labeled MHC class I tetramers (K^b) containing the stimulatory residues 4 to 11 (IMYNYPAM) of the low molecular weight protein antigen TB10.4 of Mtb (obtained from the National Institutes of Health Tetramer Core Facility) were used to detect Mtb-specific CD4⁺ and CD8⁺ T cells, respectively.

Statistical Analysis

All experiments were repeated at least twice and a minimum number of samples analyzed per each experimental group was equal to three or more. Results are expressed as mean and standard error. Unless otherwise indicated, Student’s unpaired two-tailed t test was used for comparing experimental groups, with P < 0.05 considered significant.