Alveolar macrophages generate a noncanonical NRF2-driven transcriptional response to Mycobacterium tuberculosis in vivo

See allHide authors and affiliations

Science Immunology  26 Jul 2019:
Vol. 4, Issue 37, eaaw6693
DOI: 10.1126/sciimmunol.aaw6693

Interfering with macrophage activation

One of the many reasons for Mycobacterium tuberculosis (M.tb.) persistence is that the bacterium blocks activation of M.tb.-specific immune responses. Here, using a mouse model of tuberculosis, Rothchild et al. demonstrate that M.tb.-infected alveolar macrophages in the lung show an impaired ability to drive antibacterial responses. These M.tb.-infected macrophages up-regulate an antioxidant transcription signature that is driven by the transcription factor nuclear factor erythroid 2–related factor 2 (NRF2). By infecting mice engineered to lack NRF2 in myeloid cells with M.tb., they show that loss of NRF2 in myeloid cells lowered bacterial burdens. Their studies have brought to the fore the importance of NRF2 in regulating early interactions between M.tb. and host immune cells.


Alveolar macrophages (AMs) are the first cells to be infected during Mycobacterium tuberculosis (M.tb.) infection. Thus, the AM response to infection is the first of many steps leading to initiation of the adaptive immune response required for efficient control of infection. A hallmark of M.tb. infection is the slow initiation of the adaptive response, yet the mechanisms responsible for this are largely unknown. To study the initial AM response to infection, we developed a system to identify, sort, and analyze M.tb.-infected AMs from the lung within the first 10 days of infection. In contrast to what has been previously described using in vitro systems, M.tb.-infected AMs up-regulate a cell-protective antioxidant transcriptional signature that is dependent on the lung environment but not bacterial virulence. Computational approaches including pathway analysis and transcription factor motif enrichment analysis identify NRF2 as a master regulator of the response. Using knockout mouse models, we demonstrate that NRF2 drives expression of the cell-protective signature in AMs and impairs the control of early bacterial growth. AMs up-regulate a substantial pro-inflammatory response to M.tb. infection only 10 days after infection, yet comparisons with bystander AMs from the same infected animals demonstrate that M.tb.-infected AMs generate a less robust inflammatory response than the uninfected cells around them. Our findings demonstrate that the initial macrophage response to M.tb. in the lung is far less inflammatory than has previously been described by in vitro systems and may impede the overall host response to infection.


Infection with Mycobacterium tuberculosis (M.tb.) is now the leading cause of death by a single pathogen worldwide (1). Transmission of M.tb. generally occurs via aerosols expelled by cough from an infected person. Although inhaled M.tb. is rapidly engulfed by alveolar macrophages (AMs) in the lung (2), mobilization of a robust immune response is extremely slow and the lack of a speedy response is thought to contribute to disease progression (3). In animal models, recruitment of innate cells such as neutrophils and monocytes to the lung and T cell priming in the draining mediastinal lymph nodes do not occur until 11 to 14 days after infection, by which time the bacteria have replicated nearly 1000-fold in the lung (2, 47). The mechanisms that restrain or subvert the host response in the lung immediately after infection remain largely unknown.

AMs are the first cells in the lung to be infected with M.tb. after aerosol transmission (2). AMs are the largest resident macrophage population in the lung and serve as pulmonary immune sentinels, constantly sampling the airway for foreign particles (8). In addition, AMs perform a critical homeostatic role clearing inhaled material from the airway and recycling pulmonary surfactant, which requires AMs to uncouple phagocytic function from inflammatory responses (9, 10). Because of their steady-state functions, AMs express unique transcriptional and epigenetic profiles that are highly distinct from those of other tissue-resident macrophages (1113).

Similar to other macrophage populations that are constantly exposed to environmental stimuli, AMs express a number of inhibitory receptors that have been shown to dampen their responses. However, under conditions such as acute lung injury, AMs can become highly activated and release damaging levels of pro-inflammatory mediators that contribute to airway disease (8). During other pulmonary infections such as influenza, in which AMs serve as secondary or tertiary responders, AMs exhibit pro-inflammatory transcriptional responses and can serve as an immunotherapy target to limit inflammation (14, 15). However, during these infections, the AMs are likely influenced by cytokines released into the pulmonary environment by other cell types.

Unlike most pulmonary infections, the initial infectious dose of M.tb. is very low, and for the first several days, AMs are the only cells productively infected with the bacteria (2). Therefore, M.tb. infection provides an opportunity to assess the ability of AMs to respond to direct intracellular infection rather than environmental cues. Results from two recent studies provide hints that the response of AMs to M.tb. might be suboptimal: At 2 weeks after infection, when M.tb. can be found in multiple cell populations, AMs are significantly less effective at controlling bacterial replication than interstitial macrophages (IMs), another macrophage subset found in the lung (16), and AMs were shown to facilitate relocalization of the bacteria from the airway into the lung interstitium (2).

Because of the paucity of M.tb.-infected AMs during the first few days after aerosol challenge, their presumed immediate response to the bacteria has been extrapolated from in vitro culture systems using models such as bone marrow–derived macrophages (BMDMs). These studies have identified numerous sensing pathways and cytokine responses, including tumor necrosis factor (TNF), type I interferon (IFN), and interleukin-1β (IL-1β) (1722) that have been shown to play crucial roles in host immunity in both animal and human studies (23). However, experiments using knockout mice have demonstrated that these mediators do not affect the course of disease during the first week of infection in vivo (2426), despite the fact that they are up-regulated within hours by M.tb.-infected macrophages in vitro.

To more precisely define the response of AMs to M.tb. in vivo, we developed an infection model and isolation procedure that yields sufficient numbers of infected AMs to perform global systems-level analyses at any point within the first 10 days after infection. Our results demonstrate that the rapid up-regulation of pro-inflammatory genes observed for macrophages responding to M.tb. in vitro is delayed in AMs infected in vivo, occurring days instead of hours after infection. Immediately after in vivo infection, AMs up-regulate a cell-protective antioxidant transcriptional program regulated by the transcription factor nuclear factor erythroid 2–related factor 2 (NRF2). Activation of this program is dependent on the lung microenvironment. In the absence of NRF2, AMs have an enhanced ability to control early bacterial growth.


AMs provide a replication niche for M.tb. through the first 10 days of infection

In a standard low-dose M.tb. aerosol challenge model, mice are infected with ~100 colony-forming units (CFU), which, because of the low replication rate of the bacteria, does not allow for the isolation of sufficient numbers of M.tb.-infected cells from a reasonable number of animals for analysis. Therefore, we developed a high-dose model in which mice are infected with 2 × 103 to 4 × 103 CFU of an mEmerald-expressing H37Rv strain of M.tb. Using this model, we found that, over the first 2 weeks, the vast majority of infected cells early on were AMs, with significant numbers of infected neutrophils [polymorphonuclear (PMN)] and monocyte-derived macrophages (MDMs) appearing between day 10 and day 14 (Fig. 1A, figs. S1 and S2, and table S1). The localization of the bacteria and the timing of recruitment of innate immune cells in the high-dose model are consistent with other studies that have used a low-dose model (2, 5), suggesting that the higher dose does not alter the initial immune response in the lung. At this dose, ~0.5 to 2% of AMs were infected (Fig. 1B) and the total number of M.tb.-infected AMs did not change significantly over the first 2 weeks (Fig. 1C). By day 14, the numbers of infected neutrophils and AMs were equivalent.

Fig. 1 AMs provide a replication niche for M.tb. through the first 10 days of infection.

(A) Percentage of total mEmerald+ cells, (B) percentage of mEmerald+ for each cell population, and (C) total number of mEmerald+ cells in the lung between 2 hours and 14 days after high-dose aerosol infection with mEmerald-tagged H37Rv (n = 3 mice per time point). (D) Microscopy of BAL samples 1 and 10 days after high-dose aerosol infection with mEmerald-tagged H37Rv and quantitation of bacteria per AM (n = 3 replicates per time point, each replicate was pooled from three mice). Eosin, eosinophils. Data are means ± SEM and are representative of two independent experiments.

We found that M.tb.-infected AMs were enriched within bronchoalveolar lavage (BAL) samples compared with the total lung (fig. S3) and used high-throughput microscopy to analyze BAL samples from mice after high-dose infection with mEmerald-H37Rv to quantify the number of bacteria in each infected cell. At 1 day after infection, 81.3 ± 1.8% (mean ± SEM) of infected AMs contained a single bacillus; by day 10, only 13.2 ± 2.3% (mean ± SEM) of AMs contained a single bacillus, with 58.4 ± 6.1% of AMs containing two to five bacilli and 28.4 ± 3.8% containing more than five bacilli (Fig. 1D). These data demonstrate that AMs serve as a replication niche for M.tb. through the first 10 days of infection, even in the setting of a high-dose infection, an observation that has been similarly described at 1 and 2 weeks after infection in other studies (2, 16).

M.tb.-infected AMs up-regulate an NRF2-associated cell-protective response in vivo

To characterize the early macrophage transcriptional response to M.tb. infection, we infected mice with a high dose of mEmerald-H37Rv and isolated AMs by expression of both Siglec F and CD64 (27, 28) from BAL by fluorescence-activated cell sorting (FACS) 24 hours after infection (figs. S3 and S4). Isolating AMs by BAL allowed the samples to be kept on ice throughout processing and sorting, eliminating the need for lengthy and harsh digestion of lung tissue. Three populations were sorted for analysis by RNA sequencing (RNA-Seq): M.tb.-infected AMs (mEmerald+), bystander AMs (mEmerald), and AMs from naïve mice (Fig. 2A).

Fig. 2 M.tb.-infected AMs up-regulate an NRF2-associated antioxidant gene signature.

(A) Gating scheme to sort naïve, bystander, and M.tb.-infected AMs from BAL samples after high-dose aerosol infection with mEmerald-tagged H37Rv. (B) Heatmap of gene expression (log2 fold change over average of naïve AMs) for 196 DE genes between naïve and M.tb.-infected AMs (filtering criteria: average CPM > 1, |fold change| > 2, and FDR < 0.01, Benjamini-Hochberg calculated). Columns are independent experiments (pooled mice), and rows are genes. Genes called out are known NRF2 target genes of interest as well as down-regulated pro-inflammatory genes. Colored bars to the left indicate NRF2 association as determined by three different methods: (C) IPA, (D) transcription factor promoter binding motif enrichment analysis (HOMER), and (E) ChIP-Seq peak analysis. (F) Quantitative reverse transcription PCR (RT-qPCR) validation of NRF2-associated genes for naïve, bystander, and M.tb.-infected AMs 24 hours after infection. Values are relative to Ef1a. Data are presented from three independent experiments with one-way ANOVA with Dunnett’s post hoc test. *P < 0.05, **P < 0.01, ***P < 0.001.

At 24 hours after infection, we identified 196 genes that were differentially expressed (DE) [average counts per million (CPM) > 1, |fold change| > 2, false discovery rate (FDR) < 0.01] between M.tb.-infected and naïve AMs (Fig. 2B and table S3). No genes were DE between bystander and naïve AMs at this stringency, indicating an absence of systemic changes to the lung 24 hours after infection.

To identify enriched pathways and potential transcriptional regulators, we analyzed the up-regulated (131) and down-regulated (65) genes independently using three complementary computational approaches. Ingenuity Pathway Analysis (IPA) revealed that the up-regulated genes were significantly enriched for the “NRF2-mediated oxidative stress response” pathway (P = 10–11.3) [Fig. 2, B (green squares) and C]. This pathway includes genes involved in a number of cell-protective functions, including antioxidant production (Nqo1, Cat, Prdx1, and Txnrd1), iron metabolism (Hmox1), and glutathione metabolism (Gstm1, Gclm, and Gsta3) (29). We also searched for enriched transcription factor promoter binding motifs using HOMER (Hypergeometric Optimization of Motif EnRichment) (30). The top four enriched transcription factor binding motifs in the 131 up-regulated genes (q = 0.0006) were motifs for either NRF2 or BACH1, which are known to compete for the same binding sites [Fig. 2, B (yellow squares) and D] (31). To validate these results, we used a published NRF2 chromatin immunoprecipitation sequencing (ChIP-Seq) analysis of peritoneal macrophages that were stimulated with the NRF2 agonist diethyl maleate (GSE75177) (32). Twenty-nine percent of the 131 up-regulated genes contained an NRF2 ChIP-Seq peak compared with 1.5% of background genes. The difference in proportions is highly significant [χ2 (1, N = 10,946) = 531.21, P < 2.2 × 10−16] [Fig. 2, B (orange squares) and E]. We validated the up-regulation of several of the NRF2-associated genes by quantitative polymerase chain reaction (qPCR) (Fig. 2F). These results demonstrate that the transcription factor NRF2 is associated with the up-regulated transcriptional signature in M.tb.-infected AMs 24 hours after infection in vivo. A similar analysis of the down-regulated genes did not uncover any significantly enriched pathways or candidate transcriptional regulators.

M.tb.-infected AMs do not immediately up-regulate classical pro-inflammatory genes

We examined whether, in addition to NRF2-associated genes, M.tb.-infected AMs up-regulated classical pro-inflammatory genes that have been shown to be expressed by various macrophage subsets in response to M.tb., including cytokines and chemokines (e.g., Tnf, Il1b, and Cxcl9) (23), pro-inflammatory receptors, costimulatory molecules, and Fc receptors. Very few of these genes were significantly up-regulated by M.tb.-infected AMs (fig. S5). In contrast, several genes involved in glycolytic metabolism (Hif1a, Pkm, and Aldoa) were significantly up-regulated by M.tb.-infected AMs (33). Although previous studies have shown that, at 2 weeks after infection, M.tb.-infected AMs have relatively higher expression of genes involved in fatty acid oxidation and lower expression of genes involved in glycolysis compared with M.tb.-infected IMs (16), our data demonstrate that M.tb.-infected AMs do up-regulate genes involved in glycolysis relative to their naïve state.

Virulent M.tb. is not required to activate the NRF2-associated signature generated by AMs

To determine whether the NRF2 signature was specifically induced by virulent M.tb., we repeated the high-dose infection using an mEmerald-expressing avirulent strain of M.tb. lacking the RD1 virulence locus (ΔRD1-H37Rv) (fig. S6A). The ΔRD1 strain lacks a functional ESX-1 secretion system, which has been shown to contribute to pathogenicity in vivo (3436). Macrophage in vitro studies have demonstrated that the RD1 region is required for M.tb.-mediated activation of host cytosolic receptors and production of type I IFN (17, 37) as well as inflammasome activation and IL-1β production (38, 39). As an additional control, we exposed mice to fluorescent 1-μm carboxylate-coated latex beads by aerosol treatment (fig. S6B). To confirm that the beads were not contaminated with Toll-like receptor ligands, BMDMs were treated with the beads for 8 hours and gene expression changes were measured. No significant transcriptional changes in pro-inflammatory or NRF2-associated genes were observed in vitro (fig. S7).

Both the ΔRD1-infected and bead-positive AMs displayed up-regulation of NRF2-associated genes at 24 hours, although the magnitude of the fold change for some genes was smaller than that for H37Rv-infected AMs (Fig. 3A and table S3). Unlike in vitro models that have shown a role for the ESX-1 locus, contained within RD1, in regulating the type I IFN responses of macrophages to M.tb. (17, 37, 40), there were very few significant differences between the transcriptional responses of H37Rv-infected versus ΔRD1-infected AMs 24 hours after infection (Fig. 3B). The NRF2-mediated oxidative stress response was the most highly enriched pathway in ΔRD1-infected AMs by IPA (table S2). On the basis of the stringencies imposed on the 24-hour H37Rv-infected AMs, no genes were significantly changed in the bead-positive AMs and, similarly, no pathways were enriched by IPA (Fig. 3C and table S2). However, it is worth noting that, within the most changed genes (with filtering criteria of FDR < 0.05, fold change > 1.5), NRF2-associated genes showed the greatest enrichment. Overall, these data demonstrate that AMs up-regulate the NRF2-associated signature as a response to virulent bacteria, avirulent bacteria, or inert beads, suggesting that it is a more general response by AMs to the uptake of particles. Although it is not surprising that particle uptake leads to transcriptional changes, it is notable that infection with a virulent pathogen appears to induce no additional host response within the first day of infection, suggesting that classical pathogen sensing is deficient, inhibited, or delayed in AMs.

Fig. 3 Up-regulation of the NRF2-associated signature does not require virulent M.tb. infection.

(A) Heatmap of log2 fold change gene expression over average of naïve AMs for H37Rv-infected, ΔRD1-infected, and bead-positive AMs 24 hours after treatment. Columns represent averages of three independent experiments. Rows represent 196 DE genes described in Fig. 2. Colored bars to the left indicate NRF2 association as described in Fig. 2. (B and C) Scatterplots comparing gene expression values [log2 fold change (FC) over average of naïve AMs] for H37Rv-infected versus ΔRD1-infected AMs (B) or H37Rv-infected versus bead-positive AMs (C) 24 hours after infection, with significant DE genes highlighted (average CPM > 1, |fold change| > 2, and FDR < 0.01, Benjamini-Hochberg calculated). Data are presented from three independent experiments.

M.tb.-infected AMs exhibit both NRF2-associated and pro-inflammatory responses 10 days after infection

To characterize the kinetics of the AM response to M.tb. infection, we extended our transcriptional analysis to include four additional time points: 0.5 (12 hours), 2, 4, and 10 days after infection. We identified 252 genes that were significantly up-regulated at one or more of these time points compared with naïve AMs (Fig. 4A and table S3). Many of the 131 genes up-regulated at 24 hours and associated with NRF2 by either IPA, transcription factor motif analysis, or ChIP-Seq analysis showed sustained expression through 10 days of infection (Fig. 4A, top). Furthermore, by IPA, the NRF2-mediated oxidative stress response pathway was the most highly enriched pathway at all five time points (P = 10–5.3 to 10–10.1) (table S2). HOMER analysis also pinpointed an NRF2 motif as the most enriched transcription factor motif at 2, 4, and 10 days and identified no enriched motif at 0.5 days (table S2).

Fig. 4 M.tb.-infected AMs induce both NRF2-associated and pro-inflammatory responses 10 days after infection.

(A) Heatmap of gene expression (log2 fold change over naïve AMs) for 252 genes up-regulated in M.tb.-infected AMs compared with naïve AMs for at least one of five time points (filtering criteria: average CPM > 1, |fold change| > 2, and FDR < 0.01, Benjamini-Hochberg calculated). Top 131 genes are significantly up-regulated in M.tb.-infected AMs at 1 day after infection. Bottom 121 genes are not significantly up-regulated in M.tb.-infected AMs at 1 day after infection. Colored bars indicate NRF2 association as described in Fig. 2. (B and C) GSEA and top 50 ranked leading edge genes in the “TNFA signaling via NFkB” pathway for M.tb.-infected AMs at 10 days after infection. Data are presented from three independent experiments per time point.

The kinetic analysis also identified a late pro-inflammatory response up-regulated in M.tb.-infected AMs primarily at 10 days after infection (Fig. 4A, bottom). At 10 days after infection, M.tb.-infected AMs displayed significant up-regulation of genes in the TNFA signaling via nuclear factor κB (NFκB) pathway as determined by gene set enrichment analysis (GSEA) compared with naïve AMs [normalized enrichment score (NES) = 1.8, FDR = 0.0013], a pathway not significantly enriched in M.tb.-infected AMs 1 day after infection (Fig. 4B). The ranked leading edge genes in this pathway reveal the changes in gene expression over time, including increases in expression of Il1a, Tnf, Rel, Relb, NFkb2, Ccl2, Bhlhe40, and Nfe2l2 (Fig. 4C). A number of pro-inflammatory cytokine and chemokine genes (Tnf, Il1a, Cxcl2, Cxcl3, and Ccl17) are significantly up-regulated in M.tb.-infected AMs only 10 days after infection (fig. S8).

Bystander AMs express a pro-inflammatory transcriptional signature 10 days after infection

To disentangle the responses to intracellular infection from the responses to the inflammatory milieu in the lung, we analyzed the transcriptomes of the bystander AMs (uninfected cells from infected animals) at 1 and 10 days after infection. As described above, bystander AMs displayed no significant gene expression changes compared with naïve AMs at 1 day after infection (Fig. 5A). In contrast, by 10 days after infection, bystander AMs showed abundant changes in gene expression with a total of 205 significantly changed genes (Fig. 5B). Twenty-eight of these genes (highlighted in blue) were shared with the M.tb.-infected AMs (Fig. 5C), whereas 177 of them (highlighted in red) were uniquely changed in bystander but not in M.tb.-infected AMs (Fig. 5E). Another 200 genes (highlighted in purple) were DE only in M.tb.-infected AMs (Fig. 5D). Comparison between these sets of genes demonstrates that expression changes found only in M.tb.-infected AMs are enriched for NRF2-associated genes (Fig. 5D), whereas expression changes found only in bystander AMs are enriched for inflammatory pathways (Fig. 5E). These differences are confirmed by IPA. Although NRF2-mediated oxidative stress response and phosphatase and tensin homolog (PTEN) signaling were more highly enriched in M.tb.-infected AMs compared with bystander AMs, bystander AMs showed differential expression for genes in a number of other pathways, including calcium signaling, NFAT (nuclear factor of activated T cells) regulation of the immune response, role of pattern recognition receptors in recognition of bacteria and viruses, and STAT3 (signal transducer and activator of transcription 3) pathway (Fig. 5F). Enrichment of these pathways indicates that some of the inflammatory signals received by bystander cells are either blocked or integrated differently in infected cells 10 days after infection. Overall, these data suggest that, over the first week and a half, AMs respond directly to infection as well as to systemic changes in the lung environment and that these two signals may cross-regulate.

Fig. 5 Bystander AMs express a unique transcriptional signature 10 days after infection.

(A and B) Scatterplots comparing gene expression values (log2 fold change over average of naïve AMs) for H37Rv-infected versus bystander AMs at 1 day (A) and 10 days (B) after infection, with significant DE genes highlighted (|fold change| > 2 and FDR < 0.01, Benjamini-Hochberg calculated). Data are presented from three independent experiments. (C to E) Heatmaps of log2 fold change gene expression at 10 days over average of naïve AMs. Colored bars indicate NRF2 association as described in Fig. 2. (C) Twenty-eight genes DE by both bystander and M.tb.-infected AMs. (D) Two hundred genes DE only by M.tb.-infected AMs. (E) One hundred seventy-seven genes DE only by bystander AMs. Columns represent the average of three independent experiments. Genes of interest are noted to the right. (F) IPA comparing gene expression from bystander AMs and M.tb.-infected AMs. Canonical pathways with |z scores| > 1 and P values < 0.05 were reported.

Both cell-intrinsic and environmental factors shape the AM response to M.tb.

To determine whether the AM response to M.tb. is cell intrinsic or environment dependent, we compared the response of AMs infected in vitro with the in vivo measurements described above. We isolated AMs by BAL from naïve wild-type (WT) mice, allowed them to adhere for 18 hours, infected them with H37Rv, and measured their transcriptional response and ability to control bacterial growth. In parallel, we performed identical experiments with BMDMs, which have been used extensively, including by our group, to investigate how macrophages respond to M.tb. (41). Similar to their response to M.tb. infection in vivo, AMs displayed little to no increase in pro-inflammatory gene expression (including Il1b, Il6, and Nos2) after infection in vitro, whereas BMDMs robustly up-regulated these genes (Fig. 6A). One notable exception was Tnf, which was significantly up-regulated by AMs in response to H37Rv infection in vitro. In contrast to AMs infected in vivo, neither AMs nor BMDMs infected in vitro up-regulated NRF2-associated genes (Fig. 6B). Similar gene expression results were observed for AMs cultured with or without granulocyte-macrophage colony-stimulating factor (GM-CSF) (fig. S9) and at 12- or 24-hour time points after infection, matching the time points measured in vivo (fig. S10). Overall, AMs were more permissive to bacterial growth than BMDMs, leading to a significant increase in bacterial burden as measured by CFU 5 days after infection (Fig. 6C). These results suggest that the inability of AMs to up-regulate pro-inflammatory genes in response to intracellular infection is cell intrinsic, whereas the up-regulation of the NRF2-associated pathway is dependent on signals from the lung microenvironment.

Fig. 6 The AM response to M.tb. is driven by both cell type and environment.

In vitro H37Rv infection of AMs and BMDMs. (A) RT-qPCR gene expression analysis of pro-inflammatory genes 8 hours after infection. (B) RT-qPCR analysis of NRF2-associated genes 8 hours after infection. (C) CFU assay to measure bacterial burden in each cell type over 5 days. Data are representative of three independent experiments with three technical replicates each. Multiple t tests with Holm-Sidak correction. *P < 0.05, ***P < 0.001.

Expression of NRF2 impairs the ability of AMs to control bacterial growth

Our computational analyses identified NRF2 as a potential regulator of the in vivo AM response to M.tb. infection. To test this prediction, we isolated naïve and M.tb.-infected AMs from Nrf2−/− mice 24 hours after high-dose infection with mEmerald-H37Rv and performed RNA-Seq. Deletion of NRF2 altered the expression of only a small number of genes in the basal transcriptional profile of AMs [only 11 genes were DE (|fold change| > 2, FDR < 0.05), including Nfe2l2, the gene encoding NRF2, and several known NRF2 targets (Srxn1, Hmox1, and Slc11a1)] (fig. S11 and table S4). In contrast, the response of Nrf2−/− AMs to M.tb. infection was strongly attenuated compared with that of WT AMs, and many of the genes exhibiting altered responses were associated with NRF2 by IPA, transcription factor motif analysis, or ChIP-Seq analysis (Fig. 7A, group 2, and table S4). The set of genes that responded to M.tb. infection in Nrf2−/− AM was completely contained within the set of 196 genes that responded to M.tb. infection in WT AMs (|fold change| > 2, FDR < 0.01), suggesting that NRF2 does not act to restrain the transcriptional response in this setting, as has been reported previously (42). There was a small set of pro-inflammatory genes (i.e., Mx1, Cxcr1, and Ifit2) that were up-regulated by both WT and Nrf2−/− AM (Fig. 7A, group 1).

Fig. 7 Modulation of NRF2 activity alters macrophage response and control of M.tb.

(A) Heatmap of gene expression (log2 fold change over average of respective naïve AMs) for WT (first column) and Nrf2−/− (second column) M.tb.-infected AMs, averaged from at least two independent experiments. Rows depict all genes that are DE between naïve and M.tb.-infected WT AMs, as shown in Fig. 2B (181 genes expressed out of original 196 gene list). Groups 1 and 2 are defined in the text. Colored bars indicate NRF2 association as described in Fig. 2. (B) Lung bacterial burden measured by CFU assay at 10 days after infection with low-dose H37Rv from Nrf2LysM, Nrf2CD11c, and their respective Nrf2fl littermate controls. (C) MHC II mean fluorescence intensity (MFI) of Zombie Violet (ZV) M.tb.-infected AMs (top) and percentage of dead (Zombie Violet+) M.tb.-infected AMs (bottom) as measured by flow cytometry at 10 days after infection with high-dose mEmerald-H37Rv from Nrf2LysM, Nrf2CD11c, and their respective Nrf2fl littermate controls. Data are presented from two independent experiments (A) or are representative of two independent experiments with five mice per group (B and C). Two-tailed unpaired Student’s t test. *P < 0.05.

NRF2 functions as a master regulator of an antioxidant stress response and likely plays critical roles in many cell types. Flow cytometric analysis of the lungs of naïve Nrf2−/− mice showed altered major histocompatibility complex II (MHC II) expression in AMs and DCs (dendritic cells), decreased numbers of MDMs, and an increased percentage of CD8+ T cells (fig. S12); therefore, we chose to evaluate the functional role of NRF2 in AMs during M.tb. infection by generating NRF2 conditional knockout mice. Although there is no model for AM-specific gene deletion, we used both LysMcre/+ and CD11ccre/+ strains that delete floxed genes in either macrophages and neutrophils or AMs and DCs, respectively (43), in combination with an Nrf2floxed/floxed strain. These breedings generated the following conditional knockout models: Nrf2floxed/floxed; LysMcre/+ (Nrf2LysM) and Nrf2floxed/floxed; CD11ccre/+ (Nrf2CD11c). We confirmed the deletion of Nrf2 transcript and the absence of NRF2 protein in AMs in both strains by sequencing and Western blot (fig. S13), as previously described (44, 45). Analysis of myeloid cell number and activation, T cell frequencies and activation, and pro-inflammatory cytokine protein expression identified only one significant difference between naïve lungs of Nrf2CD11c and Nrf2fl controls: increased MHC II expression in AMs lacking NRF2 expression. There was a similar trend for AMs from Nrf2LysM mice compared with Nrf2fl littermate controls, but it did not reach significance (fig. S14). We repeated the transcriptional profiling on M.tb.-infected AMs from Nrf2LysM, Nrf2CD11c, and Nrf2fl littermate controls and found similar differences in gene expression compared with what had been observed between M.tb.-infected WT and Nrf2−/− AMs (fig. S15 and table S5). We examined the bacterial burden in these mice 10 days after infection, the latest time point when the vast majority of bacteria are still within AMs, and found that both conditional knockout models had lower bacterial burdens than their Nrf2fl littermate controls (Fig. 7B). M.tb.-infected AMs lacking NRF2 expression had higher levels of MHC II surface expression, reflecting the differences observed at baseline, and were more prone to cell death as measured by increased staining with viability dye Zombie Violet (Fig. 7C). The differences in cell viability were limited to M.tb.-infected cells and not observed in bystander cells (fig. S16). These results suggest that NRF2 may impede host control by limiting baseline AM activation and by blocking cell death of M.tb.-infected AMs, which could allow the bacteria more time to replicate in a permissive cell type and prevent M.tb. from being taken up by cells with more bactericidal capacity.


Here, we describe an NRF2-driven transcriptional signature generated by M.tb.-infected AMs in vivo. This innate response to M.tb. is very different from what has been previously observed in macrophage infection studies, including our own, that have shown that M.tb.-infected macrophages in vitro up-regulate pro-inflammatory genes primarily driven by NFκB and type I IFN signaling (17, 19, 37, 41). This dissimilarity is somewhat unexpected given that AMs express almost all of the canonical pathogen-sensing machinery thought to be required to detect M.tb. infection (46). Our data suggest that the initial host response to M.tb. infection is relatively anti-inflammatory and may contribute to the extended time required for antigen transport to the draining lymph node and subsequent activation of the adaptive immune response. These results complement previous studies that have shown that the earliest events in the lung, particularly the interactions between macrophages and M.tb., have a major impact on subsequent disease outcome (2, 4, 16, 4749). Although NRF2 and several of its target genes, notably Hmox1, have been shown to be up-regulated during chronic M.tb. infection (5052), our results suggest that the early NRF2 signature in AMs is a more general response to phagocytic activity rather than to bacterial infection. In the context of a meta-analysis of 32 studies covering 77 different host-pathogen interactions, this NRF2-driven transcriptional response falls far outside the norm (53). Our results highlight the importance of using in vivo systems to dissect lung-specific immunological events and suggest that established in vitro models do not adequately replicate the initial host response to M.tb. (54).

Studies using Nrf2−/− mice have demonstrated both protective (Streptococcus pneumoniae, Staphylococcus aureus, respiratory syncytial virus, and Pseudomonas aeruginosa) and harmful (Haemophilus influenzae and Marburg virus) roles for NRF2 during infection (5561). To evaluate the functional role of NRF2 in AMs during M.tb. infection, we chose to use two conditional knockout models (LysMcre and CD11ccre). One caveat is that cre expression by these promoters also deletes NRF2 in either neutrophils and MDMs (LysM) or DCs (CD11c). During the first 10 days of infection when M.tb. almost exclusively infects AMs, this approach allows us to interrogate the direct role of NRF2 in M.tb.-infected AMs. Conversely, beyond 10 days, significant numbers of neutrophils, MDMs, and DCs begin to participate in the response, making it impossible to link changes in the outcome of infection at later times to AM-specific NRF2 expression. In addition to infection, NRF2 can be activated by lung injury caused by noncommunicable conditions such as chronic obstructive pulmonary disease, cigarette smoking, cystic fibrosis, and exposure to air pollution (62, 63). The finding that NRF2 expression hinders AMs from controlling intracellular infection suggests an intriguing mechanism by which smoking and indoor air pollution may contribute to tuberculosis (TB) risk (64, 65).

The finding that AMs do not initially mount an inflammatory response to M.tb. infection is supported by bacterial transcriptional profiling we performed in collaboration on M.tb. residing within infected AMs in vivo using a novel bead-based method, Path-Seq, to enrich for bacterial transcripts (66). The study by Peterson et al. (66) shows that, unlike infection of BMDMs in vitro, which causes M.tb. to dynamically regulate mycolic acid biosynthesis and increase expression of environmental stress regulators such as DosR, infection of AMs leads to fewer changes to the stress-related networks in the bacterial transcriptome. Targeting AMs may be one of M.tb.’s most important virulence strategies. To spread effectively after aerosol transmission of small numbers of bacteria, M.tb. targets a macrophage population that does not activate classical innate inflammatory pathways and instead responds by safeguarding against cellular damage. Our findings provide one explanation for the “macrophage paradox,” coined by Price and Vance (67), in which intracellular pathogens often infect the very cell type bestowed with effector functions to control their growth.

AMs express a number of inhibitory receptors at baseline that are thought to contribute to their immunoregulatory phenotype (8). We do not yet understand how these other pathways might affect the AM response to M.tb. or provide redundancy with the NRF2 pathway. Our data suggest that M.tb.-infected AMs undergo a shift toward an increasingly pro-inflammatory phenotype over the first 10 days of infection, but the underlying mechanisms and the subsequent events remain unknown. Is this change cell intrinsic or dependent on the uptake and recognition of bacteria by other cell types after AM cell death? Similarly, we do not yet understand why a portion of the pro-inflammatory program expressed by bystander AMs 10 days after infection is absent in infected AMs. Are there additional inhibitory mechanisms apart from NRF2 that specifically restrain M.tb.-infected AMs? Last, we do not yet know how human AMs respond to M.tb. upon initial infection in vivo. These studies were performed using mice living in specific pathogen–free facilities that are exposed to a very limited set of environmental factors, whereas humans are continuously exposed to a variety of airborne stimuli and pathogens that may alter the baseline state of AMs. How pre-exposure to different stimuli affects human AM function, and therefore TB susceptibility, is not yet understood. Further investigation into the intervening events between initial AM infection and the adaptive T cell response as well as how the AM response to M.tb. can be modified by environmental exposure may facilitate a better understanding of the early events of TB infection and inform future TB vaccine strategies.

Study design

The aim of this study was to measure the immune response of AMs to intracellular M.tb. infection in vivo. We characterized the transcriptional profile of murine M.tb.-infected AMs after aerosol infection by sorting cells and performing low-input library amplification, followed by RNA-Seq. To determine the effect of the transcription factor NRF2 on the response of AMs, we generated NRF2 conditional knockout mouse models (Nrf2LysM and Nrf2CD11c) and tested their responses against Nrf2fl littermate controls. The number of replicates for each experiment is indicated in the figure caption. No animals were excluded in these studies.



C57BL/6 and Nrf2−/− (B6.129X1-Nfe2l2tm1Ywk/J) mice were purchased from the Jackson Laboratory (Bar Harbor, ME). Nrf2floxed (C57BL/6-Nfe2l2tm1.1Sred/SbisJ), CD11ccre [B6.Cg-Tg(Itgax-cre)1-1Reiz/J], and LysMcre [B6.129P2-Lyz2tm1(cre)Ifo/J] were purchased from the Jackson Laboratory and bred to generate Nrf2CD11c and Nrf2LysM mice. Mice were housed and maintained in specific pathogen–free conditions at the Seattle Children’s Research Institute (SCRI), and experiments were performed in compliance with the Institutional Animal Care and Use Committee. Six- to 12-week-old male and female mice were used for all experiments, except for RNA-Seq, which used only female mice for uniformity. Mice infected with M.tb. were housed in a biosafety level 3 facility in an animal biohazard containment suite.

M.tb. aerosol infections and lung mononuclear cell isolation

Most aerosol infections were performed with a stock of WT H37Rv transformed with an mEmerald reporter pMV261 plasmid, provided by C. Sassetti and C. Baer (University of Massachusetts Medical School, Worcester, MA). Some infections used an M.tb. strain with a deletion of the virulence determinant RD1 region (ΔRD1), provided by D. Sherman (SCRI, Seattle, WA), and transformed with the same mEmerald expression plasmid. For both standard (~100 CFU) and high-dose (~2000 to 4000 CFU) infections, mice were enclosed in an aerosol infection chamber (Glas-Col) and frozen stocks of bacteria were thawed and placed inside the associated nebulizer. To determine the infectious dose, three mice in each infection were sacrificed 1 day later and lung homogenates were plated onto 7H10 plates for CFU enumeration, as previously described (68).

Bead aerosolization

Carboxylate 1.0-μm fluorescent beads (Thermo Fisher Scientific) were aerosolized using an LC sprint reusable nebulizer (PARI) attached to a vacuum pump and an air flow regulator, as described previously (69).

Lung single-cell suspensions

At each time point, lungs were removed and single-cell suspensions of lung mononuclear cells were prepared by Liberase Blendzyme 3 (70 μg/ml; Roche) digestion containing deoxyribonuclease I (DNase I) (30 μg/ml; Sigma-Aldrich) for 30 min at 37°C and mechanical disruption using a gentleMACS dissociator (Miltenyi Biotec), followed by filtering through a 70-μm cell strainer. Cells were resuspended in FACS buffer [phosphate-buffered saline (PBS), 1% fetal bovine serum (FBS), and 0.1% NaN3] before staining for flow cytometry.

AM isolation

BAL was performed by exposing the trachea of euthanized mice, puncturing the trachea with Vannas Micro Scissors (VWR), and injecting 1 ml of PBS using a 20-gauge, 1-inch IV catheter (McKesson) connected to a 1-ml syringe. The PBS was flushed into the lung and then aspirated three times, and the recovered fluid was placed on ice. The wash was repeated three additional times. Cells were filtered and spun down. For antibody staining, cells were suspended in FACS buffer. For cell culture, cells were plated at a density of 1 × 105 cells per well (96-well plate) in complete RPMI [RPMI plus FBS (10%; VWR), l-glutamine (2 mM; Invitrogen), and penicillin-streptomycin (100 U/ml; Invitrogen) with or without GM-CSF (10 ng/ml; PeproTech)] and allowed to adhere overnight in a 37°C humidified incubator (5% CO2). Media with antibiotics were washed out before infection.

Cell sorting and flow cytometry

Fc receptors were blocked with anti-CD16/32 (2.4G2, BD Pharmingen). Cell viability was assessed using Zombie Violet dye (BioLegend). Cells were suspended in 1× PBS (pH 7.4) containing 0.01% NaN3 and 1% FBS. Surface staining included antibodies specific for murine: Siglec F (E50-2440, BD Pharmingen), CD11b (M1/70), CD64 (X54-5/7.1), CD45 (104), CD3 (17A2, eBioscience), CD4 (RM4-4), CD8 (53-6.7, BD Biosciences), CD19 (1D3, eBioscience), CD11c (N418), CD103 (M290, BD Biosciences), I-A/I-E (M5/114.15.2), and Ly6G (1A8) (reagents from BioLegend unless otherwise noted). Cell numbers were enumerated using Polybead Polystyrene 15.0 μm Microspheres (Polysciences). Cell sorting was performed on FACSAria (BD Biosciences). Sorted cells were collected in complete medium, spun down, resuspended in TRIzol, and frozen at −80°C overnight before RNA isolation. Samples for flow cytometry were fixed in 2% paraformaldehyde solution in PBS and analyzed using an LSRII flow cytometer (BD Biosciences) and FlowJo software (Tree Star Inc.).

RNA-Seq and analysis

RNA isolation was performed using TRIzol (Invitrogen), two sequential chloroform extractions, GlycoBlue carrier (Thermo Fisher Scientific), isopropanol precipitation, and washes with 75% ethanol. RNA was quantified with the Bioanalyzer RNA 6000 Pico Kit (Agilent). Because of the low number of M.tb.-infected cells recovered (~2000 to 4000 cells total after pooling BAL from 10 to 12 mice), all complementary DNA (cDNA) libraries were constructed and amplified using the SMARTer Stranded Total RNA-Seq Kit (v1 or v2) - Pico Input Mammalian (Clontech) per the manufacturer’s instructions. Sequencing was performed on an Illumina NextSeq (2 × 75, paired-end). Stranded paired-end reads of length 76 were preprocessed: The first three nucleotides of R1 (v1 kit) or R2 (v2 kit) were removed as described in the SMARTer Stranded Total RNA-Seq Kit User Manual (v1: 112215, v2: 063017), and read ends consisting of 50 or more of the same nucleotide were removed. The remaining read pairs were aligned to the mouse (mm10) and M.tb. H37Rv genome using the GSNAP aligner (v. 2016-08-24), allowing for novel splicing. Concordantly mapping read pairs (average, 15 million per sample) that aligned uniquely were assigned to exons using the subRead program and gene definitions from Ensembl Mus_Musculus GRCm38.78 coding and noncoding genes. Only genes for which at least three samples had at least 10 counts and had an average CPM > 1.0 were retained, resulting in a total of 10,946 genes. Fold change comparisons between WT and NRF2 gene deletion strains were further filtered on genes with less than two samples with zero reads. Differential expression was calculated using the edgeR package from FDR was computed with the Benjamini-Hochberg algorithm (tables S3 to S5). Hierarchical clusterings were performed in R using “TSclust” and “hclust” libraries. Heatmap and scatterplot visualizations were generated in R using the “heatmap.2” and “ggplot2” libraries, respectively.

Ingenuity Pathway Analysis

IPA (Qiagen) was used to identify enriched pathways for DE genes between naïve and M.tb.-infected or naïve and bystander AMs (cutoff values: FDR < 0.01; |fold change| > 2). Canonical pathways with enrichment score (ES) P value of <0.05 with greater than 10 gene members are reported. IPA was also used to identify differentially enriched pathways between bystander and M.tb.-infected AMs at 10 days after infection (cutoff values: FDR < 0.05; |fold change| > 2). Canonical pathways with |z scores| > 1 and P < 0.05 were reported.

Promoter scanning (HOMER)

Promoter regions of genes that were up-regulated [|fold change| > 2, FDR < 0.01, log2(average CPM) > 1.0] were scanned for DNA protein–binding motif overrepresentation using the HOMER program (v4.9.1, (30). Promoter regions were defined as 2000 nucleotides upstream of the gene start to 1000 nucleotides downstream. Background sequences were taken from the promoter regions of expressed genes defined by Ensembl Mus_Musculus GRCm38.78 (N = 10,946). Four hundred two known motifs were scanned, and hypergeometric P values were computed.

NRF2 ChIP-Seq analysis

FASTQ files from the Gene Expression Omnibus (GEO) dataset GSE75175 were downloaded for three NRF2 ChIP-Seq experiments assaying peritoneal macrophages from WT mice, including the corresponding sequencing of input DNA (32). For each sample, single ended reads of length 101 were filtered to remove those consisting of 50 or more of the same nucleotide or of low-quality base calls. Filtered reads were aligned against mm10 using GSNAP (v. 2011-11-20) with no allowance for splicing. Uniquely mapped reads were filtered for duplicates based on alignment position. A total of 3975 peaks were called using MACS2 (v.2.1.0) for the combined ChIP-Seq samples using the input DNA libraries as the controls. Called peaks were annotated by checking for overlap with promoter regions as described above.

Gene set enrichment analysis

Input data for GSEA consisted of lists, ranked by –log(P value), comparing RNA-Seq expression measures of target samples and naïve controls including directionality of fold change. Mouse orthologs of human Hallmark genes were defined using a list provided by Molecular Signatures Database (MSigDB) (70). GSEA software was used to calculate enrichment of ranked lists in each of the respective hallmark gene lists, as described previously (71). A nominal P value for each ES is calculated on the basis of the null distribution of 1000 random permutations. To correct for multiple hypothesis testing, a NES is calculated that corrects the ES based on the null distribution. An FDR is calculated for each NES. Leading edge subsets are defined as the genes in a particular gene set that are part of the ranked list at or before the running sum reaches its maximum value.

BMDM isolation and culture

BMDMs were cultured in complete RPMI with recombinant human M-CSF (50 ng/ml; PeproTech Inc.) for 6 days. Media with antibiotics were washed out before infection with H37Rv and replaced with complete RPMI without antibiotics. M-CSF was included in the BMDM media for the entire course of the experiments.

M.tb. in vitro culture and infection

H37Rv was grown in 7H9 media at 37°C to optical density (O.D.) of 0.1 to 0.3. The final concentration was calculated on the basis of strain titer, and bacteria were added to macrophages at an effective multiplicity of infection of 0.5 for 2 hours. Cultures were then washed three times to remove extracellular bacteria. Infected macrophages were cultured for up to 5 days. For CFU measurement, cells were lysed with 1% Triton X-100/PBS and lysates from triplicate conditions were plated in serial dilutions on Middlebrook 7H10 agar plates (Thermo Fisher Scientific) and cultured at 37°C for 21 days.

Quantitative reverse transcription PCR

For gene expression analysis, 1 × 105 AMs or BMDMs were plated in 96-well plates overnight, followed by in vitro infection as described above. RNA isolation was performed as described above. RT-qPCR reactions were carried out using TaqMan primer probes (ABI) and the TaqMan Fast Universal PCR Master Mix (Thermo Fisher Scientific) in the CFX384 Touch Real-Time PCR Detection System (Bio-Rad). Data were normalized by the level of Efia expression in individual samples. Fold induction was computed with respect to the normalized expression levels of respective macrophages under unstimulated conditions within the same experiment.


Cells acquired by BAL were spun down and plated on a Lab-Tek II 8-chamber glass-bottom slide (Nunc). The slides were kept for 2 hours in a 37°C humidified incubator (5% CO2) to allow adherence before media removal and fixation with 2% paraformaldehyde in PBS. Surface staining was performed with anti-CD45 (clone 30-F11; BioLegend). ProLong Diamond Antifade Mountant with 4′,6-diamidino-2-phenylindole (DAPI) (Thermo Fisher Scientific) was used for mounting. Cells were imaged using a 100× objective [1.40 numerical aperture (NA)] on a DeltaVision Elite with the following excitation filter cubes: Cy5 (632/22), green fluorescent protein (GFP) (475/28), and DAPI (390/18) and emission cubes: Cy5(679/34), GFP (525/48), and DAPI (435/48). The entire cell volume was captured using a series of Z-stack images with a 0.2-μm step size. Numbers of Mycobacteria cells per macrophage were enumerated by manual counting using the Z-stack of images. Representative images were first deconvolved with a theoretical point spread function using SVI Huygens Essential before a maximum intensity project image was created in Imaris image analysis software (Bitplane).

Enzyme-linked immunosorbent assay

Lung homogenate was digested in 1 ml of radioimmunoprecipitation assay (RIPA) lysis and extraction buffer (Thermo Fisher Scientific) with Halt Protease Inhibitor (Thermo Fisher Scientific), followed by mechanical disruption using a gentleMACS dissociator (Miltenyi Biotec). Samples were double filtered through 0.2-μm filters to remove bacteria. Total protein concentration was measured by Pierce BCA Protein Assay (Thermo Fisher Scientific), and samples were normalized to total protein (1.5 mg/ml). Murine IL-6, IL-1β, and TNF DuoSet enzyme-linked immunosorbent assays (ELISAs) were performed in accordance with the manufacturer’s instructions (R&D Systems). Absorbance was recorded at 450 nm with wavelength correction performed by subtracting 570 nm of absorbance using a SpectraMax M2 plate reader (Molecular Devices).

Western blot

Naïve AMs were isolated by BAL and pooled from five mice, plated overnight for adherence, and stimulated for 5 hours with the NRF2 agonist dimethyl fumarate (3.65 μM; Sigma-Aldrich) before protein collection with RIPA buffer (Cell Signaling Technology) and HALT protease inhibitor (Thermo Fisher Scientific). Western blotting analyses were performed using standard techniques and transblotted onto nitrocellulose membranes. Membranes were probed with relevant primary antibodies: rabbit anti-NRF2 (D1Z9C) (1:1000; Cell Signaling Technology) and rabbit anti-mouse β-actin1–horseradish peroxidase (HRP) antibody (1:5000, Jackson ImmunoResearch). Primary antibodies were detected by a secondary rat anti-rabbit–HRP antibody (1:2000, Jackson ImmunoResearch).

Statistical analyses

RNA-Seq was analyzed using the edgeR package from, and the FDR was computed using the Benjamini-Hochberg algorithm. All other data are presented as means ± SEM. For comparison of multiple conditions, data were analyzed by one-way analysis of variance (ANOVA) (95% confidence interval) with Holm-Sidak correction (for multiple comparisons) or Dunnett’s correction (for multiple comparisons to a single control). For comparison of two conditions, data were analyzed by two-tailed unpaired Student’s t test (with Holm-Sidak correction for multiple testing). Statistical analysis and graphical representation of data were performed using either GraphPad Prism v6.0 software or R. At least three to five mice were used per group in each experiment, and experiments were performed at least two to three times, as indicated in the figure legends. To denote significance, *P < 0.05, **P < 0.01, ***P < 0.001.


Fig. S1. Lung flow cytometry gating scheme for myeloid cells.

Fig. S2. Lung flow cytometry gating scheme for detection of M.tb.-infected cells with overlay of mEmerald+ populations.

Fig. S3. BAL and lung flow cytometry gating scheme for cell sorting.

Fig. S4. Bacterial burden at the time of AM cell sorting.

Fig. S5. Gene expression heatmaps of naïve, bystander, and M.tb.-infected AMs 24 hours after aerosol infection.

Fig. S6. Gating strategy for sorting of ΔRD1-infected AMs and bead-positive AMs.

Fig. S7. Gene expression of BMDMs after treatment with fluorescent beads.

Fig. S8. Gene expression heatmaps of M.tb.-infected AMs 0.5, 1, 2, 4, and 10 days after aerosol infection.

Fig. S9. AM response to in vitro M.tb. infection with and without GM-CSF treatment.

Fig. S10. AM response to in vitro M.tb. infection at 8, 12, and 24 hours after infection.

Fig. S11. Gene expression differences between naïve WT and Nrf2−/− AMs.

Fig. S12. Baseline lung conditions of WT and Nrf2−/− mice.

Fig. S13. Gene expression and protein confirmation of NRF2 knockout in AMs from Nrf2LysM, Nrf2CD11c, and Nrf2fl littermate controls; WT; and Nrf2−/− mice.

Fig. S14. Baseline lung conditions of Nrf2LysM, Nrf2CD11c, and Nrf2fl mice.

Fig. S15. Gene expression heatmaps of M.tb.-infected Nrf2LysM, Nrf2CD11c, Nrf2fl, WT, and Nrf2−/− AMs 24 hours after aerosol infection.

Fig. S16. Viability of bystander Nrf2LysM, Nrf2CD11c, and Nrf2fl AMs 10 days after high-dose aerosol infection.

Table S1. Summary table accounting for M.tb.-infected cells in the lung 1 to 14 days after high-dose mEmerald-H37Rv infection.

Table S2. IPA and HOMER results tables.

Table S3. RNA-Seq data for WT AMs.

Table S4. RNA-Seq data for Nrf2−/− AMs.

Table S5. RNA-Seq data for Nrf2fl, Nrf2LysM, and Nrf2CD11c AMs.

Table S6. Raw data file.


Acknowledgments: We thank the staff at SCRI vivarium for animal care, P. Troisch and the Next Gen Sequencing core at the Institute for Systems Biology, and C. Sassetti and C. Baer at the University of Massachusetts Medical School for the mEmerald bacterial strains. We also thank members of the Urdahl laboratory for providing helpful discussions. Funding: This work was supported by the National Institute of Allergy and Infectious Disease of the NIH under awards U19AI106761, U19AI135976, U19AI100627, and R01AI032972 (to A.A). Author contributions: A.C.R., G.S.O., J.N., A.H.D., and A.A. designed the experiments. A.C.R., D.M., G.S.O., and J.N. conducted the experiments. L.M.A. and A.H.D. performed computational analyses. A.C.R., A.H.D., E.S.G., and A.A. wrote the paper. Competing interests: The authors declare that they have no competing interests. Data and materials availability: Raw and processed RNA-Seq data can be accessed from the National Center for Biotechnology Information (NCBI) GEO database under accession numbers GSE125287 and GSE132185.

Stay Connected to Science Immunology

Navigate This Article