Autophagy promotes efficient T cell responses to restrict high-dose Mycobacterium tuberculosis infection in mice

Abstract

Although autophagy sequesters intracellular Mycobacterium tuberculosis (Mtb) in in vitro cultured macrophages, loss of autophagy in macrophages in vivo does not result in susceptibility to a standard low-dose Mtb infection until late during infection, leaving open questions regarding the protective role of autophagy during Mtb infection. We report herein that loss of autophagy in lung macrophages and dendritic cells results in acute susceptibility of mice to high-dose Mtb infection, a model better mimicking active tuberculosis. However, rather than observing a role for autophagy in controlling Mtb replication in macrophages, we find that autophagy suppresses macrophage responses to Mtb infection that otherwise result in the accumulation of myeloid derived suppressor cells and subsequent defects in T cell responses. Our finding that the pathogen-plus-susceptibility gene interaction is dependent on dose has important implications on understanding how Mtb infections in humans lead to a spectrum of outcomes and the potential use of autophagy modulators in clinical medicine.

INTRODUCTION

In vitro experiments suggest that autophagy limits Mtb replication in macrophages through targeting the pathogen to the lysosome for degradation, a process termed xenophagy^1–4. However, Mtb replication in macrophages is only one readout of a pathogenic process that involves intricate interactions between Mtb and a variety of host cells as well as complex signaling networks between immune cells. Consequently, the translation of in vitro results to in vivo control of Mtb pathogenesis remains uncertain. In addition, loss of autophagy in innate immune cells does not lead to susceptibility to low-dose Mtb until late stages of infection^5–8 and whether this late role for autophagy in vivo involves Mtb sequestration and degradation remains unknown. Therefore, the precise mechanistic impact of autophagy on Mtb pathogenesis in vivo remains elusive.

Pulmonary Mtb infection results in latent tuberculosis (TB) infection or active TB disease, the latter of which results in clinical manifestations of TB. The outcome of Mtb infection is determined by the immune response in a given individual⁹, where the roles for particular immune pathways are most commonly studied in mouse models. Although a low-dose (50–100 Mtb) mouse model is most commonly used in the field, recent studies have used ultra-low doses (1–5 Mtb) to induce latent TB and high-doses (~1000 Mtb) to model active TB^10–12. In wild-type C57Bl/6 mice, only a high dose, but not the standard low-dose, of the Mtb HN878 strain elicited gene expression profiles with similarities to the blood signatures observed in human active TB patients¹². In addition, studies in animal models have demonstrated that higher infectious doses contribute to TB disease progression and have different requirements for immune control^10,12,13. Aerosol delivery of high-dose Mtb triggers progressive active TB in nonhuman primates, which is associated with macrophage interferon (IFN) responses¹⁰. These elevated IFN responses resemble the elevated IFN signature in the blood of active TB patients^10,14,15. These observations underscore that higher doses of Mtb inoculum in animal models stimulate immune mechanisms more akin to active TB and emphasize an association of heightened IFN signaling with active TB.

BECLIN1 and FIP200, proteins that are essential for initiating autophagy, also function independently of autophagy to dampen IFNγ responses of tissue-resident macrophages¹⁶. Given the association between heightened macrophage IFN responses and TB disease progression in high-dose Mtb infection animal models, we hypothesized that BECLIN1 could impact Mtb pathogenesis in the high-dose Mtb model, in a manner not observed in the low-dose model. We revealed that BECLIN1 was indeed required in macrophages to control high-dose Mtb infection. However, this process is autophagy-dependent and distinct from the role for BECLIN1 and FIP200 in immune quiescence. Although loss of autophagy in lung macrophages and dendritic cells (DCs) resulted in severe early susceptibility to high-dose Mtb infection, autophagy was dispensable to control Mtb replication in macrophages in vivo. Instead, autophagy suppressed macrophage responses to Mtb infection that otherwise result in the accumulation of myeloid derived suppressor cells and subsequent defects in sustained T cell responses. Our studies identify a critical and previously undescribed role for autophagy to control Mtb pathogenesis in a model that better reproduces the immune responses occurring in active TB patients.

RESULTS

Loss of autophagy results in early susceptibility to high-dose Mtb

We infected Becn1^f/f-LysM-cre mice, which delete Becn1 (gene encoding BECLIN1) in LysM⁺ cells (neutrophils, macrophages, monocytes, and some DCs), with a high dose (~1000 CFUs) of Mtb strain Erdman. We discovered that Becn1^f/f-LysM-cre mice were acutely susceptible to high-dose Mtb infection, succumbing by 61 days post infection (dpi), while the majority of Becn1^f/f control mice survived passed 140 dpi (Fig. 1a). Individual deletion of other essential autophagy genes, Atg14, Fip200 (Rb1cc1), Atg7, and Atg16l1, in LysM expressing cells also resulted in similar susceptibility during high-dose Mtb infection (Fig. 1b–1e). Deletion of these genes in mice has previously been shown to be sufficient to result in loss of autophagy in LysM⁺ cells^6,16,17. We also confirmed deletion of Atg16l1 in myeloid populations of Atg16l1^f/f-LysM-cre mice (Extended Data Fig. 1a) as well as the disruption of autophagy in alveolar macrophages by performing western blots for p62 (SQSTM1) accumulation and LC3 (MAP1LC3B) lipidation (Fig. 1f), supporting that autophagy is disrupted in the primary immune cells from Atg16l1^f/f-LysM-cre mice. Compared to prior studies in in vitro bone marrow derived macrophages⁵, alveolar macrophages from naïve Atg16l1^f/f mice exhibited more robust baseline autophagy, reflected by high levels of LC3-II (Fig. 1f), and the single copy of the LysM-cre allele from Atg16l1^f/f-LysM-cre mice resulted in efficient deletion of the Atg16l1 gene (Extended Data Fig. 1a) and loss of LC3-II in alveolar macrophages (Fig. 1f). In addition to autophagy, BECLIN1, ATG16L1, and ATG7 are essential for LC3-associated phagocytosis (LAP)¹⁸, which has also been implicated in host defense against Mtb¹⁹. Rubcn^−/− mice, lacking the RUBICON protein essential for LAP, were not more susceptible to high-dose Mtb in mice (Fig. 1g). This demonstrated that LAP was not required to control high-dose Mtb infection.

Figure 1. High-dose Mtb infection model uncovers protective role of autophagy in myeloid cells in vivo.

a-e,g, Survival of mice that harbor myeloid deficiency (LysM-cre) in autophagy genes (Becn1 (a), Atg14 (b), Fip200 (c), Atg16l1 (d), Atg7 (e) versus floxed littermate controls, and Rubicon^−/− versus Rubicon^+/+/Rubicon^+/− (WT/Het) (g) after aerosol infection with high-dose of 1000 colony-forming units (CFU) of M. tuberculosis (Mtb) strain Erdman. P values were calculated using log-rank Mantel-Cox tests.

f, Western blot analysis of p62, LC3 and Actin in alveolar macrophages from bronchoalveolar lavage fluid (BALF) of naïve mice. Representative data of n≥5 biological replicates.

h-i, Mtb CFU in lungs and spleens at 21 days post infection (dpi) of high-dose Mtb infections. Means ± s.e.m. pooled from ≥3 experiments are graphed. Atg14^f/f, n= 16 mice for lungs and n= 12 mice for spleens; Atg14^f/f-LysM-cre, n= 16 mice for lungs and n= 8 mice for spleens. P values were calculated by two-tailed Mann-Whitney tests.

j-k, Survival of mice after aerosol infection with low-dose of Mtb. P values were calculated using log-rank Mantel-Cox tests.

l, Mtb CFU in lungs at 21 dpi of low-dose Mtb infections. Means ± s.e.m. pooled from 2 experiments are graphed. Atg14^f/f, Atg14^f/f-LysM-cre, n= 5; Atg5^f/f, Atg5^f/f-LysM-cre, n= 4 mice. P value was calculated by two-tailed Mann-Whitney tests.

ns= not significant. ** for P < 0.01, *** P < 0.001, and **** P < 0.0001.

To determine how autophagy genes in myeloid cells mediate protection during high-dose Mtb infection, we further characterized the most susceptible mouse lines, Atg14^f/f-LysM-cre and Atg16l1^f/f-LysM-cre mice (median survival time of 29.5 dpi and 24 dpi, respectively), using them interchangeably based on mice availability. Both Atg14^f/f-LysM-cre and Atg16l1^f/f-LysM-cre mice displayed rapid weight loss and significantly higher lung bacterial burden without increased Mtb dissemination to the spleens compared to control mice at 21 dpi of high-dose Mtb infection (Fig. 1h–1i, Extended Data Fig 1b–1c). Our findings with high-dose Mtb contrasted with data from low-dose Mtb infection where only Atg5^f/f-LysM-cre mice exhibited increased early susceptibility (Fig. 1j), a phenotype not shared by other autophagy deficient mice^5,6,20. We validated the previously published findings that Atg14^f/f-LysM-cre mice do not succumb before 140 dpi and were able to control Mtb burden following low-dose Mtb infection (Fig. 1k–1l). These data support a protective role for multiple autophagy genes in the lungs of mice that is specifically revealed during high-dose Mtb infection.

Autophagy is not required to limit Mtb replication in macrophages

We interrogated which myeloid cell type(s) require autophagy genes to control high-dose Mtb infection by comparing the susceptibility of mice that delete autophagy factors in LysM⁺ cells, Mrp8⁺ cells (neutrophils), and CD11c⁺ cells (lung macrophages and DCs)^21,22. Deleting Becn1 or Atg14 from Mrp8⁺ cells resulted in wild-type level resistance (Fig. 2a–b), demonstrating that autophagy is not required in neutrophils during high-dose Mtb infection. Therefore, the role for autophagy during high-dose Mtb infection is genetically distinct from the autophagy-independent role for ATG5 in neutrophils during low-dose Mtb infection^3,6. In contrast, loss of BECLIN1, ATG14, or ATG16L1 in CD11c⁺ lung macrophages and DCs resulted in increased susceptibility to high-dose Mtb (Fig. 2c–e). Loss of ATG14 in CD11c⁺ cells also resulted in increased lung Mtb burden at 21 dpi (Fig. 2f), similar to that observed in Atg14^f/f-LysM-cre mice (Fig. 1h). These results indicate that autophagy is required in CD11c⁺ lung macrophages and DCs to control high-dose Mtb infection.

Figure 2. Autophagy genes in macrophages/DCs are required to control high-dose Mtb infection in mice.

a-e, Survival of mice harboring Becn1 (a, c), Atg14 (b, d), Atg16l1 (e) deletion in Mrp8⁺ or CD11c⁺ cells versus controls after aerosol infection with high-dose Mtb. P by log-rank Mantel-Cox tests.

f, Mtb CFU in lungs at 21 dpi with high-dose Mtb infection. Data are mean ± s.e.m., pooled from 2 experiments. Atg14^f/f, n= 11; Atg14^f/f-CD11c-cre, n= 10 mice. P value was calculated using two-tailed Mann-Whitney tests.

g, Survival of mice after aerosol infection with high-dose Mtb. P value was calculated by two-tailed Mann-Whitney tests.

P < 0.05 was denoted *, and **** for P < 0.0001. ns= not significant.

In cell culture, loss of autophagy in macrophages results in increased Mtb replication due to the inability of these macrophages to eliminate Mtb by xenophagy^1–5,23,24. Specifically, autophagy increases lysosome-mediated Mtb killing through recruitment of adaptor molecules such as p62 to Mtb-containing phagosomes, where loss of p62 or essential autophagy genes leads to a similar increase in Mtb burden in vitro^1,2,25. We infected p62^−/− mice and found that loss of p62 expression did not lead to increased susceptibility to high-dose Mtb infection (Fig. 2g). However, since Mtb burdens were higher in the lungs of mice deficient for autophagy in CD11c⁺ myeloid cells at 21 dpi after high-dose Mtb infection (Fig. 1h–1i, 2f) and other autophagy adaptors may compensate for loss of p62, it was still possible that xenophagy was controlling Mtb replication in macrophages in vivo. To test this possibility, we further studied the response of macrophages to high-dose Mtb infection. Alveolar macrophages are the first cells infected upon Mtb inhalation and are the initial cells wherein Mtb replicates^26,27. Aerosol infection of mice with 2000–4000 Mtb (a dose higher than used here) results in a single Mtb in >80% of infected alveolar macrophages²⁷. To determine if xenophagy degrades Mtb in alveolar macrophages, we analyzed the events early after high-dose Mtb infection, as Mtb mainly replicate in alveolar macrophages before trafficking to the interstitium and subsequently infecting other innate immune cells²⁶. Mtb CFUs in the lungs of Atg16l1^f/f-LysM-cre mice at 7 dpi were similar to those in Atg16l1^f/f controls, suggesting that there is not an overall difference in bacterial replication early during infection (Fig. 3a). To determine what cell types harbored the replicating Mtb at 7 dpi, we used an mCherry expressing Mtb to trace infected cells (Fig. S1). Similar to previous reports in wild-type mice^26,27, the majority of cells infected at 7 dpi were alveolar macrophages (Fig. 3b). The number of Mtb infected alveolar macrophages and the levels of intracellular Mtb per alveolar macrophage, measured by mean fluorescent intensity (MFI) of mCherry, in Atg16l1^f/f-LysM-cre mice were comparable to control mice at 7 dpi (Fig. 3b–3c). Together these data demonstrate that autophagy is not required in alveolar macrophages to control initial Mtb replication. The other main cell types infected by high-dose Mtb were neutrophils and non-alveolar macrophages, which includes interstitial macrophages and recruited macrophages (Fig. 3b). These cells also harbored similar levels of Mtb in Atg16l1^f/f-LysM-cre mice compared to controls at 7 dpi (Fig. 3b–3c).

Figure 3. Autophagy does not limit Mtb replication in alveolar macrophages.

a,g, Mtb CFU in the right lungs of mice at 7 dpi (a) and 14 dpi (g) of high-dose Mtb infections. Atg16l1^f/f, n= 9 mice at 7 dpi, n= 7 at 14 mice dpi; Atg16l1^f/f-LysM-cre, n= 10 mice.

b,h, Number of total mCherry+ cells in the left lungs at 7 dpi (b) and 14 dpi (h) after high-dose infection with mCherry-Mtb. Data (Means ± s.e.m.) from 2 independent experiments are graphed. Atg16l1^f/f, n= 9 mice for macrophages, neutrophils at 7 dpi, n= 5 mice for DCs, monocytes at 7 dpi, n= 7 mice for all cell types at 14 dpi; Atg16l1^f/f-LysM-cre, n= 9 mice for macrophages, neutrophils at 7 dpi, n= 5 mice for DCs, monocytes at 7 dpi, n= 8 mice for all cell types at 14 dpi.

c,i, mCherry mean fluorescent intensity (MFI) in mCherry+ cells at 7 dpi (c) and 14 dpi (i) after high-dose infection with mCherry-Mtb. Data (Means ± s.e.m.) from 2 independent experiments are graphed. Atg16l1^f/f, Atg16l1^f/f-LysM-cre, n= 9 mice at 7 dpi; Atg16l1^f/f, n= 7 mice at 14 dpi; Atg16l1^f/f-LysM-cre, n= 8 mice at 14 dpi.

d, Quantification of the number and area of LC3 puncta in Mtb+ cells and bystander cells from mice infected with. Data (Means ± s.e.m.) representative of two independent experiments of 3 biological replicates each. Atg16l1^f/f, n= 29 cells for bystander cells, n= 25 cells for Mtb+ cells; Atg16l1^f/f-LysM-cre, n= 16 cells for bystander cells, n= 33 cells for Mtb+ cells.

e, Representative images (of n>3 mice) of LC3-stained alveolar macrophages from mice at 7 dpi with high-dose mCherry-Mtb. LC3 (green), nuclear staining (blue) and Mtb (red). Scale bars, 5 μm.

f, % of Mtb coated with LC3 in Mtb infected alveolar macrophages from mice at 7 dpi with high-dose mCherry-Mtb. Atg16l1^f/f, Atg16l1^f/f-LysM-cre, n= 31 mice. Data (Means ± s.e.m.) representative of two independent experiments of 3 biological replicates each.

P values were calculated by two-tailed Mann-Whitney tests. * for P < 0.05, ** for P < 0.01, *** P < 0.001, and **** P < 0.0001. ns= not significant.

In addition to the flow cytometry experiments, we further evaluated whether Mtb colocalizes to autophagosomes in an ATG16L1-dependent manner in vivo. We isolated alveolar macrophages from bronchoalveolar lavages of infected mice at 7 dpi and analyzed LC3 puncta formation and its co-localization of Mtb using fluorescent microscopy. The number and size of LC3+ puncta in Mtb-infected Atg16l1^f/f alveolar macrophages trended higher than in bystander Atg16l1^f/f alveolar macrophages, although this difference was not significant (Fig. 3d–3e). ATG16L1 deficiency resulted in the accumulation of large LC3+ structures specifically within Mtb infected cells (Fig. 3e), where the number and size of LC3 puncta were significantly increased in infected ATG16L1-deficient cells compared to Atg16l1^f/f controls (Fig. 3d). There was no difference in the degree of LC3 co-localization with Mtb in infected alveolar macrophages from Atg16l1^f/f and Atg16l1^f/f-LysM-cre mice at 7 dpi (Fig. 3e–3f). In a complementary approach, we also isolated alveolar macrophages from naïve mice and infected them ex vivo with mCherry-Mtb. Similar to as observed in alveolar macrophages isolated from infected mice, there were an increased area of LC3+ structures formed in ex vivo infected ATG16L1-deficient alveolar macrophages at 48 hours post infection but no difference in colocalization with Mtb in the two genotypes (Extended Data Fig. 2a–2c). Since ATG16L1-deficient macrophages were defective in generating LC3-II (Fig. 1f), the LC3+ structures in these cells likely represent LC3-I sequestration to p62 aggregates, which accumulate when autophagy is impaired^28,29. The presence of these structures further supports that autophagy was induced during Mtb infection but effectively disabled in alveolar macrophages from Atg16l1^f/f-LysM-cre mice. The absence of these structures in Mtb-infected Atg16l1^f/f alveolar macrophages implies that following autophagy induction by Mtb infection, autophagosomal proteins, such as LC3 and p62, are likely being fluxed to the lysosome and degraded in alveolar macrophages before being detected as puncta. Together, the absence of differences in bacterial burdens in alveolar macrophages from Atg16l1^f/f and Atg16l1^f/f-LysM-cre mice despite the induction of autophagy indicates that autophagy is not functioning to control Mtb replication in Atg16l1^f/f alveolar macrophages. The analysis of LC3 puncta formation and LC3-Mtb colocalization was not done previously in alveolar macrophages from in vivo Mtb-infected mice until this study.

By 14 dpi, Atg16l1^f/f-LysM-cre mice had higher Mtb lung burden relative to control mice (Fig. 3g), indicating that autophagy promotes the control of Mtb replication between 7 and 14 dpi. Even fewer Atg16l1^f/f-LysM-cre alveolar macrophages were infected at 14 dpi (Fig. 3h) and there was still no difference in the intracellular Mtb burden per alveolar macrophage at 14 dpi in Atg16l1^f/f-LysM-cre comparing to Atg16l1^f/f mice (Fig. 3i), supporting that autophagy is not required to control Mtb burdens in macrophages. We also isolated alveolar macrophages from infected mice at 14 dpi to analyze by microscopy and again observed larger LC3 puncta in ATG16L1-deficient alveolar macrophages, but no difference in Mtb-LC3 colocalization between the genotypes, supporting that autophagy was induced in Mtb-infected alveolar macrophages but this did not impact Mtb burden in macrophages (Extended Data Fig. 2d–2f). At 14 dpi, there were more Mtb infected neutrophils and more Mtb per neutrophil in Atg16l1^f/f-LysM-cre mice (Fig. 3h–3i), coinciding with higher CFUs in the lungs (Fig. 3g). Together, these results indicate that the higher Mtb burdens in Atg16l1^f/f-LysM-cre mice at 14 dpi were not due to a defect in Mtb elimination in alveolar macrophages by xenophagy, but instead due to a higher level of neutrophil infection.

Autophagy in innate immune cells sustains T cells

To determine how loss of autophagy in myeloid cells could impact Mtb infection and replication in neutrophils, we investigated neutrophil inflammation following high-dose Mtb infection. Neutrophils were recruited to lungs after infection and by 14 dpi the frequency and number of neutrophils in Atg16l1^f/f-LysM-cre lungs were greater than in the control mice (Fig. 4a, S2), resulting in the increased number of Mtb-infected neutrophils in Atg16l1^f/f-LysM-cre mice at 14 dpi (Fig. 4b, 3h). This increased neutrophil accumulation and infection in Atg16l1^f/f-LysM-cre mice were even more pronounced at 21 dpi (Fig. 4a–4b). Consistent with the increased neutrophil inflammation, the lungs of Atg16l1^f/f-LysM-cre mice at 14dpi and 21 dpi contained higher levels of proinflammatory cytokines (Extended Data Fig. 3a). The increased neutrophils were associated with significantly lower numbers of antigen presenting cells, alveolar macrophages and DCs, in Atg16l1^f/f-LysM-cre mice at 21 dpi with high-dose Mtb, while numbers of non-alveolar macrophages and monocytes were similar in the lungs of Atg16l1^f/f-LysM-cre and Atg16l1^f/f mice (Fig. 4c–4f, Extended Data Fig. 3b). Histopathological analysis revealed severe inflammation with larger lesions in the infected lungs of Atg16l1^f/f-LysM-cre mice (Fig. 4g, Extended Data Fig. 4), consistent with the increased accumulation of neutrophils in these mice. Autophagy deficiency in macrophages leads to increased inflammasome activation and pyroptosis^30,31, which could explain the increased neutrophil accumulation. However, blocking inflammasome activation by deletion of Casp1/11 in Atg14^f/f-LysM-cre mice did not influence their susceptibility to high-dose Mtb infection (Fig. 4h). Loss of autophagy in macrophages has also been shown to result in increased necrosis during Mtb infection in vitro⁵, where the main necrotic cell death pathways are pyroptosis and necroptosis. Like pyroptosis, necroptosis is an inflammatory cell death pathway that could impact immune cell recruitment and infection. Therefore, we examined the role of necroptosis in the susceptibility of Atg14^f/f-LysM-cre mice to high-dose Mtb infection and found that blocking necroptosis by deleting Mlkl was unable to rescue the susceptibility of Atg14^f/f-LysM-cre mice during infection (Fig. 4i), suggesting that necroptosis is not necessary for the susceptibility.

Figure 4. Myeloid autophagy suppresses neutrophil inflammation and promotes T cell response during high-dose Mtb infection.

a,b,c,d,e,f,j, The number and percentage of immune cells in the lung of mice during high-dose Mtb infection. Data (Means ± s.e.m.) pooled from 3 independent experiments. Atg16l1^f/f, n= 15 mice at naïve condition, n= 7 mice at 14 dpi, n= 12 mice for alveolar macrophages, total neutrophils, Mtb+ neutrophils, T cells at 21 dpi, n= 9 mice for non-alveolar macrophages, monocytes at 21 dpi; Atg16l1^f/f-LysM-cre, n= 16 mice at naïve condition, n= 10 mice at 14 dpi, n= 17 mice for alveolar macrophages, total neutrophils, Mtb+ neutrophils, T cells at 21 dpi, n= 9 mice for non-alveolar macrophages, monocytes at 21 dpi. P values were calculated by two-tailed Mann-Whitney tests.

g, Representative H&E-stained sections of mouse lungs at 14 dpi and 21 dpi of high-dose Mtb of 4 biological replicates.

h,i, Survival of high-dose Mtb infected Caspase1/11 (h) or MLKL (i) deficient mice that harboring ATG14 myeloid deficiency versus floxed littermate controls. P by log-rank Mante-Cox tests.

k, Mtb CFU in mediastinal lymph nodes (mLN) lungs after high-dose Mtb infections. Data represent Means ± s.e.m. Atg16l1^f/f, n= 6 mice at 8 dpi, n= 7 mice at 11, 14 dpi; Atg16l1^f/f-LysM-cre, n= 5 mice at 8, 14 dpi, n= 6 mice at 11 dpi. P by two-tailed Mann-Whitney tests.

l,m, The percentage of P25 CD4+ T cells that have undergone proliferation measured by CellTrace dilution, number of P25 CD4+ T cells, and the percentage of Nur77-GFP negative P25 CD4+ T cells in the mediastinal lymph nodes (mLN, l) and lungs (m) after high-dose Mtb infection. Data (Means ± s.e.m.) from 2 independent experiments are graphed. Atg16l1^f/f, n= 6 mice at 8 dpi, n= 7 mice at 11, 14 dpi; Atg16l1^f/f-LysM-cre, n= 5 mice at 8, 14 dpi, n= 6 mice at 11 dpi. P values were calculated by two-tailed Mann-Whitney tests.

* for P < 0.05, ** for P < 0.01, *** P < 0.001, and **** P < 0.0001. ns= not significant.

The lower numbers of antigen-presenting cells prompted us to analyze the T cell responses in Atg16l1^f/f-LysM-cre mice infected with high-dose Mtb. We observed significantly fewer T cells and fewer activated T cells (CD44^hiCD62L^lo) in the lungs of Atg16l1^f/f-LysM-cre mice at 21 dpi (Fig. 4j, Fig. S3), indicating impaired lung T cell responses in these mice during high-dose Mtb infection. However, the percentage of antigen-specific CD4⁺ T cells was similar in the lung draining mediastinal lymph node (mLN) and the lungs of Atg16l1^f/f-LysM-cre and Atg16l1^f/f mice at 14 and 21 dpi (Extended Data Fig. 5a–5c), indicating that antigen presentation had occurred effectively. In wild-type mice, antigen-presenting cells carrying live Mtb migrate to the mLN between 6–8 dpi to activate T cells³². Mtb-specific CD4⁺ T cells then proliferate and traffic to the lung to play their central role in controlling Mtb replication³³. To determine the reason for lower numbers of CD4⁺ T cells in the lungs of Atg16l1^f/f-LysM-cre mice at 21 dpi, we probed the kinetics of antigen presentation, T cell activation, proliferation, and trafficking to the lungs of high-dose Mtb infected mice. The timing and kinetics of antigen-presenting cell trafficking of Mtb to the mLN was similar in Atg16l1^f/f-LysM-cre mice compared to controls (Fig. 4k). To dissect whether the lower numbers of T cells at 21 dpi in the lungs of Atg16l1^f/f-LysM-cre mice was due to impaired T cell priming, we monitored the activation and proliferation of adoptively transferred Mtb-specific T cells (Ag85B specific CD4⁺ T cells³², referred to hereafter as P25 cells) during high-dose Mtb infection (Fig. S4). Proliferation of P25 cells began between 8–11 dpi in the mLN of both Atg16l1^f/f-LysM-cre mice and Atg16l1^f/f mice (Fig. 4l). The frequency and number of proliferating P25 cells in the mLN were similar in Atg16l1^f/f-LysM-cre and Atg16l1^f/f mice at 11 dpi and 14 dpi (Fig. 4l). Additionally, the number and frequency of proliferated P25 cells trafficked to the lung were similar in Atg16l1^f/f-LysM-cre mice and Atg16l1^f/f mice at 11 dpi and 14 dpi (Fig. 4m). TCR-mediated cell activation, measured by Nur77-GFP positivity, was also comparable in the mLN and lungs of Atg16l1^f/f-LysM-cre and Atg16l1^f/f mice (Fig. 4l–4m). The MHC-II levels on antigen-presenting cells in the lungs were also similar between Atg16l1^f/f-LysM-cre and Atg16^f/f mice at 14 dpi (Extended Data Fig. 5d), indicating intact antigen presentation at this time point. Together, these analyses indicate that the kinetics of antigen trafficking, T cell priming, and T cell arrival in the Mtb-infected lung remain unaffected in the Atg16l1^f/f-LysM-cre mice during high-dose Mtb infection. Therefore, the lower T cell numbers in the lung at 21 dpi in Atg16l1^f/f-LysM-cre mice occurred after T cell priming and trafficking to the lung.

Autophagy in CD11c⁺ cells suppresses MDSC accumulation

We noted that some of the neutrophils in the lungs of high-dose Mtb infected mice expressed an intermediate level of Ly6G and Gr-1 (Ly6G^intGr-1^int) (Fig. 5a, S5), a characteristic of myeloid derived suppressor cells (MDSCs) known to potently suppress T cell function during cancer and chronic infections^34,35. Atg16l1^f/f-LysM-cre mice infected with high-dose Mtb accumulated a higher percentage of Ly6G^int cells at 14 and 21 dpi (Fig. 5a). This led us to hypothesize that the lower number of T cells in the Atg16l1^f/f-LysM-cre mice during high-dose Mtb infection could be due to increased neutrophil inflammation, including the accumulation of Ly6G^int cells. To determine if depletion of neutrophils would improve T cell responses and subsequent control of Mtb, we administered anti-Ly6G (clone 1A8) antibody between 8–28 dpi, following a regimen previously shown to rescue Atg5^f/f-LysM-cre and other susceptible mouse lines from Mtb infection^6,36. Anti-Ly6G treatment effectively eliminated Gr-1 high (Gr-1^hi) neutrophils in infected mice but failed to improve survival (Fig. 5b, Extended Data Fig. 6a). Anti-Ly6G antibody partially decreased Mtb burden in the lungs of mice compared to control-IgG treated mice (Extended Data Fig. 6b). This was likely due to reduced Mtb infected Gr-1^hi neutrophils during 1A8 antibody treatment (Extended Data Fig. 6c). However, anti-Ly6G antibody treatment did not affect the numbers of Gr-1^int neutrophils or Mtb-infected Gr-1^int cells in the lungs of infected Atg16l1^f/f-LysM-cre mice (Extended Data Fig. 6d–6e). Thus, the resistance of the Gr-1^int population to antibody-mediated depletion likely explains why anti-Ly6G antibody treatment failed to completely revert the Mtb burdens in Atg16l1^f/f-LysM-cre mice (Extended Data Fig. 6b). Anti-Ly6G antibody treatment also failed to rescue the decrease of alveolar macrophages, DCs, or T cells in Atg16l1^f/f-LysM-cre lungs at 21 dpi (Fig. 5c, Extended Data Fig. 6f), suggesting that accumulation of the Gr-1^int population could be contributing to the defects in innate immune cells and T cells.

Figure 5. Autophagy deficiency in myeloid cells leads to accumulation of MDSC like cells during high-dose Mtb infection.

a. Representative flow cytometry plots of live CD45⁺MerTK⁻CD64⁻ cells gated for Ly6G^high and Ly6G^int neutrophils from mice lungs. Data is representative of >3 mice.

b, Survival of high-dose Mtb infected mice that were treated with neutrophil depletion antibody (1A8) or isotype control immunoglobulin (control) every other day from 8–28 dpi. P values calculated by log-rank Mantel-Cox tests. Atg14^f/f, n= 13 mice for control and 1A8 treatment; Atg14^f/f-LysM-cre, n= 12 mice for control and 1A8 treatment; Atg16l1^f/f, n= 16 mice for control treatment, n= 14 mice for 1A8 treatment; Atg16l1^f/f-LysM-cre, n= 16 mice for control treatment, n= 15 mice for 1A8 treatment.

c, The number of T cells in lungs of high-dose Mtb infected mice treated with 1A8 or control antibody at 21 dpi. Data (Means ± s.e.m.) pooled from 3 independent experiments. Atg16l1^f/f, n= 10 mice for control treatment, n= 8 mice for 1A8 treatment; Atg16l1^f/f-LysM-cre, n= 8 mice for control treatment, n= 9 mice for 1A8 treatment. P values calculated by two-tailed Mann-Whitney tests.

d, Gene set enrichment analysis (GSEA) of bulk RNAseq of lungs from Atg14^f/f-LysM-cre versus Atg14^f/f mice at 14 dpi of high-dose Mtb infection. The green curve represents the density of the genes identified in RNA-seq analysis where normalized enrichment score (NES), P_adj, and false discovery rate (q-value) are indicated.

e,f, UMAP plot of scRNAseq results of all immune cells (e) and neutrophils (f) across all samples (lungs from naïve and 21 dpi of high-dose Mtb infected Atg14^f/f, Atg14^f/f-CD11c-cre, Atg14^f/f-LysM-cre mice, n=4 samples pooled per group). Colored according to identified clusters, and dot plot showing the expression level of markers gene S100a8.

g, Distribution of neutrophil clusters from scRNAseq analysis of lungs of naïve and 21 dpi of high-dose Mtb infected (Mtb) mice.

h, GSEA of activated-PMN-MDSC and PMN-MDSC signatures³⁹ (Fig. S6) in scRNAseq neutrophil cluster 1 and 4 (PMN1, PMN4). The green curve represents the density of the genes identified in scRNA-seq analysis.

i, Dot plot showing highly expressed genes in PMN1 and PMN4, and expression of these genes in PMN-MDSCs of mouse cancer models and human cancer patients³⁹, and active TB patients. Dot size represents prevalence of the transcript, and shade indicates expression level. The red arrows indicate genes identified among signature genes in active TB patients¹⁵.

j, Level of surface molecules on neutrophils from lungs of mice at 14 dpi of high-dose Mtb infection, measured by flow cytometory analysis. Representative data (mean ± s.e.m.) of 2 experiments. Atg16l1^f/f, n= 4; Atg16l1^f/f-LysM-cre, n= 5 mice. P values were calculated by two-tailed Mann-Whitney tests. * for P < 0.05, ** for P < 0.01, *** P < 0.001.

k, GSEA of neutrotime early gene and late gene signatures⁴⁷ in scRNAseq neutrophil 4 (PMN4). The green curve represents the density of the genes identified in scRNA-seq analysis.

l, Representative flow cytometry plots of live CD45⁺ cells from mice blood and bone marrow gated for Ly6G^high and Ly6G^int neutrophils. Data is representative of 3 mice from 2 experiments.

* for P < 0.05, ** for P < 0.01, *** P < 0.001, and **** P < 0.0001. ns= not significant.

The Ly6G^intGr-1^int cells that accumulate in infected lungs of Atg16l1^f/f-LysM-cre mice are reminiscent of polymorphonuclear MDSCs^34,35. Polymorphonuclear MDSCs are known to exacerbate disease by suppressing T cell function, including in TB models³⁷. Pathway analysis of RNA-sequencing (RNA-seq) data from total lungs revealed significant downregulation of IFNγ response pathways in Atg14^f/f-LysM-cre mice at 14 dpi (Fig. 5d), when the numbers of T cells in the lungs of autophagy-deficient mice were similar to control mice but Ly6G^intGr-1^int cells had accumulated in the lungs of autophagy-deficient mice (Fig. 5a). These data suggested that the appearance of Ly6G^intGr-1^int neutrophil population coincides with suppression of T cell function and IFNγ production. However, polymorphonuclear MDSCs lack definitive surface markers, which precludes the unequivocal identification of these cells. Therefore, to determine if the Gr-1^int neutrophils accumulating during Mtb infection share characteristics with polymorphonuclear MDSCs, we performed single cell RNA-seq (scRNA-seq) on naïve and Mtb-infected lungs from Atg14^f/f, Atg14^f/f-LysM-cre and Atg14^f/f-CD11c-cre mice. CD45⁺ cells were stratified for unbiased clustering to identify cell types (Fig. 5e). Within neutrophils, six S100a8⁺ sub-clusters (clusters PMN0-PMN5) were identified (Fig. 5f). Uninfected Atg14^f/f-LysM-cre and Atg14^f/f-CD11c-cre lungs showed altered neutrophil populations (Fig. 5g), consistent with the elevated basal lung inflammation previously reported these mice³⁸. Mtb infection had a substantial impact on the neutrophil populations (Fig. 5g). At 21 dpi, Atg14 expression affected the presence of specific neutrophil subsets, with PMN1 and PMN4 being the major clusters present in the lungs of Atg14^f/f-LysM-cre and Atg14^f/f-CD11c-cre mice, compared to PMN2 and PMN3 in Atg14^f/f mice (Fig. 5g).

To identify signatures of pathways enriched in PMN1 and PMN4, we performed Gene Set Enrichment Analysis (GSEA) of prior defined pathways and PMN-MDSC signature genes sets in murine cancer models³⁹ (Fig. S6). GSEA analysis demonstrated that the most significantly enriched pathway in PMN1, compared to all PMN clusters, was the activated-PMN-MDSC gene set from murine cancer (Fig. 5h, S7a). PMN4 was also distinctly enriched for the cancer PMN-MDSC gene set (Fig. 5h, S7b). There was a substantial overlap between differentially upregulated genes in PMN1 and activated-PMN-MDSC genes from murine cancer, as well was with PMN4 and PMN-MDSC signature genes³⁹ (Fig. 5i), indicating that neutrophils accumulating during high-dose Mtb infection in lungs of Atg14^f/f-LysM-cre and Atg14^f/f-CD11c-cre mice are similar to polymorphonuclear MDSCs in murine cancer models. This data supports prior studies that have identified an association between the presence of polymorphonuclear MDSCs and progressed active TB disease^37,40–43. Eleven enriched activated-PMN-MDSC and PMN-MDSC genes (Pglyrp1, Plscr1, Gyg, Ifitm3, Ifitm1, Il1rn, Cd274, Gadd45b, Sod2, Npc2, and Hmox-1) in PMN1 and PMN4 are among the signature transcripts within the blood of active TB patients^15,44 (Fig. 5i), suggesting that PMN1 and PMN4 may represent MDSC signatures in human TB patients and could be associated with TB progression. We investigated whether neutrophils accumulating in autophagy-deficient mice following high-dose Mtb infection exhibited markers for MDSCs by performing flow cytometry on lungs and confirmed a decrease in surface levels of Ly6G, MHC-II, and CD63 (marker for primary granule release) on neutrophils of Atg16l1^f/f-LysM-cre mice, indicating a defect in activation and/or altered functions (Fig. 5j). Neutrophils from Mtb infected Atg16l1^f/f-LysM-cre mice also exhibited increased surface expression of PD-L1 (Fig. 5j), an immune checkpoint molecule that limits T cell proliferation and function through interaction with PD-1. PD-1/PD-L1 is upregulated in non-human primates with severe TB disease⁴⁵ and limits the number of Mtb-specific CD4 T cells in mice, supporting PD-L1’s role in suppressing T cell responses during Mtb infection⁴⁶. Collectively these results suggest that the decrease in T cells in Atg16l1^f/f-LysM-cre mice could be due to accumulation of immunosuppressive polymorphonuclear MDSCs.

Neutrophil maturation mainly takes place in the bone borrow before the mature neutrophils egress into the blood and tissues⁴⁷. To investigate if the polymorphonuclear MDSCs that accumulated in the lungs were immature cells originating from altered granulopoiesis or mature neutrophils shaped in the milieu of infected lungs, we compared our scRNA-seq results with published gene signatures of neutrophil maturation⁴⁷. We found that PMN4, the PMN-MDSC cluster, was enriched for early-continuum genes, which are stage-specific genes of neutrophil precursors and immature neutrophils, while late-continuum genes associated with mature neutrophils were downregulated in the cluster (Fig. 5k). These findings suggested that the lungs of Mtb infected Atg16l1^f/f-LysM-cre mice contain increased immature neutrophils. In line with these results, we observed an accumulation of Ly6G^int cells in the blood and bone marrow of infected Atg16l1^f/f-LysM-cre mice (Fig. 5l). These results align with previous studies reporting expanded neutrophil precursors and their recruitment to the periphery during chronic inflammation⁴⁸. Taken together, our results suggest that the immature polymorphonuclear MDSCs arise in the bone marrow of high-dose Mtb infected Atg16l1^f/f-LysM-cre mice and subsequently accumulate in the lungs, where they suppress T cell responses. The increased accumulation of MDSCs occurred in both Atg14^f/f-LysM-cre and Atg14^f/f-CD11c-cre mice during high-dose Mtb infection (Fig. 5g), supporting that autophagy genes function specifically in CD11c⁺ cells to suppress the accumulation of polymorphonuclear MDSC populations.

Autophagy suppresses apoptosis and promotes T cell proliferation

The accumulation of MDSCs in the lungs of Mtb infected Atg14^f/f-LysM-cre, Atg14^f/f-CD11c-cre, and Atg16l1^f/f-LysM-cre mice is associated with a subsequent decrease in alveolar macrophages, DCs, and T cells (Fig. 4c–4d, 4j). To determine the mechanistic basis for the loss of alveolar macrophages and DCs, we analyzed the gene expression profiles of the macrophages/monocytes/DCs (MMD) clusters in the scRNA-seq data (Fig. 6a). The 6 MMD clusters were significantly shifted upon Mtb infection in all genotypes (Fig. 6b). Pathway analysis on the CD11c⁺ (Itgax) subclusters revealed that inflammatory pathways were upregulated in CD11c⁺ cells of Atg14^f/f-LysM-cre and Atg14^f/f-CD11c-cre mice at 21 dpi (Fig. 6c, S8–S9), consistent with higher bacterial burdens and inflammation in autophagy-deficient mice at that time point (Fig. 1h–1i, 2f). In addition, apoptosis was among the significantly upregulated pathways in CD11c+ cells in infected Atg14^f/f-LysM-cre mice (Fig. 6c), consistent with the decreased macrophages and DCs observed in these mice at 21 dpi (Fig. 4c–4d). Flow cytometry analysis revealed increased activated Caspase 3/7 in alveolar macrophages of infected Atg16l1^f/f-LysM-cre mice at 21 dpi but not in non-alveolar macrophages (Fig. 6d, Extended Data Fig. 7), indicating heightened apoptosis in alveolar macrophages. We did not detect a difference in the levels of alveolar macrophage apoptosis at 14 dpi in Atg16l1^f/f-LysM-cre and Atg16l1^f/f mice (Fig. 6e), indicating that the increased alveolar macrophage apoptosis occurs following the accumulation of the Ly6G^IntGr1^Int neutrophils in the lungs (Fig. 5a). These data link MDSC accumulation with inflammatory and apoptosis pathways in macrophages in response to high-dose Mtb infection.

Figure 6. Autophagy myeloid deficiency leads to apoptosis of alveolar macrophage and reduced T cell proliferation after high-dose Mtb infection.

a,f, UMAP plots of macrophages/monocytes/DCs clusters (MMD, d), and T cell clusters (h) from scRNAseq analysis colored by clusters.

b,g, Distribution of MMD (b), and T cells (g) sub-clusters.

f, significantly elevated and downregulated HALLMARK pathways found within all CD11c⁺ (Itgax) cells (see Fig. S8) from infected Atg14^f/f-LysM-cre lungs compared with infected Atg14^f/f controls as assessed by GSEA analysis.

d,e, Representative histogram and quantification of flow cytometry analysis of Caspase3/7 activity (FLICA+) in alveolar macrophages from the lungs at 14 dpi (d) and 21 dpi (e) of high-dose Mtb infection. Data (mean ± s.e.m.) pooled from 2 independent experiments. Atg16l1^f/f, n= 6; Atg16l1^f/f-LysM-cre, n= 6 mice. P values calculated by two-tailed Mann-Whitney tests.

* for P < 0.05. ns= not significant.

Both activated-PMN-MDSC and PMN-MDSC isolated from tumor-bearing mice robustly suppress T cell responses, suggesting that the polymorphonuclear MDSC populations in Atg14^f/f-LysM-cre and Atg14^f/f-CD11c-cre mice at 21 dpi could be responsible for the decreased T cell numbers in the lungs. Analysis of the T cell clusters within the scRNA-seq datasets identified 6 sub-clusters (cluster T0-T5) (Fig. 6f). Upon Mtb infection in all genotypes, T cells skewed towards Th1 with increases in the Tbx21+ cluster (T0) and decreased naïve Ccr7+ clusters (T2, T5) (Fig. 6g). However, the Mki67⁺ proliferating T cell (T1) was reduced in Atg14^f/f-LysM-cre and Atg14^f/f-CD11c-cre lungs as compared to Atg14^f/f controls at 21 dpi (Fig. 6f–6g). These data suggest that an accumulation of MDSCs is associated with decreased T cell proliferation, which could explain the lower numbers of T cells at 21 dpi in the lungs of high-dose Mtb infected Atg14^f/f-LysM-cre and Atg14^f/f-CD11c-cre mice.

DISCUSSION

Our studies have unveiled an unprecedented role for autophagy in CD11c⁺ lung macrophages and DCs that is required for control of Mtb pathogenesis in a high-dose murine model. Contrary to prior in vitro studies, our results indicate that loss of autophagy in macrophages does not affect cell intrinsic control of Mtb replication in vivo. Instead, loss of autophagy in CD11c⁺ cells results in excessive accumulation of polymorphonuclear MDSCs, which shares similarity to observations in active TB patients. The accumulation of polymorphonuclear MDSCs that harbor high bacterial burden is followed by a reduction of antigen-presenting cells and decreased T cell proliferation in the lung, thwarting sustained T cell responses and derailing the control of Mtb infection.

Research on autophagy during Mtb infection has primarily centered on studying xenophagy in vitro in cell culture^1–4. Deletion of autophagy genes in myeloid cells leads to a transient increase in lung Mtb burdens at 21dpi with low-dose Mtb⁵, however, this is not associated with higher bacterial burdens in autophagy-deficient macrophages and is instead associated with increased neutrophil accumulation^5,7. Furthermore, deletion of autophagy genes in myeloid cells does not result in susceptibility to low-dose Mtb infection until late during infection^5,7, where the role of autophagy at these later time points is unknown. This late susceptibility is observed when the target autophagy genes are deleted using two copies of the LysM-cre allele⁵ or one copy of the CD11c-cre allele⁷, which suggests that the role for autophagy could be in lung macrophages where CD11c driven expression of Cre is more efficient at targeting genes in alveolar macrophages than LysM-driven expression of Cre²². Herein, we report early susceptibility to high-dose Mtb infection in mice deleted for autophagy genes using either one copy of LysM-cre or CD11c-cre, which underpins the critical role of autophagy in alveolar macrophages during high-dose Mtb infection.

We also report the first cytological analysis of autophagy-deficient alveolar macrophages from Mtb-infected mice, revealing that autophagy is indeed induced in alveolar macrophages during infection, but does not contribute to controlling Mtb replication in alveolar macrophages in vivo. These data in primary alveolar macrophages are in contrast to previous reports using cultured macrophages that support a role for xenophagy in controlling Mtb replication, which highlights cell-type specific impacts of autophagy and likely stems from the diverse and specialized functions of tissue-resident macrophages shaped by distinct tissue microenvironments⁴⁹. Distinct from the role of xenophagy in cultured macrophages, autophagy as an evolutionarily conserved pathway has other critical immune regulatory functions in vivo during Mtb pathogenesis. Our results emphasize the importance of utilizing in vivo systems to elucidate lung-specific immunological responses during Mtb infection, challenging the adequacy of in vitro models in replicating the host’s complex reactions to Mtb²⁷.

It is unclear why Mtb infection only manifests as active TB disease in a fraction of individuals with similar genetic susceptibility⁵⁰. Using a higher Mtb infectious dose in mice, which better recapitulates the immune responses during human active TB disease, we demonstrate that the outcome of a pathogen-plus-susceptibility gene interaction is dependent on pathogen burden. The in vivo role for multiple core autophagy factors for early protection against Mtb pathogenesis were revealed during defense against high-dose Mtb in mice. Future in-depth investigations of host responses to high-dose Mtb will reveal the mechanism of this variance in gene function that is dependent on pathogen burden. This will have important implications both for understanding the molecular roles of autophagy gene in infection and inflammation, and for the potential use of autophagy modulators in clinical medicine.

MATERIALS AND METHODS

Mice

All flox mice (Atg5^{f/f 51}, Atg7^{f/f 52}, Atg16l1^{f/f 53}, Atg14^{f/f 17,54}, Fip200^f/f, Becn1^f/f, Rubcn^−/−) used in this study have been described previously and colonies are maintained in an enhanced barrier facility^6,16. LysM-cre (Jax #004781), CD11c-cre (Jax #007567), Mrp8-cre (Jax #021614), Casp1/11^−/− (Jax #004781) from the Jackson Laboratory were crossed to specific flox mice. P25/Nur77-GFP/CD45.1 mice³³ were a gift from Dr. Joel Ernst. Mlkl^−/− mice were from Dr. James Murphy⁵⁵. Male and female littermates (aged 8–12 weeks) were used and were subject to randomization. Statistical consideration was not used to determine mouse sample sizes. The mice were housed and bred at Washington University in St. Louis and Tsinghua University in specific pathogen-free conditions in accordance with federal and university guidelines, and protocols were approved by the Animal Studies Committee of Washington University and Tsinghua University.

M. tuberculosis infection in mice

Mycobacterium tuberculosis Erdman, GFP-Mtb Erdman⁵⁶, and mCherry-expressing Mtb Erdman (mCherry-Mtb) were used in all experiments. mCherry-Mtb was generated by transforming the wild-type strain Erdman with the pCherry3 plasmid⁵⁷. Mtb was cultured at 37°C in 7H9 (broth) or 7H11 (agar) (Difco) medium supplemented with 10% oleic acid/albumin/dextrose/catalase (OADC), 0.5% glycerol, and 0.05% Tween 80 (broth). Cultures of GFP-Mtb and mCherry-Mtb were grown in the presence of kanamycin or hygromycin, respectively, to ensure plasmid retention. Mtb cultures in logarithmic growth phase (OD600 = 0.5–0.8) were washed with PBS + 0.05% Tween-80, sonicated to disperse clumps, and diluted in sterile water before delivering 1000 CFUs of aerosolized Mtb per lung using an Inhalation Exposure System (Glas-Col). Within 2 hours of each infection, lungs were harvested from at least two control mice, homogenized, and plated on 7H11 agar to determine the input CFU dose. At each time point after infection, Mtb titers were determined by homogenizing the superior, middle, and inferior lobes of the right lung and plating serial dilutions on 7H11 agar. Colonies were counted after 3 weeks of incubation at 37°C in 5% CO2.

Western blot and PCR

Bronchoalveolar lavage fluid (BALF) from Atg16l1^f/f and Atg16l1^f/f-LysM-cre mice were extracted, centrifuged and then resuspended in 2× Laemmli buffer, resolved using 4–20% polyacrylamide gels (BioRad), transferred to PVDF membranes (BioRad) and detected using the following antibodies: anti-LC3b (CST, 2775S), anti-p62/SQSTM1 (Sigma, P0067) and anti-actin-HRP (CST, 5125s) or secondary goat-anti-Rabbit-HRP (CWBIO, 01334). HRP was detected using ECL (Epizyme, SQ201). For PCR, live lung alveolar macrophages, neutrophils, DCs and monocytes were sorted by BD FACSARIA^™ III. Genome of these 4 types of lung myeloid cells were extraced by universal genome DNA kit (CWBIO, CW2298s). Atg16 flox allele deletion was determined from genome DNA amplified by PCR with the following primers: FLOX-5F: CTTTCTCAACAGAACCAGCAGTAC; and Deletion-3 R: AGAAGTGCCTGTGGGTGGGT. Floxed allele results in 1260 bp band and excised allele results in 494 bp band. Products were amplified with 94°C melt (×5 min), 94°C (×30 sec), 60°C (×30 sec), 72°C (×1 min) for 30 cycles.

Immune fluorescent microscopy

For staining of macrophages from in vivo infection, alveolar macrophages from BALF of mCherry-Mtb infected mice were adhered to coverslips by centrifugation at 400 xg for 5 min to isolate alveolar macrophages followed by incubation for 20 min before being fixed with 4% paraformaldehyde. For ex vivo infected alveolar macrophage staining, alveolar macrophages from BALF of naïve mice were adhered to coverslips in a 24-well plate overnight at 37°C in RPMI with 10% FBS. Normal mouse serum opsonized mCherry-Mtb was added at an MOI of 10 and the plate was spun at 200 xg for 10 minutes, the media removed and replaced with fresh media, and the cells were incubated for 48 hours at 37°C before being fixed in 4% paraformaldehyde. Staining was performed with antibodies specific for LC3(MBL Life Sciences, Cat. #PM036). DNA was labeled with 5 g/ml Hoechst 33342 (Molecular Probes). mCherry-Mtb was used to detect cells with Mtb infection. A Nikon A1Rsi confocal microscope coupled with NIS software was used to take Z-stack images with a 60x oil immersion objective. Z-Stack confocal images were merged using FIJI software. Fields that were imaged and used for quantification were selected randomly based on areas containing around 100 cells to keep the cell number consistent between groups or treatments. Ridge detection analysis was performed on 8-bit images.

T cell transfer and neutrophil depletion

For tracing Mtb specific T cells, 4–5 ×10⁶ P25 TCR-Tg CD4⁺ T cells (CD45.1) were labeled with CellTrace (Invitrogen) and adoptively transferred by tail vein injection in 100 μL of sterile PBS 24 hours (h) before mice were infected by Mtb. For neutrophil depletion, mice were intraperitoneally injected every 48 h with 200 μg monoclonal anti-Ly6G antibody (clone 1A8; Leinco) or 200 μg polyclonal rat serum IgG (Sigma-Aldrich) diluted in sterile PBS beginning at 8 dpi and ending at 28 dpi.

Flow cytometry, quantification of apoptosis, and cytokine analysis

Mouse lungs (left lobe) were excised after PBS perfusion, placed in DMEM containing 10% FBS, minced finely, and digested at 37°C for an hour with mechanical disruption with a stir bar and enzymatic digestion with Liberase Blendzyme III (Roche), hyaluronidase (Sigma), and DNase I (Sigma). Mediastinal lymph nodes were digested with collagenase B and DNase I (Sigma). Digested tissues were treated with ACK buffer to remove red blood cells and passed through a 70 μm cell strainer to generate single cell suspensions. Cells were stained with Zombie-violet (Biolegend, Cat. #423114) or Zombie-NIR (Biolegend, Cat. #423106) before resuspending in PBS with 2mM EDTA, 3% FBS, and anti-FcγRII/III (Biolegend, Cat. #101302) for blocking. Cells were then labeled with specific antibodies against CD45 (Biolegend, Cat. #147716), Ly6G (Biolegend, Cat. #127618), CD11c (Biolegend, Cat. #147716), CD64 (Biolegend, Cat. #161006), MerTK (Biolegend, Cat. #151506), Ly6C (Biolegend, Cat. #128041), CD11b (Biolegend, Cat. #101230), I-A/I-E (BD Biosciences, Cat. #AB_2741827), siglecF (BD Biosciences, Cat. #AB_2740118), Gr-1 (Biolegend, Cat. #108457), TCRb (Biolegend, Cat. #109222), CD4 (Biolegend, Cat. #100432), CD8a (BD Biosciences, Cat. #AB_2732919), CD62L (Biolegend, Cat. #161204), CD44 (Biolegend, Cat. #103012) following manufacture recommendation. Flow cytometric analysis was performed on an LSRFortessa (BD Biosciences) and Aurora (Cytek), and data was analyzed with FlowJo software (Tree Star). Total cell number was multiplied by the percentage of specific cell type in total single cells, as analyzed by flow cytometry.

For analyzing apoptosis, single-cell suspensions from mouse lungs were stained with Zombie-NIR and washed with PBS before being resuspended in PBS with FLICA substrate (Bio-Rad, Cat. #ICT9125) according to the manufacturer’s protocol and incubated for 15 mins at room temperature. Cells were then washed with the apoptosis wash buffer provided in the kit before prior to being stained with several specific antibodies as described above. Flow cytometric analysis was described as above.

For determining T cell antigen specificity, singe-cell suspensions from mouse lungs were incubated with APC-conjugated ESAT-6 (QQWNFAGIEAAASA), Ag85 (FQDAYNAAGGHNAVF), or negative control human CLIP (PVSKMRMATPLLMQA) tetramer (NIH Tetramer Core Facility) for 75 minutes at room temperature before being stained with Zombie-NIR and several specific antibodies as described above. Flow cytometric analysis was performed as described above.

For cytokine analysis, lung homogenates were filter sterilized and then analyzed with Mouse Chemokine/Cytokine Panel 1A (ThermoFisher) on a Luminex xMAP FlexMAP3D (Luminex Corp) with Millipore Sigma Belysa software.

Bulk RNA-seq

For RNA-seq, lung tissues from Atg14^f/f and Atg14^f/f-LysM-cre mice at 14 dpi of high-dose Mtb were collected. RNA was isolated from cells using the RNeasy minikit (QIAGEN) in accordance with the manufacturer’s instructions. An mRNA Illumina sequencing library was generated and run on an Illumina HiSeq as previously described³⁸. Each group contained n = 4 samples. DESeq2 was used for differential gene expression analysis⁵⁸, and the data from which was used as a ranked list in pre-ranked gene set enrichment analyses to identify pathway enrichment by using cluster Profiler 4.0⁵⁹.

Single-cell RNAseq

scRNA-seq sample preparation was done according to the manufacturer instructions (10x Genomics). Briefly, single cells from naïve or infected lungs were enriched for live cells using dead-cell depletion kit (Miltenyi). Then single cell suspensions were subjected to droplet-based massively parallel scRNA-seq using the Chromium Single Cell 3′ (v3) Reagent Kit in the BSL-3 laboratory as per manufacturer’s instructions (10x Genomics). Cell suspensions were loaded at 1,000 cells/μl to capture 10,000 cells/lane. The 10x Chromium Controller generated GEM droplets, where each cell was labeled with a specific barcode, and each transcript was labeled with a unique molecular identifier (UMI) during reverse transcription. The barcoded cDNA was isolated and removed from the BSL-3 space for library generation. The cDNA underwent 11 cycles of amplification, followed by fragmentation, end repair, A-tailing, adaptor ligation, and sample index PCR as per the manufacturer’s instructions. Libraries were sequenced on a NovaSeq S4 (Illumina), targeting 50,000 read pairs/cell.

The Cell Ranger Single-Cell Software 3.0 available on the 10x Genomics website was used to perform sample demultiplexing. We aligned the resulting fastq files on mouse genome (Ensembl 98) with Cellranger counts. Cellranger counts were processed through the R package Seurat 4⁶⁰. We filtered cells that (1) had more than 5% of mitochondrial gene content, (2) had less than 500 detected genes. Data was log-normalized with a scale factor of 10,000. Dimensionality reduction and clustering were done by detecting the most variable genes using the FindVariableFeatures function. Latent variables (number of UMI’s and mitochondrial content) were regressed out using a negative binomial model (function ScaleData). A UMAP dimensionality reduction was performed on the scaled matrix using the first 30 PCA components to obtain a two-dimensional representation of the cell states. For clustering, we used the functions FindClusters (resolution 0.5). To identify marker genes, we used FindAllMarkers to compare each cluster against all other clusters. For each cluster, only genes that were expressed in more than 25% of cells with at least 0.25-logfold differences were considered. Clusters were annotated using SingleR⁶¹. Non-immune cells that had low CD45 (Ptprc) expression were excluded, and doublets were removed based on scDblFinder⁶². Designated cell types were sub-clustered using Findclusters (resolution 0.3 for PMN, 0.1 for MMD and T cells), and FindMarkers were used to identify differentially expressed genes (DEgenes). DEgenes of PMN1 or PMN4 were defined by comparing these sub-clusters to all PMNs. Pathway analysis with GSEA on DEgenes of neutrophil subclusters PMN1 and PMN4 was done by using cluster Profiler 4.0⁵⁹. Specifically, (HALLMARK, KEGG and REACTOME), as well as 2 custom defined MDSC gene sets (Fig. S7) from published dataset (GSE163834)³⁹ and 2 defined neutrotime gene sets from published dataset (GSE165276)⁴⁷. Top DEgenes of PMN1 and PMN4 were interrogated in MDSC clusters of published cancer database³⁹ to generate dot plots using ggplot2 in R. These DEgenes were also compared to published marker genes in active TB patients and hits were indicated as arrows on dotplot¹⁵. CD11c+ subclusters were chosen from MMD on resolution 0.3 based on log2FC of CD11c (Itgax) (Fig. S9). DEgenes of infected Atg14^f/f-LysM-cre or Atg14^f/f-CD11c-cre were defined by comparing to infected Atg14^f/f within CD11c+ MMDs. Pathway analysis with GSEA of DEgenes was done by using cluster Profiler 4.0⁵⁹, to compare between infected groups.

Statistical analysis for biological experiments

All data are from at least two independent experiments. Samples represent biological (not technical) replicates of mice randomly sorted into each experimental group. No blinding was performed during animal experiments. No animal or data point was excluded from the analyses. Statistical differences were calculated using Prism (9.0; GraphPad Software) using log-rank Mantel-Cox tests (survival), unpaired two-tailed Student’s t tests, or unpaired two-tailed Mann-Whitney tests. Sample sizes were sufficient to detect differences as small as 10% using the statistical methods described. When used, center values and error bars represent means ± s.e.m. P < 0.05 was considered significant. P >0.05 was denoted *, ** for P < 0.01, *** P < 0.001, and **** P < 0.0001. Data are presented as mean ± s.e.m. Notable comparisons that were not significantly different are designated as not significant (ns).

Extended Data

Extended Data Fig. 1: Analysis of Atg16l1 gene deletion efficiency from primary cells and weight loss of mice infected with high-dose Mtb.

a, PCR assay to confirm deletion of Atg16l1 in myeloid cells from the lung of mice. 2 representative samples of n = 4. b-c, Weight loss of mice after aerosol infection with high-dose Mtb. Data are presented as mean ± s.e.m.

Extended Data Fig. 2: Ex vivo and in vivo analysis of autophagy during Mtb infection in alveolar macrophages.

a, Representative images of LC3-stained alveolar macrophages isolated from naïve mice and then infected ex vivo with mCherry-Mtb for 48 h. LC3 (green), nuclear staining (blue) and Mtb (red). Scale bars, 2 or 3 μm. b,c, Quantitative analysis of the area of LC3 puncta (b) and % of Mtb colocalized with LC3 (c) in mCherry-Mtb+ CD11c+ alveolar macrophages infected ex vivo with mCherry-Mtb. Atg16l1^f/f, n = 14 cells for area of LC3 puncta, n = 15 cells for % of LC3+ puncta Mtb area; Atg16l1^f/f-CD11c-cre, n = 21 cells for area of LC3 puncta, n = 26 cells for % of LC3+ puncta Mtb area. d, Representative images of LC3-stained alveolar macrophages from mice at 14 dpi with high-dose mCherry-Mtb. LC3 (green), nuclear staining (blue) and Mtb (red). Scale bars, 2 μm. e,f, Quantitative analysis of the area of LC3 puncta (e) and % of Mtb colocalized with LC3 (f) in mCherry-Mtb+ CD11c+ alveolar macrophages from BALF of mice at 14 dpi of high-dose Mtb infection. Atg16l1^f/f, n = 27 cells for area of LC3 puncta, n = 31 cells for % of LC3+ puncta Mtb area; Atg16l1^f/f-CD11c-cre, n = 29 cells for area of LC3 puncta, n = 31 cells for % of LC3+ puncta Mtb area. Data representative (Means ± s.e.m.) of n = 3 biological repeats. P values calculated by two-tailed Mann-Whitney tests. * for P < 0.05, and **** P < 0.0001. ns = not significant.

Extended Data Fig. 3: Heightened inflammation in lungs of autophagy deficient mice after high-dose Mtb infection.

a, Concentration of cytokines in high-dose Mtb infected lungs as detected by multiplex cytokine panel. Data (Means ± s.e.m.) pooled from 2 independent experiments. Atg16l1^f/f, n = 7 mice at 14, 21dpi; Atg16l1^f/f-LysM-cre, n = 9 mice at 14 dpi, n = 7 mice at 21 dpi. P values were calculated by two-tailed t-tests. b, The number of total immune cells in the lung of mice during high-dose Mtb infection. Data (Means ± s.e.m.) pooled from 3 independent experiments. Atg16l1^f/f, n = 15 mice at naïve condition, n = 7 mice at 14 dpi, n = 12 mice at 21 dpi; Atg16l1^f/f-LysM-cre, n = 16 mice at naïve condition, n = 10 mice at 14 dpi, n = 17 mice at 21 dpi. P values were calculated by two-tailed Mann-Whitney tests. * for P < 0.05, ** for P < 0.01, *** P < 0.001, and **** P < 0.0001. ns = not significant.

Extended Data Fig. 4: Histology analysis of naïve mouse lungs.

a, Representative H&E-stained sections of naïve mouse lungs. Data represent n = 3 mice each group.

Extended Data Fig. 5: Analysis of antigen specific T cells and MHC-II level on innate immune cells.

a,b,c Quantification (a,b) and representative flow plot (c) of Ag85a and ESAT6 positive CD4+ T cells from lungs and mLNs of mice at 14 dpi (a) and 21 dpi (b,c) of high-dose Mtb infection. Atg16l1^f/f, n = 6; Atg16l1^f/f-LysM-cre, n = 6 mice. d, MHC-II mean fluorescent intensity (MFI) in alveolar macrophages, non-alveolar macrophages, DCs, and monocytes from lungs at 14 dpi of high-dose infection with high-dose Mtb. Atg16l1^f/f, n = 6; Atg16l1^f/f-LysM-cre, n = 5 mice. Data (Means ± s.e.m.) from 2 independent experiments are graphed. P values calculated by two-tailed Mann-Whitney tests. ns = not significant.

Extended Data Fig. 6: Accumulation of Ly6G^intGr-1^int neutrophils in autophagy deficient mice is associated with susceptibility and high Mtb burden.

a,d,f, The number and percentage of Gr-1 high (Gr-1^hi) neutrophils (a), Gr-1 int (Gr-1^int) neutrophils(d), and alveolar macrophages and DCs (f) in lungs of high-dose Mtb infected mice treated with neutrophil depletion antibody (1A8) or isotype control immunoglobulin (control) at 21 dpi of high-dose Mtb infection. Atg16l1^f/f, n = 9 mice for control treatment, n = 10 mice for 1A8 treatment; Atg16l1^f/f-LysM-cre, n = 9 mice for control treatment, n = 10 mice for 1A8 treatment. b, Mtb CFU in lungs at 21 dpi of high-dose Mtb infections with 1A8 or control antibody treatment. Atg16l1^f/f, n = 10 mice for control treatment, n = 11 mice for 1A8 treatment; Atg16l1^f/f-LysM-cre, n = 10 mice for control treatment, n = 12 mice for 1A8 treatment. c,e, number of Mtb infected cells, measured as GFP+ cells in the lungs at 21 dpi with 1A8 or control antibody treatment. Atg16l1^f/f, n = 9 mice for control treatment, n = 10 mice for 1A8 treatment; Atg16l1^f/f-LysM-cre, n = 9 mice for control treatment, n = 10 mice for 1A8 treatment. Data (Means ± s.e.m.) pooled from 3 independent experiments. P values calculated by two-tailed Mann-Whitney tests. * for P < 0.05, ** for P < 0.01, *** P < 0.001, and **** P < 0.0001. ns = not significant.

Extended Data Fig. 7: Non-alveolar macrophages in the lungs did not exhibit increased apoptosis at 21 dpi.

a, Representative histogram and quantification of flow cytometry analysis of FLICA+ non-alveolar macrophages from lungs of mice at 21 dpi of high-dose Mtb infection. Grey histogram indicates isotype control. Data are presented as mean ± s.e.m. Atg16l1^f/f, n = 6; Atg16l1^f/f-LysM-cre, n = 6 mice. P values calculated by two-tailed Mann-Whitney tests. ns = not significant.

Data availability

scRNA-seq data have been deposited in the NCBI Gene Expression Omnibus (GEO) database are accessible through accession number GSE201410. RNA-seq data have been deposited in the NCBI GEO database are accessible through accession number GSE245206. The raw data is provided in the source data table. The method used for analyzing sequencing data is detailed in the materials and methods session. All other relevant data are available from the corresponding author upon reasonable request.