Autophagy prevents early proinflammatory responses and neutrophil recruitment during Mycobacterium tuberculosis infection without affecting pathogen burden in macrophages

The immune response to Mycobacterium tuberculosis infection determines tuberculosis disease outcomes, yet we have an incomplete understanding of what immune factors contribute to a protective immune response. Neutrophilic inflammation has been associated with poor disease prognosis in humans and in animal models during M. tuberculosis infection and, therefore, must be tightly regulated. ATG5 is an essential autophagy protein that is required in innate immune cells to control neutrophil-dominated inflammation and promote survival during M. tuberculosis infection; however, the mechanistic basis for how ATG5 regulates neutrophil recruitment is unknown. To interrogate what innate immune cells require ATG5 to control neutrophil recruitment during M. tuberculosis infection, we used different mouse strains that conditionally delete Atg5 in specific cell types. We found that ATG5 is required in CD11c+ cells (lung macrophages and dendritic cells) to control the production of proinflammatory cytokines and chemokines during M. tuberculosis infection, which would otherwise promote neutrophil recruitment. This role for ATG5 is autophagy dependent, but independent of mitophagy, LC3-associated phagocytosis, and inflammasome activation, which are the most well-characterized ways that autophagy proteins regulate inflammation. In addition to the increased proinflammatory cytokine production from macrophages during M. tuberculosis infection, loss of ATG5 in innate immune cells also results in an early induction of TH17 responses. Despite prior published in vitro cell culture experiments supporting a role for autophagy in controlling M. tuberculosis replication in macrophages, the effects of autophagy on inflammatory responses occur without changes in M. tuberculosis burden in macrophages. These findings reveal new roles for autophagy proteins in lung resident macrophages and dendritic cells that are required to suppress inflammatory responses that are associated with poor control of M. tuberculosis infection.


Introduction
According to the World Health Organization, 10 million people fell ill with Mycobacterium tuberculosis infection and 1.5 million people died of tuberculosis (TB) in 2020, marking the first increase in TB-associated deaths in over a decade [1]. Whether a person controls the initial M. tuberculosis infection or develops active TB disease is directly impacted by the type of immune response elicited in the infected individual [2]. Therefore, better understanding of what constitutes a protective versus non-protective immune response to M. tuberculosis infection is critical for developing better therapies and prevention measures to fight this deadly disease. Genetic mouse models have provided invaluable insight into the immunological processes that are required for control of M. tuberculosis infection. Infection of mice through the aerosol route leads to phagocytosis of M. tuberculosis by alveolar macrophages, initiating an inflammatory response and recruitment of innate immune cells to the lung [2]. M. tuberculosis replicates within these innate immune cells until antigen specific T cells traffic to the lung where they activate the innate immune cells to restrain M. tuberculosis replication and suppress inflammation. M. tuberculosis establishes a chronic infection in wild-type (WT) mice, which survive for over a year with this infection.
ATG5 is required for the intracellular pathway of autophagy, a process by which cytoplasmic contents are targeted to the lysosome for degradation [18,19]. Initiation of autophagy involves phagophore formation from the endoplasmic reticulum, which is mediated by the ULK1 complex (ULK1/ULK2, ATG13, FIP200, and ATG101) and the PI3 kinase complex (ATG14L, BECLIN1, VPS15, and VPS34) [20,21]. Elongation of the autophagosomal double membrane depends on 2 ubiquitin-like conjugation systems. In the first system, ATG12 is activated by ATG7, transferred to ATG10, and covalently attached to ATG5. The second ubiquitin-like component is LC3 (microtubule-associated protein 1 light chain 3), which is conjugated to phosphatidylethanolamine, generating the membrane bound form called LC3-II through the actions of ATG7 and ATG3. ATG5-ATG12 facilitates LC3 lipidation through its interactions with ATG3, while ATG16L1 specifies the localization of LC3 conjugation to the autophagosome membrane [18,19,21,22]. The autophagosome membrane is then completed and targeted for fusion with the lysosome where the autophagosome cargo are degraded. In addition, ATG5 also functions outside of autophagy, including during M. tuberculosis infection [3], although these activities remain poorly understood. Recent work supports an autophagy-dependent role for ATG5 in LysM + innate immune cells in suppressing neutrophil recruitment to the lungs during M. tuberculosis infection [23]. Using in vitro cell culture experiments, multiple groups have reported that macrophages require autophagy to control M. tuberculosis replication by targeting the pathogen to the lysosome (xenophagy) as well as to prevent necrosis following days of infection in culture [5,[23][24][25][26][27][28][29]. However, to date there is no evidence that xenophagy functions in this capacity in vivo. Therefore, the mechanistic basis for how loss of ATG5 results in early and exaggerated recruitment of neutrophils during M. tuberculosis infection in vivo remains unknown.
In this manuscript, we dissect the role for ATG5 in regulating neutrophil recruitment and accumulation during M. tuberculosis infection in vivo. We find that ATG5 functions with other autophagy proteins specifically in CD11c + lung macrophages and DCs to limit the production of cytokines and chemokines that otherwise promote neutrophil influx to the lung early in M. tuberculosis infection. We demonstrate that loss of autophagy in macrophages and DCs does not affect M. tuberculosis burden in these cell types in vivo and instead changes the inflammatory response to the infection. In addition, ATG5 is required in lung macrophages and DCs to limit IL-17A production from CD4 + T cells. Together, our studies reveal new roles for ATG5 and other autophagy proteins in regulating inflammatory responses during infection, which with further dissection could provide insight into pathways that may be targeted to effectively promote protective immune responses during TB.

ATG5 is required in CD11c + lung macrophages and DCs to control neutrophil recruitment and accumulation early during M. tuberculosis infection in vivo
M. tuberculosis infection of Atg5 fl/fl -LysM-Cre mice results in the recruitment of a higher number of neutrophils in the lungs at 14 dpi as compared to Atg5 fl/fl controls, despite equivalent bacterial burdens at this time point [3]. There are also no differences in the abundance of nonneutrophil cell types in Atg5 fl/fl -LysM-Cre and Atg5 fl/fl mice at 14 dpi [3]. This indicates that specifically neutrophils are accumulating in M. tuberculosis-infected Atg5 fl/fl -LysM-Cre mice due to a defect in inflammatory responses to infection and not due to higher burden. To determine which LysM + cells required ATG5 to control the early influx of neutrophils into the lungs during M. tuberculosis infection, we compared bacterial burdens and neutrophil inflammation in Atg5 fl/fl -LysM-Cre, Atg5 fl/fl -Mrp8-Cre (deletion in neutrophils), Atg5 fl/fl -Cd11c-Cre (deletion in lung macrophages and DCs), and Atg5 fl/fl controls at 14 dpi. At 14 dpi, the Atg5 fl/fl -LysM-Cre and Atg5 fl/fl -Cd11c-Cre mice, but not Atg5 fl/fl -Mrp8-Cre mice, had higher levels of neutrophil inflammation in the lungs as compared to Atg5 fl/fl controls (Fig 1A and 1B). The degree of increased neutrophil frequency was similar in Atg5 fl/fl -LysM-Cre and Atg5 fl/fl -Cd11c-Cre mice, indicating that loss of Atg5 in CD11c + cells, but not neutrophils, leads to the early influx of neutrophils into the lungs during M. tuberculosis infection. At 14 dpi, none of the mouse strains harbored increased M. tuberculosis burden in their lungs (Fig 1C), indicating that the increase in neutrophil abundance in Atg5 fl/fl -Cd11c-Cre mice is not due to elevated bacterial burden and reflects a dysregulated inflammatory response to infection.
To determine if the higher levels of neutrophils in the lungs of M. tuberculosis-infected Atg5 fl/fl -CD11c-Cre mice was due to elevated neutrophil abundance in circulation prior to or during infection, we monitored neutrophil frequency in the blood in uninfected and 14 dpi Atg5 fl/fl and Atg5 fl/fl -CD11c-Cre mice. There was no significant difference in the frequency of neutrophils in the blood of uninfected or 14 dpi Atg5 fl/fl and Atg5 fl/fl -CD11c-Cre mice (S1A Fig), suggesting that the accumulation of neutrophils in the lungs of M. tuberculosis-infected Atg5 fl/fl -CD11c-Cre mice was due to specific recruitment of neutrophils to the site of infection or an inability to clear neutrophils from the lung. To begin to investigate this latter possibility, we monitored whether dead neutrophils were accumulating in the lungs of Atg5 fl/fl -CD11c-Cre mice during M. tuberculosis infection by analyzing neutrophil viability at 14 dpi by flow cytometry (S1B Fig). We did not observe a significant difference in the frequency of viable neutrophils at 14 dpi in Atg5 fl/fl and Atg5 fl/fl -CD11c-Cre mice, indicating that increased neutrophil inflammation in the lungs of Atg5 fl/fl -CD11c-Cre mice was not due to differences in neutrophil viability.
The higher levels of neutrophils in the lungs of Atg5 fl/fl -LysM-Cre and Atg5 fl/fl -Cd11c-Cre mice were sustained through 21 dpi (Fig 1D). However, only Atg5 fl/fl -LysM-Cre mice, and not Atg5 fl/fl -Cd11c-Cre mice, had higher bacterial burdens in the lungs at 21 dpi (Fig 1E), similar to as previously reported [3]. Loss of Atg5 in neutrophils results in increased susceptibility to M. tuberculosis infection in some, but not all, Atg5 fl/fl -Mrp8-Cre mice [3]. The susceptible Atg5 fl/fl -Mrp8-Cre mice accumulate higher neutrophil numbers and bacterial burdens in their lungs at 21 dpi (Fig 1D and 1E) [3]. Therefore, loss of Atg5 in neutrophils is likely contributing , Atg5 fl/fl -LysM-Cre (n = 6), Atg5 fl/fl -CD11c-Cre (n = 6), and Atg5 fl/fl -Mrp8-Cre (n = 7) mice. Neutrophil frequency is reported as a ratio relative to the average neutrophil frequency in Atg5 fl/fl control mice at 21 dpi within a given experiment. Susceptible and healthy Atg5 fl/fl -Mrp8-Cre mice are defined as done previously where susceptible Atg5 fl/fl -Mrp8-Cre mice have lost more than 5% of their pre-infection body weight by 20 dpi and healthy Atg5 fl/fl -Mrp8-Cre mice have lost less than 5% of their pre-infection body weight at 20 dpi [3]. (E) Lung burden from the right lung at 21 dpi in Mtb-GFP infected Atg5 fl/fl (n = 19), Atg5 fl/fl -LysM-Cre (n = 6), Atg5 fl/fl -CD11c-Cre (n = 6), and Atg5 fl/fl -Mrp8-Cre (n = 7) mice. (F) Kaplan-Meier curve of survival proportions during Mtb-GFP infection of Atg5 fl/fl (n = 9) and Atg5 fl/fl -CD11c-Cre (n = 9) mice. Statistical differences were determined by a logrank Mantel-Cox test (F) or one-way ANOVA and Šídák multiple comparison test (B-E). * P < 0.05, ** P < 0.01, *** P < 0.001, **** P < 0.0001. Differences that are not statistically significant are designated as ns. Pooled data from at least 2 separate experiments is graphed where each data point is from 1 biological replicate. The individual numerical values used to generate the graphed data in Fig 1, the statistical analyses performed to analyze these data, and the p values from these statistical tests are in S1 Data. dpi, days postinfection. https://doi.org/10.1371/journal.pbio.3002159.g001 to the higher burdens in the lungs of Atg5 fl/fl -LysM-Cre mice at 21 dpi. These data indicate that ATG5 has a role in CD11c + lung macrophages and DCs to regulate early recruitment of neutrophils, but not the control of M. tuberculosis replication during M. tuberculosis infection at 14 and 21 dpi.
To determine how the loss of Atg5 in CD11c + cells and the resulting early influx of neutrophils into the lungs affected host susceptibility, we monitored survival in M. tuberculosisinfected Atg5 fl/fl -Cd11c-Cre mice as compared to Atg5 fl/fl controls. Atg5 fl/fl -Cd11c-Cre mice succumbed to M. tuberculosis infection between 100 and 150 dpi, which was significantly earlier than Atg5 fl/fl controls (median survival time of 259 dpi) (Fig 1F), but not as early as Atg5 fl/fl -LysM-Cre mice (succumb 30 to 40 dpi [3]). These data demonstrate that ATG5 is required in CD11c + lung macrophages and DCs to control early neutrophil recruitment and promote survival following M. tuberculosis infection.

The role for ATG5 in lung macrophages and DCs in regulating neutrophil recruitment is dependent on other autophagy proteins
Deletion of multiple different autophagy genes in all LysM + innate immune cells can result in increased neutrophil recruitment to the lung during M. tuberculosis infection [23]. However, we previously showed that at least 1 role for ATG5 in LysM + innate immune cells in controlling M. tuberculosis infection is autophagy independent [3]. Therefore, it is not known if the role for ATG5 specifically in CD11c + lung macrophages and DCs is the same as described when broadly deleting Atg5 in all LysM + cells. To determine whether the regulation of neutrophil recruitment by ATG5 in CD11c + lung macrophages and DCs was dependent on other autophagy proteins or represented the autophagy-independent role for ATG5 during M. tuberculosis infection, we monitored neutrophil abundance in the lungs of mice lacking expression of another essential autophagy protein, BECLIN 1, in CD11c + cells (Becn1 fl/fl -Cd11c-Cre) at 14 dpi by flow cytometry. Similar to Atg5 fl/fl -Cd11c-Cre mice, Becn1 fl/fl -Cd11c-Cre mice also exhibited elevated neutrophil frequency in the lung at 14 dpi relative to Becn1 fl/fl control mice (Fig 2A), despite no difference in bacterial burden (Fig 2B). In addition, analysis of M. tuberculosis-infected Atg16l1 fl/fl -LysM-Cre and Becn1 fl/fl -LysM-Cre mice also revealed higher levels of neutrophils in the lungs at 14 dpi relative to controls, without higher bacterial burdens (Fig 2C and 2D).
In addition to their role in canonical autophagy, the proteins ATG5, BECLIN 1, and ATG16L1 are also required for the process of LC3-associated phagocytosis (LAP), where LC3 is recruited to the phagosome, resulting in LC3 + single membrane vesicles that traffic to the lysosome for degradation. LAP can dampen inflammatory responses through efferocytosis, pathogen removal, stimulating inhibitory immune-receptor signaling, and reducing auto-antigen levels [30][31][32][33]. In contrast to canonical autophagy, LAP uses RUBICON and UVRAG instead of ATG14 in the PI3K complex and does not depend on ULK1 [33,34]. To distinguish between whether ATG5, BECLIN 1, and ATG16L1 were functioning through autophagy or LAP to regulate neutrophil recruitment during M. tuberculosis infection, we infected mice lacking RUBICON expression, a protein specifically required for LAP. Rubicon -/mice had no difference in neutrophil accumulation or bacterial burdens as compared to WT controls following M. tuberculosis infection (Fig 2C and 2D), indicating that LAP is not required to control neutrophil inflammation during M. tuberculosis infection. Importantly, BECLIN 1 and ATG5 function at different steps of autophagy. Therefore, the requirement of both BECLIN 1 and ATG5 suggests that both the initiation and elongation steps of autophagy are required in CD11c + cells to control neutrophil recruitment early during M. tuberculosis infection. Targeting of pathogens to the lysosome via autophagy is termed xenophagy, and multiple studies The proportion of viable (Zombie -) alveolar macrophages, eosinophils, neutrophils, inflammatory monocytes, CD11b + DC, CD103 + DC, and non-alveolar macrophages in the lung in Atg5 fl/fl (n = 8) and Atg5 fl/fl -CD11c-Cre (n = 7) mice at 14 dpi. Statistical differences were determined by Student t test to compare the LysM-Cre or CD11c-Cre mice to their respective floxed control and Rubicon -/mice to WT C57BL/6J mice (A-G).* P < 0.05, ** P < 0.01, *** P < 0.001, **** P < 0.0001. Differences that are not statistically significant are designated as ns. Pooled data from at least 2 separate experiments is graphed where each data point is from 1 biological replicate. The individual numerical values used to generate the graphed data in Fig 2, the statistical analyses performed to analyze these data, and the p values from these statistical tests are in S2 Data. DC, dendritic cell; dpi, days postinfection; WT, wild-type. have reported roles for xenophagy in controlling M. tuberculosis replication in macrophages and DCs in cell culture in vitro [5,[23][24][25][26][27][28][29]. However, in addition to there being no differences in bacterial burden in the lungs at 14 dpi (Figs 1C and 2D) [3], there was no significant difference in the proportion of macrophages, eosinophils, neutrophils, inflammatory monocytes, or DCs that were infected with M. tuberculosis ( Fig 2E) and no difference in the M. tuberculosis burden in autophagy-deficient cells in the lungs of Atg5 fl/fl -LysM-Cre mice compared to Atg5 fl/fl controls at 14 dpi (Fig 2F). Loss of xenophagy in bone marrow-derived macrophages (BMDMs) in vitro has also been associated with increased necrosis during M. tuberculosis infection [23]. However, we did not detect a difference in the viability of macrophages, inflammatory monocytes, eosinophils, neutrophils, or DCs in the lungs of Atg5 fl/fl and Atg5 fl/fl -Cd11c-Cre mice at 14 dpi (Figs 2G and S1B), although we cannot rule out effects on the balance of different cell death pathways. These data support that the role for autophagy in CD11c + lung macrophages and DCs early during M. tuberculosis infection in vivo is independent of xenophagy regulating M. tuberculosis replication. In addition, these data show that the elevated neutrophil abundance in M. tuberculosis-infected Atg5 fl/fl -LysM-Cre mice is not due to differences in the cell viability or the cell types infected with M. tuberculosis but instead is driven by an imbalanced inflammatory response.

M. tuberculosis infection
We have previously shown that lungs of Atg5 fl/fl -LysM-Cre mice at 14 dpi contain higher levels of G-CSF and IL-17A than control mice [3], cytokines that promote neutrophil development and recruitment. At this time point, the primary CD11c + cell types that are infected by M. tuberculosis are the lung macrophages, encompassing alveolar and interstitial macrophages [35,36]. Therefore, we hypothesized that autophagy could be suppressing the production of these cytokines from M. tuberculosis-infected macrophages. We tested this hypothesis by culturing BMDMs from Atg5 fl/fl , Atg5 fl/fl -LysM-Cre, Atg16l1 fl/fl , Atg16l1 fl/fl -LysM-Cre, Becn1 fl/fl , and Becn1 fl/fl -LysM-Cre mice and infecting with M. tuberculosis in vitro before monitoring cytokine and chemokine production using a cytokine bead array (Bio-Rad) on the supernatants from infected cultures (Figs 3A-3F and S2). Of the 23 cytokines tested, we detected significantly higher levels of IL-1β, G-CSF, KC, TNF-α, and RANTES from the Atg5 fl/fl -LysM-Cre macrophage cultures compared to controls at 24 hpi ( Fig 3A-3E), despite no difference in bacterial burdens or BMDM viability at this time point (Figs 3F and S2A), indicating that the heightened proinflammatory response of Atg5 -/-BMDMs was not in response to increased antigen or macrophage cell death. The levels of these cytokines were only different following M. tuberculosis infection and not in mock-infected cultures, indicating that the increased proinflammatory responses were infection induced. The higher levels of G-CSF and KC, both proinflammatory signals associated with neutrophil inflammation [37-39], produced from Atg5-deficient macrophages in response to M. tuberculosis infection were dose dependent and confirmed by ELISA (Fig 3H and 3I). Similar to Atg5 fl/fl -LysM-Cre BMDMs, Atg16l1 fl/fl -LysM-Cre BMDMs also produced higher levels of IL-1β, G-CSF, and TNF-α following M. tuberculosis infection in vitro (Fig 3A, 3B and 3D). Becn1 fl/fl -LysM-Cre BMDMs also produced more IL-1β and G-CSF following M. tuberculosis infection in vitro compared to control cells (Figs 3A, 3B and S2) despite no difference in M. tuberculosis burden at this time point (Fig 3F). In addition, Becn1 fl/fl -LysM-Cre BMDMs produced higher levels of IL-6, MIP-1α, MIP-1β, and MCP-1 following M. tuberculosis infection (Figs 3G and S2). IL-6 in particular is associated with neutrophil recruitment [40][41][42][43] and also trended higher in M. tuberculosisinfected Atg5 fl/fl -LysM-Cre BMDMs (Fig 3G), so we further analyzed the levels of IL-6 produced by M. tuberculosis-infected Atg5 fl/fl -LysM-Cre BMDMs using an ELISA (Fig 3J). These analyses confirmed higher levels of IL-6 secretion from M. tuberculosis-infected Atg5 fl/fl -LysM-Cre BMDMs compared to controls. We were able to detect more cytokines and chemokines being differentially produced by M. tuberculosis-infected BMDMs than what was previously detected in the lungs of Atg5 fl/fl -LysM-Cre mice at 14 dpi [3]. This is likely due to the dilution of signals produced by macrophages in the context of the total lung homogenate, decreasing our sensitivity to detect macrophage-specific responses. In addition, it is possible that the inflammatory responses of BMDMs differ from lung macrophages and DCs during M. tuberculosis infection. Nonetheless, loss of expression of the autophagy proteins ATG5, ATG16L1, or BECLIN 1 in macrophages both in vivo and in vitro results in higher levels of  cytokines and chemokines that are associated with neutrophil recruitment and accumulation following M. tuberculosis infection relative to controls, indicating that canonical autophagy is required in macrophages to control proinflammatory responses during M. tuberculosis infection without affecting M. tuberculosis burden in macrophages.
To determine whether the increased neutrophil inflammation following M. tuberculosis infection in autophagy-deficient mice results from increased inflammasome activation, we crossed  and 4B), indicating that increased neutrophil recruitment in the absence of autophagy is occurring independent of CASPASE1/11. Caspase1/11 deletion also did not extend the survival of Atg5 fl/fl -LysM-Cre mice, indicating that increased inflammasome activation does not contribute to the early susceptibility of these mice (Fig 4C).
Autophagy has also been shown to suppress inflammatory responses via the process of mitophagy, where autophagy targets old and damaged mitochondria to the lysosome for degradation [56,57]. The build-up of damaged or dysfunctional mitochondria in the absence of autophagy results in loss of mitochondrial membrane potential and the release of reactive oxygen species (ROS), mitochondrial DNA, and ATP to the cytosol where it can lead to oxidative damage, inflammasome activation, and proinflammatory cytokine production [57][58][59][60][61]. Mitophagy requires the canonical autophagy proteins, including ATG5, ATG16L1, and BECLIN 1, as well as PARKIN and PTEN-induced kinase 1 (PINK1) [62]. PINK1 accumulates on damaged mitochondria and activates the mitochondrial E3 ubiquitin ligase, PARKIN, to ubiquitinylate damaged mitochondria [58,61]. Optineurin and NDP52 are the main mitophagy receptors that interact with the ubiquitinylated mitochondria and LC3, leading to autophagosome engulfment of the mitochondria [58,63]. To investigate whether loss of mitophagy could contribute to higher neutrophil accumulation in the lungs following M. tuberculosis infection, we measured neutrophil frequency in the lungs at 14 dpi by flow cytometry in Parkin -/and Pink1 -/mice relative to WT controls. There was no difference in neutrophil abundance or bacterial burdens in M. tuberculosis-infected Parkin -/or Pink1 -/mice relative to WT mice at 14 dpi (Fig 4D and 4E), indicating that mitophagy is not required to control neutrophil recruitment early during M. tuberculosis infection. To determine if mitophagy is required in macrophages to control proinflammatory cytokine and chemokine production during M. tuberculosis infection, we generated BMDMs from Parkin -/-, Pink1 -/-, and WT mice and infected the macrophages with M. tuberculosis for 24 h. We measured cytokine and chemokine levels from mock and M. tuberculosis-infected cultures using the cytokine bead array (Bio-Rad). Unlike in autophagy-deficient BMDMs, there were no differences in IL-6, IL-1β, G-CSF, KC, TNF-α, or RANTES production by M. tuberculosis-infected Parkin -/and Pink1 -/macrophages at 24 hpi compared to WT macrophages (S3 Fig), nor any differences in bacterial

ATG5 is required to suppress early T H 17 responses in the lungs during M. tuberculosis infection
The higher levels of IL-17A observed in the lungs of Atg5 fl/fl -LysM-Cre mice at 14 dpi with M. tuberculosis relative to controls was not reproduced by BMDMs infected with M. tuberculosis for 24 h (S2 Fig). Although there are many reasons to explain this, one possibility is that the macrophages were not the source of IL-17A in vivo. We investigated what cell type was expressing higher levels of IL-17A in the  Fig 5A) [3]. However, a greater frequency and number of the CD4 + T cells in the lungs of Il17a-GFP/ Atg5 fl/fl -LysM-Cre mice at 14 dpi were IL-17-GFP + compared to Il17a-GFP/Atg5 fl/fl mice (Fig 5B and 5C). These data indicate that CD4 + T cells contribute to the higher levels of IL-17A in the lungs of M. tuberculosis-infected Atg5 fl/fl -LysM-Cre mice and ATG5 is required in innate immune cells to negatively regulate T H 17 responses during M. tuberculosis infection.
IL-17A drives neutrophil influx by promoting the production of neutrophil chemokines MIP-1α and KC and through activation of endothelial cells [64][65][66]. Therefore, the increased T H 17 responses could be responsible for the early influx and accumulation of neutrophils in the lungs of M. tuberculosis-infected Atg5 fl/fl -LysM-Cre mice. To determine if the increased expression of IL-17A by T cells was responsible for the enhanced influx of neutrophils at 14 dpi in Atg5 fl/fl -LysM-Cre mice, we depleted CD4 + T cells by administering antibodies specific for CD4 from day −2 to 14 dpi (Fig 5D and 5E). At 14 dpi, we harvested the lungs for enumeration of M. tuberculosis burden and neutrophil abundance and found that there was no effect of CD4 + T cell depletion on either readout (Fig 5F and 5G). In addition, blocking IL-17A signaling by administering an anti-IL-17A antibody from day −1 to 14 dpi (Fig 5H) did not affect M. tuberculosis burden (Fig 5I) or neutrophil abundance (Fig 5J) in Atg5 fl/fl -LysM-Cre mice and Atg5 fl/fl mice at 14 dpi. Therefore, although ATG5 is required in innate immune cells to suppress IL-17A expression in T cells, this role does not contribute to regulating neutrophil accumulation early during M. tuberculosis infection.

Discussion
Atg5 fl/fl -LysM-Cre mice are extremely susceptible to M. tuberculosis infection, where neutrophils accumulate in the lungs of infected Atg5 fl/fl -LysM-Cre mice by 14 dpi and are sustained at high levels until the mice succumb to the infection between 30 and 40 dpi [3]. It was previously unknown how ATG5 imparted control of neutrophil recruitment to the lungs during M. tuberculosis infection. We have discovered that ATG5 is required in CD11c + lung macrophages and DCs to regulate proinflammatory cytokine production and neutrophil accumulation in the lungs early during M. tuberculosis infection. This role for ATG5 is shared with ATG16L1 and BECLIN 1, but not RUBICON, indicating it is autophagy dependent and does not involve LAP. We were able to reproduce the heightened proinflammatory responses in M. tuberculosis-infected autophagy-deficient BMDMs in vitro, suggesting that autophagy specifically suppresses inflammatory responses from macrophages during M. tuberculosis infection, although this does not rule out a similar role in DCs in vivo. Alveolar macrophages are among the first cells to encounter M. tuberculosis in the airways and orchestrate the initial response to infection, recruiting other innate immune cells to the lung [35,36]. We postulate that similar to the BMDMs, autophagy-deficient alveolar macrophages overproduce proinflammatory signals during M. tuberculosis infection, leading to increased neutrophil recruitment early during infection.
The increased levels of cytokines and chemokines produced by autophagy-deficient macrophages was dependent on M. tuberculosis infection, demonstrating that pathogen detection was required. However, the heightened proinflammatory responses in autophagy-deficient macrophages occurred in the absence of differences in bacterial burden, indicating that the enhanced inflammatory response is not due to increased antigen. Although multiple studies have shown effects of loss of autophagy on M. tuberculosis burden in macrophages and DCs in vitro in cell culture [5,[23][24][25][26][27][28][29], there was no difference in M. tuberculosis burden in these cell types in vivo when we detected higher levels of proinflammatory cytokine production and neutrophil influx. Therefore, the role for autophagy in regulating proinflammatory cytokine production from CD11c + lung macrophages and We ruled out the involvement of CASPASE1/11-dependent inflammasome activity and mitophagy in the autophagy-dependent regulation of neutrophil accumulation during M. tuberculosis infection, leaving open the question of how autophagy regulates macrophage proinflammatory responses during infection. Our understanding of the number of cellular pathways regulated by autophagy continues to grow, and there are many possible mechanisms by which autophagy could regulate inflammatory responses in macrophages [18]. One possibility is that targeting of autophagy components to intracellular M. tuberculosis prevents sensing of M. tuberculosis by cytosolic pattern recognition receptors and subsequent proinflammatory cytokine and chemokine production, without impacting M. tuberculosis replication. Autophagy can also regulate levels of IL-1β [44,62,67], IL-6 [32, 68,69], and TNF-α [68,69] downstream of pathogen sensing, in part by negatively regulating NFκB activation [70][71][72]. Autophagy can both directly promote autophagic cell death and apoptosis as well as negatively regulate RIPK3-dependent necrotic cell death, all which can impact inflammatory responses [23, [73][74][75][76][77]. Although we did not detect overall differences in the viability of autophagy-deficient cells at the time points we observed effects on cytokine and chemokine production in vivo or in vitro, the mechanism of cell death in Atg5 fl/fl and Atg5 -/cells during M. tuberculosis infection may vary and could impact the inflammatory responses [73,74,78]. Autophagy may also be required in macrophages for the efficient removal of dead or infected neutrophils from the lung and resolution of inflammation. The best-studied process for removal of dead cells by phagocytes is efferocytosis [79]. LAP has specifically been shown to contribute to efferocytosis [80], and although we rule out a role for LAP in regulating neutrophil accumulation during M. tuberculosis infection, in remains possible that processing of dead cells may be attenuated in the absence of autophagy. Another possible mechanism for how autophagy regulates inflammatory responses from macrophages during M. tuberculosis infection involves the process of ER-phagy. ER-phagy is induced under conditions of ER stress, accumulation of unfolded proteins, and during infection [81,82]. ER-phagy restrains ER stress responses by targeting excess or damaged endoplasmic reticulum to autophagosomes for degradation [81], but in the absence of autophagy, ER stress activates NFκB-dependent transcription of inflammatory cytokines, such as IL-1β, IL-6, IL-18, and TNF-α [82].
We also discovered that ATG5 was required in innate immune cells to suppress early T H 17 responses during M. tuberculosis infection. At this point, we do not know if the increased abundance of IL-17 + CD4 + T cells in M. tuberculosis-infected Atg5 fl/fl -LysM-Cre mice is due to an autophagy-dependent or independent role for ATG5, but there are multiple other reports of autophagy regulating T H 17 responses. The effect of loss of Atg5 in innate immune cells on IL-17A expression from T cells may be explained by the requirement for autophagy in DCs to negatively regulate surface expression of disintegrin and metalloproteinase domain-containing protein 10 (ADAM10). ADAM10 cleaves its substrate ICOSL and lower levels of ICOSL leads to decreased ICOSL-ICOS interactions between DCs and T cells, resulting in less CD25 hi CD4 + T regulatory cells and more IL-17 + CD4 + T cells [83]. In addition, autophagy negatively regulates T H 17 differentiation by reducing IL-23 and IL-1β levels, which promote T H 17 differentiation and IL-17A secretion [4,[84][85][86]. Nonetheless, depleting CD4 + T cells or blocking IL-17A did not rescue the increased neutrophil accumulation in the lungs of M. tuberculosisinfected Atg5 fl/fl -LysM-Cre mice at 14 dpi, suggesting that the hyper-inflammatory responses from infected autophagy-deficient macrophages and DCs is sufficient to recruit excessive neutrophils early during infection. However, it is possible that the heightened T H 17 responses in M. tuberculosis-infected Atg5 fl/fl -LysM-Cre mice have a longer-term impact on the increased susceptibility of these mice and could contribute to the susceptibility of Atg5 fl/fl -Cd11c-Cre mice during the chronic phase of infection.
Our data support a role for autophagy in restricting neutrophil accumulation in the lung but does not rule out additional contributions from autophagy-independent roles for ATG5, BECLIN 1, and ATG16L1. In particular, loss of Atg5, Becn1, or Atg16l1 in LysM + cells led to different degrees of elevated neutrophil frequency in the lung at 14 dpi relative to control mice, where loss of Atg5 results in the greatest increase in neutrophil abundance. This could in part be due to differences in efficiency of gene deletion in the different lines [23]. In addition, we have previously identified an autophagy-independent role for ATG5 in neutrophils, where loss of ATG5 in neutrophils can result in early lethality during M. tuberculosis infection [3]. Therefore, loss of Atg5 in LysM + cells, which includes neutrophils, may result in higher neutrophil frequency in the lung at 14 dpi than loss of Becn1 or Atg16l1 because of this autophagy-independent role for ATG5 in neutrophils. We hypothesize that the combination of the newly discovered role for autophagy in CD11c + lung macrophages and DCs to regulate inflammatory responses and an autophagy-independent role for ATG5 in neutrophils collectively allow for control of M. tuberculosis infection, where loss of both functions results in the extreme susceptibility of Atg5 fl/fl -LysM-Cre mice to M. tuberculosis infection. Higher abundances of neutrophils have been associated with poor disease prognosis and treatment outcomes in TB patients [14][15][16]. Therefore, our new findings and future dissection of the ATG5-dependent mechanisms of regulating neutrophil recruitment to the lungs during M. tuberculosis infection will provide critical insight into how to promote protective immune responses to TB.

Infection of mice with M. tuberculosis and measurement of bacterial burden in the lungs
M. tuberculosis Erdman expressing GFP (Mtb-GFP [10,92]) was used in all experiments except experiments with the Il-17a-GFP/Atg5 fl/fl -LysM-Cre reporter mice when WT Erdman was used. M. tuberculosis 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 expressing M. tuberculosis were grown in the presence of kanamycin (20 μg/mL) to ensure plasmid retention. M. tuberculosis 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 100 CFUs of aerosolized M. tuberculosis per lung using an Inhalation Exposure System (Glas-Col). Within 2 h of each infection, lungs were harvested from at least 2 control mice, homogenized, and plated on 7H11 agar to determine the input CFU dose. At 14 dpi, M. tuberculosis 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% CO 2 .

Flow cytometry from blood and infected lungs
Blood was collected by cardiac puncture into K2EDTA anticoagulant tubes (BD, 365974). Red blood cell lysis was performed by adding 2 mL of ACK lysis buffer (Gibco, A10492-01) per 100 μL of EDTA-treated blood for 5 min at room temperature. Cells were pelleted at 500×g for 5 min and then resuspended in 50 μL of PBS + 2% HI-FBS + 2 mM EDTA in the presence of Fc receptor blocking antibody (BioLegend, 101302). Cells were labeled with antibodies as described below. Lungs were perfused with sterile PBS and digested for 1 h with 625 μg/mL collagenase D (Roche 11088875103) and 75 U/mL DNase I (Sigma D4527). Cells were quenched with PBS + 2% heat-inactivated (HI)-FBS, + 2 mM EDTA and passed through a 70 μm filter. Cells were stained with Zombie-violet or Zombie-NIR in PBS at 1:2,000 dilution for 5 min at room temperature prior to resuspending in PBS + 2% HI-FBS + 2 mM EDTA in the presence of Fc receptor blocking antibody (BioLegend, 101302) for blocking. Cells were labeled with antibodies at a 1:200 dilution using the following mouse markers: CD11b_BV605

Culturing and infection of bone marrow-derived macrophages (BMDMs)
BMDMs were generated by flushing femurs and tibias of mice and culturing the cells in DMEM, 20% HI-FBS, 10% supernatant from 3T3 cells overexpressing M-CSF + 1% MEM non-essential amino acids (Cellgro 25-025-CI), 2 mM L-glutamine, 100 U/mL penicillin, and 100 μg/mL streptomycin (Sigma P4333) at 37˚C in 5% CO 2 in non-TC treated plates. After 6 days, non-adherent cells were removed and 1 × 10 6 adherent macrophages were seeded per well in 6-well non-TC treated plates in DMEM, 10% HI-FBS, 1% MEM non-essential amino acids, and 2 mM L-glutamine. BMDMs were rested overnight at 37˚C in 5% CO 2 . M. tuberculosis was grown to an OD of 0.6 to 0.8, washed with PBS twice, sonicated to disperse clumps, centrifuged at 55×g for 10 min to remove clumps and resuspended in antibiotic-free BMDM media. Macrophages were infected at an MOI of 10 by centrifuging the cells at 200×g for 10 min. BMDMs were washed with PBS twice to remove free M. tuberculosis and fresh BMDM media was added to each well. Cells were incubated at 37˚C and 5% CO 2 for 24 h. To determine CFU counts, the cells were lysed with 0.05% triton X-100, serially diluted, and plated onto 7H11 agar and incubated for 21 days when bacterial colonies were counted. At 24 hours postinfection (hpi), supernatants were stored at −80˚C for cytokine analysis. To assess BMDM viability, cells were harvested at 24 hpi by gently scrapping, washed with PBS twice, stained with Zombie-violet in PBS at 1:2,000 dilution for 5 min at room temperature, washed once with PBS, and fixed in 4% PFA for 20 min at room temperature before performing flow cytometry. Flow cytometry was performed on an LSR-Fortessa (BD Bioscience) and analyzed using FlowJo software (Tree Star). The gating strategy to identify viable BMDMs at 24 hpi is depicted in S1D Fig

IL-17A blocking and T cell depletion
To neutralize IL-17A, 100 μg of InVivo monoclonal anti-IL-17A (Bio X Cell, BE0173) neutralizing antibody was administered to Atg5 fl/fl and Atg5 fl/fl -LysM-Cre mice by intraperitoneal (i.p.) injection every other day starting at 1 day prior to infection, with the final dose delivered at 13 dpi, similar to published protocols [93]. Control mice received 100 μg of IgG from mouse serum (Sigma, I5381) by i.p. injection every other day starting 1 day prior to infection and finishing on 13 dpi. To deplete CD4 + T cells from mice, 250 μg of anti-mouse CD4 (Leinco Technologies, C1333) was administered by i.p. injection at 2 days prior to infection, 5 dpi, and 12 dpi. Control mice received 250 μg of IgG from rat serum (Sigma, 18015) i.p. on 2 days prior to infection, 5 dpi, and 12 dpi.

Data and statistics
All experiments were performed at least twice. When shown, multiple samples represent biological (not technical) replicates of mice randomly sorted into each experimental group. No blinding was performed during animal experiments. Animals were only excluded when pathology unrelated to M. tuberculosis infection was present (i.e., bad teeth leading to weight loss). Determination of statistical differences was performed with Prism (GraphPad Software) using log-rank Mantel-Cox test (survival), unpaired two-tailed t test (to compare 2 groups with similar variances), or one-way ANOVA with Šídák multiple comparison test (to compare more than 2 groups). When used, center values and error bars represent the mean +/-SEM. In all figures, all significant differences are indicated by asterisks: * P < 0.05, ** P < 0.01, *** P < 0.001, **** P < 0.0001. Nonsignificant comparisons of particular interest are noted as ns. and Atg5 fl/fl -CD11c-Cre (n = 4) mice at 14 dpi and uninfected (naïve) mice. Statistical differences were determined by Student t test comparing the genotypes within a particular cell type or treatment group. * P < 0.05, ** P < 0.01, *** P < 0.001, **** P < 0.0001. Differences that are not statistically significant are designated as ns. Pooled data from at least 2 separate experiments are graphed where each data point is from 1 biological replicate. The individual numerical values used to generate the graphed data in S1 Fig