Research Article

mTOR regulates metabolic adaptation of APCs in the lung and controls the outcome of allergic inflammation

See allHide authors and affiliations

Science  08 Sep 2017:
Vol. 357, Issue 6355, pp. 1014-1021
DOI: 10.1126/science.aaj2155

Metabolic programming of tissue APCs

Antigen-presenting cells (APCs) are scattered throughout the body in lymphoid organs and at the portals of pathogen entry, where they act as sentinels of the immune system. Sinclair et al. demonstrate that APCs at different sites have distinctive metabolic signatures and that the development and function of these cells are determined not only by their transcriptional program, but also by their metabolic state (see the Perspective by Wiesner and Klein). The authors identified a central role for mTOR in mediating the metabolic adaptation of such tissue-resident APCs by influencing the immunological character of allergic inflammation. Thus, tissues endow resident APCs with distinctive metabolic characteristics that control APC development and function.

Science, this issue p. 1014; see also p. 973

Abstract

Antigen-presenting cells (APCs) occupy diverse anatomical tissues, but their tissue-restricted homeostasis remains poorly understood. Here, working with mouse models of inflammation, we found that mechanistic target of rapamycin (mTOR)–dependent metabolic adaptation was required at discrete locations. mTOR was dispensable for dendritic cell (DC) homeostasis in secondary lymphoid tissues but necessary to regulate cellular metabolism and accumulation of CD103+ DCs and alveolar macrophages in lung. Moreover, while numbers of mTOR-deficient lung CD11b+ DCs were not changed, they were metabolically reprogrammed to skew allergic inflammation from eosinophilic T helper cell 2 (TH2) to neutrophilic TH17 polarity. The mechanism for this change was independent of translational control but dependent on inflammatory DCs, which produced interleukin-23 and increased fatty acid oxidation. mTOR therefore mediates metabolic adaptation of APCs in distinct tissues, influencing the immunological character of allergic inflammation.

A fundamental property of the immune system is its ability to sense pathogens through pattern recognition receptors. Emerging evidence demonstrates that immune cells also sense environmental perturbations that induce cellular stress and metabolic changes (1, 2). However, the mechanisms underlying such environmental sensing are poorly understood. The amino acid sensor GCN2 plays a role in programming dendritic cells (DCs) to respond to viral vaccination (3, 4) and controls intestinal inflammation by suppressing inflammasome activation in the gut (5). In addition to GCN2, the mechanistic target of rapamycin (mTOR), a central orchestrator of cellular metabolism, regulates DC function (6).

mTOR functions in two supramolecular complexes termed mTORC1 and mTORC2, engaged to differing degrees by diverse stimuli, including Toll-like receptor signaling, growth factors, and nutrient availability. mTOR signaling affects diverse downstream cellular processes, including cell cycle progression, cell growth, differentiation, survival, and metabolism (7). Although mTOR is required for effector T cell expansion (8) and germinal center B cell responses (9), it restrains T memory generation (10). Within DCs, mTOR promotes type I interferon production by plasmacytoid DCs (11, 12) but limits proinflammatory cytokine production by classical DCs (13). Moreover, mTOR promotes steady-state DC accumulation downstream of homeostatic cytokine signaling (1416). The precise functions of mTOR have been proposed to be DC subset dependent (17), but the consequence of mTOR signaling in DCs and the effect on immune responses in vivo are poorly understood.

Results

mTOR promotes lung APC homeostasis in vivo

We bred mice bearing floxed Mtor with a CD11c-Cre deleter strain (mTORΔAPC), resulting in specific genetic ablation, reduced protein abundance, and loss of mTOR pathway activity in CD11c+ cells (fig. S1, A to D). mTOR inhibition augments DC proinflammatory cytokine production (13, 17), which we observed in mTORΔAPC bone marrow (BM)–derived DCs and primary splenic DCs (fig. S1E), confirming functional deletion of mTOR.

DCs can be subdivided into CD8α+/CD103+ and CD11b+ subsets. Globally, DCs require Flt3L and granulocyte macrophage colony–stimulating factor (GM-CSF) signals for homeostatic maintenance (18, 19), both of which activate mTOR (14). In addition, antigen-presenting cells (APCs) occupy distinct microenvironments where mTOR may exhibit site-specific functions (15, 20). mTORΔAPC ablation did not exert global effects on APCs but instead revealed defects in APC accumulation at specific anatomical sites. There was negligible requirement for cell-intrinsic mTOR in most tissues examined; however, splenic CD11b+ DCs were modestly reduced and there was a skewed ratio of CD103+ to CD11b+ DCs in thymus (fig. S2). We also confirmed a substantial reduction in Langerhans cell frequency (15) (fig. S3). However, the most striking effect of mTORΔAPC deletion was observed in the lung. Here, CD103+ DC and CD11c+ alveolar macrophage (AM) populations were significantly reduced (Fig. 1A). Mixed BM chimeras showed that homeostatic defects were cell intrinsic and additionally revealed a minor impairment in CD11b+ lung DC accumulation (Fig. 1B).

Fig. 1 Critical requirement for mTOR signaling to maintain lung innate immune homeostasis.

Lungs from naïve WT (black) and mTORΔAPC (red) mice were analyzed by flow cytometry. (A) Bar charts show mean frequencies ± SEM of indicated lung APC populations; contour plots show representative flow cytometric data. Red arrows indicate gating hierarchies. (B) Mixed chimeras were established by reconstituting B6 CD45.1+/CD45.2+ hosts with donor mTORΔAPC (CD45.2) and B6 CD45.1 BM at a ~3:1 ratio. Bar chart shows the percentage contribution to chimerism of mTORΔAPC-derived donor cells after exclusion of residual host cells. Gray bars indicate control (CD11c populations); red bars indicate CD11c+ APCs. (C) Bar charts show normalized read count for indicated genes in WT (black) and mTORΔAPC (red) within the CD11c+CD11b+ lung DC population. (D) Contour plot shows expression of Sirpα and HSA. Pie chart shows average frequencies of indicated populations from one representative experiment (n = 4). (E) Bar chart shows the chimeric contribution of mTORΔAPC-derived donor cells. Gray bars denote control CD11c populations; red bars indicate CD11c+ APCs. Data represent three or more independent experiments.

mTOR may simply regulate numerical accumulation of APCs or, alternatively, may be central to controlling lung APC programming. Transcriptomes derived from mTORΔAPC lung APCs versus wild-type (WT) controls clustered distinctly by principal components analysis (PCA), whereas mTORΔAPC splenic populations clustered closely with their WT counterparts (fig. S4). Previous studies have identified core sets of lineage-specifying genes, enriched in either CD11b+ lung APCs or AMs, relative to other APC subsets (21, 22). mTORΔAPC AMs exhibited a broad loss of AM molecular identity (fig. S5A). Rather than disrupting the CD11b+ APC cellular identity, mTOR ablation skewed the balance of CD11c+CD11b+ lung APCs toward a macrophage/monocytic composition (fig. S5B) (22), typified by reduced heat-stable antigen (HSA), Flt3, and a trend of lower Irf4 expression, with elevated MerTK and CD64 transcription (Fig. 1C). Using an alternative gating scheme (22), we found that CD11c+MHC-II+Sirpα+HSA+ DCs were reduced in mTORΔAPC mice (Fig. 1D) due to cell-intrinsic defects in accumulation (Fig. 1E). Thus, mTOR plays a central, cell-intrinsic role in APC homeostatic accumulation and subset composition in the lung.

Steady-state DC homeostasis depends on cell development, proliferation, migration, and death rates. Ablation or inhibition of mTOR increased death rates of lung APCs (Fig. 2A and fig. S6, A and B), whereas precursor frequency, proliferation rates, and migration were unaffected (fig. S6, C to G). There was no evidence that GM-CSF, M-CSF, or Flt3L expression was perturbed in mTORΔAPC mice, excluding a role for other cell-extrinsic homeostatic factors (fig. S6H). mTOR associates with two molecular complexes (mTORC1/2). Whereas mice lacking the critical mTORC1 protein Raptor phenocopied homeostatic defects seen in mTORΔAPC lung, APCs were unaffected in mTORC2-deficient RictorΔAPC mice (fig. S7A). mTORC1 signaling targets the translational regulator S6K1/2 and deactivates the translational repressor 4E-BP1/2 (fig. S7, B and C) (7). Curiously, S6k1−/−S6k2−/− mice exhibited normal lung APC homeostasis, and genetic deletion of 4ebp1/2 did not rescue the survival of mTORΔAPC DCs (fig. S7, D and E). Thus, mTORC1 controls APC cell survival through a translation-independent mechanism.

Fig. 2 mTOR regulates lung APC survival via a metabolic mechanism.

Lung from WT (black) or mTORΔAPC (red) mice was examined by flow cytometry and by cellular metabolomic profiling. (A) Histograms show representative staining of Live/Dead Aqua on indicated ex vivo populations from WT (gray) and mTORΔAPC (red) lung. (B) PCA plot shows distinct clustering of lung and splenic APC metabolomes from WT mice. (C) Heat maps show differentially abundant metabolite features between WT and mTORΔAPC subsets. The metabolite numbers include redundant mass spectral peaks. (D) Dot graph shows metabolic pathways distinctly perturbed by mTOR deficiency, with amino acid metabolism pathways highlighted in red. Dot size relates to the number of differentially expressed metabolites. (E) Network analysis reveals a higher representation of tentative amino acid metabolites within the mTORΔAPC CD103+ DC subset. Each link is from a known metabolic reaction. The network and tentative metabolite annotation was produced by the Mummichog software. Data represent three or more independent experiments or two to three biological replicates.

APCs in the lung display a distinctive mTOR-dependent metabolic signature

The pronounced phenotype of APCs in mTORΔAPC lung relative to other tissues was striking, particularly considering that development of CD103+ DCs in both lung and other tissues is regulated by the common transcription factor Batf3 (23). We hypothesized that the lung microenvironment may imprint a distinctive metabolic phenotype with heightened sensitivity to mTOR ablation. To investigate this, comprehensive metabolic profiling was performed on lung and splenic APCs from WT mice. Lung and spleen APCs were metabolically distinct (Fig. 2B), exhibiting differences in amino acid and nucleic acid metabolic intermediates (fig. S8, A and B), which were also reflected within observed transcriptional signatures (fig. S8, C and D). We next examined mTORΔAPC metabolomes, and the degree of metabolic perturbation in mTOR-deficient APC subsets correlated with the extent of their homeostatic disruption (Fig. 2C). Despite the skewing in the phenotypic composition of the lung CD11b+ DCs (Fig. 1D), there was a minor perturbation in the metabolome (Fig. 2C). Pathway analysis of differentially abundant metabolites revealed that several amino acid metabolic pathways were disrupted in mTORΔAPC APCs (Fig. 2, D and E). mTORΔAPC AMs also exhibited alterations in lipid metabolite expression and corresponding changes in expression of the sterol regulatory binding protein (Srebp1/2) target genes (24) (fig. S9). Srebp1 orchestrates lipid anabolism and is a downstream target of mTOR, which directly links mTOR deficiency and metabolic perturbations in lung APCs (25, 26). To test this experimentally, we administered the Srebp1/2 inhibitor fatostatin, which substantially decreased AM frequency, a phenotype that was recapitulated by rapamycin treatment (fig. S10A). However, CD103+ DCs remained unaltered by either inhibitor, suggesting that inhibitor administration was insufficient to entirely phenocopy mTORΔAPC mice, or that the Srebp pathways are not critical for CD103 DC survival. Finally, we investigated the role of another mTOR target, the glycolytic regulator Hif1α, observing that lung APC homeostasis was independent of Hif1α (fig. S10B). Thus, lung APCs have distinct mTOR-dependent metabolic signatures, and there appears to be a direct link between mTOR and Srebp-dependent lipid metabolism in AMs.

Ablation of mTOR in APCs reprograms allergic inflammation in the lung

We next investigated whether lung immunity was functionally affected in mTORΔAPC mice. Mice were intranasally challenged with house dust mite (HDM) allergen, which is an established murine model for asthma (27, 28) (Fig. 3A). Despite reduced frequencies of select lung APC subsets, humoral responses in mTORΔAPC mice were unexpectedly increased (Fig. 3B and fig. S11A). Surprisingly, we observed a striking reduction in T helper cell 2 (TH2) cells producing interleukin-4+ (IL-4+) and IL-5+/IL-13+ and enhanced frequencies of TH17 cells in the lung and bronchoalveolar lavage (BAL) (Fig. 3C and fig. S11B).

Fig. 3 mTOR ablation in APCs results in neutrophilic lung inflammation.

House dust mite (HDM) extract was administered to WT (black) or mTORΔAPC (red) mice. (A) Diagram shows the allergen administration and analysis schedule. i.n., intranasal. (B) Scatter plots show serum antibody titers in indicated mouse strains. (C) Bar charts show the frequency of CD4+IL-4+ TH2 cells or CD4+IL-17A+ TH17 cells in indicated organs. Contour plots show representative flow cytometric data. (D) Bar charts show the frequency of CD11b+Siglec-F+ eosinophils or CD11b+Ly-6G+ neutrophils. Pseudocolor plots show representative flow cytometric data. Data represent three or more independent experiments.

TH2-polarized allergy is typified by robust recruitment of eosinophils, whereas IL-17 can enhance neutrophils. Neutrophilia occurs in ~50% of human asthma cases, correlating with steroid resistance and worsening prognosis (29, 30). Remarkably, we observed pronounced neutrophilia in mTORΔAPC lung and BAL (Fig. 3D). Skewed TH2 to TH17 modality was further observed in an alternative papain-induced model (fig. S12), whereas an anatomically discrete TH17-dependent experimental autoimmune encephalomyelitis model was not significantly affected by mTOR deficiency in APCs (fig. S13). TH17 allergic phenotypes were driven by mTORC1 ablation (fig. S14A) but were independent of Hif1α (fig. S14B) or Srebp1/2 (fig. S14C). mTORC1 thus regulates innate programming of T helper responses in lung, controlling the balance between eosinophilic and neutrophilic mediators of allergic inflammation.

mTOR-deficient CD103+ DCs and AMs are dispensable for neutrophilic allergy

We next determined the APC subset responsible for promoting TH17/neutrophilic allergy in mTORΔAPC mice. CD103+ DCs have important antiviral functions due to a superior ability to cross-present antigens (31). Consistent with this, mTORΔAPC mice exhibited impaired CD8+ T cell responses to live-attenuated influenza vaccine challenge (Fig. 4A). If CD103+ DCs restrain TH17/neutrophilic allergy, then we should observe a similar phenotype in Batf3−/− mice, which lack a transcription factor required for CD103+ DC development (Fig. 4B). However, Batf3−/− mice did not exhibit differential severity or modality of inflammation (Fig. 4, C to E), suggesting that impaired development of CD103+ DC was not necessary to promote neutrophilic inflammation.

Fig. 4 mTOR-dependent CD103+ DCs are dispensable for neutrophilic lung inflammation but required for antiviral immunity.

(A) Flumist Quadrivalent vaccine (25 μL) was administered intranasally to WT (black) or mTORΔAPC mice (red). Vaccine response was examined 7 days after infection. Bar charts show the frequency of CD8+IFNγ+ cells in indicated organs. Contour plots show representative flow cytometric data. (B to E) HDM was chronically administered to WT (black) and Batf3−/− (gray) mice. (B) Contour plots show the absence of CD103+ and CD8α+ DCs in lung and spleen of Batf3−/− mice. (C) Serum enzyme-linked immunosorbent assays (ELISAs) showing total IgE or HDM-specific IgG1 antibody titers. (D) Bar charts show TH2 (IL-4+) and TH17 (IL-17A+) frequencies in lung and BAL. (E) Bar charts show frequencies of eosinophils and neutrophils in lung and BAL. (A) Data represent three or more independent experiments. [(B) to (E)] n = 5 mice per group.

AMs reside in the lung alveolar lumen and play a primary role in phagocytizing particulates and turning over lung surfactant. Defects in AM result in accumulated lung surfactant protein, termed pulmonary alveolar proteinosis (PAP) (32). Histological analysis of naïve mTORΔAPC lung showed airspace filling by eosinophilic material (Fig. 5A), elevated mucus (Fig. 5B), accumulation of lung surfactant protein (Fig. 5C), and increased debris and “foamy” macrophages in BAL (Fig. 5D). Increased proteinaceous matter was also evident in lung preparations by flow cytometry (Fig. 5E). These data represent hallmark features of PAP disease (33), representing a first link between mTOR signaling and PAP. We next determined whether functional loss of AMs also contributes to neutrophilic lung inflammation. Using a previously described protocol (34), we observed ~75% depletion of AMs 3 days after a single intratracheal (i.t.) dose of clodronate liposomes but not control liposomes (Fig. 5F). After HDM challenge, humoral immunity was enhanced when AMs were depleted, in concordance with the phenotype observed in mTORΔAPC mice (Fig. 5G). However, TH17 polarization and neutrophilia were not present under these conditions (Fig. 5, H and I), consistent with observations that AMs cannot present antigen (28). Thus, absence of AM likely contributes to an enhanced magnitude of humoral immunity in mTORΔAPC mice but cannot account for TH17/neutrophilic allergy.

Fig. 5 mTOR ablation in AMs results in PAP and enhances the antibody magnitude but not modality of allergy.

(A to C) Images show histopathological sections of WT or mTORΔAPC lung stained with (A) hematoxylin and eosin, (B) PAS, (C) immunohistochemical antibodies against surfactant protein D. (D) BAL cellular fractions were prepared by cytospin and analyzed after Diff-Quik staining. Slides are shown at [(A) to (C)] x200 magnification, where scale bars represent 100 μm and (D) x400 magnification, where scale bars represent 50 μm. (E) Pseudocolor plots show the frequency of Live/Dead Aqua+ FSClo (dead cell/debris) events as determined by flow cytometry. (F to I) HDM was administered to indicated groups for 17 days. Mice were treated with indicated clodronate or control liposome on days –3, 3, and 10. (F) AM frequency on day 0 before HDM administration. (G) Serum ELISAs showing total IgE or HDM-specific IgG1 antibody titers. (H) Bar charts show the frequency of IL-4+ and IL-17+ T cells in indicated groups and organs. (I) Bar charts show the frequency of eosinophils and neutrophils in indicated groups and organs; pseudocolor plots show representative data. Data represent [(A) to (D)] three or more slides from two or more biological replicates, [(E) to (I)] two or more representative experiments.

mTOR deficiency reprograms inflammatory CD11b+ DCs to promote neutrophilic allergy

CD11b+ DCs directly promote TH2 allergy after HDM insult (28, 35) but can also prime TH17 immunity when activated by alternative stimuli (22, 36). To address whether mTORΔAPC CD11b+ DCs promoted neutrophilic allergy in a dominant manner, we first generated mixed chimeras where donor WT and mTORΔAPC bone marrow was injected at a 1:10 ratio to allow repopulation of CD103+ DC and AM populations by WT cells, whereas ~80% of CD11c+CD11b+ DCs remained mTOR-deficient (fig. S15, A and B). HDM promoted elevated humoral responses and pronounced neutrophilia in chimeras (fig. S15, C and D). Moreover, primed CD11b+ DCs from mTORΔAPC mice directly promoted TH17/neutrophilic allergy in WT recipients (fig. S16). Thus, neutrophilic inflammation was dominantly conferred by mTORΔAPC CD11b+ DCs.

It was recently shown that mice deficient in Kruppel-like factor 4 (Klf4) have reduced frequencies of the Sirpα+HSA+ subset in lung CD11b+ DCs and display impaired TH2 responses to HDM (37). We thus investigated how mTORΔAPC CD11b+ lung DCs differentially responded to allergy. HDM was instilled i.t., and mice were examined after 3 days. mTORΔAPC CD11b+ migratory DCs accumulated with enhanced frequency in mediastinal lymph node (MedLN) (fig. S17A), and cellular inflammation was enhanced in the lung (fig. S17B). During allergic airway inflammation, a distinct CD11b+ Csf-1–dependent inflammatory DC population develops from monocytes (28, 38). Inflammatory DCs could be identified in allergic lung based on their surface CD11c+MHC-II+CD11b+ phenotype and Csf1 dependence (fig. S18, A and B) (28) and were elevated in mTORΔAPC mice (Fig. 6A). Lung inflammatory DCs were confirmed to be distinct from lung-resident interstitial macrophages, AMs, and classical DCs. They exhibited lower expression of F4/80, higher expression of MHC-II and CD11c compared to interstitial macrophages (fig. S18C), lower Siglec-F expression than AMs (fig. S18D), and high expression of CD64 and Ly6C relative to resident CD11b+ DCs (fig. S18E). Compared with WT, mTORΔAPC inflammatory DCs had higher expression of lineage-associated markers CD64 and Ly6C and elevated CD80 and CD86 costimulatory molecule expression (fig. S18, E and F), consistent with a more progressive differentiation/activation status. To directly test the requirement for inflammatory DCs in TH17/neutrophilic inflammation, we ablated this population by administering αCsf1r neutralizing antibodies, which reverted mTORΔAPC allergy to TH2/eosinophilic modality (Fig. 6B and fig. S19).

Fig. 6 CD11b+-intrinsic mTOR signaling restrains TH17/neutrophilic polarized HDM allergic responses.

HDM (75 μg) was administered i.t. to WT and mTORΔAPC mice on days 0 and 2. Lungs were analyzed on day 3. (A) Bar chart shows total inflammatory DC cells in lung. (B) HDM was administered to mTORΔAPC mice for 3 weeks, along with αCsf1r neutralizing monoclonal antibody or control. Bar charts show the frequency of eosinophils or neutrophils in lung or BAL. (C) Dot graph shows significantly altered metabolic pathways between WT and mTORΔAPC inflammatory DCs, with amino acid pathways highlighted in red and fatty acid metabolism pathways highlighted in blue. Dot size corresponds to the number of differentially expressed metabolites for a given pathway. The x axis shows the log10 P value. Heat map shows the relative abundance of metabolites involved in FAO pathways, based on tentative annotations produced by the Mummichog software. (D) WT and mTORΔAPC mice were treated with etomoxir (etxr) and HDM for 3 days. Bar chart shows CD86 mean fluorescence intensity (MFI); histogram shows representative flow cytometry data. (E) mTORΔAPC mice were treated with the Cpt1 inhibitor etomoxir and HDM for 17 days. Bar chart shows expression of IL-17A+ T cells in lung; contour plots show representative flow cytometric data. (F) HDM was chronically administered to mTORΔAPC mice for 3 weeks, along with IL-12p70 control (gray) or IL-12/23p40 (blue) neutralizing antibodies. Bar charts show the frequency of eosinophils or neutrophils in lung or BAL. Data represent two or more independent experiments.

Given the observed mTOR-dependent metabolic phenotypes in steady-state lung APCs, we asked whether there was a metabolic basis for the TH17 allergic phenotype. Despite minimal differences between resting classical CD11b+ DCs in lungs of WT and mTORΔAPC mice, there was a striking perturbation in the metabolome of inflammatory DCs (fig. S20). In addition to differences in amino acid metabolism found in other DC subsets, we detected elevated fatty acid (FA) metabolites (Fig. 6C). mTOR inhibition in BMDCs can enhance fatty acid oxidation (FAO) and thereby regulate DC life-span/cytokine expression profiles (39, 40). However, a physiological role for this mechanism in primary DCs has not been described. To directly investigate whether elevated FAO underlies enhanced inflammatory DC activity, we treated mTORΔAPC mice with etomoxir, an inhibitor of the rate-limiting FAO enzyme carnitine palmitoyltransferase 1 (Cpt1) (41). Etomoxir significantly reduced the activation phenotype of mTORΔAPC inflammatory DCs without affecting WT (Fig. 6D) and also reduced TH17 polarization in mTORΔAPC lung (Fig. 6E). These observations are consistent with a model in which mTOR deletion in inflammatory lung DCs increases FAO to promote TH17-polarizing activity.

Finally, we investigated whether dysregulated cytokine production by inflammatory DCs influences allergic polarity. The proinflammatory cytokines IL-6 and IL-1β function upstream of IL-23 to promote a TH17-polarizing cascade. Analysis of CD11b+ DCs in lung and MedLN 3 days after administration of HDM revealed evidence of elevated IL-12/23p40, IL-6, and IL-1β, all of which are associated with TH17 polarization and neutrophilia (fig. S21, A and B) (29). IL-23 neutralization reverted lung neutrophilia toward an eosinophilic phenotype (Fig. 6F), concomitant with restoration in T cells of TH2 polarity (fig. S21C), identifying this cytokine as a key immunological player in mTORΔAPC-dependent neutrophilic inflammation.

Discussion

Here, we demonstrate a striking heterogeneity in the metabolic profiles of APCs in distinct tissues and reveal a key role for mTOR in programming the homeostasis and function of APC subsets in the lung. In the steady state, mTORΔAPC mice exhibited tissue-dependent phenotypes with reduced numbers of AMs and CD103 DCs in the lung and skin, but not elsewhere, unlike more global phenotypes observed when GM-CSF or Flt3l signaling is impaired (14, 18, 38). These effects were mediated by Raptor-dependent mTORC1 signaling but, surprisingly, were independent of translational regulation of mRNA, the canonical signaling pathway downstream of mTOR. Instead, our data reveal a role for Srebp1/2 signaling downstream of mTOR in homeostasis of lung APCs in the steady state (fig. S10). The selective effects of mTOR deficiency in CD103 DCs in the lung and skin were surprising, because the development of this DC subset in diverse tissues is controlled by the transcription factor Batf3. Thus, these results suggest that the homeostasis of DCs in tissues is orchestrated by the superimposition of a metabolic program on the transcriptional network that regulates development.

In addition to these effects in the steady state, mTOR deficiency also altered the character of allergic inflammation induced by lung CD11b+ DCs, skewing it toward the TH17/neutrophilic phenotype. This phenotype closely parallels aspects of human disease, in which neutrophilia associates with higher clinical severity and worse prognosis in asthma (29). Interestingly, the allergic phenotype observed in mTORΔAPC mice was associated with elevated immunoglobulin E (IgE). Such elevation in serum IgE is also observed in neutrophilic asthma patients (42) and can be enhanced by IL-23 (43), which is consistent with the IL-23 dependence of the mTORΔAPC neutrophilic inflammation phenotype. In normal circumstances, mTOR signaling in activated BMDCs suppresses oxidative phosphorylation and is in part responsible for a switch to glycolytic metabolism, which rapidly depletes intracellular energy reserves and associates with a shortened cellular life span (1). In contrast, mTOR inhibition allows BMDCs to supply their energetic demands through a combination of glycolysis and FAO, increasing cell survival and proinflammatory phenotypes (39). We observed elevated FAO and increased production of several proinflammatory cytokines in mTORΔAPC lung. Indeed one of these cytokines, IL-23, plays an obligate role in TH17/neutrophilic inflammation. Together, our study reveals a key role for mTOR in programming the homeostasis and function of APCs in the lung. In the steady state, mTOR supports accumulation of lung subsets by promoting lipid synthesis (anabolism) (fig. S22A). In the context of allergic inflammation, mTOR restrains catabolic processes such as FAO, preventing excessive inflammation (fig. S22B). The results herein identify tissue-restricted metabolic adaptations of APCs as a critical determinant of immune regulation, raising the prospect of metabolic reprogramming of such tissue-resident APCs for therapeutic benefit in various inflammatory disorders.

Supplementary Materials

www.sciencemag.org/content/357/6355/1014/suppl/DC1

Materials and Methods

Figs. S1 to S22

Table S1

References (4458)

References and Notes

Acknowledgments: We thank V. Bliss and P. Sharma for histological support, G. Tharp and N. Patel for genomics analysis, V. Tran for metabolomics, B. Cervasi and K. Gill for fluorescence-activated cell sorting (FACS), and H. Hartweger (The Rockefeller University) and E. Schweighoffer (Francis Crick Institute) for advice on B cell FACS staining. We also thank C. Wongtrakool for advice and discussion on lung inflammation models and readouts. Finally, we thank Pulendran laboratory members for discussion and M. Sinclair for proofreading the manuscript. This work was supported by funding from Action Cycling Atlanta and from the People Programme (Marie Curie Actions) of the European Union’s Seventh Framework Programme (FP7/2007-2013) (C.S.) and from the U.S. National Institutes of Health (grants R37 DK057665, R37 AI048638, U19 AI090023, and U19 AI057266) (B.P.). We are not aware of any conflicts of interest. Additional data presented in this manuscript are available in the supplementary materials.
View Abstract

Navigate This Article