The Energy Sensor AMPKα1 Is Critical in Rapamycin-Inhibition of mTORC1-S6K-Induced T-cell Memory

Energy sensors mTORC1 and AMPKα1 regulate T-cell metabolism and differentiation, while rapamycin (Rapa)-inhibition of mTORC1 (RIM) promotes T-cell memory. However, the underlying pathway and the role of AMPKα1 in Rapa-induced T-cell memory remain elusive. Using genetic and pharmaceutical tools, we demonstrate that Rapa promotes T-cell memory in mice in vivo post Listeria monocytogenesis rLmOVA infection and in vitro transition of effector T (TE) to memory T (TM) cells. IL-2- and IL-2+Rapa-stimulated T [IL-2/T and IL-2(Rapa+)/T] cells, when transferred into mice, differentiate into short-term IL-7R−CD62L−KLRG1+ TE and long-lived IL-7R+CD62L+KLRG1− TM cells, respectively. To assess the underlying pathways, we performed Western blotting, confocal microscopy and Seahorse-assay analyses using IL-2/T and IL-2(Rapa+)/T-cells. We determined that IL-2(Rapa+)/T-cells activate transcription FOXO1, TCF1 and Eomes and metabolic pAMPKα1(T172), pULK1(S555) and ATG7 molecules and promote mitochondrial biogenesis and fatty-acid oxidation (FAO). We found that rapamycin-treated AMPKα-deficient AMPKα1-KO IL-2(Rapa+)/TM cells up-regulate transcription factor HIF-1α and induce a metabolic switch from FAO to glycolysis. Interestingly, despite the rapamycin treatment, AMPKα-deficient TM cells lost their cell survival capacity. Taken together, our data indicate that rapamycin promotes T-cell memory via transcriptional FOXO1-TCF1-Eomes programs and AMPKα1-ULK1-ATG7 metabolic axis, and that AMPKα1 plays a critical role in RIM-induced T-cell memory.


Introduction
During an acute infection, antigen-specific CD8 + cytotoxic T lymphocytes (CTLs) activated by antigen-presenting cells rapidly proliferate and differentiate into effector T (T E ) cells [1]. These T E cells constitute an important arm of adaptive immunity and provide protection against pathogens [2]. After pathogen clearance, the majority (90-95%) of these T E cells are eliminated by apoptosis during the contraction phase; however, a small fraction (5-10%) survive and further differentiate into CD8 + memory T (T M ) cells. These CD8 + T M cells respond robustly upon reencountering the same antigen and result in a recall response that efficiently prevents pathogen-induced re-infection [2]. Therefore, a greater understanding of the mechanisms of T-cell survival and memory formation is critical for developing vaccine and immunotherapeutic strategies against cancer and infectious diseases [3].
Growing evidence indicates T-cell metabolism is an important factor in T-cell differentiation and immunity [4][5][6]. During T-cell activation and differentiation, cellular metabolism undergoes dynamic changes [4][5][6]. Naive T-cells exhibit basal levels of nutrient uptake and mainly use mitochondrial oxidative phosphorylation (OXPHOS) and fatty acid oxidation (FAO) for energy production [4][5][6]. Upon activation, most T-cells meet the energetic demands of proliferation and performing their effector functions as short-term T E cells by switching from catabolic FAO to aerobic glycolysis [4][5][6]. Fuel preference is subsequently altered in a small subpopulation of these T-cells, gradually resetting back to a more catabolic FAO state to support T-cell memory formation in the memory phase [7,8]. However, the underlying mechanism is not fully understood.
CD8 + T M cells have two unique characteristics; namely, expression of the T M cell phenotype (CCR7 + CD62L + IL-7R + ) and the capacity for recall responses upon secondary stimuli [9,10]. The T M cell phenotype plays an important role in the survival and homeostasis of these cells. For example, the chemokine receptor CCR7 and adhesion molecule CD62L control T M cell migration and homing to lymphoid organs, where CCR7 + CD62L + IL-7R + CD8 + T M cells are capable of inducing robust recall responses when supported by stromal cell-derived IL-7 [10][11][12]. The transcriptional factor FOXO1 (forkhead box protein-O1) and its two down-stream targets TCF1 (T-cell factor-1) and Eomes (eomesodermin) regulate the T M cell phenotype and differentiation [13][14][15][16][17].
It is well established that mammalian target of rapamycin complex-1 (mTORC1) and adenosine monophosphate-activated protein kinase alpha-1 (AMPKα1) are two evolutionally conserved energy sensors that regulate cellular metabolism and differentiation [18,19]. mTORC1 modulates various cellular processes including T-cell metabolism, proliferation, migration, survival, and differentiation by regulating the activity of downstream targets ribosomal S6K (S6 kinase) and eIF4E (eukaryotic translation initiation factor-4E) [2,20]. S6K in turn regulates its substrate ribosomal S6, which is routinely used as a proxy of mTORC1 activity. Ahmed's group originally reported in 2009 that rapamycin (Rapa)-inhibition of mTORC1 (RIM) promoted CD8 + T-cell memory [21]. More and more evidence has since been generated to support this finding, showing that RIM maintains T-cell plasticity and promotes antigen-specific CD8 + central memory in T-cells [22][23][24]. However, the transcriptional and metabolic pathways by which RIM promotes long-term CD8 + T-cell memory are not well established. AMPKα1 regulates catabolic processes by activating the essential autophagy kinase Unc-51-like autophagy-activating kinase-1 (ULK1), and stimulating mitochondrial biogenesis which leads to preferential use of FAO metabolism for energy production [25]. Autophagy is a cytosolic self-recycling process in which proteins and organelles are degraded via lysosomes to provide essential metabolic intermediates to cells [26,27]. Two major components of the autophagy pathway are ULK1 and autophagy-related gene-7 (ATG7) [28]. Phosphorylation of AMPKα1 (pAMPKα1, T 172 ) leads to activation of ULK1 and ATG7, which in turn promote autophagy [29] and mitochondrial biogenesis as well as FAO metabolism [25]. The energy sensor AMPKα1 has also been found to regulate T-cell memory [30,31]. However, its critical role in RIM-induced T-cell memory formation is elusive.
In this study, we sought to elucidate the molecular mechanism by which RIM induces T-cell memory formation by applying state-of-the-art genetic and pharmaceutical tools in various in vivo and in vitro experiments. To assess Rapa-promoted T-cell memory, we conducted kinetic flow cytometry analysis in C57BL/6 (B6) mice infected with recombinant Listeria monocytogenes rLmOVA to demonstrate that in vivo treatment with Rapa promotes long-term T-cell survival and IL-7R + CD62L + KLRG1 − T M cell formation. We also prepared in vitro IL-2-and IL-2+Rapa-stimulated T [IL-2/T and IL-2(Rapa+)/T] cells for further systematic characterization. We show that after being adoptively transferred into B6.1 mice, IL-2/T and IL-2(Rapa+)/T-cells display high and low levels of mTORC1-S6K signaling and become short-lived IL-7R − CD62L − KLRG1 + T E cells and long-lived IL-7R + CD62L + KLRG1 − T M cells, respectively. We also demonstrate that in vitro Rapa-induced IL-2(Rapa+)/T M cells up-regulate CD45RA, a stem cell-like memory T (T SCM ) cell marker [32]. To elucidate molecular pathways by which Rapa-induced IL-2(Rapa+)/T M cell formation, we performed Western blot, confocal and electron microscopy and Seahorse assay analyses using IL-2/T and IL-2(Rapa+)/T-cells. We establish that IL-2(Rapa+)/T M cells activate the FOXO1-TCF1-Eomes transcriptional programs controlling T M -cell differentiation and upregulate the AMPKα1-ULK1-ATG7 metabolic axis regulating mitochondrial biogenesis and FAO metabolism. To assess the critical role of the energy sensor AMPKα1 in RIM-induced T-cell memory, we genetically engineered AMPKα1 knock-out (KO)/OTI mice and isolated AMPKα1 KO IL-2(Rapa+)/T M cells from these animals. We show that in vitro the AMPKα1 deficiency up-regulates the expression of the transcription factor hypoxia-inducible factor-1 (HIF-1α) and induces a metabolic switch from FAO to glycolytic metabolism. These AMPKα1 KO IL-2(Rapa+)/T M cells also lose long-term survival upon their post-adoptive transfer into B6.1 mice.

Rapamycin Promotes T-cell Survival and Memory Formation In Vivo Post Infection with Recombinant Listeria Monocytogenes rLmOVA
To confirm that Rapa promotes T-cell memory, B6 mice were infected with recombinant Listeria monocytogenes rLmOVA and treated with rapamycin daily (days −1 to 7 postinfection). OVA-specific CD8 + T-cell responses were measured by flow cytometry analysis of mouse peripheral blood stained with PE-labelled H-2K b /OVA SIINFEKL (OVAI) peptide (PE-tetramer) and FITC-labelled anti-CD8 + antibody (FITC-CD8) at 7, 15, 30 and 60 d post infection. We found a similar frequency of OVA-specific CD8 + T-cells in mice treated with or without Rapa at the peak of CD8 + T-cell responses 7 days post-infection, but a decreased contraction of the CD8 + T-cell response in the Rapa-treated group when compared to the untreated group ( Figure 1A). In fact, the higher frequency of CD8 + T-cells was maintained at days 30 and 60, when CD8 + T-cells differentiated into T M cells ( Figure 1A). At the peak of CD8 + T-cell responses, T E cells consist of two subsets: IL-7R − CD62L − KLRG1 + short-lived effector cells (SLECs) and IL-7R + CD62L − KLRG1 − memory precursor effectorcells (MPECs). The SLECs are terminally differentiated and poised for apoptosis while the MPECs show a greater propensity to survive and differentiate into T M cells [33,34]. Therefore, we next examined the phenotype of CD8 + T-cells in infected mice by gating PE-tetramer + CD8 + T-cells 7 days post-infection for further measurement of the above three markers. We demonstrated that CD8 + T-cells in Rapa-treated mice expressed more of the T M cell markers IL-7R and CD62L, but less of the T E cell marker KLRG1, whereas the reciprocal expression pattern of these cell surface markers was observed in CD8 + Tcells derived from untreated mice ( Figure 1B). These data indicate that Rapa treatment induces CD8 + T-cells to differentiate into MPECs, leading to more T-cell memory formation post-infection with recombinant Listeria monocytogenes rLmOVA ( Figure 1A). Rapamycin not treated mice received saline and served as controls. Longitudinally tracked kinetics of OVA-specific CD8 + T-cells in blood using flow cytometry. The line graph shows the frequencies of OVA-specific CD8 + T-cells in rapamycin-treated (red) and not treated (blue) groups. (B) Cell phenotype of the OVA-specific CD8 + T-cells in the peripheral blood was analyzed using cell surface memory markers by flow cytometry. Data presented as means ± SEM are representative of two independent experiments with similar results (n = 2-3/group/experiment). * p < 0.05, ** p < 0.01 by two-tailed Student's t-test.

Rapamycin Promotes the Transition of T E into Long-Term CD45RA + Stem Cell-Like T M Cells In Vitro
Rapa is reportedly capable of promoting the transition of T E into T M cells in vivo [21]. Based upon this capability, we developed an in vitro culture approach using Rapa treatment to promote the transition of T E into T M cell precursors, similar to past protocols that have used the pro-survival cytokines IL-7 and IL-15 for the same purpose [35][36][37]. Our in vitro culture-prepared T E and T M cells approximate the T E and T M cell differentiation program observed in vivo. Briefly, naïve OTI CD8 + T-cells were cultured in a medium containing OVAI peptide and IL-2 for 3 d, and then these active CD8 + T-cells were cultured for another 2 d in medium containing IL-2 in the absence or presence of Rapa to generate IL-2/T and IL-2(Rapa+)/T-cells, respectively ( Figure 2A). Subsequent flow cytometry analysis found that IL-2(Rapa+)/T-cells up-regulated IL-7R and CD62L and down-regulated KLRG1, while the reciprocal response was observed for IL-2/T-cells ( Figure 2A). To measure survival of these T-cell subsets in vivo, we adoptively transferred equal amounts of in vitro-prepared CD45.1 + /2 + IL-2/T and CD45.2 + IL-2(Rapa+)/T-cells into CD45.1 + B6.1 mice ( Figure 2B). This approach allows us to separately analyze the survival of IL-2/T and IL-2(Rapa+)/Tcells in B6.1 mouse peripheral blood by kinetic flow cytometry analyses post-cell transfer ( Figure 2C), as previously described [36]. We found significantly more IL-2(Rapa+)/T than IL-2/T donor cells in host mice at different time points post-cell transfer. At 30 days post-cell transfer, IL-2(Rapa+)/T-cells were 13-fold more abundant than IL-2/T donor cells in host mice, indicating IL-2(Rapa+)/T-cells survive much longer than IL-2/T-cells. We then characterized these T-cells by measuring expression of the T M cell marker CD62L and the T SCM cell marker CD45RA [32]. Interestingly, we found that all IL-2/T and IL-2(Rapa+)/T-cells expressed CD62L while 58% of the IL-2(Rapa+)/T but none of the IL-2/T-cells expressed CD45RA at days 14 and 30 post-cell transfer ( Figure 2D   . (E) Graph shows recall response of rapamycin-treated (CD45.2 + ) or not treated (CD45.1 + /2 + ) T-cells 30 days after the adoptive transfer and 5 days post-rLmOVA (5000 CFU) intravenous challenge (n = 4). Data presented as means ± SEM are representative of two independent experiments with similar results (n = 2-3/group/experiment). * p < 0.05, ** p < 0.01 by two-tailed Student's t-test.

IL-2(Rapa+)/T-cells Suppress mTORC1/S6K Signaling and Activate the FOXO1-TCF1-Eomes Transcriptional Pathway
To further characterize the IL-2/T and IL-2(Rapa+)/T-cells, we performed Western blotting analyses to assess whether Rapa treatment inhibits mTORC1 activity in our experimental system ( Figure 3A,B). We assessed the expression of ribosomal S6 which is routinely used as a proxy of mTORC1 activity [38]. We found Rapa treatment abolished the phosphorylated form of S6 (pS6, S 235/236 ) in IL-2(Rapa+)/T, but not IL-2/T, cells ( Figure 3B). Thus, our data indicate Rapa treatment inhibits mTORC1 activity. The transcription factors FOXO1, Id3, TCF1 and Eomes are known to regulate IL-7R + CD62L + KLRG1 − T M cell differentiation, while Id2 and T-bet control IL-7R − CD62L − KLRG1 + T E cell differentiation [39]. Among these factors, FOXO1 is upstream of TCF1 and Eomes and controls their activity [13][14][15][16][17], while Id3 is a downstream target of TCF1 that impinges upon T-cell memory [40]. We therefore measured the relative abundance of these proteins by Western blot analysis and found that IL-2(Rapa+)/T-cells displayed higher levels of FOXO1, TCF1, Eomes and Id3, but lower levels of Id2 and T-bet ( Figure 3B). The reciprocal transcription factor expression profile was observed for IL-2/T-cells ( Figure 3B). Therefore, our data indicate IL-2(Rapa+)/T-cells activate the FOXO1-TCF1-Eomes transcriptional pathway for T M cell differentiation, while IL-2/T-cells trigger the T-bet transcriptional pathway for T E cell differentiation. To further confirm that the IL-2(Rapa+)/T-cells prepared in vitro indeed activate the FOXO1-TCF1-Eomes transcriptional pathway, we adoptively transferred CD45.1 + /2 + OTI naïve CD8 + T-cells into CD45.2 + B6 mice (to increase the frequencies of OTI T-cells) and then infected them with rLmOVA in the presence or absence of Rapa treatment ( Figure 3C), as described in Figure 1A. The intracellular expression of transcription factors or kinases was determined through flow cytometry by gating on the CD8 and CD45.1 double-positive T-cell population ( Figure 3C). We found that transferred OTI T-cells in Rapa-treated mice expressed higher levels of the transcription factors FOXO1 and Eomes, but lower levels of the pS6 (S 235/236 ) kinase and transcription factor T-bet compared to untreated mice ( Figure 3C), indicating Rapa also activates the T-cell transcriptional FOXO1 pathway in vivo. Active FOXO1 localizes to the nucleus while its phosphorylated form (pFOXO1) is inactive, and re-localizes to the cytosol where it is subsequently polyubiquitinated and degraded [17,41]. TCF1, a downstream target of Wnt signaling, plays a regulatory role in T-cell memory through its nuclear localization [17]. To visualize the subcellular localization of FOXO1 and TCF1, we performed confocal microscopy and found greater nuclear enrichment for both factors in IL-2(Rapa+)/T-cells compared to IL-2/T-cells ( Figure 3D).  CD8 + T-cells (CD45.1 + /2 + ) into C57BL/6 (CD45.2 + ) mice (n = 5), and 24 hr later infection (intravenous) with rLmOVA (2000 CFU/mice). Mice were treated with rapamycin from day −1 to day +5. Rapamycin not treated mice received saline and served as controls. Five days post-infection, adoptively transferred T-cells were tracked using CD45.1 + T-cells in total host CD8 + T-cells. The expression of signaling molecules and transcription factors was examined using intracellular staining by flow cytometry. Histograms ( Figure 3C, upper panels) and bar diagrams ( Figure 3C, lower panels) indicate relative expression of molecules in rapamycin-treated (red) and not treated (blue) groups. (D) Confocal microscopy for the cellular localization of FOXO1 (green) ( Figure 3D, upper panels) and TCF1 (green) ( Figure 3D, lower panels) in nuclear (blue) of IL-2/T E and IL-2(Rapa+)/T M cells. These T-cells were stained with PE-anti-CD8 antibody (red). Data presented as means ± SEM are representative of two independent experiments with similar results (n = 2-3/group/experiment). * p < 0.05, ** p < 0.01 by two-tailed Student's t-test.

IL-2(Rapa+)/T-cells Activate the AMPKα1-ULK1-ATG7 Metabolic Axis
AMPKα1 is an evolutionarily conserved energy-sensing kinase that regulates catabolic processes by activating the autophagy-related molecules ULK1 and ATG7 [28] and stimulating mitochondrial biogenesis to drive the FAO required for T-cell memory [28]. The transcription factor HIF-1α is a master regulator of glycolytic gene expression and stimulates flux through this pathway [42]. In contrast, the repressor Bcl-6 down-regulates glycolytic flux [43]. We therefore assessed the relative abundance of pAMPKα1 (T 172 ), pULK1 (S 555 ), ATG7, Bcl-6 and HIF-1α in IL-2/T and IL-2(Rapa+)/T-cells by Western blot analysis ( Figure 4A). We demonstrated that IL-2(Rapa+)/T-cells have higher levels of pAMPKα1 (T 172 ), pULK1 (S 555 ) and ATG7 ( Figure 4B), suggesting that activation of the AMPKα1-ULK1-ATG7 metabolic axis promotes FAO and underlies the switch in fuel preference. Conversely, IL-2/T-cells exhibit the reciprocal expression profile for these factors and instead harbor more of the transcription factor HIF-1α, a master regulator for glycolysis. We also measured the relative abundance of pAMPKα1 (T 172 ) in OTI T-cells and found that those isolated from Rapa-treated mice had higher levels of pAMPKα1 (T 172 ) ( Figure 3C). This finding indicates that Rapa also activates the AMPKα1 metabolic pathway in T-cells in vivo.  and OCR values above this line is a mitochondrial spare respiratory capacity (SRC) in respective groups; Line diagram shows OCR measured in real-time under basal conditions and in response to indicated mitochondrial inhibitors (red-rapamycin-treated; blue-not treated with rapamycin); p < 0.0001 after FCCP injection. Data from two to three independent experiments are presented as means ± SEM (n = 4/group). * p < 0.05, ** p < 0.01, *** p < 0.001 by two-tailed Student's t-test.

IL-2(Rapa+)/T-cells Promote Mitochondrial Biogenesis
AMPK-peroxisome proliferator-activated receptor-gamma coactivator-1α (PGC-1α) is a critical regulator of mitochondrial biogenesis [44]. To assess whether enhanced activity of the AMPKα1-ULK1-ATG7 metabolic axis affects the expression of PGC-1α to modulate organelle content, we performed Western blot analysis. Indeed, we found IL-2(Rapa+)/Tcells displayed higher levels of PGC-1α compared to IL-2/T-cells ( Figure 4B). To further confirm this finding, we performed flow cytometry and confocal microscopy analyses using MitoTracker Green, which binds to the mitochondrial membranes and stains the organelle [45]. We demonstrated that IL-2(Rapa+)/T-cells contain a fragmented reticulum with many more individual mitochondria than IL-2/T-cells ( Figure 4C,D). Mitochondrial morphology has previously been shown to affect fuel preference in T-cells, with small, round organelles supporting glycolysis in T E cells and long, tubular mitochondria being central to FAO in T M cells [46]. Consistent with this idea, our electron microscopy analysis showed that mitochondria were elongated and tubular in IL-2(Rapa+)/T-cells while they were small and spherical in IL-2/T-cells ( Figure 4E).

IL-2(Rapa+)/T-cells Have Substantial Mitochondrial SRC and Rely on FAO
Mitochondria are bioenergetic organelles controlling energy homeostasis that contribute to T M cell survival [47] via OXPHOS, by providing the spare respiratory capacity (SRC) essential for FAO [48,49]. Therefore, to assess the effect of Rapa treatment on mitochondrial metabolism in T-cells, we measured the bioenergetic profiles of IL-2/T and IL-2(Rapa+)/T-cells under basal conditions and following the addition of various agents to block flux through the electron transport chain (ETC) or impair ATP synthesis. We demonstrated that IL-2(Rapa+)/T-cells produced more ATP than IL-2/T-cells ( Figure 4F). Moreover, IL-2(Rapa+)/T-cells had a higher rate of O 2 consumption (OCR) when compared to the rate of extracellular acidification (ECAR) (Figure 4F), which is a marker of FAO and indicates preferential usage of fatty acids over sugars as a fuel source. In contrast, IL-2/Tcells had a lower OCR and an elevated ECAR, indicating they relied more on glycolytic flux to maintain energy homeostasis compared to IL-2(Rapa+)/T-cells ( Figure 4F). Finally, the maximal OCR following FCCP injection was significantly higher in IL-2(Rapa+)/T-cells, consistent with the idea that the SRC allows these cells to produce more ATP via OXPHOS than IL-2/T-cells and utilize FAO metabolism to preserve energy homeostasis.

AMPKα1 Deficiency Down-Regulates CD45RA Expression in IL-2(Rapa+)/T-cells and Abolishes Their Long-Term Survival
To measure cell survival, an equal number of CD45.1 + /2 + IL-2(Rapa+)/T and AMPKα1 KO CD45.2 + IL-2(Rapa+)/T-cells prepared in vitro were adoptively transferred into CD45.1 + B6.1 mice, and cell survival and expression of the T SCM marker CD45RA were kinetically monitored by flow cytometry (Figure 5E), as described in Figure 2B. These analyses demonstrated that the survival IL-2(Rapa+)/T-cells (1.8%) was more than 10-fold higher than that of AMPKα1 KO IL-2(Rapa+)/T-cells 30 days post-cell transfer ( Figure 5F), indicating AMPKα1 is required for the ability of Rapa to promoter long-term survival of T M cells. In addition, AMPKα1 deficient IL-2(Rapa+)/T-cells were almost devoid of CD45RA expression even at 7 days post-cell transfer ( Figure 5G), indicating the kinase is essential to sustain expression of this cell surface molecule.

Discussion
Rapa inhibits the kinase activity of mTOR by forming complexes with its intracellular receptor FK506-binding protein-12 (FKBP12), which then bind to the N-terminal domain of mTOR and interfere with mTORC1 and mTORC2 assemblies [50]. Depending on the nature of the Rapa exposure, it can uniquely inhibit mTORC1 and mTORC2 formations owing to their differential sensitivities. For example, low Rapa concentrations or transitional exposure to Rapa only inhibit mTORC1 assembly, while high Rapa concentrations or long-term exposure to Rapa are capable of limiting mTORC2 formation [50].
Rapa-inhibition of mTORC1 (RIM) promoted CD8 + T-cell memory [21][22][23][24]. However, the transcriptional and metabolic pathways by which RIM promotes long-term CD8 + T-cell memory are not well elucidated. In this study, we demonstrate that Rapa promotes T-cell memory in vivo post-infection of mice with Listeria monocytogenesis rL-mOVA and stimulates the transition of effector T (T E ) to memory T (T M ) cells in vitro. IL-2-and IL-2+Rapa-stimulated T [IL-2/T and IL-2(Rapa+)/T] cells with high and low levels of mTORC1-S6K signaling differentiate into short-term IL-7R − CD62L − KLRG1 + T E and long-lived IL-7R + CD62L + KLRG1 − T M cells, respectively. To assess the underlying pathways, we performed Western blotting, confocal microscopy and Seahorse-assay analyses using IL-2/T E and IL-2(Rapa+)/T M cells. We determined that IL-2(Rapa+)/T M cells activate the FOXO1, TCF1 and Eomes transcription factors and the metabolic regulators pAMPKα1(T 172 ), pULK1(S 555 ) and ATG7 to promote mitochondrial biogenesis and FAO, while down-regulating the transcription factors T-bet and HIF-1α. We observed the reciprocal expression pattern for these molecules in IL-2/T E cells. Therefore, our data indicate that Rapa-inhibition of mTORC1-S6K promotes T-cell memory via activation of the FOXO1-TCF1-Eomes transcriptional pathway and the AMPKα1-ULK1-ATG7 metabolic axis, whereas IL-2-stimulated strong mTORC1-S6K signaling controls T E cell formation via activation of the transcriptional T-bet and metabolic HIF-1α pathways ( Figure 6). It has been demonstrated that the FOXO1-TCF1-Eomes transcriptional pathway is indispensable for T M cell differentiation and formation [17]. In this study, we used IL-2(Rapa+)/T-cells derived from AMPKα1 KO/OTI mice to demonstrate that AMPKα1 deficiency ablates the AMPKα1 pathway, blunts mitochondrial biogenesis, induces lower rates of OXPHOS metabolism and abolishes long-term T-cell survival, indicating that the AMPKα1-ULK1-ATG7 metabolic axis is also indispensable for CD8 + T M cell formation. Therefore, we conclude that RIM promotes T-cell memory by coordinate regulation of the FOXO1-TCF1-Eomes transcriptional program and the AMPKα1-ULK1-ATG7 metabolic axis ( Figure 6).
The CD8 + T M cell population was originally divided into two subsets: CCR7 − CD62L − IL-7R + effector memory T (T EM ) and CCR7 + CD62L + IL-7R + central memory T (T CM ) cells [10,11]. Due to the lymph node homing receptors CCR7 and CD62L, CCR7 + CD62L + IL-7R + T CM cells mainly reside in lymph nodes while CCR7 − CD62L − IL-7R + T EM cells are mostly localized to peripheral tissues [10,11]. Recently, a new subset of the T M cell population known as CD45RA + IL-7R + CD62L + T SCM cells has been defined, and shows increased proliferative activity, superior antitumor immunity, and the capacity to differentiate into T EM and T CM cells [32]. These T SCM -cells express two naïve T-cell markers CD45RA and IL-7R [32], which are critical for T-cell differentiation and homeostasis as well as for the regulation of T-cell signaling threshold [12,51,52]. Up-regulation of IL-7R expression in T M cells is directly regulated by the transcription factor FOXO1 [53]. However, except for a few reports, molecular regulation of T-cell CD45RA expression is less well studied. It was reported that IL-7 drives the phenotypic revision of IL-7R + CD62L + T CM and IL-7R + CD62L − T EM cells into CD45RA + IL-7R + CD62L + T SCM cells by facilitating the epigenetic reorganization essential for secondary responses [54]. Metabolic control of T-cell immunity via its intermediary metabolites is a newly emerging area of fundamental immunology [55]. In this study, we demonstrate that Rapa treatment promotes the transition of IL-2/T E cells into IL-2(Rapa+)/T M cells in vitro. Interestingly, we also observe that AMPKα1 deficiency in IL-2(Rapa+)/T-cells leads to the down-regulation of CD45RA, but not IL-7R, expression. This finding suggests a potential association between AMPKα1-regulated FAO metabolism and CD45RA expression, and indicates Rapa-promoted CD45RA + IL-7R + CD62L + T SCM differentiation is mediated by AMPKα1 activity. Therefore, the underlying mechanism warrants further study. mTORC1 and AMPKα1 are two evolutionally conserved signaling molecules that regulate cell metabolism and differentiation [18,19]. However, their interplay in controlling CD8 + T-cell differentiation is less understood. In this study, we demonstrate that IL-2 stimulates strong mTORC1-S6K activity but very weak AMPKα1 signaling, leading to IL-2/T E cell differentiation. However, when mTORC1 is inhibited by Rapa, IL-2+Rapa stimulation induces IL-2(Rapa+)/T M cell formation via activation of the metabolic AMPKα1 pathway. This finding argues that mTORC1 suppresses AMPKα1 in IL-2/T E cells, perhaps via Raptordependent phosphorylation of AMPKα1 at S 347 to block its phosphorylation at T 172 and inhibit its subsequent activation [56]. However, when AMPKα1 is absent, IL-2(Rapa+)/Tcells derived from AMPKα1 KO/OTI mice lose T M cell formation and long-term T-cell survival. Instead, the mTORC1-S6K-controled transcription factor HIF-1α [42] was upregulated in IL-2(Rapa+)/T-cells lacking AMPKα1, which leads to a metabolic switch from FAO to glycolysis. This finding indicates that AMPKα1 normally inhibits mTORC1 in IL-2(Rapa+)/T-cells and is supported by earlier studies which showed that AMPKα1 activators such as 5-aminoimidazole-4-carboxamide-1-β-4-ribofuranoside (AICAR) and metformin enhance the inhibitory effect of Rapa on mTORC1 [57,58]. The inhibitory effect of AMPKα1 on mTORC1 may be due to its ability to phosphorylate the mTORC1 binding partner Raptor at S 722 and S 792 [59] or to activate the mTORC1 suppressor TSC2 [38,60,61]. Therefore, our data collectively indicate a negative feedback mechanism exists between these two energy sensors in CD8 + T-cell differentiation and suggest that their relative activities play an important role in immune responses and immune diseases [2,62].
The theory of yin (negative regulation) and yang (positive regulation) with a negative feedback interplay represents one of the most fundamental principles in traditional Chinese medicine. This theory has been applied to interpret the immune system. For example, CD28 and CTLA-4 have been described as the yin and yang of T-cell co-stimulation [63]. The yin and yang interplay of cytokine IFN-γ in inflammation and autoimmune diseases has also been reported [64]. It has also been suggested that Th1 cells belong to yang whereas Th2 and Treg cells belong to yin, where yang represents immune initiation and response while yin represents immune regulation and tolerance [65]. Recently, it was reported that AMPK represents the yin that signals a lack of nutrients and inhibits cell growth, whereas mTORC1 represents the yang that signals the availability of nutrients and promotes cell growth [66]. However, how this theory informs immune responses at the molecular level is still unknown. Our study is significant in this regard for it presents the first evidence showing how mTORC1 acts as the yang and AMPK acts as the yin in generating signals that control CD8 + T-cell differentiation into effector and memory T-cells ( Figure 6). Specifically, (i) IL-2-stimulated strong mTORC1 (yang) signaling in IL-2/T-cells controls cell expansion with glycolysis metabolism for cell energy and differentiation into IL-2/T E cells with effector function via activation of transcriptional T-bet and metabolic HIF-1α (yang) pathways; (ii) IL-2+Rapa-stimulation results in weak mTORC1 (yang) but strong AMPKα1 (yin) signaling and regulates cell differentiation into IL-2(Rapa+)/T M cells that are quiescent and uses FAO metabolism for cell energy (Figure 6), and (iii) AMPKα1 (yin) and mTORC1 (yang) interplay under a negative feedback mechanism in CD8 + T-cell differentiation.
Inhibition of mTORC1 can alter the intrinsic properties of CD8 + T-cells to favor their differentiation into memory T-cells and, as such, may be an attractive strategy to enhance therapeutic anti-tumor immunity. Therefore, elucidation of the exact molecular signaling through mTORC1 that controls the transcriptional FOXO1 and metabolic AMPKα1 path-ways for T-cell memory may open up new avenues for developing new therapeutics to enhance T M cell responses in humans. An additional therapeutic target of interest in this context is phospholipase D (PLD), an enzyme that catalyzes the hydrolysis of the major phospholipid phosphatidylcholine within cell membrane [67]. It has been reported that the phosphatidic acid (PA) produced by PLD plays an important role in T-cell receptormediated activation via its activation and stabilization of mTORC1 and mTORC2 [50,67]. Therefore, we speculate that inhibition of PLD may also promote T-cell memory through down-regulation of mTOC1 and mTORC2 [17,21].
The ultimate goal of vaccine development is to stimulate hosts to produce a large number of good quality T M cells against cancer and infectious diseases [3]. This goal will be greatly impacted by our new finding that Rapa promotes CD8 + T-cell memory via the coupled regulation of the FOXO1-TCF1-Eomes transcriptional pathway and the AMPKα1-ULK1-ATG7 metabolic axis ( Figure 6). Accumulating evidence highlights the integration of transcriptional and metabolic pathways, with transcription factors shaping metabolic programming of T-cell memory and cellular metabolites reciprocally regulating the expression of genes involved in T-cell memory programming at the epigenetic and transcriptional levels [8,68]. Therefore, ongoing investigation of new metabolic or epigenetic signals critical to T-cell memory programs is warranted and will greatly advance mTORC1-targeted immunotherapeutics and vaccine development for cancer and infectious diseases.
Taken together, we provide the first evidence that RIM promotes CD8 + T-cell memory by coupling the FOXO1-TCF1-Eomes transcriptional pathway and the AMPKα1-ULK1-ATG7 metabolic axis and that the energy sensor AMPKα1 plays a critical role in RIMinduced CD8 + T-cell memory. Our findings collectively provide novel mechanistic insight into how inhibition of the mTORC1 pathway impinges upon CD8 + T-cell memory that will greatly impact vaccine development and immunotherapy for cancer and infectious diseases.

Splenocytes and Peripheral Blood Mononuclear Cell Preparation
Single-cell suspensions from spleens were prepared by mashing up the organs with a 10-mL syringe plunger against a 70-µm cell strainer into 10-mL RPMI 1640 medium (RPMI) supplemented with 10% (v/v) fetal calf serum (FCS) and 20 µg/mL gentamicin. Peripheral blood samples were collected into tubes containing heparin (BD Biosciences, San Jose, CA, USA) by nicking the lateral tail vein. Red blood cells in splenocytes and peripheral blood were lysed with ammonium-chloride-potassium (ACK) lysing buffer (150 mM NH4Cl, 10 mM KHCO3, and 0.1 mM EDTA) by incubating for 5-min at room temperature, followed by quenching with RPMI medium, centrifugation, and re-suspension in 5 mL of RPMI medium.

Flow Cytometry
Surface staining was performed by incubating the cells with fluorescently labeled antibodies for 30 min on ice in PBS supplemented with 2% FCS and 0.1% sodium azide. The following antibodies used for cell surface staining were purchased from Biolegend (San Diego, CA, USA): PE594 Cy5-CD8 (clone53-6.7), FITC-CD45.1 (clone A20), Alexa Fluor 700-CD45.2 (clone 104), APC KLRG1 (clone 2F1), Brilliant Violet 510-IL-7Ra (clone A7R34), and APC-A750-CD62L (clone MEL-14). Brilliant Violet 421-CD45RA (clone 14.8) antibody was obtained from BD Biosciences. Intracellular staining was done as per the manufacturer's (BD Bioscience) protocol for flow cytometry. Briefly, after cell-surface staining, T-cells were fixed and permeabilized in 1 mL of fixation and permeabilization working solution for 20 min on ice, followed by a wash in 1X permeabilization solution and incubation with primary antibodies against T-bet, FOXO1, and pS6 (S 235/236 ) (Cell Signaling, Danvers, MA, USA) in 100 µL of 1X permeabilization buffer. Cells were washed once with 1 mL of 1X permeabilization solution, followed by incubation with PE-goat anti-rabbit IgG secondary antibody (Biolegend) in 100 µL of 1X permeabilization buffer. The samples were then washed twice with 1X permeabilization solution and re-suspended in PBS supplemented with 2% FCS and 0.1% sodium azide for flow cytometry analyses. The cell mitochondrial mass was measured by staining cells with MitoTracker Green (Life Technologies, Carlsbad, CA, USA). Briefly, cells were stained with MitoTracker Green at 10 nM for 15 min at 37 • C according to the manufacturer's manual. The cells were washed three times with PBS and then analyzed by flow cytometry, as we previously described [45]. Flow cytometry data were acquired using CytoFLEX (Beckman Coulter, Brea, CA, USA) and analyzed with FlowJo software (TreeStar, Ashland, OR, USA).

Confocal and Electron Microscopy Imaging
Mito Tracker Green (Life Technologies) was used to measure mitochondrial mass. In vitro Rapa-treated (Rapa+) and untreated (Rapa−) T-cells were stained with 50 nM MitoTracker Green or 10 µM TMRM for 15 min at 37 • C, according to the manufacturer's manual. Stained cells were washed three times with PBS, followed by incubation with 5 µg/mL of Hoechst 33342 solution (Life Technologies) for 20 min at room temperature. Confocal images were acquired using the Zeiss LSM700 confocal microscope (Carl Zeiss, Oberkochen, BW, Germany) with ×20 objective [45]. For electron microscope imaging, cell pellets (2 × 10 6 T-cells/each) were fixed in 2% paraformaldehyde and 2.5% glutaraldehyde in 100 mM sodium cocodylate. After fixation, samples were washed in cacodylate buffer and fixed in 1% osmium tetroxide. After extensive washing in H 2 O, samples were stained with 1% aqueous uranyl acetate for 1 h and washed again. Samples were dehydrated in ethanol, embedded in Eponate 12™ resin (Ted Pella, Redding, CA, USA), and sectioned for imaging [70]. Images were acquired using a JOEL 1200 EX transmission electron microscope.

Seahorse Assays
The OCR and ECAR of in vitro-activated CD8 + T-cells were measured in XF RPMI media containing 25 mM glucose, 2 mM L-glutamine, and 1 mM sodium pyruvate (Agilent, Lexington, MA, USA). Cells were then plated onto XF8 cell culture microplates (1.5 × 10 5 cells per well) coated with poly-D-lysine (Sigma-Aldrich) to facilitate T-cell attachment. A mitochondrial stress test was performed by measuring OCR (pmol min-1) at the basal level and after sequential injection of oligomycin (1.5 µM), FCCP (2.5 µM), and rotenone/antimycin A (0.5 µM) (Agilent, CA, USA), and run on an Agilent Seahorse XFp analyzer (Seahorse Bioscience, Agilent, CA, USA). The following assay conditions were used for the experiments with the Seahorse system: 3 min mixture; 0 min wait; and 3 min measurement.

Statistical Analysis
Statistical analyses were performed using unpaired t-test (two-tailed) or analysis of variance for comparison of means using GraphPad Prism6 software (GraphPad, La Jolla, CA, USA) [44,69]. Values of p < 0.05 and <0.01 were considered significant and very significant, respectively. Author Contributions: J.X. conceived the project, designed the experimental work and provided supervision; A.A. designed and performed the experiment work; A.X., M.F.I. and Z.W. generated B6.1/OTI and AMPK KO/OTI mice and provided lab support; S.C.L. provided technical help in cell metabolism analysis; R.C. provided technical help in electron microscopy analysis; K.A.A. provided help in performing Sea-Horse assays; J.X. and A.A. wrote the manuscript. All authors have read and agreed to the published version of the manuscript.

Conflicts of Interest:
The authors declare no competing financial interest. T-cell factor-1 T E cell effector T-cell T M cell memory T-cell IL-2/T-cell IL-2-stimulated T-cell IL-2(Rapa+)/T-cell IL-2+Rapa-stimulated T-cell ULK1