Cutting Edge: mTORC2 Regulates CD8⁺ Effector and Memory T Cell Differentiation through Serum and Glucocorticoid Kinase 1

CD8⁺ T cells are a major component of the adaptive immune response in response to acute infection. During the peak immune response, two distinct subsets of CD8⁺ T cells are observed based on surface expression of the IL-7R (CD127) and killer cell lectin-like receptor subfamily G, member 1 (KLRG1) (1–3). Short-lived terminal effector CD8⁺ T cells are defined as CD127⁻ and KLRG1⁺, whereas memory precursor CD8⁺ T cells are defined as CD127⁺ and KLRG1⁻. A significant amount of CD8⁺ T cells found within the early memory precursor population will generate a stable, long-lived pool of memory T cells, poised to respond on Ag rechallenge. Several studies over the years have defined critical signaling pathways and transcriptional mediators that influence the decision-making process of how naive CD8⁺ T cells differentiate into effector or memory T cells (4). Mechanistic target of rapamycin (mTOR) is an evolutionarily conserved signaling pathway that integrates environmental cues to regulate cellular metabolism, protein synthesis, differentiation, survival, and growth (5, 6). Numerous studies have now demonstrated that both mTOR complex 1 (mTORC1) and mTOR complex 2 (mTORC2) regulate effector and memory CD8⁺ T cell differentiation (7–9). Two studies, including ours, revealed an important role for mTORC2 signaling because mTORC2-deficient (Rictor knockout, T-Rictor^−/−) CD8⁺ T cells generated more long-lived memory CD8⁺ T cells (8, 9).

Because of the critical role of mTORC2 in regulating CD8⁺ T cell responses, we hypothesized that a direct downstream target of mTORC2, SGK1, might also be an important regulator of effector and memory CD8⁺ T cell differentiation because our group has previously shown that SGK1 is necessary for Th2 CD4⁺ T cell differentiation (10). We hypothesized that SGK1 might be responsible for the enhanced memory phenotype that we observed in T-Rictor^−/− mice (10, 11). To this end, we used our SGK1-deficient (T-SGK1^−/−) mice to study the role of SGK1 in CD8⁺ effector and memory T cell differentiation.

Mice were kept in accordance with guidelines of the Johns Hopkins University Institutional Animal Care and Use Committee. T cell conditional knockout Sgk1 and Rictor mice were previously described (10). C57BL/6J, CD4-Cre, and CD8⁺ OT-I transgenic mice were obtained from Jackson Laboratories and bred to CD90.1 backgrounds.

Murine CD8⁺ T cells were purified by negative selection using a CD8⁺ isolation kit and MACS Cell Separation (Miltenyi Biotec/BioLegend). Immunoblotting for mTOR signaling was performed as previously described (10). OT-I CD8⁺ splenocytes were stimulated with 100 ng/ml OVA I (Anaspec) for 48 h, then expanded and rested in either IL-2 or IL-7/15 (10 ng/ml; PeproTech). Viable CD8⁺ T cells were enriched through Ficoll gradient for functional analysis. Human PBMCs were obtained from healthy control leukopaks.

For acute infection experiments, 5e3 to 1e6 naive CD8⁺ OT-1 T cells congenically marked with Thy1.1 were transferred i.v. to Thy1.2 C57/BL6 recipients by retroorbital injection. Recipients were challenged with Vaccinia-OVA, Listeria-OVA, or X31 influenza (Charles River). Mice were s.c. implanted with 1 × 10⁶ EL4-OVA (ATCC). Eleven days after tumor inoculation, mice received 7.5 × 10⁵ day 4 IL-2–activated wild type (WT) or T-SGK1^−/− OT-I CD8⁺ T cells. Mice were randomized into groups on day of therapy. Tumor end point and volume were calculated as previously described (8).

Chromatin immunoprecipitation (ChIP) assay was done using the MAGnify ChIP kit (Applied Biosystems) according to the manufacturer’s instructions. In brief, 10e6 Thy1.1⁺ adoptively transferred and sorted OT-I CD8⁺ T cells were used. Anti-mouse Foxo1 Ab (2880; Cell Signaling) or rabbit IgG bound to magnetic beads. Immunoprecipitates and input fraction were analyzed by quantitative PCR (qPCR) using SYBR green PCR Master Mix (Applied Biosystems) and the following primers: IL7r forward, 5′-ACCTCATCAGCCTTTCATGG-3′; IL7r reverse, 5′-ATCCCCTGAGCAAACTAGCA-3′; Eomes forward, 5′-CAAAGAGGGCTCGTTGAGAG-3′; and Eomes reverse, 5′-CCTAATTCGCGTGCTTCTTT-3′.

Flow cytometry was previously described using a BD FACSCalibur or Celesta and analyzed using FlowJo7.6 (8). Gates were determined by using unstimulated controls or isotype controls where appropriate. Cells were sorted using a BD FACSAria.

RNA was isolated and reverse transcribed into cDNA using the SuperScript III First-Strand Synthesis System for RT-PCR (Invitrogen) according to the manufacturer’s protocol. Primer and probe sets used to detect Cd127, Eomes, and 18S rRNA were from Applied Biosystems with TaqMan Universal Master Mix II using a Step One Plus Real Time PCR System.

Prism version 7.0 (GraphPad Software) was used to perform statistical analyses, including unpaired Student t test and two-way ANOVA. A p value <0.05 was statistically significant.

To analyze SGK1 activity on CD8⁺ T cell activation, we followed the kinetics of SGK1 activity by measuring the phosphorylation of its downstream target N-myc downstream-regulated gene 1 (NDRG1) (12). WT CD8⁺ T cells activated with TCR (signal 1) and costimulation (signal 2) rapidly increased SGK1 activity, as observed by an increase in phosphorylation of p-NDRG1 (T346) over time, which closely resembles the kinetics of other mTORC1/2 downstream targets (Supplemental Fig. 1A). Likewise, T cell activation of mTORC2-deficient (T-Rictor^−/−) knockout CD8⁺ T cells led to a significant defect in SGK1 signaling, further confirming SGK1 is indeed a downstream target of mTORC2 signaling (Supplemental Fig. 1B). Loss of SGK1 activity in T-SGK1^−/− CD8⁺ T cells did not abolish mTORC1 signaling, but unlike T-Rictor^−/− CD8⁺ T cells, T-SGK1^−/− CD8⁺ T cells still retained mTORC2 activity as measured by p-Akt (S473). In addition, we confirmed that SGK1 was active and downstream of mTORC2 in vivo by observing phosphorylation of NDRG1 in sorted effector WT CD8⁺ T cells, but not in both T-Rictor^−/− and T-SGK1^−/− CD8⁺ T cells (Supplemental Fig. 1C). Therefore, we have established SGK1 as a downstream target of mTORC2 signaling on activation of CD8⁺ T cells in vitro and in vivo.

Next, we wanted to determine whether mTORC2 activity through SGK1 is important in regulating CD8⁺ T cell differentiation during an acute in vivo immune response. To determine whether cell-intrinsic SGK1 expression regulates the CD8⁺ T cell response, we cotransferred WT and T-SGK1^−/− transgenic CD8⁺ T cells specific for the class I OVA peptide (OT-I) with distinct congenic markers into naive WT recipients and subsequently infected with Listeria monocytogenes expressing OVA (LM-OVA). On day 7, we analyzed donor CD8⁺ T cells in the spleen and peripheral lymph nodes, and we observed no significant differences in numbers between WT and T-SGK1^−/− CD8⁺ T cells (Fig. 1A). Phenotypically, T-SGK1^−/− CD8⁺ T cells adopted a less short-lived effector phenotype (CD127⁻KLRG1⁺) but an earlier memory precursor phenotype (CD127⁺KLRG1⁻ and CD122⁺) compared with WT CD8⁺ T cells (Fig. 1B, 1C). T-SGK1^−/− CD8⁺ T cells also adopted a central memory-like phenotype because they expressed higher amounts of CD62L (Fig. 1D). Cytokine analysis on peptide rechallenge revealed T-SGK1^−/− CD8⁺ T cells to be both potent and polyfunctional, because they produced higher amounts of IFN-γ, TNF, IL-2, and perforin compared with WT CD8⁺ T cells (Fig. 1E). Similarly, we used an SGK1 inhibitor, EMD638683, to determine whether pharmacologic targeting of SGK1 could enhance CD8⁺ T cell differentiation (13). SGK1 inhibition in both murine and human CD8⁺ T cells enhanced memory generation and cytokine polyfunctionality on restimulation (Supplemental Fig. 2). These data demonstrate downstream mTORC2 signaling through SGK1 inhibits the generation of memory CD8⁺ T cells.

Because we observed an early memory precursor phenotype in the T-SGK1^−/− mice, we hypothesized that loss of SGK1 would ultimately lead to enhanced memory CD8⁺ T cell formation on contraction. We infected WT and T-SGK1^−/− mice with various acute infectious pathogens to analyze the formation of memory CD8⁺ T cells by monitoring Ag-specific CD8⁺ T cells through tetramer analysis. We first harvested Vaccinia-OVA–immunized mice to analyze the OVA-specific CD8⁺ T cells in the memory pool. In support of the early display of memory precursor CD8⁺ T cells, T-SGK1^−/− mice contained more Ag OVA-specific CD8⁺ T cells in the spleen compared with WT mice (Fig. 2A). Similarly, using the LM-OVA and the X31 influenza models, T-SGK1^−/− mice had more Ag-specific CD8⁺ T cells in the memory pool demonstrating increased memory generation in various infectious models (Fig. 2B, 2C). A critical feature of memory cells is their prolonged survival, and indeed we observed better survival of activated T-SGK1^−/− CD8⁺ T cells compared with WT CD8⁺ T cells in the absence of any exogenous survival cytokines, which may be associated with higher expression of the antiapoptotic factors Bcl2 and Bcl-xL (Supplemental Fig. 3).

Finally, we wanted to determine the recall response of memory T-SGK1^−/− CD8⁺ T cells, so we rechallenged Vaccinia-OVA immunized WT and T-SGK1^−/− mice with LM-OVA to elicit an OVA CD8⁺ T cell–specific recall. On day 6 after reinfection, T-SGK1^−/− mice had more OVA-specific CD8⁺ T cells compared with WT mice, thereby displaying potent recall capacity of memory CD8⁺ T cells (Fig. 2D). Because increased memory potential of donor T cells is highly desired for optimal antitumor efficacy, we wanted to determine how T-SGK1^−/− CD8⁺ cells would perform in a model of adoptive T cell therapy. We implanted naive WT mice with EL4-OVA and then transferred IL-2–generated WT or T-SGK1^−/− effector OT-I CD8⁺ T cells into tumor-bearing hosts and measured tumor growth over time. Indeed, T-SGK1^−/− CD8⁺ cells showed superior antitumor control with adoptive cell therapy (Fig. 2E). To support these observations and the role of memory CD8⁺ T cells in tumor immunity, we cotransferred equal numbers of congenically distinct activated WT and T-SGK1^−/− OT-I CD8⁺ T cells into tumor-bearing mice to analyze the response of donor CD8⁺ T cells within the same host. On day 4 after T cell transfer, we observed significant enrichment of T-SGK1^−/− OT-I CD8⁺ T cells compared with WT counterparts in the tumor draining lymph node, spleen and within the tumor microenvironment (Fig. 2F). Thus, CD8⁺ T cells lacking SGK1 signaling demonstrate enhanced differentiation into long-lived memory CD8⁺ T cells, which can also generate elevated responses on Ag rechallenge.

Because Foxo1 is a direct downstream target of SGK1 and previous studies showed that activation of Foxo in the absence of Rictor resulted in enhanced memory formation (8, 9), we hypothesized that loss of SGK1 resulted in enhanced Foxo activity for memory differentiation. Foxo1 transcriptional activity is mediated by various phosphorylation sites primarily by Akt but also SGK1. Thus, we analyzed a known SGK1-dependent phosphorylation site (S256) on Foxo1 that controls the translocation of Foxo1 into the cytoplasm on activation, leading to transcriptional inactivation of Foxo1. We separated nuclear and cytoplasmic compartments of CD8⁺ T cells to determine how phosphorylation and nuclear localization of Foxo1 were regulated on TCR stimulation in previously activated WT and T-SGK1^−/− CD8⁺ T cells. Stimulated WT CD8⁺ T cells showed increased phosphorylation of Foxo1 in the cytoplasm compared with T-SGK1^−/− CD8⁺ cells. In contrast, T-SGK1^−/− CD8⁺ cells demonstrated elevated total Foxo1 protein within the nucleus because Foxo1 could not traffic into the cytoplasm (Fig. 3A). We specifically observed less phosphorylation of Foxo1 (S256) in in vivo effector T-SGK1^−/− OT-I CD8⁺ cells compared with WT OT-I CD8⁺ T cells in response to Vac-OVA (Fig. 3B). Foxo1 is critical in regulating expression of critical transcription factors involved in memory CD8⁺ T cell formation and maintenance, such as Eomes and TCF-1 (14, 15). Indeed, on day 7 of LM-OVA infection, we observed T-SGK1^−/− OT-I CD8⁺ T cells expressing more Foxo1⁺Eomes⁺ and TCF-1 compared with WT OT-I CD8⁺ T cells in the spleen and lymph nodes (Fig. 3C).

Because we observed an increase in these critical transcription factors, we wanted to determine whether Foxo1 itself was controlling gene expression of known targets, such as CD127 and Eomes. mRNA gene expression analysis of T-SGK1^−/− CD8⁺ T cells revealed that these targets are also upregulated at the transcriptional level, consistent with our in vivo findings showing elevated CD127 and Eomes protein levels (Fig. 3D). Furthermore, ChIP analysis of Foxo1 in in vivo–activated WT and T-SGK1^−/− CD8⁺ cells revealed enhanced binding of Foxo1 to the Il7r and Eomes promoters (Fig. 3E). These findings demonstrate that SGK1 controls CD8⁺ memory T cell differentiation by regulating Foxo1-mediated transcriptional programing.

In conclusion, we define a previously undescribed role for the mTORC2-SGK1-Foxo1 signaling axis as an essential pathway to control the transcriptional programming for CD8⁺ T cell effector and memory differentiation. By defining SGK1 as a downstream target of mTORC2, our data allow for the specific inhibition of SGK1 rather than upstream inhibition of mTORC2 and/or AKT as a means of enhancing memory formation. In addition to enhancing immunotherapy for cancer, our data have important and timely implications for enhancing the efficacy of preventative vaccines. For example, our data support a strategy whereby elderly or immunosuppressed patients might receive COVID-19 vaccination in the setting of an SGK1 inhibitor to boost the generation of memory T cells for more long-term immunity.

We thank the Powell lab for all helpful suggestions. We thank the National Institutes of Health Tetramer Core Facility for generously providing flow tetramers.

J.D.P. is a cofounder and equity holder of Dracen Pharmaceuticals. C.H.P. and J.D.P. are currently employees of Calico LLC. The other authors have no financial conflicts of interest.