Foxo1 Serine 209 Is a Critical Regulatory Site of CD8 T Cell Differentiation and Survival

In response to bacterial or viral infection, CD8 T cells undergo cell fate decisions to appropriately balance pathogen clearance with the generation of long-lived memory to ensure both immediate host survival and durable immunity against subsequent infections. Some transcription factors act at crucial nodes to integrate pathogen and host cues and instruct CD8 T cell differentiation. One such transcription factor, Foxo1, is necessary for the establishment and maintenance of CD8 T cell memory (1–5). Foxo1-regulated genes promote memory T cell development by increasing T cell survival, homing to secondary lymphoid organs, and expression of downstream transcription factors that further establish the memory state (1, 2, 6). Expression of these genes is severely impaired in the absence of Foxo1, resulting in markedly reduced CD8 T cell persistence and expansion during recall responses (1–3).

Foxo1 transcriptional activity is tightly regulated by posttranslational modifications including phosphorylation by the kinases Akt and Mst1 (7–10). It is well established that Akt-dependent phosphorylation of Foxo1 at threonine (T)24, serine (S)253, and S316 residues prevents Foxo1 transcriptional activity by promoting Foxo1 association with 14-3-3 proteins, leading to its cytoplasmic sequestration (7, 8, 11). In contrast, Mst1 phosphorylates Foxo1 at a distinct residue located within its forkhead DNA-binding domain: S212 in humans and S209 in mice (10). The first studies characterizing Mst1 phosphorylation of Foxo1 and its paralog Foxo3 in neurons reported that Mst1 positively regulates Foxo transcription factors by causing dissociation from 14-3-3 proteins, promoting Foxo nuclear translocation (10, 12). Conversely, x-ray crystallography data show that S209 directly participates in binding to cognate insulin-response elements and Daf-16 family–binding elements (DBE) (9, 13, 14). Consequently, phosphorylation of the Foxo1 forkhead domain at S209 completely abrogates binding to cognate DNA in vitro (9). Moreover, Mst1 deficiency in humans and mice causes an immunodeficiency syndrome characterized by impaired naive T cell survival and trafficking and associated autoimmunity (15–22). However, whether this immunodeficiency is a consequence of defective Foxo1 S209 phosphorylation in T cells is unclear. Additionally, the impact of Foxo1 S209 phosphorylation on the expression of canonical Foxo1 target genes and CD8 memory T cell development is unknown.

In this study, we investigated how Foxo1 S209 regulates effector versus memory CD8 T cell differentiation. We generated Foxo1 variants in which S209 was mutated to alanine (S209A [SA]) or aspartic acid (S209D [SD]) to determine how Foxo1 behaves in the nonphosphorylated or phosphorylated states, respectively. SA-Foxo1 exhibited Akt-dependent nuclear trafficking and largely recapitulated wild-type (WT)–Foxo1 transcriptional activity but unexpectedly was defective in promoting survival of memory-like T cells in vitro. Moreover, SA-Foxo1–reconstituted CD8 T cells favored memory precursor versus short-lived effector differentiation yet were defective in generating central memory T cells (TCM) in response to Listeria monocytogenes infection in vivo. In contrast, SD-Foxo1 was exported out of the nucleus and was unable to transactivate canonical target genes, consistent with the structural analyses implicating S209 in direct DNA contact. However, SD-Foxo1–reconstituted T cells had a unique transcriptional signature distinct from WT or Foxo1-deficient T cells, including genes that mediate cell survival and CD8 effector function. Furthermore, SD-Foxo1 enhanced CD8 T cell survival in vitro. Finally, we found that SD-Foxo1–reconstituted CD8 T cells favored short-lived effector differentiation and survival but were unable to form CD8 memory precursor cells and memory cells in vivo in part because of active suppression of CD127 expression.

Together, our findings demonstrate that phosphorylation of Foxo1 at S209 is neither activating nor inhibitory per se. Instead, modification of Foxo1 at S209 promotes its nuclear exclusion and simultaneously results in expression of alternative transcriptional targets to promote effector function and cell survival at the expense of memory T cell differentiation. Thus, these studies have dissociated the signals for T cell survival from those for memory differentiation.

WT-Foxo1-GFP (no. 17551; Addgene) and ADA-Foxo1-GFP (no. 35640; Addgene) mammalian expression vectors were purchased from Addgene (23, 24). A nonsynonymous substitution in the Addgene vectors located within the Foxo1 transactivation domain that changes a conserved leucine at position 619 to proline was reverted to the native leucine by PCR site–directed mutagenesis. The respective Foxo1-GFP fragments were inserted into the pCigar retroviral vector by PCR, restriction digest, and ligation according to standard protocols. PCR site–directed mutagenesis was used to introduce the SA and SD mutations into Foxo1 at the Mst1 phosphorylation site, and Gibson Assembly (no. E2611; New England Biolabs) was used to replace the homologous sequences in Foxo1 with the amplified DNA fragments. To generate the Foxo1 quad mutant (QM; QM-Foxo1: T24A, S209D, S253D, and S316A), gene block DNA fragments containing the indicated mutations were synthesized by Integrated DNA Technologies and inserted into the retroviral Foxo1-GFP vector by standard restriction digest and ligation. For each retroviral construct, the entirety of the Foxo1 gene was sequenced in the Dartmouth Molecular Biology Core Facility to verify the presence of the intended mutation(s) and the absence of any unintended mutations.

For 293T studies, cells were transfected by calcium phosphate transfection with retroviral plasmids encoding Foxo1-GFP variants and a mammalian expression vector encoding histone H2B fused to red fluorescent mCherry (H2B-mCherry; no. 20972; Addgene) (25) as a nuclear marker. At 24 h posttransfection, 293T cells were seeded on eight-well Nunc Lab-Tek II Chambered Coverglass (no. 155409; Thermo Fisher Scientific) at a density of 8 × 10⁴ cells per well in complete DMEM media. At 24 h postseeding, 293T cells were treated with leptomycin B (no. L2913; MilliporeSigma) or vehicle control (methanol/H₂O) for 6 h to inhibit exportin/nuclear export sequence–mediated nuclear export. For the final 45 min of leptomycin B treatment, cells were treated with 10 μM MK2206 Akt allosteric inhibitor (no. sc-364537; Santa Cruz Biotechnology) or vehicle control. The 293T cells were visualized on a Zeiss Axio Observer Z1 microscope at 40× magnification. Foxo1-GFP fluorescence intensity in the nuclear and cytoplasmic compartments was measured using FIJI and the nuclear-to-cytoplasmic (N/C) ratio of Foxo1-GFP was calculated on a per-cell basis.

Retroviruses encoding Foxo1-GFP variants were packaged in 293T cells. Whole splenocytes from OT1 TCR transgenic mice were cultured in complete RPMI 1640 media with 0.4 mg/ml chicken OVA (no. A5503; MilliporeSigma) and 50 ng/ml recombinant mouse IL-2 (no. 575406; BioLegend). At 24–32 h postactivation, OT1 T cells were transduced with Foxo1 retroviral supernatants containing 8 μg/ml polybrene and centrifuged at 1,500 × g for 1.5 h. After centrifugation, OT1 T cells were rested at 37°C for 1 h before replacing the viral supernatant with fresh complete RPMI 1640 media containing 50 ng/ml recombinant mouse IL-2. At 48 h posttransduction, Foxo1-GFP–positive OT1 T cells were sorted by FACS and seeded on eight-well chambered coverglass (no. 155409; Thermo Fisher Scientific) at a density of 2.1 × 10⁵ cells per well in complete RPMI 1640 media with 50 ng/ml recombinant mouse IL-2. At 24 h postseeding, OT1 T cells were treated with 2 μM MK2206 or vehicle control (sterile H₂O) for 45 min and visualized on a Zeiss Axio Observer Z1 microscope at 63× magnification. Foxo1-GFP fluorescence intensity in the nuclear and cytoplasmic compartments was measured using FIJI and the N/C ratio of Foxo1-GFP was calculated on a per-cell basis.

OT1 T cells were transduced with Foxo1-GFP variants as described above. At 48 h posttransduction, transduced OT1 T cells were centrifuged in a Ficoll gradient to remove dead cells, treated with 2 μM MK2206 or vehicle control for 45 min at 37°C, and lysed using an NE-PER Nuclear and Cytoplasmic Extraction Kit (no. 78833; Thermo Fisher Scientific). Immunoblotting was performed according to standard protocols. The following Abs were purchased from Cell Signaling Technologies: Foxo1 (clones C29H4 catalog no. 2880 and D7C1H catalog no. 14952), vinculin (clone E1E9V, catalog no. 13901), and lamin A/C (clone 4C11, catalog no. 4777). Immunoblots were visualized on a FluorChem Q Multi Image III analyzer (ProteinSimple), and protein band intensity was measured using α View FluorChem Q software (ProteinSimple). The N/C ratio of Foxo1-GFP was calculated using the protein band intensity of Foxo1-GFP in the nuclear lysate fraction versus the protein band intensity of Foxo1-GFP in the cytoplasmic lysate fraction.

Foxo1^fl/fl, granzyme B–Cre, CD45.1, OT1 TCR transgenic (Foxo1–conditional knockout [cKO]–OT1) mice were generated by crossing granzyme B–Cre, CD45.1, OT1 mice provided by E. Usherwood with Foxo1^fl/fl mice obtained from The Jackson Laboratory (no. 024756) and originally generated by R. DePinho (26). Foxo1-cKO-OT1 mice were backcrossed to B6/CD45.1/OT1 mice for six generations to obtain a background of >95% WT B6. Foxo1^fl/fl, distal Lck-Cre, P14 TCR transgenic (Foxo1-cKO-P14) mice were provided by S. Hedrick (6). Whenever possible, age- and sex-matched controls were used for experiments. Male and female Foxo1-cKO cohoused mice were used between 6 and 12 wk of age for all experiments. Where indicated, WT B6 or B6/CD45.1 mice obtained from The Jackson Laboratory or Charles River Laboratories were used as recipients of T cell adoptive transfers. All recipient mice were used between 7 and 10 wk of age. All mice were cared for according to the guidelines of Dartmouth College, and the Institutional Animal Care and Use Committee approved all experimental protocols.

Retroviruses encoding Foxo1-GFP variants were packaged in 293T cells. Whole splenocytes from Foxo1-cKO-OT1 (OT1 TCR transgenic) mice were cultured in complete RPMI 1640 media with 0.4 mg/ml chicken OVA (no. A5503; MilliporeSigma). At 24–32 h postactivation, Foxo1-cKO-OT1 T cells were transduced with Foxo1 retroviral supernatants containing 8 μg/ml polybrene and centrifuged at 1500 × g for 1.5 h. After centrifugation, Foxo1-cKO-OT1 T cells were rested at 37°C for 1 h before replacing the viral supernatant with fresh complete RPMI 1640 media. At 24 h posttransduction, Foxo1-cKO-OT1 T cells were cultured in fresh complete RPMI 1640 with the addition of either 50 ng/ml recombinant mouse IL-2 (no. 575406; BioLegend) or 50 ng/ml recombinant mouse IL-15 (no. 566302; BioLegend) to polarize toward effector-like or memory-like phenotypes, respectively. T cell cultures were passaged every 48 h with fresh complete RPMI 1640 and cytokines. After 4 d, rmIL-2 and rmIL-15 polarized Foxo1-cKO-OT1 T cells were collected and processed for flow cytometry and quantitative RT-PCR (qRT-PCR).

Retroviruses encoding Foxo1-GFP variants were packaged in 293T cells. CD8 T cells were purified from Foxo1-cKO-P14 (P14 TCR transgenic) mice using an EasySep Mouse CD8⁺ T Cell Isolation Kit (no. 19853; STEMCELL Technologies). Purified Foxo1-cKO-P14 T cells were activated in complete RPMI 1640 media with Dynabeads Mouse T-Activator CD3/CD28 beads (no. 11456D; Thermo Fisher Scientific). At 24–32 h postactivation, Foxo1-cKO-P14 T cells were transduced with Foxo1 retroviral supernatants containing 8 μg/ml polybrene and centrifuged at 1,500 × g for 1.5 h. After centrifugation, Foxo1-cKO-P14 T cells were rested at 37°C for 1 h, dissociated from magnetic DynaBeads, and reconstituted in ×1 PBS without calcium and magnesium. A total of 5 × 10⁴ transduced Foxo1-cKO-P14 T cells were adoptively transferred by i.v. injection into CD45.1 recipient mice. On the same day as adoptive T cell transfers, 5 ml of brain heart infusion broth (no. 53286; MilliporeSigma) was inoculated with gp33-expressing L. monocytogenes (Lm33; provided by J. Obar and originally generated by H. Shen) (27) and incubated overnight at 37°C, 250 rpm. After overnight culture, the Lm33 was added to 45 ml of additional brain heart infusion broth and cultured until reaching an OD600 of 0.797. Five milliliters of OD 0.797 Lm33 was centrifuged at 16,000 × g for 15 min and reconstituted in 1 ml of 1× PBS without calcium and magnesium for a concentration of 5 × 10⁹ CFU/ml. The Lm33 was further diluted in 1× PBS to a final concentration of 5 × 10⁵ CFU/ml, and mice were infected by i.v. injection with 5 × 10⁴ CFU of Lm33. Spleens were harvested at 7, 14, and 40 d postinfection and processed for flow cytometry.

Spleens were homogenized in complete RPMI 1640 media to generate single-cell suspensions. RBCs were lysed using RBC Lysis Buffer (no. 420301; BioLegend). Alternatively, CD8 T cells were used directly from in vitro cultures. Whole splenocytes or CD8 T cells were stained with Zombie Viability Dye (BioLegend) in 1× PBS without calcium and magnesium according to the manufacturer’s protocol. Cells were blocked using anti-CD16 and anti-CD32 Abs (no. BE0307; Bio X Cell) and stained with fluorochrome-conjugated Abs diluted in 1× PBS without calcium and magnesium containing 2% FBS and 1 mM EDTA. The following Abs and dyes were obtained from BioLegend, unless otherwise indicated: CD8 (clone 53-6.7, catalog no. 100734 and 100722), CD62L (clone MEL-14, catalog no. 104436), KLRG1 (clone 2F1/KLRG1, catalog no. 138414 and 138416), CD127 (clone A7R34, catalog no. 135010), CD45.2 (clone 104, catalog no. 109837 and 109815), P2RX7 (clone 1F11, catalog no. 148705), Zombie Viability Dyes (catalog no. 423105 and 423113), CellTrace Violet (catalog no. C34557; Thermo Fisher Scientific), and annexin V (catalog no. 640941 and 640908). All flow cytometry data were analyzed using an MACSQuant Analyzer 10 (Miltenyi Biotec) and FlowJo software.

Foxo1-cKO-OT1 T cells were transduced with Foxo1 variants and polarized to memory-like T cells in IL-15 cultures, as described above. Transduced T cells were cultured in IL-15–replete media for 3 d, stained with Zombie Violet and annexin V APC live/dead stains, and further processed for flow cytometry. For proliferation studies, transduced T cells were stained with CellTrace Violet (no. C34557; Thermo Fisher Scientific) and cultured in IL-15–replete media prior to further processing for flow cytometry.

T cells were sorted by FACS into Foxo1-GFP⁺ and Foxo1-GFP⁻ populations. RNA was isolated using TRIzol Reagent (no. 15596026; Thermo Fisher Scientific) according to the manufacturer’s protocol. Isolated RNA was reverse transcribed using qScript cDNA SuperMix (no. 95048; Quantabio) according to the manufacturer’s protocol. Transcripts were quantified using PerfeCTa SYBR Green SuperMix for iQ (no. 95054; Quantabio) and run in duplicate on a CFX96 Touch Real-Time PCR Detection System (Bio-Rad Laboratories). Relative transcript quantities were calculated by the ΔΔCq method and normalized to β-actin/Actb using CFX Maestro software (Bio-Rad Laboratories). The following primer sequences were used as published by Kim et al. (1) or obtained from PrimerBank (http://pga.mgh.harvard.edu/primerbank/): Actb (5′-TTGCTGACAGGATGCAGAAG-3′, 5′-ACATCTGCTGGAAGGTGGAC-3′), Bcl2 (5′-ATAACCGGGAGATCGTGATG-3′, 5′-CAGGCTGGAAGGAGAAGATG-3′), Ccr7 (5′-AAAGCACAGCCTTCCTGTGT-3′, 5′-GACCTCATCTTGGCAGAAGC-3′), Foxo1 (5′-CCGGAGTTTAACCAGTCCAA-3′, 5′-TGCTCATAAAGTCGGTGCTG-3′), Il7r (5′-TGGCTCTGGGTAGAGCTTTC-3′, 5′-GTGGCACCAGAAGGAGTGAT-3′), Sell (5′-CTCGAGGAACATCCTGAAGC-3′, 5′-AGCATTTTCCCAGTTCATGG-3′), Tcf7 (5′-CAATCTGCTCATGCCCTACC-3′, 5′-CTTGCTTCTGGCTGATGTCC-3′), Klf2 (5′-CTCAGCGAGCCTATCTTGCC-3′, 5′-CACGTTGTTTAGGTCCTCATCC-3′), and P2rx7 (5′-AGACAAACAAAGTCACCCGG-3′, 5′-GCTCACCAAAGCAAAGCTAATG-3′).

Live, CD8⁺, Foxo1-GFP⁺ T cells from IL-15 cultures were sorted using an FACSAria II. Total RNA was purified using an RNeasy Mini Kit (no. 74134/74136; QIAGEN) according to the manufacturer’s protocol. The 3′ mRNA libraries were prepared using TruSeq Library Prep Kits (Illumina) according to the manufacturer’s protocol and sequenced using a NextSeq 500 sequencer (Illumina). Reads were quality trimmed using CLC Genomics Workbench Tool (v11.0), and sequences were mapped to the Mus musculus GRCm38/mm10 reference genome using default parameters.

Significant differentially expressed genes (DEGs; false discovery rate [FDR] < 5%) were identified between SD-Foxo1 and all other variants using DESeq2 (28). DEGs and Foxo1-variant samples were clustered using Pearson uncentered correlation with average linkage. The heatmap color scale was generated based on log2 transformed transcripts per million counts +1 and median centered on a per gene basis.

For pathway analysis, significant DEGs (FDR <5%) were identified between SD-Foxo1 and all other variants using DESeq2. Significant DEGs were annotated with Gene Ontology (GO) functional terms via g:Profiler (29), and GO terms with p < 0.05 corrected for multiple testing via default g:SCS method were treated as significant. Pathways with an FDR <0.05 were considered significant. To compare pathway enrichment across cell types, we compiled a list of biologically relevant pathways that were significantly different between SD-Foxo1 and at least one other Foxo1 variant and determined their FDR-corrected p values and direction of regulation in all Foxo1 variants. The direction of regulation for each pathway was based on the mean fold change in gene expression relative to SD-Foxo1 for all gene hits within the specified pathway. Visualizations of these data were created using ggplot2 (v. 3.2.1).

All statistical analyses were performed using Prism 7 or 8 software (GraphPad). Where noted, data were analyzed with two-tailed, unpaired Welch or Student t tests or one- or two-way ANOVA with Tukey post hoc tests to determine whether the means of two or more groups were significantly different. A p value <0.05 was considered to be statistically significant.

In neurons, Foxo1 nuclear localization is promoted by Mst1 phosphorylation and inhibited by Akt phosphorylation via control of Foxo1 binding to 14-3-3 cytosolic adapter proteins (7, 11, 12). To determine the effect of Mst1 on Foxo1 subcellular localization in T cells, we generated a series of GFP-tagged Foxo1 variants. The phospho-null (SA) and phosphomimetic (SD) Foxo1 variants harbor serine 209 to alanine or aspartic acid substitutions to prevent or mimic Mst1 phosphorylation, respectively (Fig. 1A). GFP-tagged WT-, SA-, and SD-Foxo1 were retrovirally introduced into OT1 TCR transgenic T cells under T cell activating conditions that result in robust Akt activity. Fluorescence microscopy revealed predominantly cytosolic localization of all three variants, including SD-Foxo1, contrary to expectations based on the neuronal studies (Fig. 1B, 1C). Conversely, an Akt-insensitive ADA-Foxo1 variant had constitutively nuclear localization, as previously reported (Fig. 1A–C) (30). When Akt activation was pharmacologically inhibited with MK2206, WT- and SA-Foxo1 localized to the nucleus, consistent with release from 14-3-3. However, SD-Foxo1 remained in the cytosol despite Akt inhibition, indicating that an Akt-independent mechanism controlled its cytosolic localization (Fig. 1B, 1C).

To validate our findings, T cells expressing the Foxo1 variants were lysed, and the quantity of Foxo1 protein was assessed in the nuclear and cytoplasmic fractions. Western blot analysis showed that retrovirally introduced Foxo1 variants were expressed in the bulk T cell population at levels comparable to endogenous Foxo1 (Foxo1^E). In agreement with our fluorescence microscopy data, SD-Foxo1 was resistant to nuclear localization upon Akt inhibition with an ∼4-fold reduction in N/C ratio compared with WT- and SA-Foxo1 (Fig. 1D, 1E). Nevertheless, a small but detectable fraction of SD-Foxo1 remained in the nucleus (Fig. 1B–E). Importantly, Foxo1^E translocated into the nucleus upon Akt inhibition, which confirmed that MK2206 treatment had the expected effect on Foxo1 trafficking (Fig. 1F) (7, 8). Thus, SD-Foxo1 is resistant to nuclear localization, suggesting that, unlike in neurons (10, 12), S209 phosphorylation promotes cytoplasmic localization of Foxo1 in CD8 T cells.

We next sought to determine whether SD-Foxo1 localization was due to impaired nuclear import or enhanced nuclear export. To distinguish between these possibilities, we generated an additional Foxo1 variant that combined the SD mutation with the Akt phospho-null ADA mutations (QM, QM-Foxo1), which would allow us to genetically isolate the effect of SD from that of Akt phospho-residues (Fig. 2A). Expression of WT- and SD-Foxo1 in 293T cells cultured in serum replete media showed similar subcellular localization to that observed in OT1 T cells, indicating that regulation of Foxo1 at residue 209 is conserved between these cell types (Fig. 2B, 2C, 2E).

Although Akt inhibition released WT-Foxo1 from 14-3-3 cytosolic adapters, allowing nuclear localization of WT-Foxo1, SD-Foxo1 and QM-Foxo1 remained cytoplasmic with a significantly reduced N/C ratio compared with WT-Foxo1 (Fig. 2D, 2E). This reinforced the observation that the SD mutation promotes cytoplasmic localization of Foxo1 independent of Akt phosphorylation. Treatment with the nuclear export inhibitor leptomycin B alone had little impact on WT- or SD-Foxo1 localization because both were already predominantly cytoplasmic. However, leptomycin B was sufficient to promote nuclear accumulation of QM-Foxo1 and a significantly higher N/C ratio (Fig. 2B–E). The capacity of leptomycin B to promote nuclear accumulation of the Akt-insensitive QM mutant, but not the Akt-sensitive SD mutant, strongly suggests that in the absence of Akt phosphorylation, mimicking S209 phosphorylation promotes cytoplasmic localization by enhancing Foxo1 nuclear export. This conclusion is further supported by the observation that treatment with MK2206 and leptomycin B together caused nuclear accumulation of all three Foxo1 variants (Fig. 2B–E).

Together, these data suggest that phosphorylation of Foxo1 at S209 promotes its cytoplasmic localization in primary murine CD8 T cells in a nonredundant parallel pathway to Akt. Our data in 293T cells also suggest that the cytoplasmic localization of SD-Foxo1 is due to enhanced nuclear export rather than impaired nuclear translocation.

Foxo1 target genes including Klf2, Sell, and Ccr7 are critical for the development and maintenance of memory CD8 T cells (1, 2, 4, 31). To determine whether Foxo1 variants are capable of rescuing Foxo1 target gene transcription, we generated Foxo1 cKO mice in which a floxed Foxo1 allele is deleted in activated CD8 T cells expressing a Granzyme B promoter–driven Cre recombinase (GzmB-Cre). OT1 transgenic TCR expression allowed tracking of the Ag specific response. These mice and their CD8 T cells are hereafter referred to as Foxo1-cKO-OT1. Foxo1-cKO-OT1 cells were reconstituted with WT (WT^R T cells)–, SA (SA^R T cells)–, or SD-Foxo1 (SD^R T cells) and cultured in IL-15–containing media (Fig. 3A). Culturing CD8 T cells in vitro in the presence of IL-15 generates memory-like T cells that express Ccr7 and Sell, which encodes CD62L (32, 33). Under these conditions, both WT^R and SA^R T cells expressed CD62L at equivalent levels to control Cre-negative T cells that retain Foxo1^E (Fig. 3B, 3C). In contrast, SD^R T cells had significantly lower surface expression of CD62L, which was comparable to the level observed in Foxo1-cKO-OT1 T cells (cKO; Fig. 3B, 3C).

Additionally, the expression of Klf2, Sell, and Ccr7 transcripts was significantly reduced in SD^R T cells compared with WT^R T cells (Fig. 3D). SA^R T cells expressed Sell and Ccr7 at levels comparable to WT^R T cells and significantly higher than SD^R T cells but, interestingly, were unable to induce expression of Klf2 (Fig. 3D). The observed differences in target gene transactivation were not due to differential expression of the Foxo1 variants.

In contrast to IL-15, culturing CD8 T cells in vitro in the presence of IL-2 generates effector-like T cells that express much lower levels of Foxo1 target genes (32, 33). Under these conditions, WT^R T cells expressed CD62L/Sell, Klf2, and Ccr7 at significantly higher levels than Cre-negative control T cells, whereas SD^R T cells failed to express these genes (Supplemental Fig. 1A–D). Interestingly, CD62L/Sell expression in SA^R T cells was augmented above WT^R T cells, but SA^R T cells did not express Klf2 or Ccr7 in IL-2 culture conditions (Supplemental Fig. 1B–D).

Together, these data suggest that phosphorylation of Foxo1 at S209 impairs the ability of Foxo1 to transactivate canonical target genes in CD8 T cells. Although these results are consistent with nuclear exclusion of SD-Foxo1, they do not exclude other negative regulatory mechanisms such as impaired DNA binding (9).

Foxo1 can promote the transcription of alternative target genes in the absence of direct DNA binding (34, 35). We thus wanted to identify the transcriptome-wide effect of Foxo1 phosphorylation at residue S209. For this purpose, we performed bulk 3′ RNA sequencing (RNAseq) on WT^R, SA^R, SD^R, and Foxo1-cKO T cells from IL-15 cultures and identified DEGs (Fig 4A). To determine which functional pathways were differentially regulated by the Foxo1 variants, the significant DEGs between SD-Foxo1 and all other Foxo1 variants were assigned GO functional terms. This analysis revealed that several GO pathways were differentially expressed by SD^R cells compared with cKO, WT^R, or SA^R cells (Fig. 4B). Interestingly, regulation of the GO pathways “regulation of translation,” “regulation of protein deacetylation,” and “regulation of cellular amide metabolic process” was significantly different between SD^R and cKO T cells, suggesting functional differences between SD-Foxo1 and Foxo1 deficiency (Fig. 4B). Additionally, DEGs assigned to the GO functional term “positive regulation of immune system process” had significantly higher expression in SA^R T cells as compared with SD^R T cells. This pathway includes the canonical Foxo1 target genes Il7r and Sell, which are essential for CD8 T cell memory establishment (1, 2) (Fig. 4C). In contrast, DEGs assigned to the GO functional term “leukocyte mediated cytotoxicity” are significantly upregulated in SD^R T cells as compared with WT^R and SA^R T cells and include Gzmm, which is also expressed at significantly higher levels in SD^R T cells as compared with cKO T cells (Fig. 4D). Surprisingly, we observed higher mRNA expression of the purinergic receptor P2RX7 (P2rx7) that is required for the survival of memory CD8 T cells (36) in cKO and SD^R cells compared with SA^R and WT cells. Identification of P2rx7 prompted us to independently analyze known Foxo1 target genes that also regulate cell survival. Among these, Bcl2l11, which encodes the proapoptotic protein Bim, had the highest expression in SA^R T cells and the lowest expression in SD^R T cells (Fig. 4E). Conversely, expression of antiapoptotic Bcl2 was not significantly different between Foxo1 variants (data not shown).

Together, these data suggest that phosphorylation of S209 simultaneously inhibits expression of canonical Foxo1 target genes and supports expression of alternative genes and functional pathways, including those involved in CD8 effector function and cell survival. Furthermore, differentially regulated pathways identified between Foxo1-cKO T cells and SD-Foxo1 T cells suggest that SD-Foxo1 is functionally distinct from knockout of Foxo1.

Foxo transcription factors exert distinct effects on cell survival in a cell type and context-dependent manner. Foxo1 is critical for the survival of naive CD4 and CD8 T cells as well as memory CD8 T cells (1, 2, 31, 37), effects that are partly attributable to induction of CD127 expression and upregulation of Bcl-2. The surprising observation that SD^R T cells express high levels of P2rx7 led us to hypothesize that this variant may promote a distinct mode of CD8 T cell survival.

We first sought to independently validate the augmented P2rx7 expression observed in SD^R T cells in our RNAseq experiments. We measured the transcript and protein levels of P2rx7/P2RX7 using the in vitro IL-15 culture system. Activated Foxo1-cKO-OT1 cells were reconstituted with Foxo1 variants and allowed to grow in IL-15 containing media for 3 d, followed by P2rx7/P2RX7 expression analysis by qRT-PCR (Fig. 5A) and flow cytometry (Fig. 5B). Similar to the RNAseq data, P2rx7 transcript and protein expression were higher in SD-Foxo1 cells as compared with SA-Foxo1 and WT-Foxo1.

We next sought to determine whether SD-Foxo1 confers a survival advantage to CD8 T cells as compared with WT- and SA-Foxo1. Three days after IL-15 culture, annexin V and Zombie Violet staining revealed a significant increase in the survival of SD^R memory-like T cells compared with SA^R (Fig. 5C, 5D). Moreover, the survival of SA^R T cells was significantly lower than both WT^R and SD^R T cells despite their robust expression of canonical Foxo1 target genes. Surprisingly, SD^R T cells also had significantly increased survival compared with Foxo1-cKO-OT1 T cells (Fig. 5C, 5D) despite similar expression of P2rx7/P2RX7 (Fig. 5A, 5B) and Bcl2l11 (Fig. 4E). Stratifying CD8 T cells by their level of Foxo1-GFP expression revealed that the survival differences caused by Foxo1 variants were maintained over a wide range of Foxo1-GFP expression (Fig. 5E, 5F). Thus, these differences in survival were independent of corresponding differences in Foxo1 variant expression. In addition, tracking CD8 T cell proliferation with CellTrace Violet showed no differences in proliferation between Foxo1 variants, demonstrating that differential proliferation did not contribute to the observed differences in the viability of CD8 T cells (Fig. 5G, 5H).

Together, these data suggest that phosphorylation at residue S209 promotes Foxo1-dependent CD8 T cell survival through a distinct biological mechanism. The differences in survival between Foxo1 variants may involve P2RX7 and/or Bim. However, this survival advantage is unlikely to be solely dependent on P2rx7 or Bcl2l11 expression because expression of these genes is similar between SD^R and Foxo1-cKO T cells despite significantly higher survival of the former.

Foxo1 is crucial for the generation and maintenance of CD8 memory T cells. Given the failure of SD-Foxo1 to transactivate canonical target genes in vitro, we hypothesized that phosphorylation at residue S209 might be antagonistic to CD8 T cell memory development in vivo. To investigate this, we used Foxo1 cKO mice that express distal Lck-Cre and the P14 transgenic TCR that recognizes lymphocytic choriomeningitis virus gp33 (gp33) (2, 4) and (Foxo1-cKO-P14). Foxo1-cKO-P14 (cKO) T cells reconstituted with Foxo1 variants (WT^R, SA^R, and SD^R) ex vivo were adoptively transferred into congenically marked recipients that were subsequently infected with Lm33 (Fig. 6A).

Il7r is an Foxo1 target gene that encodes the IL-7R α-chain (IL-7Rα/CD127) and is expressed in both memory precursor effector T cells (MPECs) and long-lived memory T cells. At the peak of CD8 T cell expansion, 7 d postinfection with Lm33, WT^R, and SA^R T cells in the spleen had comparable percentages of CD127⁺ cells, whereas SA^R T cells produced a significantly higher percentage of CD127⁺/KLRG1⁻ MPECs (Fig. 6B). In contrast, SD^R T cells had a significant deficit in CD127 expression and MPEC formation compared with WT^R and SA^R T cells and, instead, had increased percentages of CD127⁻/KLRG1⁺ short-lived effector cells (SLECs; Fig. 6B). This was consistent with the RNAseq analyses, showing enhanced expression of cytotoxic genes and pathways in SD^R T cells (Fig. 4). Surprisingly, the generation of CD127⁺ cells, including MPECs, was also significantly impaired in SD^R T cells compared with cKO T cells, whereas SD^R T cells produced a significantly higher percentage of SLECs.

During the contraction phase, 14 d after Lm33 infection, SD^R T cells remained unable to effectively express CD127. Only 5% of SD^R T cells expressed CD127, whereas ∼35% of WT^R, 55% of SA^R, and 16% of cKO T cells were CD127⁺ (Fig. 6B). Importantly, this ∼1.6-fold increase in the percentage of CD127⁺ SA^R T cells compared with WT^R T cells suggests that the Foxo1 S209 residue is actively phosphorylated during the contraction phase. These differences corresponded with a significant reduction in MPECs among SD^R T cells as compared with WT^R, SA^R, and cKO T cells. Conversely, SLECs remained significantly enriched among SD^R T cells as compared with WT^R, SA^R, and cKO T cells (Fig. 6B).

Analysis of memory T cells in the spleen at 40 d after Lm33 infection revealed that SD^R T cells had a persistent deficit in CD127 expression as compared with WT^R, SA^R, and cKO T cells (Fig. 6C). Although the percentage of CD127⁺ cKO T cells remained significantly lower than that observed in WT^R and SA^R T cells, these cells did partially recover CD127 expression in an Foxo1-independent manner, which has been shown by others (2). Surprisingly, SD^R T cells failed to recover CD127 expression to the same extent as cKO T cells (Fig. 6C). Despite persistent differences in CD127 expression, both SA^R and SD^R T cells failed to generate long-lived CD62L⁺/CD127⁺ TCM, displaying significantly decreased percentages of TCM compared with their WT-Foxo1 counterparts (Fig. 6C). This suggests that dynamic regulation of Foxo1 by S209 phosphorylation may be required for commitment to the TCM lineage with both phosphorylated and unphosphorylated forms of Foxo1 serving complementary roles. However, because SA-Foxo1 induces comparable or augmented expression of surface CD62L and Sell transcripts compared with WT-Foxo1 in vitro (Figs. 3C, 3D, 4C), the lack of CD62L expression by SA^R T cells in vivo is unlikely to be due to an intrinsic deficit in the ability of SA-Foxo1 to transactivate target genes. In contrast to TCM formation, SA^R T cells generated a significantly higher percentage of CD62L⁻/CD127⁺ effector memory T cells than WT^R, SD^R, and cKO T cells (Fig. 6C).

The surprising deficit in CD127⁺ memory precursor and long-lived memory formation in SD^R T cells as compared with cKO T cells led us to hypothesize that SD-Foxo1 actively suppresses CD127 expression. To investigate this possibility, we stratified day 14 WT^R, SA^R, and SD^R T cells by their level of Foxo1-GFP expression. Interestingly, this revealed that although CD127 mean fluorescence intensity (MFI) increased in proportion to the level of WT- and SA-Foxo1, CD127 MFI decreased in proportion to the level of SD-Foxo1 (Fig. 6D; dotted line represents the mean CD127 MFI in cKO T cells). Together, these data suggest that rather than functioning as a simple null mutant, SD-Foxo1 actively suppresses at least one canonical target gene in vivo. This phenomenon may underlie the distinct capacities of SD^R and cKO T cells to form MPECs, SLECs, and long-lived memory cells.

We next sought to determine how these Foxo1 variants impacted the numbers and survival of CD8 T cells throughout their response to Lm33 infection. At the peak of expansion, there were significantly more SD^R T cells as compared with WT^R and SA^R T cells (Fig. 6E). Although the total number of SD^R T cells remained elevated at 14 and 40 d postinfection, the number of these cells was more similar to that of WT^R and SA^R T cells. Interestingly, there were significantly fewer deaths of SD^R SLECs as compared with cKO SLECs at both 7 and 14 d postinfection (Fig. 6F). The enhanced survival of SD^R SLECs likely contributes to the increased percentage of these cells in vivo and agrees with the enhanced survival of SD^R T cells in the in vitro experiments (Fig. 5). However, this survival advantage appears to be limited to SLECs in vivo as other subsets of SD^R T cells had comparable survival to their WT^R and SA^R counterparts (data not shown).

Altogether, these results suggest that phosphorylation of Foxo1 at S209 represses CD8 memory T cell formation in vivo by actively inhibiting the expression of CD127. Our data also suggest that S209 phosphorylation simultaneously promotes the preferential survival of SLECs.

Because Foxo transcription factors were first identified as Mst1 substrates in neurons (10, 12), phosphorylation of Foxo1 by Mst1 has been widely viewed as an activating event (38–41). Deficiency of Mst1 or Foxo1 in T cells results in partially overlapping phenotypes, including decreased numbers of naive T cells in secondary lymphoid organs and increased T cell susceptibility to apoptosis (31, 37, 42–44). However, in contrast to the profound deficiency of Klf2, CD62L, CCR7, and CD127 expression in Foxo1 cKO T cells (1, 2, 31, 37), loss of Mst1 in T cells does not diminish expression of these Foxo1 targets (42, 43, 45). To date, the extent to which these distinct and overlapping features are mediated through an Mst1-Foxo1 transcriptional axis or by the independent functions of either protein has not been directly examined in T cells. In this study, we performed a comprehensive analysis to determine how the Mst1 target residue S209 impacts Foxo1 transcription in CD8 T cells using phospho-null SA-Foxo1 and phosphomimetic SD-Foxo1 variants. These studies yielded several unexpected and, to our knowledge, novel findings: we found that phosphomimetic SD-Foxo1 fails to induce the canonical Foxo1 transcription program required for CD8 memory T cell development and maintenance (1, 2). Instead, SD-Foxo1 was associated with a distinct transcriptional signature from that of WT-Foxo1, SA-Foxo1, and Foxo1-cKO; furthermore, SD-Foxo1 actively suppressed the expression of CD127 during the CD8 T cell response to infection in vivo. Importantly, SD-Foxo1 promoted CD8 T cell survival in vitro and SLEC survival in vivo despite the absence of canonical target gene expression. Surprisingly, we found that SD-Foxo1 is resistant to nuclear localization in CD8 T cells, whereas SA-Foxo1 retains typical Akt-dependent nuclear transport.

Our in vivo studies demonstrate that SD^R T cells have a profound deficit in their ability to generate memory precursors and long-lived memory cells, even compared with Foxo1-cKO T cells. This corresponds with dose-dependent suppression of CD127 expression by SD-Foxo1 and an increased percentage and survival of SD^R SLECs compared with their cKO counterparts. These findings suggest that phosphorylation of Foxo1 at S209 actively promotes the effector response at the expense of memory formation. Surprisingly, SA^R T cells were also defective in TCM generation, despite their robust expression of CD127 and capacity to form memory precursors and effector memory T cells. This observation suggests that—although phosphorylation of Foxo1 at S209 may be antagonistic to memory precursor formation—both phosphorylated and unphosphorylated Foxo1 are necessary for the expression of CD62L and generation of TCM in vivo. Whether this reflects distinct temporal regulation of Foxo1 phosphorylation or direct cooperation between the phosphorylated and unphosphorylated forms of Foxo1 is an important avenue for future investigation. The differences we observed in the ability of SA-Foxo1 to induce Klf2 and Sell/CD62L expression in vitro and in vivo are also intriguing. However, because these analyses were performed at 6 d after T cell activation versus 40 d after Lm33 infection, respectively, and because Klf2 expression was not measured in vivo, it is not possible to directly relate these findings based on the available data.

Together, our RNAseq analysis, survival data, and in vivo studies highlight novel SD-Foxo1 functions that cannot be explained by the simple loss of function observed in Foxo1-cKO T cells. RNAseq analysis revealed that SD^R T cells had a partially distinct transcriptional signature from Foxo1-cKO T cells. Furthermore, examining the relationship between SD-Foxo1 expression and CD127 expression in vivo revealed that SD-Foxo1 actively suppresses CD127 expression. We also found that SD-Foxo1 enhances the survival of IL-15–cultured CD8 T cells in vitro and SLECs in vivo. This was surprising given that SD^R T cells did not express canonical Foxo1 targets. We identified high P2RX7 expression and low Bcl2l11 expression in SD^R T cells, which may contribute to the survival advantage of these cells. However, the increased survival of SD^R T cells compared with cKO T cells both in vitro and in vivo suggests that other survival mechanisms are also important. These findings may be partly explained by our fluorescence microscopy and Western blot studies quantifying Foxo1 subcellular localization, which show a residual pool of SD-Foxo1 in the nucleus despite its predominantly cytoplasmic localization. Thus, it is possible that the interaction of SD-Foxo1 with other transcription factors in the nucleus can promote expression of noncanonical target genes (46–49). In addition, residual SD-Foxo1 in the nucleus may indirectly affect gene expression by inhibiting other transcription factors, similar to Foxo1-mediated repression of T-bet (6, 50). Together, these data suggest that phosphorylation of Foxo1 at S209 does not simply inactivate Foxo1 but instead diverts its activity toward distinct cellular functions.

The similarity we identified between WT- and phospho-null SA-Foxo1 nuclear trafficking agrees with work by Du and colleagues (18), which suggests that Foxo1 localization is similar in Mst1-sufficient and Mst1-deficient murine CD4 T cells. However, this finding, in combination with the unexpected observation that phosphomimetic SD-Foxo1 is resistant to nuclear localization, is in marked contrast to previous findings in neurons. Those studies indicated that Mst1 promotes the nuclear localization of Foxo1 and its close mammalian paralog Foxo3 by causing their dissociation from 14-3-3 adaptors (10, 12). Our finding that phosphomimetic SD-Foxo1, and its Akt-insensitive counterpart QM-Foxo1, have enhanced nuclear export through an Akt-independent mechanism does not exclude the ability of Mst1 to cause Foxo1 dissociation from 14-3-3. However, it does suggest that in Ag- and cytokine-stimulated CD8 T cells, the cumulative effect of Foxo1 phosphorylation at S209 is increased cytoplasmic localization. Thus, our data suggest a unique role for S209 phosphorylation in regulating the N/C shuttling of Foxo1 in CD8 T cells. Whether this directly causes impaired transactivation of Foxo1 target genes or is a consequence of altered DNA targeting is an important avenue for future investigation.

To our knowledge, Mst1 is the only known kinase that phosphorylates Foxo1 at S209. Nevertheless, we cannot exclude the possibility that this residue is a substrate for other kinases, and the discovery of additional kinases that phosphorylate Foxo1 at this site would be a significant finding. Complementary studies are warranted to further elucidate the findings described in this study. In particular, the marked difference we observed between WT- and SA-Foxo1 transactivation of CD127 during the CD8 T cell contraction phase makes this an exciting physiologic context for future investigation.

In summary, our findings suggest that phosphorylation of S209 allows Foxo1 to mediate unique functions in different phases of the CD8 T cell response. Future studies may support the role of Mst1 as neither a simple activator nor repressor of Foxo1 or identify additional kinases that regulate Foxo1 through the S209 residue. Our work suggests that dynamic modulation of Foxo1 activity via phosphorylation at S209 is critical for balancing CD8 T cell differentiation and survival with the biological context.

We thank S. M. Hedrick (University of California, San Diego) for Foxo1-cKO-P14 mice, G. Ward in the DartLab FACS core facility for cell sorting, the Dartmouth microscopy core facility for training in fluorescence microscopy, F. W. Kolling IV for RNAseq, and H. Nguyen and other members of the Huang laboratory for discussions and technical assistance.

The authors have no financial conflicts of interest.