Cellular Size as a Means of Tracking mTOR Activity and Cell Fate of CD4+ T Cells upon Antigen Recognition

mTOR is a central integrator of metabolic and immunological stimuli, dictating immune cell activation, proliferation and differentiation. In this study, we demonstrate that within a clonal population of activated T cells, there exist both mTORhi and mTORlo cells exhibiting highly divergent metabolic and immunologic functions. By taking advantage of the role of mTOR activation in controlling cellular size, we demonstrate that upon antigen recognition, mTORhi CD4+ T cells are destined to become highly glycolytic effector cells. Conversely, mTORlo T cells preferentially develop into long-lived cells that express high levels of Bcl-2, CD25, and CD62L. Furthermore, mTORlo T cells have a greater propensity to differentiate into suppressive Foxp3+ T regulatory cells, and this paradigm was also observed in human CD4+ T cells. Overall, these studies provide the opportunity to track the development of effector and memory T cells from naïve precursors, as well as facilitate the interrogation of immunologic and metabolic programs that inform these fates.


Introduction
The evolutionarily conserved serine/threonine kinase mammalian Target of Rapamycin (mTOR) is a central integrator of environment cues and intracellular stress signals, dictating the course of cellular growth, proliferation and differentiation [1]. The mTOR kinase forms two distinct signaling complexes, mTORC1 and mTORC2, with distinct upstream activators and discrete downstream targets [2]. mTORC1, characterized by the scaffolding proteins Raptor and PRAS40 along with mLST8 and Deptor, is canonically activated upon stimulation of PI3-kinase [3]. This occurs in part through the phosphorylation and inactivation of the mTORC1-inhibitory proteins TSC1/2 [2,4,5]. These events result in the phosphorylation of the canonical mTORC1 substrates S6 kinase and 4EBP1, causing enhanced translation of mRNA transcripts with TOP or TOP-like motifs, increased mitochondrial biogenesis, and enhanced expression of the critical metabolic transcription factors Myc and HIF-1α [6][7][8][9]. mTORC1 activity plays a critical role in regulating cell size, particularly in promoting the increase in cell size upon activation [10].
In T cells, the kinase activity of mTOR has been shown to be modulated by numerous immunological stimuli including TCR ligation, co-stimulation/co-inhibition, cytokine/chemokine exposure, and adhesion molecule engagement [11]. Indeed, mTOR has emerged as an important integrator of cues from the immune microenvironment to guide T cell effector and memory differentiation [12,13]. In the case of CD4+ T cells, pharmacological inhibition of mTOR signaling or genetic deletion of the mTOR gene results in a severe defect in the ability of a naïve CD4+ T cell to adopt an effector phenotype following activation [14][15][16][17][18]. Instead, CD4+ T cells activated in the absence of mTOR signaling adopt a default Foxp3+ regulatory cell phenotype [14]. In contrast, studies utilizing CD4+ T cells lacking components of either the mTORC1 or mTORC2 complex have revealed that mTORC1 signaling is required for the development of IFN-gamma producing Th1 cells and IL-17 producing Th17 cells, while mTORC2 is required for the development of IL-4 producing Th2 cells [19,20]. In addition to regulating CD4+ T cell effector differentiation, mTOR has been shown to play a central role in regulating the development of long-lived memory CD8+ T cells [21][22][23].
The data derived from knockout mice and pharmacologic inhibitors suggests that the differential activation of mTOR signaling during an immune response plays an important role in dictating fate decisions for activated T cells. To this end, we sought to identify cells based on their level of mTOR activity and then track their fate. In this study, we demonstrate that cell size can be used as a surrogate indicator of mTOR activity in recently activated T cells, allowing for the isolation of mTOR hi and mTOR lo populations with divergent differentiation, metabolic and survival potential. CD4+ mTOR hi T cells preferentially develop into short-lived, terminally differentiated effector cells with a robust metabolic phenotype and large proliferative capacity. Alternatively, mTOR lo CD4+ T cells are less proliferative, less glycolytic, and demonstrate a long lived phenotype. We also find that enriched within this population of mTOR lo cells, are suppressive Foxp3+ regulatory T cells, and this finding was recapitulated in human CD4+ T cells.

Materials and Methods Animals
Mice were kept in accordance with guidelines of the Johns Hopkins University Institutional Animal Care and Use Committee. 5c.c7 Rag -/and OT-II mice were purchased from Taconic. C57BL/6 mice, CD90.1 mice, CD4 Cre recombinase mice, Foxp3 GFP+ mice, and mice with loxP-flanked Raptor alleles were obtained from Jackson Laboratories. Mice with loxP-flanked Tsc2 alleles were generated by the laboratory of M. Gambello [24]. 5c.c7 Rag -/mice were derived from a B10.A background, while all other WT and OT II+ mice described in this manuscript originate from the C57BL/6 strain.

Proliferation assay
Sorted cells were cultured for 24hrs in complete media supplemented with 1ng/ml IL-2, then equivalent cell numbers were loaded in a 96 well plate and treated with 1μCi 3 H-thymidine for 18hrs. 3 H incorporation was measured using a PerkinElmer 1450 Microbeta counter.

Cell cycle analysis
For the determination of intracellular DNA content, cells were washed with PBS and fixed for 30min at 4°C in ice-cold 70% EtOH. Cells were then washed with PBS, and treated with 50U of RNase 1F (New England Biolabs M0243) for 10minutes at room temperature, and suspended in a 50ug/ml solution of propidium iodide. Cell cycle progression was determined by flow cytometry.

Flow cytometry and cell sorting
All flow cytometry experiments were performed on a BD FACS Calibur or LSR II, and analyzed using FlowJo software. Cell sorting was performed on a BD FACS Aria II.

Suppression Assays
Splenocytes from a WT C57BL/6 mouse were stimulated for 16 hours with 0.1ug/ml anti-CD3 then sorted into mTOR hi and mTOR lo populations and cultured with 1ng/ml IL-2 for 72 hrs. Cells from the mTOR hi and mTOR lo cultures were mixed at a 1:2 ratio with naïve WT CD90.1+ CD4+ T cells stained with CFSE and stimulated with 1ug/ml anti-CD3 and irradiated APCs. The proliferation of the naïve CD4+ CD90.1+ T cells was measured by CFSE dilution by flow cytometry. For the transwell assays, mTOR hi or mTOR lo cells were segregated from the CFSE labeled naïve WT CD90.1+ CD4+ T cells by a Transwell permeable support 0.4uM polycarbonate membrane (Corning). Both the segregated suppressor and responder populations were co-cultured with 1ug/ml anti-CD3 and irradiated APCs.

Human CD4+ T cell experiments
Human samples were obtained through an institutional review board approved protocol. Blood was drawn from a total of 8 healthy volunteers in 3 independent experiments. PBMCs were extracted after Ficoll separation during centrifugation (Ficoll-Paque PLUS, GE Healthcare). CD4+ T cells were purified by magnetic bead separation according to the manufacture's instructions (Human CD4+ T cell enrichment kit, Stem Cell Technologies). Purified CD4+ T cells were stimulated 20 hrs with dynabeads human T-activator anti-CD3/anti-CD28 beads (Gibco). The 15% biggest 'mTOR hi ' and smallest 'mTOR lo ' CD38+, CD4+ T cells were sorted on a BD FACS Aria II. Sorted cells were cultured in media supplemented with 3,000units/ml human IL-2 (provided by the NCI) for 72 hrs. Phenotype of sorted cells was determined by flow cytometry.

Statistical Analysis
Data are presented as means and standard deviation. Statistical results were generated using Graphpad Prism. Comparisons were calculated using unpaired T tests, and results with a pvalue < 0.05 were considered statistically significant: Ã p<0.05, ÃÃ p<0.01, ÃÃÃ p<0.001.

Cell size indicates mTOR activity in recently activated CD4+ T cells
Following TCR stimulation and appropriate co-stimulation, naïve CD4+ T cells dramatically increase in size after 24 hours (Fig. 1A). The cellular processes that regulate cell size are in part controlled by mTOR activity [10,26]. To determine if mTOR regulates the size of activated T cells, we stimulated T cells in the presence of mTOR kinase inhibitor, PP242 [27]. TCR-induced blastogenesis is significantly inhibited by the addition PP242 (Fig. 1A). Furthermore, T cell activation induces mTORC1 activity as indicated by phosphorylation of ribosomal-S6 at S235/236 and S6 Kinase at T389 (Fig. 1B). As expected, phosphorylation of these proteins is diminished with addition of PP242. We consistently find that mTORC1 activity is markedly enriched in the T cells that have increased size (Fig. 1B). The increase in size and mTORC1 activity is best observed after 15 hours of stimulation (S1 Fig.). Thus, there is a direct correlation of cellular size with mTORC1 activity upon activation of a clonal population of naïve T cells.
To further examine the relationship of mTORC1 activity with cellular size, we examined TCR-induced blastogenesis in Raptor deficient T cells (T-Raptor -/-), which have dramatically reduced mTORC1 activity. After 24 hours of stimulation, the vast majority (80%) of the T-Raptor -/-T cells remain small in size compared with~30% of the WT T cells (Fig. 1C). Consistent with the role of mTORC1 in regulating T cell size, the largest cells (labeled 'Big') detected after stimulation of WT T cells display the highest mTORC1 activity, while the smallest cells (labeled 'Small') demonstrate the lowest mTORC1 activity (Fig. 1C). Likewise, WT cells with the highest mTORC1 activity (labeled 'pS6+') are the largest in size (Fig. 1D). Examination of cell size at the 45 th , 25 th and 10 th percentiles reveals a graded effect of mTORC1 activity on cell size, whereby regardless of size percentile, mTORC1 activity is consistently highest in the largest cells (Fig. 1E). The largest difference in mTORC1 activity is most evident in the 10% smallest and largest size percentiles (Fig. 1E). Thus, we chose to utilize the 10% size cut off in all following experiments to demonstrate the most robust differences between activated cellular populations. This idea was validated by assessing size of WT, T-Raptor -/-, and T-TSC2 -/-(T-TSC2 -/-T cells have hyperactive mTORC1 activity) OT-II+ splenocytes after stimulation with OVA 323-339 peptide. Again, we observed that over half of the activated T-Raptor -/-CD4+ T cells are detected in the gate set from the 10% smallest activated WT cells, while less than 3% of T-TSC2 -/are detected in the same gating strategy (Fig. 1F). Based on these results, we propose that cell size can be used as a surrogate indicator of mTORC1 activity in recently stimulated T cells, allowing for easy identification and tracking of both mTOR hi and mTOR lo cells from an initially monoclonal culture of naïve T cells.
We propose that for an individual cell, antigen recognition (Signal 1) will occur in the context of differential mTORC1 activation depending upon accessory signals (Signal 2) from the environment. However, it was critical to demonstrate that the differential activation of mTORC1 was not simply due to lack of antigen recognition. That is, we wanted to ensure that even the cells with decreased mTOR activity still received full antigen receptor engagement.
To this end, WT, T-Raptor -/and T-TSC2 -/-OT-II + splenocytes were stimulated with OVA 323-339 peptide for 24 hours. As expected, this led to differential activation of mTORC1, whereby T-TSC2 -/-T cells have increased activity while T-Raptor -/-T cells have diminished activity compared to WT cells ( Fig. 2A). Notably, there was robust and roughly equivalent upregulation of CD69 and downregulation of CD62L in T cells from all 3 genotypes (Fig. 2B). That is, diminished or enhanced mTORC1 activity did not affect antigen-induced activation. Next, we wanted to demonstrate that such was also the case for the small (mTOR lo ) and big (mTOR hi ) populations of WT cells. Splenocytes from 5c.c7 Rag -/-TCR transgenic mice were stimulated with Pigeon Cytochrome C (PCC) peptide. Upon stimulation, there is robust activation as indicated by CD69 upregulation (Fig. 2C). Starting from this clonotypic population of naïve T cells, we observe that there emerges a heterogeneous population of antigen experienced cells that differ dramatically in size. Importantly, this difference in cell size was not due to a lack of activation as both large and small cells express high levels of the early T cell activation marker CD69, and demonstrate robust downregulation of CD62L relative to naïve T cells (Fig. 2C). We did note, however, that the levels of CD69 and CD62L were slightly decreased in the T-Raptor -/and mTOR lo cells when compared to WT and mTOR hi T cells ( Fig. 2B and 2C). The significance of this finding is unclear. mTOR hi and mTOR lo CD4+ T cells display differential proliferative, metabolic and survival fates In light of the relationship we observed between cell size and mTOR activity in recently activated CD4+ T cells, we sought to track the fate of T cells with differing levels of mTOR activity, as indicated by cell size, following antigen recognition. To this end, naïve 5c.c7 TCR transgenic Rag -/splenocytes were stimulated with PCC peptide for 24 hours. To ensure that only stimulated CD4+ T cell were included in future analysis, cells were stained for CD4 and the mTOR independent early activation marker CD69, then sorted into CD4+CD69+ FSC/SSC "big" mTOR hi and CD4+CD69+ FSC/SSC "small" mTOR lo populations (Fig. 3A). To further confirm the relationship of mTORC1 activity with cell size, sorted cells were immediately assessed for mTORC1 activity. As expected, the largest population (labeled 'Big') cells demonstrate the highest phosphorylated S6 (pS6) expression ( Fig. 3B). At the time of sorting, the process of cellular division had not yet been initiated, with~90% of both the mTOR hi and mTOR lo populations still in G 0 stage of the cell cycle as indicated by propidium iodide staining  Cell size can be used to isolate recently stimulated CD4+ T cells populations with distinct proliferative, metabolic and survival profiles. a) Splenocytes from 5c.c7 Rag -/mice were stimulated with 0.5ug/ml PCC peptide for 24hrs then sorted into CD4+CD69+ populations with FSC/SSC "big" mTOR hi or FSC/SSC "small" mTOR lo profiles. b) mTORC1 activity was assessed from "Big" and "Small" cells immediately after sorting. pS6 MFI is shown in the upper corner of the plot. c) Sorted mTOR hi cells form large homotypic aggregates when cultured in media containing IL-2 for 4hrs after sorting. d-h) Sorted cells were cultured in media containing IL-2 for 24hrs. d) mTOR hi cells exhibit a higher proliferative rate-as measured by 3 Hthymidine incorporation-than their mTOR lo counterparts. e) Sorted mTOR hi cells show robust cell cycle progression 24hrs post sorting, while sorted mTOR lo cells arrest in the G 0/1 stage. f) mTOR hi cells display a significantly higher Extracellular Acidification Rate (ECAR) and Oxygen Consumption Rate (OCR) than their sorted mTOR lo counterparts. The increased metabolic activity observed in the sorted mTOR hi cells is accompanied by a higher mitochondrial load than sorted mTOR lo cells as indicated by mitotracker green staining (g). Mitotracker green MFI shown in upper corner of FACs plot. h) Sorted mTOR lo CD4+ T cells exhibit a significantly higher Spare Respiratory Capacity (SRC) than sorted mTOR hi cells. i) Sorted mTOR hi cells exhibit a higher rate of cell death when cultured for 5 days in media supplemented with IL-2 than sorted mTOR lo cells. The bar graph to the right depicts % AnnexinV+7-AAD+ cells ± S.D. from 3 independent experiments. Three days after culture in media + IL-2, sorted mTOR lo cells show significantly higher levels of (j) Bcl-2 mRNA and (k) protein expression than sorted mTOR hi cells. mRNA transcript expression shows mean relative expression ± S.D. from 3 independent experiments, with the mTOR hi values set to 1. Bcl-2 protein expression was normalized to Actin and fold difference in expression is shown above the protein band. The data are representative of at least 3 independent experiments.  Fig.). After the sort, cells were cultured in media supplemented with 1ng/ml IL-2. Sorted mTOR hi cells rapidly form large homotypic aggregates in culture after 4 hours post sorting, which did not occur in sorted mTOR lo cultures (Fig. 3C). Twenty four hours after culture, sorted mTOR hi and mTOR lo cells were treated with tritiated ( 3 H-) thymidine to monitor proliferation and cultured for an additional 16 hours. Sorted mTOR hi CD4+ T cells exhibit a much higher rate of proliferation than sorted mTOR lo cells, as indicated by 3 H-thymidine uptake (Fig. 3D). Additionally, CD4+ mTOR hi cells display robust cell cycle progression relative to sorted mTOR lo cells, which arrest in G 0/1 following the sort (Fig. 3E). CFSE labeling of 5c.c7 splenocytes prior to stimulation confirmed that neither mTOR hi and mTOR lo sorted cellular populations have initiated division at the time of sorting, but mTOR hi cells have enhanced proliferation as indicated by CFSE dilution after 3 days of culture in IL-2 supplemented media (S2B Fig.).
mTOR plays an important role in the regulation of cellular metabolism [13,28]. As such, we next assessed the metabolic profile of the cells. Sorted mTOR hi cells are highly metabolically active, exhibiting a significantly higher Extracellular Acidification Rate (ECAR)-an indirect indicator of glycolysis-than their mTOR lo counterparts (Fig. 3F). In addition, the mTOR hi cells initially demonstrate a higher Oxygen Consumption rate (OCR) compared to mTOR lo cells (Fig. 3F). This higher rate of metabolic activity observed in the sorted mTOR hi cells is accompanied by an increase in mitochondria content (Fig. 3G). The higher ECAR observed in sorted mTOR hi CD4+ T cells is consistent with the role of mTOR in promoting glycolysis [28]. Further, the decreased ECAR observed in the small, mTOR lo cells phenocopies the metabolic profile of rapamycin treated CD4+ T cells (S3A Fig.). Thus, within a clonal population of T cells stimulated with their cognate antigen, those cells with high levels of mTORC1 activity demonstrate increased proliferation and metabolic activity when compared to those cells with relatively low mTORC1 activity.
We have shown that sorted mTOR lo CD4+ T cells exhibit a markedly decreased glycolytic metabolic profile than sorted mTOR hi cells. However, sorted mTOR lo cells possess increased Spare Respiratory Capacity (SRC) when compared to mTOR hi cells from the same culture (Fig. 3H). A similar finding was also observed for CD4+ T cell stimulated in the presence of rapamycin (S3B Fig.). SRC is calculated as the differential between a cell's basal rate of oxygen consumption and the maximal rate of oxygen consumption. SRC is a metric indicating how close a cell is currently functioning to its maximal metabolic potential. Rapidly proliferating cells, such as terminally differentiated effector T cells and tumor cells, display a very low SRC [29,30]. Conversely, long-lived cells, such as stem cells and memory T cells have been shown to have a large SRC [29]. This correlation between longevity and SRC is thought to be due to the need for long-lived cells to survive periods of metabolic stress, which would prove fatal to more terminally differentiated cell types. In light of the previously described association between SRC and longevity, we assayed the in vitro survival of sorted mTOR hi and mTOR lo CD4+ T cells. From a clonal population of activated T cells, the mTOR hi cells undergo increased apoptosis compared to sorted mTOR lo counterparts when cultured in media supplemented with IL-2 for 5 days after sorting (Fig. 3I). The increased survival observed in the mTOR lo cells was accompanied by significantly higher expression of the anti-apoptotic protein Bcl-2 at both a transcriptional (Fig. 3J) and protein (Fig. 3K) level.

mTOR lo cells express increased levels of CD25
While mTORC1 activity is reduced in the sorted small cells compared to big cell counterparts, Bcl-2 levels are highest in small, mTOR lo cells. Many factors have been shown to control Bcl-2 expression in CD4+ T cells, including cytokine receptor signaling [31]. To this end, we tested for expression of the high affinity IL-2 receptor CD25 in our sorted populations. Interestingly, CD25 expression at the transcript and protein level was higher on the mTOR lo cells (Fig. 4A  and 4B). To determine if this phenotype was graded across a spectrum of cell size, we analyzed CD69+ T cells from 4 distinctly sized populations: the 10% smallest (Q1: mTOR lo ), 10% largest (Q4: mTOR hi ) populations as previously assessed, in addition to the next 25% smallest (Q2) or largest (Q3) groups of cells (Fig. 4C). These studies revealed that CD25 protein and transcript levels, CD62L protein expression, and Bcl-2 transcript levels are inversely related to cell size and hence mTORC1 activity (Figs. 4D and 4E and S4A). Thus, from a metabolic and survival perspective, antigen recognition in the setting of strong mTORC1 activation generates cells with increased glycolytic flux that are destined to be short-lived. Alternatively, antigen recognition in the context of decreased mTORC1 activation leads to the generation of long-lived cells with increased SRC, increased Bcl-2 and increased CD25 (IL-2Ra) expression.

Short-lived, metabolically active, mTOR hi cells demonstrate an effector phenotype
Our data thus far demonstrate that for a population of clonal T cells, metabolic and survival potential is dictated by their level of mTOR activity after activation. Within this context, we hypothesized that the short-lived mTOR hi cells represent the activated cells destined to become effector cells. Consistent with this hypothesis, recently stimulated and sorted mTOR hi CD4+ T cells express significantly higher levels of IFN-gamma, IL-17a and IL-17f transcript than sorted mTOR lo cells (Fig. 5A). Additionally, sorted mTOR hi cells express much higher cell surface levels of the chemokine receptor CXCR3 (associated with pro-inflammatory effector cells) [32] than sorted mTOR lo cells (Fig. 5B). As a negative control, no difference was observed in the expression of the LFA-1 subunit CD11a.
Upon antigen recognition, mTOR lo cells demonstrate metabolic programs consistent with long-lived cells. However, we have previously demonstrated that TCR engagement in the absence of mTOR induces the generation of Foxp3+ regulatory T cells [14]. Furthermore, others have shown that T cell activation in the presence of the mTOR inhibitor rapamycin promotes the generation of Foxp3+ regulatory T cells [15][16][17][18]. Therefore, we wondered if a sub-population of the mTOR lo cells might be Foxp3+ regulatory T cells. Consistent with this hypothesis, we observed that the mTOR lo population was enriched for Foxp3+ T cells ( Fig. 5C and 5D). That is, under normally activating culture conditions (antigen activation followed by IL-2 addition), 9.4% of the mTOR lo cells express Foxp3, while only 0.16% of the mTOR hi cells expressed Foxp3 (Fig. 5D). This difference was also observed and was amplified when TGF-ß was added to the cultures (22.8% versus 4.3%). Foxp3 expression is upregulated in every division of activated mTOR lo cells after culture in IL-2, and a graded decrease in Foxp3 expression is observed in each population of increased cell size/mTORC1 activity (S4B Fig.). The preferential induction of Foxp3 expression in sorted mTOR lo cells is not limited to cultures derived from homogeneous Rag -/-TCR transgenic splenocytes, but can also be observed in cultures generated from anti-CD3 stimulated polyclonal WT CD4+ T cells (S5A Fig.). Not unexpectedly, the mTOR lo CD4+ T cells cultured in IL-2 alone were better able to suppress the proliferation of activated CD4+ T cells compared to their cultured mTOR hi counterparts (Fig. 5E). This enhanced suppression was most likely due to the increase of Foxp3+ T cells generated from the mTOR lo sorted population compared to the mTOR hi cells (S5B Fig.). The ability of the mTOR lo T cells to suppress responder T cells was markedly diminished when the responders and suppressors were separated in a transwell, indicating that the suppression is primarily contact dependent (Fig. 5E). A question remained if Foxp3+ T cells were generated de novo from the activated mTOR lo cells, or if the enhanced generation of Foxp3+ T cells observed in the sorted mTOR lo populations was simply a consequence of sorting on a high frequency of already established natural Tregs. In order to answer this question, we utilized mice which express green fluorescent protein (GFP) under the control of the Foxp3 promoter (Foxp3 GFP+ ) [33]. Splenocytes from WT Foxp3 GFP+ mice were stimulated with 0.1ug/ml anti-CD3 for 20 hrs. Activated (CD69+) CD4+ GFP-cells were sorted from the 10% biggest 'mTOR hi ' and smallest 'mTOR lo ' populations (S6A Fig.). This approach ensured that all GFP+ natural Tregs would be removed from our sorted populations. Phenotypic assessment of sorted mTOR hi and mTOR lo populations 72 hrs after culture in IL-2 revealed that suppressive Foxp3+ T cells were generated de novo from the in vitro stimulated splenocytes (S6B-C Fig.). These observations were consistent with our previous results demonstrating that mTOR lo CD4+ T cells have a greater propensity to differentiate into Foxp3+ T cells (Fig. 5C-5E).
Our data thus far suggests that under normally activating conditions, the recognition of antigen in the context of low mTOR activation promotes the de novo generation of regulatory T cells. Nonetheless, we could still detect a small percentage of Foxp3+ T cells which displayed high mTOR activity. Therefore, we sought to further phenotype the Foxp3+ T cells generated from the sorted mTOR lo and mTOR hi populations. First, Foxp3+ T cells derived from either mTOR lo and mTOR hi populations have higher expression of the known T-regulatory cell markers, CD25, CTLA-4, and GITR when compared to Foxp3 negative counterparts ( Fig. 6A  and 6B). Interestingly, Foxp3+ mTOR lo T cells have enhanced expression of CD25 and GITR compared to mTOR hi derived Foxp3+ cells (Fig. 6B). Alternatively, Foxp3+ mTOR hi T cells have enhanced expression of CTLA-4 compared to Foxp3+ cells derived from the mTOR lo population (Figs. 6B and S5). Thus, while the generation of Foxp3+ T cells is clearly enriched in the mTOR lo population of antigen-activated T cells, there also exists a small but distinct population of Foxp3+ CTLA-4 hi T cells which emerges from the mTOR hi population of antigen activated cells.

Divergent phenotypic properties of mTOR hi and mTOR lo Foxp3+ regulatory T cells
Our data are consistent with previous findings that diminished mTOR activity in T cells promotes the generation of regulatory T cells [14][15][16][17][18]. However, we also find a small, but distinct population of Foxp3+ T cells generated from mTOR hi cells. Interestingly, one of the most common techniques to generate regulatory T cells in vitro is to induce robust activation in the presence of high concentrations of TGF-ß [34]. In fact, stimulation under these conditions results in robust mTOR activation (Figs. 7A and S7). That is, while strong TCR engagement in the presence of TGF-ß promotes the generation of Foxp3+ T cells, it also results in robust mTORC1 activation. Given the highly divergent metabolic profiles and survival potentials of sorted mTOR lo and mTOR hi conventional CD4+ T cells, we wanted to determine if Foxp3+ CD4+ T cells generated in the presence of high or low mTOR activity during TCR stimulation demonstrate similar metabolic and survival potentials. To this end, WT splenocytes were represent ± S.D, n = 3 (a) or n = 6 (b). c-d) 5c.c7 Rag -/splenocytes were stimulated with PCC for 24hrs and activated (CD69+) CD4+ cells were sorted based on 4 size profiles. Cells were sorted into 10% smallest (Q1) and largest (Q4) populations in addition to the next 25% smallest (Q2) and largest (Q3) populations. c) Representation of sorting scheme utilized. d) After the sort, cells were cultured separately in IL-2 supplemented media for 3 days. On day 3, expression of CD25 and CD62L was assessed by flow cytometry. An illustration of cell size after sort is depicted above flow plots for clarification of sorted quartile populations. e) Purified CD4+ T cells from 5c.c7 Rag -/mice were stimulated by plate bound anti-CD3 and anti-CD28 for 24hrs. Cells were sorted as indicated in c, and mRNA transcript expression of CD25 and Bcl-2 was determined from sorted populations after a 3 day culture in IL-2 supplemented media. Bar graphs show relative expression ± S.D. from 3 independent experiments. The data are representative of 3 independent experiments or show cumulative results.  stimulated with 1ug/ml anti-CD3 and 1ng/ml IL-2 in the presence or absence of either 10ng/ ml TGF-ß or 500nM rapamycin. Anti-CD3 was washed out at 24 hrs, and the cultures were expanded in media supplemented with IL-2 alone, IL-2 + TGF-ß, or IL-2 + rapamycin for 72 hrs before harvest for flow-cytometric analysis. As expected, the Foxp3+ T cells from the rapamycin treated cultures had reduced mTORC1 activity (mTOR lo ) compared to the Foxp3+ T cells generated in the presence of TGF-ß (Fig. 7A). As was the case for the mTOR hi CD4+ sorted T cell populations, CD4+ T cells cultured with TGF-ß demonstrate enhanced ECAR compared to the cells activated in the rapamycin treated conditions (Fig. 7B), but reduced expression of Bcl-2 (Fig. 7C). These data suggest that culture in TGF-ß promotes an 'effector' phenotype with increased glycolysis but diminished survival. Activation of T cells in the presence of rapamycin or TGF-ß enhanced Foxp3+ T reg generation compared to untreated, activated cells (Fig. 7D), and we further assessed the phenotype of Foxp3+ cells generated from these culture conditions (Fig. 7E). Notably, phenotypic analysis of the Foxp3+ cells generated from the rapamycin versus TGF-ß culture conditions recapitulated the results observed from the sorted mTOR lo and mTOR hi Foxp3+ populations. Foxp3+ T cells generated in the presence of rapamycin have enhanced expression of CD25, GITR, CD71, and Bcl-2 compared to Foxp3+ T cells derived from TGF-ß cultures (Fig. 7E). Alternatively, Foxp3+ cells generated in the presence of TGF-ß have enhanced CTLA4 and CD39 expression compared to those generated in rapamycin (Fig. 7E). These data suggest that regulatory T cells generated in the presence of robust mTOR activity (by culture in high concentrations of TGF-ß) have differential survival potential and are metabolically distinct from those generated in the setting of low mTOR activity.
Differential fates of mTOR hi and mTOR lo human CD4+ T cells Next, we wanted to determine if we could use cell size/mTOR activity to track the fate of TCRactivated human T cells. CD4+ T cells from fresh PBMCs were isolated and the T cells were activated with human anti-CD3/anti-CD28 activator beads. Similar to our observations set to 1. d) Foxp3 protein expression was determined from sorted mTOR lo and mTOR hi populations 3 days after culture in IL-2 alone or IL-2 + TGF-ß. Bar graphs to the right depict % Foxp3+ cells ± S.D. generated in each population from 3 independent experiments. e) mTOR lo Foxp3+ regulatory T cells possess the ability to suppress the proliferation of naïve CD4+ T cells. Sorted mTOR hi or mTOR lo populations were cultured in media supplemented with IL-2 for 3 days prior to use in the suppression assay. The percentage of non-divided responders was determined by CFSE dilution 72hrs after culture of 1:2 suppressors: responders in direct contact or in transwell suppression assays. The graphs to the right depict the percentage of non-divided responders after culture with mTOR hi or mTOR lo 'suppressors' from 3 independent experiments. Connecting lines show values from sorted populations from the same experiment. The data are representative of at least 3 independent experiments or show cumulative results.
doi:10.1371/journal.pone.0121710.g005 employing murine T cells, mTORC1 activation correlates with cellular size in activated human CD4+ T cells (Fig. 8A). Upon sorting and expanding activated mTOR hi and mTOR lo T cells for 72 hrs in IL-2, mTOR lo T cells have an increase in percentage of CD127 lo CD25 hi CD4+ T cells compared to mTOR hi counterparts (Fig. 8B). Furthermore, from the CD127 lo CD25 hi population, there is an increase in Foxp3+ CTLA4+ T cells detected in the mTOR lo T cell cultures (Fig. 8C). Thus, similar to murine T cells, TCR-stimulation of human CD4+ T cells in the presence of low mTORC1 activity enhances T reg cell differentiation. Rag -/-TCR transgenic mice (a) or WT C57BL/6 mice (b-e) were stimulated with 1ug/ml soluble anti-CD3 and 1ng/ml IL-2 in the presence or absence of either 10ng/ml TGF-ß or 500nM rapamycin for 24hrs. a) mTORC1 activity was measured by flow cytometric analysis of phosphorylated-S6 expression 24hrs after stimulation. Histograms were gated from activated CD4+CD69+ populations within each culture condition. b-e) After 24hrs of stimulation, cells were expanded in media supplemented with IL-2 alone, IL-2 + TGF-ß, or IL-2 + rapamycin for an additional 72hrs before phenotypic assessment. b) Cells were run on an extracellular flux analyzer and the extra cellular acidification rate, ECAR, was determined over time.

Discussion
We have observed that within a clonal population of newly activated T cells there exists significant heterogeneity of mTORC1 activity. This heterogeneity leads to cells with differential metabolic programs, survival potential and ultimately phenotypic fate. Our data demonstrate that cell size can be utilized as a surrogate indicator of mTOR activity in recently stimulated CD4+ T cells. This observation provides a simple tool allowing for the isolation of T cell populations with distinct metabolic, proliferative, immunologic and survival potential from a heterogeneous population of activated cells.
We demonstrate a direct correlation between the level of mTOR activity upon antigen recognition by CD4+ T cells and the regulation of metabolism and survival. In part, this appears to be mediated by differential expression of anti-apoptotic members of the Bcl-2 protein family facilitated by expression of pro-survival cytokine receptor CD25. While mTOR activity is strongly induced by many pro-survival growth factors such as IL-2 [11], long term cellular survival necessitates the adoption of a quiescent metabolic and proliferative phenotype [13,35]. For example, T cells lacking the ability to enter a state of quiescence due to genetic deletion of the mTOR inhibitor TSC1 are incapable of long term survival in vivo [36]. A unique metabolic feature of long-lived stem cells and memory T cells is the presence of an increased SRC [29]. Indeed, our data demonstrate that the mTOR lo T cells possess both a greater SRC (Fig. 3H), and survival potential than mTOR hi T cells (Figs. 3I-3K). In this context, it is interesting to mTOR Activity as a Means of Tracking T Cell Fate note that decreased mTOR activity has not only been associated with increased cellular survival but also with increased longevity at the level of the whole organism [37].
We postulate that there are many factors that contribute to the heterogeneous activation of mTORC1 upon antigen recognition. For example, TCR signal strength has been shown to be important for the polarization of effector lineages in CD4+ and CD8+ T cells [38][39][40][41]. While TCR avidity was the same in our clonal population of 5c.c7 T cells, the observed modulation in mTORC1 signaling may be due to competition for peptide on the APCs, the relative activation status of a particular APC, the relative upregulation and encounter with inhibitory ligands and even the local exposure to various activating and inhibitory cytokines. Indeed, mTOR is activated by and thus integrates multiple inputs from the immune microenvironment, which in turn influence the fate of an individual cell's encounter with antigen [42].
Previous studies have demonstrated the differences in metabolic function between effector and memory CD8+ T cells [29,43,44]. That is, effector CD8+ T cells are shown to be highly glycolytic while memory CD8+ T cells are less glycolytic and demonstrate increased SRC. Likewise, several studies have demonstrated the ability of the mTOR inhibitor rapamycin to promote the generation of CD8+ memory T cells [21,23]. Our current studies employing CD4+ T cells are supported by these previous observations. Our ability to track cells based on the level of mTOR activation has enabled us to link mTOR activity with the metabolic and survival programs necessary to support effector versus memory cells. Thus, we speculate that this system could also distinguish distinct CD8+ T cell populations.
The pharmacologic inhibition or the genetic deletion of mTOR leads to the generation of Foxp3+ regulatory T cells under otherwise normally activating conditions [14][15][16][17][18]. Consistent with these reports, we observed that under activating conditions, Foxp3+ T cells were enriched within the population of mTOR lo CD4 cells. Furthermore, these data are consistent with previous works demonstrating that low doses of peptide can promote Foxp3+ T regulatory cell generation even in the absence of exogenous TGF-ß [45]. Indeed, we have observed that peptide concentrations that promote T regulatory cell generation fail to induce robust mTOR activation (data not shown). Alternatively, a common technique for generating Foxp3+ regulatory T cells in vitro is to robustly stimulate T cells with either high doses of peptide or anti-CD3 in the presence of exogenous TGF-ß [46]. Under such conditions we observe high mTOR activity (Figs. 7A and S7). T cells generated in TGF-ß culture conditions have differential metabolic and survival potentials than T cells generated in the presence of rapamycin (Fig. 7). Furthermore, Foxp3+ T cells generated from rapamycin versus TGF-ß cultures have distinct phenotypes (Fig. 7E). These data lead us to propose that Foxp3+ regulatory T cell fate/ function may be determined by level of mTOR activity. We suggest that Foxp3+ mTOR hi cells resemble short-lived "effector regulatory" T cells characterized by increased glycolysis, increased CTLA-4 and CD39 expression, and decreased Bcl-2. Alternatively, we propose that mTOR lo Foxp3+ cells behave more like long-lived "memory regulatory" cells characterized by decreased glycolytic flux, but enhanced CD25, GITR and Bcl-2 expression. Such a model is consistent with recently published reports describing effector and central regulatory T cell populations [47]. Furthermore, this distinction might help reconcile seemingly conflicting reports in the literature regarding the necessity of mTOR inhibition versus mTOR activation in T regulatory cell [48,49]. Interestingly, the ability to distinguish subsets of regulatory T cells based on size/ mTOR activity was recapitulated in human cells CD4+ T cells as well (Fig. 8).
Overall, our ability to track the fate of mTOR hi and mTOR lo cells following the activation of naïve T cells reveals a new mTOR/metabolism-centric model of T cell activation [42]. In this model, within a clonal population of T cells, the level of mTOR activation, in part through the upregulation of selective metabolic programs, helps to determine the effector versus memory fate of T cells upon antigen recognition. We believe that this new perspective helps to better define the fate and function of memory CD4+ T cells. Likewise, our model posits the existence of distinct subsets of regulatory T cells.
Supporting Information S1 Fig. mTORC1 activity correlates with cellular size upon TCR stimulation in a time dependent manner. Splenocytes from a 5c.c7 Rag -/-TCR transgenic mouse were stimulated with 0.5ug/ml PCC peptide. A time course assessment of forward scatter (FSC) and mTORC1 activity, indicated by phosphorylated levels of S6 Kinase or ribosomal S6, was measured by flow cytometry. Plots were gated from CD4+ T cells. The data are representative of two independent experiments. (TIFF) S2 Fig. mTOR lo and mTOR hi CD4+ T cells display no significant difference in cell cycle stage at the time of isolation. a) Splenocytes from 5c.c7 Rag -/mice were stimulated with 0.5ug/ml PCC peptide for 24hrs and then sorted into CD4+CD69+ populations and further separated into FSC/SSC "big" mTOR hi or FSC/SSC "small" mTOR lo populations. Cells were immediately fixed, and DNA content was determined by PI staining. mTOR hi and mTOR lo CD4+ T cells exhibited no significant difference in cell cycle stage at this time post stimulation. b) CFSE labeled 5c.c7 Rag -/splenocytes were stimulated with PCC for 24hrs and activated (CD69+) CD4+ cells were sorted based on 4 size profiles as in Fig. 4C. An illustration of cell size after sort is depicted above the flow plots for clarification of the sorted quartile populations. Top panels depict CFSE vs FSC of each of the 4 sorted populations immediately following the sort. Gates show smallest and largest populations based on Quartiles 1&4 immediately following the sort. Bottom panels depict CFSE vs FSC for each population after culture in IL-2 supplemented media for 3 days. Gate shows percentage of cells with highest CFSE expression. The data are representative of 3 independent experiments. (TIFF) S3 Fig. Rapamycin treated CD4+T cells exhibit a lower Extracellular Acidification Rate (ECAR) but higher Spare Respiratory Capacity (SRC) than untreated controls. 5c.c7 Rag -/-CD4+ T cells stimulated with PCC peptide and treated with 500nM rapamycin exhibit a lower ECAR (a), but higher SRC (b) than untreated controls after 48hrs of stimulation. (TIFF) S4 Fig. Activated mTOR lo cells have a reduced proliferative capacity compared to mTOR hi cells but can generate Foxp3+ cells in any division. a-b) CFSE labeled 5c.c7 Rag -/splenocytes were stimulated with PCC for 24hrs and activated (CD69+) CD4+ cells were sorted based on 4 size profiles as in Fig. 4C. An illustration of cell size after sort is depicted above flow plots for clarification of sorted quartile populations. After the sort, cells were cultured in IL-2 supplemented media for 3 days, and a) CD25, or b) Foxp3 protein levels were detected by surface or intracellular staining and plotted against CFSE dilution. Gates were determined based on the isotype control staining (left panels). The data are representative of 3 independent experiments. (TIFF) S5 Fig. Sorted mTOR lo CD4+ T cells from non-TCR transgenic mice preferentially develop into Foxp3+ regulatory cells. a-b) Splenocytes from a C57BL/6 mouse were stimulated with 1ug/ml anti-CD3 for 24 hrs before being sorted into CD4+CD69+ FSC/SSC "big" mTOR hi and "small" mTOR lo populations. Sorted cells were cultured in media supplemented with IL-2 for 4 days, and then analyzed for CD4 and Foxp3 expression by flow cytometry. b) The FACs plots show the percentage of Foxp3+ cells obtained from the sorted mTOR hi and mTOR lo 'suppressor' populations used in the suppression assay depicted in Fig. 5E. C) The histograms depict the CTLA-4 expression of the Foxp3+ or Foxp3-populations gated in b. The CTLA-4 MFI is shown in the upper corner of the FACs plot. The data are representative of 3 independent experiments. (TIFF) S6 Fig. Foxp3+ T cells are de novo generated upon TCR stimulation in conditions of low mTORC1 activation. a-d) Splenocytes from WT Foxp3GFP+ mice were stimulated with 0.1ug/ml anti-CD3 for 20 hrs, and then sorted into CD4+CD69+GFP negative FSC/SSC "big" mTOR hi and "small" mTOR lo populations. a) A schematic of the sorting strategy used. b-d) Sorted cells were cultured in media supplemented with IL-2 for 3 days. b) Foxp3+ expression was determined by flow cytometry. c) FACs plots show the dilution of eFluor670 labeled naive CD4+ 'responder' cells after 72hrs of stimulation in co-culture (2:1 responder: suppressor) with mTOR hi or mTOR lo cells. Splenocytes from a 5c.c7 Rag -/mouse were stimulated with anti-CD3 and anti-CD28 for 4hrs in the presence or absence of 5ng/ml TGF-ß or 500nM rapamycin. Cells were then lysed, and mTOR activity was determined by western blot. (TIFF)